Project Erinyes [Archive] - RRTech...The SOURCE for Extreme Liquid Cooling

rubidium

12-08-2008, 07:34 AM

Just out of curiousity, what kind of mobo do you use to build something like this, and what (if any) programs can take full advantage of the 16 cores?

The mobo is the recent Tyan S4989 Quad Thunder (6550EX), using the nVidia NPF3600+NPF3050 chipset. It accomodates 4 1207/Socket-F1 processors and 32 DDR2 DIMMs (supporting 128GB max). In terms of I/O, there are 2 PCI-X 133MHz slots, 2 8x PCIe slots, and 2 16x PCIe slots that support SLI. I'm ultimately targeting Shanghai processors sometime next year, but have obtained 4 new OEM 8346HE Barcelonas for $50 apiece to get me started now. They may only be 1.8GHz processors, but for less than the cost of a couple of PA120.3's I can begin to navigate the overclocking potential of this chipset and the 3rd gen Opterons. Later getting something like 2.7GHz Shanghais will give a serious performance increase.

This is a workstation intended for scientific computing at home. To see 4 sockets in the first place with Windows, one has to go beyond XP or Vista and into one of the enterprise server OS's, like Server 2008 Enterprise. XP and Vista have been emasculated by Microsoft :rant: to only show 2 physical processors (and any number of cores). I'm not that much in love with Microsoft to want to drop several grand on their server crap, so I'll be going with Ubuntu. Ubuntu sees all 16 cores and 64GB of RAM that I have right out of the box - and for free. :up: In terms of programming, one package that I use a lot for digital signal processing and image processing is Matlab, which supports multiple cores when evaluating certain key computations like the fast Fourier transform (FFT). I also want to participate in F@H. On a seperate note, I'm also including a GTX-280 in this build (and possibly 2) to develop some CUDA-based software that takes advantage of the large degree of parallelism of these GPUs.

rubidium

rubidium

12-08-2008, 07:37 AM

a very big extended motherboard. if you've seen his worklog over there, soon to be posted here...

let us just wait for the pictures to come.

Yea. The board is 16"x13" and makes the MM Ascension look cramped.:D

Pictures of the build won't get posted here for a while. Sorry, but I'll be leaving town soon and returning last week in December.

Kayin

12-08-2008, 07:46 AM

The big question is wil ya let us map our block design flow on it?

Good to see an AMD build. ESPECIALLY one this daring.

rubidium

12-08-2008, 08:07 AM

The big question is wil ya let us map our block design flow on it?

Good to see an AMD build. ESPECIALLY one this daring.

Sure, I'm willing to generate some heat ... What did you have in mind? Tell me what software you're running and the scale of the problem.

You'll have to wait until some time in January so I can get a couple of RMA issues resolved before I can claim to have a solid working rig. The system *is* working and has sustained stresses like 16 concurrent instances of Prime95 for 48 continuous hours without cutting a sweat (these are HE processors), but there are some issues. My PSU is showing problems with cold startups (hardware and OS-initiated restarts are no problem), and I discovered a bad DIMM slot (in a strategic place for my fill of RAM) on the board the other day. Both are unsat at these prices, and both are going back for replacement once I return from vacation the last week in December.

rubidium

Septim

12-09-2008, 06:19 AM

cool, you should leave us a parting pictures of where your rig is at, am sure this will generate more interest.

RedRaider

12-09-2008, 06:29 AM

I anxiously await your return...:D

rubidium

12-09-2008, 02:16 PM

cool, you should leave us a parting pictures of where your rig is at, am sure this will generate more interest.

OK, here's a photo of the rig as it stands today:
http://farm4.static.flickr.com/3226/3095516897_79df994848_o.jpg
showing how the motherboard is eating up the real estate in a MM Ascension! :rofl: 16-cores, 64GB RAM, GTX-280 (currently air cooled, but I'll be going to water next). More details on the entire build process when I return.

Oh, and I posted my procedure for constructing water temperature sensors, based on the Dallas DS18B20 one-wire, in a seperate thread to leave you all with a little something for the holidays.:topofmorn:

rubidium

Xilikon

12-09-2008, 02:22 PM

For a cruncher, it's a real drooling setup to see :D

bigslappy

12-09-2008, 02:22 PM

I see V2's in there .... WoW he's a real asset to the forum!!

smee

12-09-2008, 02:28 PM

For a cruncher, it's a real drooling setup to see :D

Yea wow thats an Awesome setup man!!!
Thats...... i mean... just.... :omg:

Hey btw, are we going to have a Folding team for the forum??
I'd fold. :)

bigslappy

12-09-2008, 02:44 PM

Yea wow thats an Awesome setup man!!!
Thats...... i mean... just.... :omg:

Hey btw, are we going to have a Folding team for the forum??
I'd fold. :)

YUP ! need someone (alot of PPL) who has experience in that field , i suggest a few of you start a Social Group to do some head banging to get it up & going NO NEED to start a forum section till we're up to speed .. we want to start that off right the 1st time

we can make a few teams look like idiots right off if ya do it right & hit them like a wave

Xilikon

12-09-2008, 02:58 PM

YUP ! need someone (alot of PPL) who has experience in that field , i suggest a few of you start a Social Group to do some head banging to get it up & going NO NEED to start a forum section till we're up to speed .. we want to start that off right the 1st time

we can make a few teams look like idiots right off if ya do it right & hit them like a wave

Unless you try to kick the [H]orde off the pedestral :hide:

I'm definitely a heavy hitter (69th worldwide in F@H) but I have loyalty to show :D

bigslappy

12-09-2008, 03:02 PM

Xilikon

12-09-2008, 03:08 PM

that's cool if yer involved on another team , im not asking u to jump ship

by just having are own team here it puts our forum out there for everyone to see ... just we need to hit it all at the same time under the RRR banner

It is a great idea for those who are looking for a team. I won't come here to pick people and there's plenty of great setups around to push some nice output. All I do care is to have more people folding for a cure :multijump:

I'm even willing to give some help for the new users and link guides that I wrote (it's the same ones as the official Stanford guides since I'm the author) if a new sub-forum is there.

breezyjr

12-09-2008, 09:01 PM

I'd like to know what problems you're having with the PSU... since we both share the same one. If there's a problem, I'd like to be ready.

No, I haven't powered mine up yet. waiting for my custom rear Ascension panel.

enjoy the time off.... at least I hope it's time off.....

RedRaider

12-15-2008, 07:53 AM

Rubidium.... Post us junkies some more pics of this insane build...:king:

This build gets RedRaider's approval rating is pending until more pics are posted...:shake:

rubidium

01-03-2009, 09:38 AM

OK, now that I'm back from vacation, it's time to walk through the Project Erinyes build from the top.

My 6-7 years of experience with building watercooled rigs began to focus on multiple socket motherboards several years ago, when I built my first high-performence workstation based upon on the Tyan S2895 dual-socket board with Opteron-280's and 16GB of DDR1. From that build, which has essentially been running 24/7 with a 20% overclock without a hitch through to this very day, I learned a lot about watercooling and overclocking "non-enthusiast" boards, which in conventional circles are generally not targeted for watercooling nor overclocking.

Back in October of last year, I became more adventurous, and decided to take on a 4-socket board and build a 16-core juggernaut for scientific computing applications. So we begin with the Tyan S4989, which accomodates 4 AMD socket-F1 (1207 pin) family processors, 32 DDR2 slots for up to 128GB of RAM, 2 16x PCIe slots with SLI support, 2 8x PCIe slots, a single 133MHz PCI-X slot, and a host of other goodies. The chipset is the nVidia 3600/3050 pair.

This board is huge: 16.2" x 13", and really eats up the available real estate in a MountainMods Ascension case:
http://farm4.static.flickr.com/3121/3162586615_3ccf37979a_o.jpg

It is also poised to eat up power, with the normal 24-pin ATX connector supplemented by 2 8-pin and 1 4-pin 12-volt lines:
http://farm4.static.flickr.com/3124/3163418024_4345fe4187_o.jpg

The size of this board warranted a custom motherboard tray and rear panel from MM, and the density of the parts layout presented several challenges from a watercooling perspective, which I'll get into next.

rubidium

rubidium

01-03-2009, 09:59 AM

When I originally placed the order for the MM Ascension, the S4989 had just entered the market and I was in the backorder queue. Without a board in my posession to take measurements from, I had to go through the lengthy, but eventually successful, ordeal of talking the Tyan engineers in Taiwan into sending me some AutoCAD files of the board. With that, I was able to get Ben at MM the proper coordinates for all mounting holes, and convince myself that everything would align properly in the Ascension. The photo below was taken at the MM fabrication site, showing my "bastard child" tray at the lower-left corner of the pallet amid all the other "normal" trays:
http://farm4.static.flickr.com/3095/3163555398_dd2277273f_o.jpg

rubidium

rubidium

01-03-2009, 10:34 AM

Based on a variety of calculations and components trades, I elected to watercool the four processors via two independent cooling circuits, each involving a PA120.3 radiator, an MCP-355 pump with EK-DDC top, and two D-Tek Fuzion v2's.

As a general rule, I always like to sketch up key subcomponents of the build to ensure that various parts will properly fit together - or at least to prepare me for the mods that I'll have to make in order to get them to properly fit together. In the case of the CPU cooling blocks, the close proximity of the RAM slots to the CPU sockets on the S4989 caused interference with the Fuzion brackets:

http://farm4.static.flickr.com/3084/3163418058_4a0a1ca861_o.jpg

This required me to plan on removing the extremities of the brackets:

http://farm4.static.flickr.com/3108/3163418098_0b48e2c6e6_o.jpg

in order for everything to fit properly:

http://farm4.static.flickr.com/3079/3162586545_34bbd79881_o.jpg

The brackets were apparently case-hardened, and required a water-cooled diamond cutoff wheel to make the cuts clean and accurate:

http://farm4.static.flickr.com/3107/3162586711_c3aec479ea_o.jpg

The Fuzion v2 mounting brackets are configured to essentially pull on the waterblock from below, rather than push the block from above. The bracket surfaces at which spring pressures are applied are elevated above the points attached to the bottom of the block, thus creating a considerable lever arm that appears to result in some lateral “splaying” at the spring pressure points. The effect may be minor, but since I could observe it with even light spring pressure, I decided to eliminate this splaying by bridging the two upper spring pressure surfaces. I used hinged 6-32 nickel-plated brass standoffs and 6-32 stainless studs to form the bridges, and attached them through existing holes in the brackets with 6-32 stainless cap screws - everything from McMaster-Carr. The brackets are rock-solid now, so that the spring pressure always remains truly perpendicular to the plane of the motherboard. No compromises!

http://farm4.static.flickr.com/3113/3162586775_42f64a2983_o.jpg

http://farm4.static.flickr.com/3110/3162586747_fd88ccdb43_o.jpg

http://farm4.static.flickr.com/3118/3163418260_592d1fcd18_o.jpg

The backing plates that came with the Tyan motherboard are really nice, rigid, hardened steel pieces that I definitely wanted to keep. Unfortunately, the supporting hardware that came with the D-Tek brackets involved 4-40 phillips-head screws and thumbnuts, while the Tyan backing plates are set up for 6-32 studs. So I ordered some stainless 6-32 studs and knurled thumbnuts from McMaster-Carr, and I’m using the original D-Tek springs along with some broad flat stainless washers I had lying around. I also used Bitspower 3/8” ID / 1/2” OD G1/4 compression fittings. The lower knurled rings on these fittings barely clear each other – and I mean barely – but barely is good enough. (A warning to D-Tek Fuzion fans wanting to use 1/2” ID compression fittings. They won’t fit.) Here is the finished block with all mounting hardware:

http://farm4.static.flickr.com/3264/3162586893_3c3195cbde_o.jpg

http://farm4.static.flickr.com/3116/3163418230_12c3f4f9e9_o.jpg

http://farm4.static.flickr.com/3209/3162586913_6b974a4d28_o.jpg

Here are some shots of the blocks installed on the S4989:

http://farm4.static.flickr.com/3125/3163607831_b3c06479e5_o.jpg

http://farm2.static.flickr.com/1077/3164449938_66439eacc8_o.jpg

http://farm4.static.flickr.com/3257/3164450578_d264bc5d96_o.jpg

rubidium

lyl

01-03-2009, 11:50 AM

Cool mounting hardware...

That mobo makes it seem like theres tons of space wasted on regular ATX :haha:

Eddie3dfx

01-03-2009, 11:56 AM

Any luck on the Rma?
God I hate rmas.. It's worse than waiting for sex!

rubidium

01-03-2009, 12:42 PM

Any luck on the Rma?
God I hate rmas.. It's worse than waiting for sex!

I've been on vacation, so there's been no RMA action up until yesterday. There are two issues that need to be resolved. The first and most serious is the difficulty in POSTing from a cold start. Once the system has properly POSTed (after about 10 on/off cycles of the AC power switch), it boots into an OS flawlessly, will run indefinitely under full load, and will re-POST/re-boot without problems from a hardware or software restart. We (Tyan and I) believe that we have traced this problem to the FUFUFUFUFUFUFUFU PSU-1300ATX-12N, and so I have obtained an RMA with FUFUFUFUFUFUFUFU and the PSU was just shipped back yesterday. [Coincidentally, note that the FUFUFUFUFUFUFUFU web site began indicating "Revision 1.5 of the FUFUFUFUFUFUFUFU PSU-1300ATX-12N has improved motherboard compatibility and no side windows." in late November, so I think my experience might be related to some bug in the startup circuit of the original edition of the PSU.]

Once I get a working PSU back and hopefully put the cold start problem to rest, I'll work out the second and less serious issue: the motherboard itself appears to have a faulty RAM slot. It's not a show stopper in terms of running the rig, but I do want it solved.

rubidium

rubidium

01-03-2009, 04:02 PM

Speaking of PSU's, one of my objectives in this build right up there next to performance was quiet. My S2895 uses a PC Power & Cooling PSU, which has been running solidly for years, but sounds like a fighter jet in afterburner. The rat's nest wiring is also difficult to manage.

In this build, I went with the watercooled FUFUFUFUFUFUFUFU PSU-1300ATX-12N:

http://farm4.static.flickr.com/3093/3163629843_86cf74e181_o.jpg

It supports modular wiring, which makes for easler cable routing:

http://farm2.static.flickr.com/1007/3163629903_febe986ea5_o.jpg

One attribute that I would not have personally put into the design is the rear porting for the cooling water:

http://farm4.static.flickr.com/3102/3164465308_5c897a40c7_o.jpg

http://farm2.static.flickr.com/1195/3163629937_eef1c0454c_o.jpg

Since I'm dedicating a 3rd cooling loop (PA120.3, MCP-355 with EK-DDC top) to this PSU, I would have preferred the ports to remain inside the case. So be it. Given the 16lb weight of this supply, I have it located in a lower rear corner of the case. After lots of trial and error paper exercises, I finally came up with a combination of Bitspower fittings (and a small fortune to pay for them) to support a scheme of getting the cooling lines into the case that is both secure mechanically and looks good:

http://farm2.static.flickr.com/1048/3163629963_19a2e77093_o.jpg

http://farm2.static.flickr.com/1172/3163630003_78542dba37_o.jpg

http://farm2.static.flickr.com/1367/3163630047_8ebd4eaaa3_o.jpg

rubidium

Baleful

01-03-2009, 10:40 PM

Absolutely sick. Words don't even describe what you are doing!!!! Awesome job man.

Kayin

01-03-2009, 10:43 PM

Fuck a duck. That's badass.

Langer

01-03-2009, 10:45 PM

Christ that's cool.

bianco

01-04-2009, 01:18 AM

nice to see your project here rubidium.
looking forward to the completion of your project and also showcase your pump mount! :D

Septim

01-04-2009, 02:54 AM

this is one bad ass build you're doing... nice mounts...
i hope you complete it soon... good luck on RMAs...
i imagine the crunching power/folding power of this beast should be untouchable...

rubidium

01-04-2009, 07:43 AM

To the pumps ...

There are three independent cooling loops, each containing a PA120.3, MCP-355 pump with EK-DDC G-1/4 top, and a Bitspower 5-1/4 bay-mounted reservoir. Two of the loops each serve pairs of CPU's, while the third loop serves the liquid-cooled power supply. Later, I plan on watercooling a GTX-280 and possibly two of them.

I thought a lot about mounting schemes for the three MCP355 pumps. I know there are off-the-shelf mounts out there, but I wanted something different. I also wanted to beat a personal nemesis of mine: rat’s nest wiring. Much of the wiring disappointments I’ve had in previous builds stem from those anachronistic 4-pin Molex connectors that are both clunky and not very friendly when it comes to powering multiple items in close proximity to each other. I decided to make three individual “pump stations” using some nice looking 0.25” translucent red acrylic, that I originally drew up to look like this:

http://farm4.static.flickr.com/3263/3163632299_345feafd4f_o.jpg

I’m using aluminum standoffs as legs at the four corners, and elsewhere as binding posts for wire connections. An MCP355 mounts to the acrylic base plate via adhesive-backed 0.325” neoprene sheet that’s great for vibration absorption. I also modified the pumps to allow the option to run them as full-up 355’s, or more quietly at reduced power as 350’s, depending on my cooling requirements, by undoing an internal solder bridge and running a 4th wire outside of the housing from the previously bridged connector. That 4th wire is switchable to 12 volts via a mini toggle switch mounted on the acrylic base plate.

Initially, I planned on having three such pump station assemblies lined up side-by-side in the Ascension, like this:

http://farm4.static.flickr.com/3261/3163632341_db7ff39fe2_o.jpg

Power (12v) and ground is supplied to each pump station by two busses that I fabricated out of 0.25” wide x 0.;065” thick brass strip stock. This gives me the tidy look I want when it came to powering these neighboring pumps. The red acrylic also looks pretty good against brushed black-anodized aluminum. Here are some photos of the results of that original concept, before wire sleeving:

http://farm2.static.flickr.com/1135/3164467572_23a869fcd7_o.jpg

http://farm4.static.flickr.com/3091/3163631775_7d1dcb6eb2_b.jpg

http://farm4.static.flickr.com/3256/3164467510_13420f5793_o.jpg

http://farm4.static.flickr.com/3104/3164467542_4bec1372a7_o.jpg

http://farm2.static.flickr.com/1076/3164467220_4260a0c6d6_o.jpg

http://farm2.static.flickr.com/1109/3163631863_87d48aa4d6_o.jpg

http://farm2.static.flickr.com/1036/3164467416_1cbcea908e_o.jpg

http://farm4.static.flickr.com/3086/3164467302_700922bc51_o.jpg

http://farm4.static.flickr.com/3264/3164467386_6191b4809e_o.jpg

Once I received the Ascension and started filling it up, I began to realize that a better arrangement than having 3 pump stations ganged in a row was to have a pair of them mounted in front of the motherboard to serve the CPU cooling loops:

http://farm2.static.flickr.com/1393/3164467736_9383ec6f59_o.jpg

http://farm4.static.flickr.com/3108/3164467840_39f8203966_o.jpg

and have the third station mounted behind the motherboard on top of the PSU, using an additional translucent red acrylic sheet:

http://farm2.static.flickr.com/1121/3164467686_55e8110f84_o.jpg

http://farm2.static.flickr.com/1127/3164467756_a9616a6fcb_o.jpg

http://farm2.static.flickr.com/1277/3163632531_719fa4602e_o.jpg

The front pair of stations are joined by 12v and ground brass bus bars (as per the original concept, but not yet installed when the photos of the installed pump stations were taken), and these are joined to the rear station by a sleeved twisted wire pair (also not shown). Also on each installed station I placed a FUFUFUFUFUFUFUFU G-1/4 quick-disconnect fitting (article with the red o-ring) that serves as a drain port for the respective cooling loop. [Note that these FUFUFUFUFUFUFUFU fittings appear to be nicely made and the spring-loaded quick-disconnect nozzles seal nicely. However, the G-1/4 threaded ends have an o-ring recess that is cut too deeply into the body of the fitting, and the fitting can be too easily rotated at the threaded connection. This was remedied by adding a second o-ring to the G-1/4 threaded ends, but FUFUFUFUFUFUFUFU should consider making the o-ring recess shallower.]

rubidium

Septim

01-04-2009, 07:54 AM

nice wiring of those 3 pumps...

rubidium

01-04-2009, 08:01 AM

Each of the three cooling loops contains a Bitspower 5-1/4" bay-mounted reservoir:

http://farm2.static.flickr.com/1292/3165991075_51cbe9ac21_o.jpg

http://farm4.static.flickr.com/3161/3165995761_17f5d9b4d3_o.jpg

FUFUFUFUFUFUFUFU quick-disconnect fittings are used as fill-ports:

http://farm2.static.flickr.com/1052/3166829256_bf587e7d3c_o.jpg

and as I mentioned in the previous post, the o-rings were doubled on the G-1/4 ends of these fittings to prevent them from rotating loose from the reservoirs when making or breaking the quick-disconnect bayonette connections.

rubidium

bianco

01-04-2009, 08:12 AM

awesome rubidium
that is just brilliant.

Xilikon

01-04-2009, 08:14 AM

It's a very nice build with a lovely attention to details. Bet this will cost you a pretty penny in BP fittings but it's worth the end result :up:

rubidium

01-04-2009, 08:25 AM

It's a very nice build with a lovely attention to details. Bet this will cost you a pretty penny in BP fittings but it's worth the end result :up:

I'm certainly not alone here in getting entrained by the relentless BP fitting "vortex" :rolling::
1. You place an order for what you think you need.
2. You discover you need more.
3. Rinse and repeat.

rubidium

Kayin

01-04-2009, 10:44 AM

Love the buss bars. Too few people realize how great they are...

bentleya

01-04-2009, 02:36 PM

Love how you done them pumps :)

rubidium

01-04-2009, 04:10 PM

The last element in each cooling circuit is a PA120.3. Each G-3/8 inlet and outlet on a radiator is reduced to G-1/4 with a reducing bushing from McMaster-Carr. Then a Bitspower Rotary-Q fitting is placed at each port in order to accomodate a Dallas one-wire temperature sensor. A seperate thread (http://www.realredraider.com/vbulletin/showthread.php?t=169) presents the results of a significant research effort on building reliable and accurate immersed temperature probles, and provides a detailed description of their construction.

The PA120.3's are mounted horizontally, with the two lower radiators that serve the CPU cooling loops arranged with their ports on the left side of the case:

http://farm4.static.flickr.com/3095/3168315228_6ccd1a70a4_o.jpg

and the upper radiator that serves the PSU cooling loop arranged with its ports on the right side of the case:

http://farm4.static.flickr.com/3126/3167483969_ffb1ae4dd9_o.jpg

The PA120.3's are prepared with 6-32 rivet nuts, and mounted to the Ascension front panel with black oxide stainless steel oval head Allen cap screws from McMaster-Carr. Sandwiched between the rads and the front panel are A.C. Ryan 120mm Meshx FanGuards (ACR-FG0724):

http://farm2.static.flickr.com/1164/3168315352_d4c328227c_o.jpg

Three Noctua NF-P12 fans per radiator have their power and ground lines connected in parallel. The RPM sense line of just the center fan on each radiator is also forwarded with the power and ground lines to one of four fan-control circuits on a Crystalfontz CFA635 LCD display, equipped with the system cooling accessory board (SCAB). In the photo below, the three composite fan cables, each corresponding to 3 fans on 1 PA120.3, plug into the SCAB on the left, while the connection entering the SCAB on the right is the Dallas one-wire bus to which all six immersed thermal probes are connected in parallel:

http://farm4.static.flickr.com/3115/3167483849_e607626e96_o.jpg

With the CrystalControl2 software supplied by Crystalfontz, the CFA635 can control the power to each radiator's fan ensemble based on the inlet temperature, outlet temperature, or combination of the two according to a user-specified curve.

rubidium

breezyjr

01-05-2009, 09:12 AM

rubidium

01-05-2009, 02:39 PM

This build still amazes me! I've only seen a select few with such attention to detail...
On my previous builds, I've always tried, then got impatient and threw the rest of the system together within an hour...hehe

Not this time though. Now that I have a working computer to game on, and such, I'm taking my time and going part by part trying to make everything perfect. Did my first three rad fans soldered together and sleeved into one cable, just 10 more fans on four different rads to go!!!! hehe

I'm curious about the PSU. If you remember, we have the same one (glad you were the Guinea pig, not me... hehe) any problems with cable length. Mine is going on the bottom middle, and I worry about the top 5 1/4" bay devices... I'm prepared to lengthen if necessary, but of course would like to avoid if possible.

Keep up the great/awesome/awe inspiring/neato/incredible work :up: !!!!

Joe_Breezy

PS. Mcmaster Carr is becoming my best friend as well....
PPS. what Rivet nut gun did you use? The one I got from McMaster Carr (I shoulda listened to R3) threads are starting to warp, either from POS, or my not knowing how to use it, at first. Now, I got the hang of it... got to set the "depth" (if that's what it's called) D'oh

From my own personal experience, I think it takes a build or two to finally shed that impatient nature that we all have to some degree. My first build back in its day was embarrassingly full of inaccurately-placed or extraneous holes because I was too quick to reach for a drill or a saw and "shoot from the hip." I spend much more time now measuring and sketching, double and triple checking, and when I am finally ready to drill or cut something I also spend a lot of time anticipating the consequences of that machining action and doing everything I can to avoid collateral damage. I no longer get frustrated by the prospect of having to disassemble something for the 13th time to make an additional mod if that's what it takes to get it right and not later regret that I ruined something in the process.

Regarding your questions:
1. The big power cables to the motherboard and to graphics cards are just right. However, the cables to power peripherals like bay devices and HDD's, i.e. the ones with the 4-pin molex connectors and the ones with the SATA connectors are too short for my taste. While they are long enough to reach anything in the case, they're too short for the kind of extended routing you want for tidy cable management. Specifically:
- the cables with the 4-pin molex plugs have the following connectors spaced at the following distances from the 6-pin plug that goes into the PSU: (molex@20inches), (molex@26inches), (molex@32inches), (molex@38inches), (floppy@44inches)
- the cables with the SATA power plugs have connectors spaced at the following distances from the 6-pin plug that goes into the PSU: (SATA@20inches), (SATA@26inches), (SATA@32inches), (SATA@38inches)
Since these are the easiest to make, I'm simply going to prepare some custom ones that have the plugs where I want them.

2. The rivet tool I used is McMaster-Carr 94370A260, and the 6-32 rivets I used are 93482A605. No problems at all with either after practicing with 2 or 3 rivets on some scrap sheet metal. And never, ever, ever drill into your rad flange without a backing plate. ;)

breezyjr

01-05-2009, 03:13 PM

And never, ever, ever drill into your rad flange without a backing plate. ;)

Yup, yup found that out the hard way.... ruined a 120.3 cause I thought I could stop the drill in time.... WRONG!:mad:

Found a piece of angled Al+ and used it, works perfectly, and stays in place pretty well too.....

Thanks for the info on the the PSU Molex and SATA connectors...

Oh, and as for my Mcmaster Rivet nut gun, I didn't know there was any kind of "depth" adjustment, so when I was practicing on a piece of scrap, I would squeeze the handle till the mandrel "popped" out.... Then found out, need to set the "depth" so the mandrel "DIDN'T" pop out, but just squeezed into place. So, now I have to use a set of pliers to get the rivets screwed on...... I guess I could fix it, by picking up one of those screw making doo hickies, like a tap, but makes the marks on the "outside." Just need a 6/32 one, and "re-tap" it..... Hmmmm... maybe a trip to the hardware store is in order..... I certainly couldn't make it any worse.

Later,
Joe_Breezy

Jtrain

01-07-2009, 08:59 PM

Wow this build is just amazing :up:

rubidium

01-11-2009, 08:45 AM

Just a quick status report to note that the build is on hold for a little while, awaiting the completion of two concurrent processes:
- RMA of the PSU (waiting for replacement v1.5 from FUFUFUFUFUFUFUFU)
- "RMA" of a fucking hernia that was repaired at age 5, but came back :( to haunt me a couple of weeks ago (surgery tomorrow :moon:) - although warranty long expired

I'm including some "tickler" photos of my initial firing up of the rig before returning the PSU for replacement.

All 16 cores and 64GB RAM properly showing up, according to Windows Server 2008:

http://farm4.static.flickr.com/3320/3187128779_ae96e9b7a3_b.jpg

Even the Crystalfontz CFA635 thinks so:

http://farm4.static.flickr.com/3301/3187128847_cfb2b0f69f.jpg?v=0

A little CPU stress testing - 48 continuous hours of 16 instances of Prime-95 - went perfectly. Only a little more than 3 gigs of RAM used, but max stress placed on the PSU and cooling system :woot::

http://farm4.static.flickr.com/3477/3187967938_a8106c2138_b.jpg

http://farm4.static.flickr.com/3101/3187968020_b0d5f715fb_b.jpg

later,
rubidium

Septim

01-11-2009, 09:32 AM

definitely nice...

how about a little crunching or folding...

rubidium

01-11-2009, 11:01 AM

definitely nice...

how about a little crunching or folding...

Yea ... once it's all back up again ...

rubidium

RedRaider

01-11-2009, 12:42 PM

Yea ... once it's all back up again ...

rubidium

We would love to have you on the team....:king:

andersson.j

01-11-2009, 01:48 PM

Truly awesome build, subscribed! :cheers:

...
In terms of programming, one package that I use a lot for digital signal processing and image processing is Matlab, which supports multiple cores when evaluating certain key computations like the fast Fourier transform (FFT). I also want to participate in F@H.
...
What kinda science are you into?
I been using Matlab in just about every class I've taken for the past three years. Mostly basic stuff though, nothing requiring a beast like your rig!

Wouldn't you get electrocuted if you touch both rails simultaneously? If so that's a pretty dangerous toggle switch!
I'm loving the look though! :up:
http://farm4.static.flickr.com/3091/3163631775_7d1dcb6eb2_b.jpg

rubidium

01-11-2009, 02:40 PM

Truly awesome build, subscribed! :cheers:

What kinda science are you into?
I been using Matlab in just about every class I've taken for the past three years. Mostly basic stuff though, nothing requiring a beast like your rig!

Wouldn't you get electrocuted if you touch both rails simultaneously? If so that's a pretty dangerous toggle switch!
I'm loving the look though! :up:

Thanks. The rails represent more than a just a cosmetic / artistic statement. When summer comes, I want to prevent flies and moths from taking up residence in the rig. :rofl::rofl::rofl:

rubidium

02-03-2009, 07:04 PM

Update:

For those wondering ":wtf: regarding this build?", I RMA'd my FUFUFUFUFUFUFUFU 1300W watercooled PSU at the start of the year and have been waiting :juggle: and waiting :juggle: and waiting :juggle:. To make a long story short, the improved Rev 1.5 of this PSU is on its way to me - ETA end of the week. So this weekend we'll give it a go and see if that bad boy will POST.

I hate RMA's ...

rubidium

ballz0r

02-03-2009, 07:53 PM

looks awesome mate :up:

have you done any flow tests yet?

one suggestion... perhaps add another FUFUFUFUFUFUFUFU QD to each res for a bleed line (so you have water in and air out)

Baleful

02-03-2009, 08:11 PM

Can't wait for some updates. This thing has seriously made me want to look into multi-socket setups now.

Septim:

definitely nice...

how about a little crunching or folding... :D

Would love to have you on the team.

rubidium

02-07-2009, 09:09 AM

Boy, talk about starting out a day with a bright outlook and good intentions, and ending up taking it from behind. :hump:
On Thursday, my RMA with FUFUFUFUFUFUFUFU finally came to an end, and my replacement 1300W PSU arrived. This one is the latest v1.5, which supposedly fixes compatibility issues with certain motherboards.

Well, this morning I got up, had a good breakfast, and couldn't wait to get this unit installed, fire up the rig, and hopefully find that the cold-start POSTing problems that I previously experienced with the older version were gone. Didn't get that far ...

Comparing the previous unit, here:

http://farm2.static.flickr.com/1195/3163629937_eef1c0454c_o.jpg

and the v1.5 replacement, here:

http://farm4.static.flickr.com/3515/3259757877_9a954beefc_o.jpg

shows that the newer edition underwent a juxtaposition of the power switch and AC receptacle. WTF??? So now, the FUFUFUFUFUFUFUFU bracket that I bought previously needs to be replaced, and I need to remod the rear panel on the Ascension. Here comes another week+ of waiting. The only consolation is that the plumbing arrangement with a magical combination of BP fittings that I was so happy with appears to still be applicable. Fuck! :mad1:

rubidium

RedRaider

02-09-2009, 06:53 AM

What the FUCK...:wth:

FUFUFUFUFUFUFUFU changed the layout of the rear of the PSU, why?

Will you be able to modify your rear panel, without it looking like ass??

breezyjr

02-09-2009, 08:18 AM

RedRaider

02-09-2009, 09:29 AM

breezyjr

02-09-2009, 09:40 AM

Now that is a fine example of how we roll here at RedRaider Technology...:up:

BTW.... Breezy... I'm almost ready for some SunFire love...:)

Boy, that last sentence would sure sound weird to someone who doesn't know what the heck it means.... hehe

That would be the Bob Carver signature Sunfire cinema grand amplifier That he's getting from me... just so no one gets the wrong idea....

[/thread hijack]

Later,
joe_breezy

rubidium

02-09-2009, 12:28 PM

Hey Rub....
Would things be any easier if I gave you the bracket I have with my FUFUFUFUFUFUFUFU PSU? It's the same bracket, but I figure you could use one of them to hack up, and see if you can make it work....
I won't be using the bracket with my build... so not a problem...

just let me know, and I'll send it out... Don't worry about cost, it was only a few dollars anyway, and I'll send it priority mail which shouldn't cost too much either...

Later,
Joe_Breezy

breezyjr (at) gmai1 d0t c0m

Hey breezy, thanks for the kind offer, but last night - after enjoying a day outdoors in the unusual spring-like weather that hit the DC area, and a cooler head prevailed - I decided to take the dremel and a file and do exactly what you suggested. ;) Only a minor adjustment to the Ascension back panel was needed, and most of the reworking was with the FUFUFUFUFUFUFUFU bracket. Once I've looked at it for a while, I can decide whether I like the look of the modded bracket or should go and get the new one. The important thing was that I was able to install the PSU without having to totally butcher anything. I'll post a photo tonight.

The good news is that my original plumbing arrangement still works out. The really good news is that the mobo POSTs properly now from a cold start without a hitch. The problem was indeed with the earlier version of the PSU. The v1.5 appears rock solid in all regards. :up:

rubidium

Kayin

02-09-2009, 02:38 PM

Congrats, now we can proceed apace.

I'll have to post my pedestal just to keep up now.

rubidium

02-09-2009, 06:26 PM

Here's what I did with the old FUFUFUFUFUFUFUFU bracket that I got for my original PSU (v1.0), but no longer fit the updated version (1.5) of the PSU that was returned in the RMA:

http://farm4.static.flickr.com/3405/3267349767_7f585b7ed5_o.jpg

http://farm4.static.flickr.com/3524/3268176150_785df6a938_o.jpg

I still need to drill a hole in the left side of the bracket to view the green LED that indicates DC power status. I still have to wonder why they dicked with the positions of the AC switch and receptacle. Some engineer needs an all-expense-paid trip to the woodshed.:slapass:

rubidium

RedRaider

02-10-2009, 09:15 AM

I'm glad to see you got it going...:up:

I will probably always be leery of FUFUFUFUFUFUFUFU, no plans to ever buy their products. But that's just me...

Cutless009

02-10-2009, 09:26 AM

its ok red, Im going to be putting FUFUFUFUFUFUFUFU through their paces very soon hopefully ;)

rubidium

02-10-2009, 09:48 AM

From what I have experienced, they seem to have good fundamental designs, but seem to drop the ball when it comes to the last 5% of execution.

This 1300W PSU is very well built. The fit and finish are to a high standard, all voltages are spot on and appear to remain so under load, and the cooling approach is performing nicely - both in the thermal management sense and in being dead quiet. To have experienced my original compatibility problem with the motherboard doesn't bother me so much. I'm guessing that some overcurrent protection logic wasn't properly tweaked to support the high cold-start surge that my rather unique board presents, and in the end the revisions made have reconciled the issue. But to have screwed around with the back panel layout in the process is an example of that 5% horseshit that I'm talking about.

I also like their quick-disconnect fittings for fill and drain functions in my build. But again, they made the G-1/4 end o-ring recess a tad too deep and/or supplied o-rings that were too thin or too supple. I had to go out and get my own o-rings to prevent the entire fitting from rotating loose when disconnecting from it. Another issue falling into that 5% "blind spot."

In my experience, these observations usually point to inadequate attentiveness on the part of the team lead, chief engineer, or what have you.

rubidium

rubidium

02-27-2009, 09:00 PM

Update: Some more housecleaning via RMA's. My new power supply has been solid as a rock - so that's solved. :up: But two other major items are being replaced as we speak.

About a week ago, my GTX-280 started displaying hash marks in BIOS, and hash marks and vertical streaks in Windows. Eventually, the video completely disappeared in Windows, yet was able to display while in BIOS. Having an integrated graphics controller on the motherboard, I was able to get back into Windows and revert the nVidia driver back to the default Windows driver. After doing this, the GTX-280 again functioned under Windows, but continued to display hash marks and vertical streaks throughout the screen, and wouldn't handle the nVidia drivers when I tried to reinstall them. So I set up the RMA with EVGA, and my new card is due here on Tuesday. BTW, the RMA with EVGA was the most painless I've ever experienced over the course of many years. :up:

With that in progress, I decided to RMA the S4989 motherboard. Back late last year when I first fired up the rig, I noticed one of the (32!) DIMM slots was giving errors in MEMTEST86 and it was not due to the RAM stick itself. I put off an RMA because I wanted to get the rig up and going a bit further before being without a board. But now's the time. My replacement is due next Friday, and get this - it's being replaced by the S4992, which is the S4989 upgraded to Hypertransport 3.0 and PCIe-2.0, with all else the same. :woot:

Prior to all this - for all you math freaks - I did a "heavy calculation" that would be next to impossible without this hardware. I filled a square matrix with random numbers and inverted it, using MATLAB 2008b which is multithreaded. Simple - except that the matrix was 31624 x 31624, or contained about 1 billion double-precision floating-point numbers. All 16 cores and about half of the RAM were going at it for about an hour:

http://farm4.static.flickr.com/3199/3314642443_9fc2651df1_b.jpg

I can't wait to get everything back together again and do a little folding ...
rubidium

Snyxxx

02-27-2009, 09:34 PM

Wow, I did not think I would ever see that many CPUs in task manager.

I really like your style.

rubidium

02-27-2009, 09:56 PM

Upcoming Istanbul cores could raise the core count another 50% with no other hardware change. :D

yeow_z

02-27-2009, 11:17 PM

holy mutha of gawd!:eek::cussing::wth::youtheman: this monster looks sizzling sexy:bow:

subscribed

Burnout21

04-03-2009, 10:10 AM

Upcoming Istanbul cores could raise the core count another 50% with no other hardware change. :D

Good lord 24 cores!!

Oh and before anyone asks, Langer set me here from over the pond in the country 'O' bit-tech.

And i think i just found cheesecake in this thread!

Baleful

04-03-2009, 10:39 AM

Good lord 24 cores!!

Oh and before anyone asks, Langer set me here from over the pond in the country 'O' bit-tech.

And i think i just found cheesecake in this thread!

LMAO....

Welcome to the land of Texans, RRTech!

@Rubidium
Any updates? I'm dying to know how things are going, and see what kind of crunching power this baby has.

Burnout21

04-03-2009, 10:51 AM

must say 16 core would be wasted on me, i struggle to get my Quad maxed out in Solidworks atm. Flow Xpress seems to slow a little, but i usually have about 10 or more parts open at once, Ctrl tabbing between them. lol!

Baleful

04-03-2009, 11:01 AM

Floyd

04-03-2009, 11:06 AM

This is shaping up to be a nice build indeed!

rubidium

04-03-2009, 11:54 AM

LMAO....

Welcome to the land of Texans, RRTech!

@Rubidium
Any updates? I'm dying to know how things are going, and see what kind of crunching power this baby has.

You could always crunch for us, that would sure put that quad to work ;)

The latest update is that, as of mid-week, I now have 4x 8384 Shanghai processors in hand (2.7 GHz). I'm going to install them tomorrow, and retire the 1.8 GHz Barcelona's after they performed their short, but important, stint of guiding this rig through a debugging phase involving several RMA's and eventually getting it all up and running. After a short period of experimentation with the new chips, I'll get geared up to crunch.

rubidium

Kayin

04-03-2009, 12:00 PM

Damn, 4 homeless Barcelonas...

*tear*

Langer

04-03-2009, 12:01 PM

Very much looking forward to it!

knightmaric

04-03-2009, 12:36 PM

rubidium

04-03-2009, 01:04 PM

I've just grazed the surface with this project but WOW!
Looked like a beast at the start and I can't wait to see what it turns out like.
First person i've seen to use one of those FUFUFUFUFUFUFUFU PSU's to =]
They're really nice and those bitspower fittings at the back look incredible.

Very excited =]

Had to go through an RMA with it, and a PITA episode of refitting the revised v1.5 to the opening in my case, but in the end the unit is proving to be solid and barely heating up the water - even with 16 instances of Prime95 running. Of course I'll be pushing it a bit harder with the new processors, as the outgoing Barcelona's were HE versions (ADP=55W), but I'm not expecting any issues.

voklskier4452

04-03-2009, 01:08 PM

my gtx 285 just had the same problem that your 280 has.. did you ever found out why that was happening?

rubidium

04-03-2009, 01:28 PM

my gtx 285 just had the same problem that your 280 has.. did you ever found out why that was happening?

No. When I described the problem to EVGA, they issued an RMA, and within days I had a new unit and it's been working perfectly ever since. I suspect the failure of the original card was simply a case of infant mortality, given that it crapped out after only a few hours. It was the most painless RMA I've ever experienced.

voklskier4452

04-03-2009, 02:01 PM

oh well same here.. should have a new one soon.. and your build looks freakin amazing

Baleful

04-03-2009, 02:04 PM

The latest update is that, as of mid-week, I now have 4x 8384 Shanghai processors in hand (2.7 GHz). I'm going to install them tomorrow, and retire the 1.8 GHz Barcelona's after they performed their short, but important, stint of guiding this rig through a debugging phase involving several RMA's and eventually getting it all up and running. After a short period of experimentation with the new chips, I'll get geared up to crunch.

rubidium

Man, I've heard nothing but good things about those Shanghai's. What a beast this thing will be.

rubidium

04-03-2009, 04:08 PM

Did I hear somebody say crunch?

http://farm4.static.flickr.com/3372/3410403590_6f93bf845d_b.jpg

Baleful

04-03-2009, 04:10 PM

Oh man, I would stand up and applaud you but I think I just popped a woody.:crazyclap::crazyclap:

Snyxxx

04-03-2009, 04:57 PM

No. When I described the problem to EVGA, they issued an RMA, and within days I had a new unit and it's been working perfectly ever since. I suspect the failure of the original card was simply a case of infant mortality, given that it crapped out after only a few hours. It was the most painless RMA I've ever experienced.

I had the same experience with EVGA RMA. Painless and professional. I will only buy EVGA Nvida video cards in the future, period.

voklskier4452

04-03-2009, 05:30 PM

its true what they say.. once you go EVGA you never go back

Cutless009

04-03-2009, 05:31 PM

I wished Id gone EVGA as soon as I got the shitty PNY's in my hand and realized how effing cheap they were.

RedRaider

04-03-2009, 06:21 PM

The latest update is that, as of mid-week, I now have 4x 8384 Shanghai processors in hand (2.7 GHz). I'm going to install them tomorrow, and retire the 1.8 GHz Barcelona's after they performed their short, but important, stint of guiding this rig through a debugging phase involving several RMA's and eventually getting it all up and running. After a short period of experimentation with the new chips, I'll get geared up to crunch.

rubidium

Oh Yeah Baby...:crack:

Now that's what I like to hear...:D

Chilly

04-03-2009, 06:36 PM

Any overclocking yet? :coke:

rubidium

04-03-2009, 06:52 PM

Any overclocking yet? :coke:

I just did some light experimenting with the 8346HE's thus far - nothing serious, 5%OC, 16x Prime95 stable. I have to find a way to be able to boost CPU voltage in order to get more serious. The BIOS is not friendly in that regard, so I need to discover some tools to get me access to those voltages through the OS. Clockgen is not one of them. There are lots of people who have been recently hacking the NFP 3600/3050 chipset pair (AKA MCP55-Pro-V/IO55). I just have to find time to read up on it all. This board is completely new OC territory for me, and I would say that there's lots of fun awaiting in the discovery of it all.

rubidium

MomijiTMO

04-03-2009, 07:03 PM

Looks tasty dude.

Chilly

04-03-2009, 08:01 PM

I just did some light experimenting with the 8346HE's thus far - nothing serious, 5%OC, 16x Prime95 stable. I have to find a way to be able to boost CPU voltage in order to get more serious. The BIOS is not friendly in that regard, so I need to discover some tools to get me access to those voltages through the OS. Clockgen is not one of them. There are lots of people who have been recently hacking the NFP 3600/3050 chipset pair (AKA MCP55-Pro-V/IO55). I just have to find time to read up on it all. This board is completely new OC territory for me, and I would say that there's lots of fun awaiting in the discovery of it all.

rubidium

Any possible "pin" mods? I know with Intel CPU's(even the i7's) you can cover up/connect a few pads with electrical tape/conductive pen and change what the motherboard detects as the default voltage.

Kayin

04-03-2009, 08:22 PM

BSEL mods. It should be possible, depending on where the pads are that determine voltage.

knightmaric

04-04-2009, 08:22 AM

You are one lucky sob.
That machine is going to be an absolute beast when it's done >_>
Shame about the bad experience with the original FUFUFUFUFUFUFUFU, but a least they sent you a newer version of it.
God knows why the changed the layout though.

Those CPU's are giving me the chills lol
You have those and i'm sat here dreaming of when I can build my AM3 rig with an 810 PII =P

rubidium

04-05-2009, 07:47 AM

Man, I've heard nothing but good things about those Shanghai's. What a beast this thing will be.

I'm startin' to like these Shanghai's. :up: After installing them late Friday night, I burned them in for 12 hours with Prime95 at stock frequency (2.7GHz). Then I simply gave them an 11% bump with Clockgen to 3.0GHz, and the rig has been 16-instance Prime95 stable for the last 16 hours - going for 24 hours. Makin' a little heat now. This has me inspired to figure out a good way to get Vcore up, 'cause I know there's more to be had out of these chips. :wink:

From Sandra, at 3.0GHz:
Dhrystone ALU: 169.648 GIPS
Whetstone iSSE2: 135.61 GFLOPS
Memory bandwidth: 27261 MB/s int, 27326 MB/s float (32GB, 50%, included in test)

rubidium

Worthy

04-05-2009, 08:18 AM

Rubidium, that is truly an awesome machine to behold.
Love the layout. That is truly a huge amount of horsepower.
I would expect nothing less from you. hehe

knightmaric

04-11-2009, 09:41 PM

I'm really wanting updates of this lol

ILikeCosmosS

04-11-2009, 10:11 PM

knight dont worry i dream too trying to get away from a pentium 4+ geforce 6200
btw i love your avatar he looks funny looking forgot what show he was on

knightmaric

04-11-2009, 10:21 PM

He's from invader zim =]
His name is Gir.
He's the most useless evil sidekick in the world lol

I'm using a Celeron D 351 and nvidia MX-4 xD

lonewolf

04-13-2009, 09:13 AM

Amazing! This build is in a league of it's own.

Shingoshi

07-20-2009, 12:27 PM

Reserved. :eek: Just joined the club.

This space will eventually walk through my recent build of a 4-socket (16-core) 3rd-gen Opteron workstation. Watercooled, of course. Mad science to follow. :frankenstein:

rubidium

Hey rubidium, I just joined too. Now we can chat, one quad owner to another.

I had the Tyan S4985, which I RMA'd. I'm now thinking of getting the S4992. Would like to hear more about your experiences with the S4989.

Shingoshi

rubidium

07-20-2009, 12:57 PM

Hey rubidium, I just joined too. Now we can chat, one quad owner to another.

I had the Tyan S4985, which I RMA'd. I'm now thinking of getting the S4992. Would like to hear more about your experiences with the S4989.

Shingoshi

Shingoshi,
Welcome to RRTech. I think you'll like it here - no BS.

Back in late February, I had to RMA the S4989. When I first fired up the rig in late 2008, I noticed one of the (32!) DIMM slots was giving errors in MEMTEST86 and it was not due to the RAM stick itself. I put off an RMA because I wanted to get the rig up and going a bit further before being without a board. But in February I finally sent it to Tyan and they replaced it with the S4992. It's been running 24/7 ever since, and most of that time it's crunching WCG at 100% loading of all 16 cores, with average power consumption of the entire rig at 635 Watts. I'm running 8384 Shanghai's.

rubidium

Shingoshi

07-20-2009, 01:29 PM

rubidium

07-20-2009, 02:00 PM

How the hell did you get those original Shanghais for $50 a piece?
That's what I really want to know! Dumpster Diving???

Seriously, am I reading the specs on the Tyan S4985 correctly. When it lists having x4 pci-e, does that mean it is limited to only x4 in those slots, and not the physical x16?

Shingoshi

My original Barcelona's (8346HE) were $50 apiece. The Shanghai's (8384)that I got in the spring were had for $500 each.

On the S4985 there are four PCIe x16 physical sockets, organized as pairs. One pair is driven by the nForce Pro 2200, and the other pair is driven by the nForce Pro 2050. Within each pair, one socket supports a full x16 signaling, while the other is limited to x4 signaling. Thus in terms of signal lanes you have two x16 and two x4 signal buses. Then there is the old-style PCI 32-bit/33MHz slot. The diagram on the second page of this (ftp://ftp.tyan.com/datasheets/d_s4985_100.pdf) illustrates this more clearly.

Shingoshi

07-20-2009, 02:22 PM

My original Barcelona's (8346HE) were $50 apiece. The Shanghai's (8384)that I got in the spring were had for $500 each.

On the S4985 there are four PCIe x16 physical sockets, organized as pairs. One pair is driven by the nForce Pro 2200, and the other pair is driven by the nForce Pro 2050. Within each pair, one socket supports a full x16 signaling, while the other is limited to x4 signaling. Thus in terms of signal lanes you have two x16 and two x4 signal buses. Then there is the old-style PCI 32-bit/33MHz slot. The diagram on the second page of this (ftp://ftp.tyan.com/datasheets/d_s4985_100.pdf) illustrates this more clearly.

That's what I thought all along. Because I had a similar issue with the Tyan S3992, thinking the physical x16 were 16 lane. Only to find out they were really only x8 electrical.

That's why I think the S4992 is the better option for me now. My only concern is for the spacing of the cpus that you have mentioned. The S4985 took my FUFUFUFUFUFUFUFU blocks without any trouble. But I'm concerned the S4992 will be as limited as you said was the case for the S4989. The one big difference between our cooling blocks though, is that mine sit higher than your's do. So I'm not sure if I really will have a problem or not. They also fit just fine on my S4980. I really don't want to change my blocks. I like them too much!

Shingoshi

Shingoshi

07-20-2009, 02:36 PM

By the way, we chose different paths when it came to dealing with the size of our motherboards. And much of my decision was influenced by the fact that I have four Tyan S2912s besides the mainboard. For that reason, I chose a path to provide for all my boards in a single case. You can see my project here:
http://www.hardforum.com/showthread.php?t=1428264

Yeah, I'm really quite crazy. But economics and practicality justifies my being so.

Shingoshi

rubidium

07-20-2009, 05:49 PM

On both the S4989 and S4992, the nominal clearance between the extremeties of the two nearest memory slots surrounding any CPU socket is 3.06 inches, and specifically +/-1.53 inches about a center line drawn between the two hold-down screws to the backing plates. The clearance between the extremeties of black components on the two nearest ram sticks surrounding any CPU is slightly more than that.

If your CPU water blocks with mounting hardware are narrower than 3.06 inches, then you should have no problem mounting them. As discussed in post #25 in this thread, my original D-tek Fuzion II brackets were 3.23 inches wide and so had to be cut down. I cut their widths down to 2.6 inches to avoid having cuts pass through existing holes in the Fuzion's universal mounting brackets.

Septim

07-27-2009, 10:06 AM

Shingoshi

07-27-2009, 01:39 PM

quick question:
shingoshi is your project done or nearing completion or simply waiting for more parts... seems i am gettin an eyestrain reading that [H] thread, or it just lacks some pictures...

maybe you could also post some build log over here... on your own thread that is...

Hi Septim,
I haven't really even started the project yet. I'm waiting to get more parts. I got my case last Friday. But now I have to wait until the first to get more parts. Aside from that, I'm beginning to wonder if I'm completely NUTS!

p0Pe

07-27-2009, 01:43 PM

sorry for off topic, but this was to great a pic to let go un noticed! got it from that thread you reffered to shingoshi

http://i50.photobucket.com/albums/f347/Naja002/Smilies/thread_is_worthless.jpg

nicky.82uk

07-27-2009, 01:52 PM

shit this is so nice

rubidium

09-04-2009, 02:44 PM

For a while I’ve been working toward upgrading the cooling system in this rig. After about six months of steady operation, I convinced myself that there are several aspects of the original layout that I want to change:

1. I’m presently using three EK 5-1/4” bay-mounted reservoirs to feed my three cooling loops. These are nice, but have proven to be difficult to fill and bleed when stacked on top of each other in a column, and they’re taking up half of my front panel real estate that I’d like to have back for other things. Solution: I’m replacing the bay-mounted reservoirs with solid copper Aquatubes.

2. The 3/8” ID x 1/2” OD primochill tubing that I’m using has been prone to collapsing, even in relatively straight sections. Throughout a 2-week period on average, several sections of my tubing runs noticeably flatten out. Although the effect is moderate and not really worrisome, it is, shall we say, irritating. When I hook up a container of water to a reservoir through quick-disconnect fittings, water is sucked in, the vacuum is relieved, and the flattened regions of the tubing return to being round. This leads me to believe that there is a slow diffusion of water through the tubing that subsequently evaporates into the atmosphere, leaving a partial vacuum behind. This is not unreasonable given the permeability figures of the polymer used, the lengths of the tubes, and their 1/16” wall thickness. Solution: I’m moving to 1/2" ID x 3/4" OD Tygon R-3400.

3. I too have finally fallen victim to the Heatkiller 3.0 addiction. Enough lurking! I give up! I’m hooked! Solution: I took delivery of 4 CPU blocks from Gary, along with a shitpile of Bitspower fittings to accommodate the move to the larger tubing.

Here are some of the goodies that will fit in the field of view of a 24mm lens:

http://farm3.static.flickr.com/2594/3887949858_50fbdb1818_b.jpg

http://farm4.static.flickr.com/3508/3887155295_c7f89b7070_b.jpg

Now there’s still one more element of motivation: the pleasure of being creative. That’s a big reason why we’re all into this, right? ;) Recently, an acquaintance of mine bought a very nice Proxxon table-top CNC milling station. Add in the fact that he owes me a big favor. Further add in the fact that I’m an accomplished machinist living life without machine tools of my own. Solution: I want to have some fun and design and machine something relevant to this cooling system upgrade. :crack:

I like the brushed stainless steel against polished copper look of the Heatkiller 3.0, and so decided to continue that theme in the design of a “reservoir station.” I’ll be mounting the 3 copper Aquatube reservoirs on a brushed stainless steel beam that spans the upper interior of my MM Ascension case from front to back. An added extravagance is that the beam will also include 3 built-in digital displays of my own design to show the respective flow rate in each loop. (You did notice the aquacomputer flow sensors, didn't you?) I'll be showing the design of the stainless steel support shortly.

rubidium

Baleful

09-04-2009, 02:52 PM

I’ll be mounting the 3 copper Aquatube reservoirs on a brushed stainless steel beam that spans the upper interior of my MM Ascension case from front to back. An added extravagance is that the beam will also include 3 built-in digital displays of my own design to show the respective flow rate in each loop. (You did notice the aquacomputer flow sensors, didn't you?) I'll be showing the design of the stainless steel support shortly.

rubidium

Oh my :slobber:

I'm really excited to see what you come up with!

Snyxxx

09-04-2009, 03:38 PM

rubidium,

Looking at your pictures, the copper color of the reservoirs have a different look. Did you apply your preservation method to these?

rubidium

09-04-2009, 03:53 PM

rubidium,

Looking at your pictures, the copper color of the reservoirs have a different look. Did you apply your preservation method to these?

No. They've just been handled a lot so they're a bit dingy looking. I just took delivery of a copper polishing kit from Caswell today. I'll be polishing these reservoirs out soon.

Baleful

09-04-2009, 04:02 PM

rubidium,

Looking at your pictures, the copper color of the reservoirs have a different look. Did you apply your preservation method to these?

I noticed the discoloration as well, the aquatubes almost appear to be a "burnt" red color.

Rubidium, after you clean all your new copper parts with the cleaning kit, will you be applying the preservation to them?

DexNfX

09-04-2009, 04:22 PM

That's one helluffa setup you got there man! NICE!!!!

rubidium

09-04-2009, 04:23 PM

I noticed the discoloration as well, the aquatubes almost appear to be a "burnt" red color.

Rubidium, after you clean all your new copper parts with the cleaning kit, will you be applying the preservation to them?

They're simply oxidized after all of the handling. That what the "burnt red" is. No big deal. A quick wipe with some Brasso or Tarn-X and the copper matches the HK in appearance. I'm still experimenting with that preservative, but once I get it down I would like to use it.

Langer

09-04-2009, 05:10 PM

You continue to be my hero. =)

Awaiting your results eagerly.

rubidium

09-04-2009, 05:18 PM

Now for my design for the reservoir support. Having taken the day off from work, I actually completed the machining of it today and will show some actual photographs of it later. But first, some 2-D drawings.

I start with a 4” (wide) x 2” (high) x 14” (long) section of Type-304 stainless rectangular tubing, having 0.25” wall thickness. Front view (14" x 2") shows 3 windows (2.5" wide x 1.25" high) cut out of the tubing to support 4-digit, 7-segment, LED displays for indicating the flow rates in each of my 3 cooling loops:

http://farm4.static.flickr.com/3514/3887982448_675c209bb1_b.jpg

The top of the beam will support the 3 copper Aquatubes (fill windows oriented up) above the display windows, via 5 pentagonally-arranged holes for each reservoir as shown in this top view (14" x 4"):

http://farm3.static.flickr.com/2529/3887983792_dc775d5fdc_b.jpg

The oval cutouts in the top of the beam precisely match the oval recesses in the bases of Aquatubes. The two R-3400 lines to each reservoir will attach via Bitspower compression fittings to the outboard two of the three bottom Aquatube G-1/4 ports, and these fittings will protrude through these oval cutouts in the top of the beam, pass through the interior of the beam, and out through the bottom via holes seen in this bottom view:

http://farm3.static.flickr.com/2457/3887186421_08a3647e65_b.jpg

All of this is perhaps better visualized with a few 3-D renders. Here's a perspective showing the front display windows, top of the beam to which the Aquatubes will be mounted, and the bottom holes for 3/4" OD tubing to pass through and on to the rest of the cooling system:

http://farm3.static.flickr.com/2627/3888000728_bd484ecdf9_b.jpg

Taking a closer look, we see that I'm planning to have some fun and carefully mill a “RUBIDIUM” logo into the front beam wall, using a 0.039” (1 mm) tungsten carbide end mill, lots of coolant, the luxury of computer control, and plenty of patience:

http://farm3.static.flickr.com/2490/3887998528_46ca5027f3_b.jpg

Another view from the side this time:

http://farm3.static.flickr.com/2570/3887987098_8f4f9d88a5_b.jpg

To prevent chafing, an o-ring grove is milled into the side of each exit hole (using a 0.125” face width Woodruff key seat cutter) through which each tube will pass at the bottom of the beam. Also, the bottom edge of each exit hole is given a radius, as seen here:

http://farm3.static.flickr.com/2669/3887204745_65dd893423_b.jpg

here:

http://farm4.static.flickr.com/3483/3887203903_7945ac299c_b.jpg

and here:

http://farm3.static.flickr.com/2460/3887192771_d7df6fbfe6_b.jpg

So much for illustrating the design through renders! My next post will include some photos of the actual beam that - with the exception of some final finishing work (deburring, etc.) - was completed earlier today. ;)

rubidium

rubidium

09-04-2009, 05:31 PM

You continue to be my hero. =)

Awaiting your results eagerly.

Jesse my good man. Good to see you coming up for air. Somehow I suspected that the mention of copper might get your attention. :D

Septim

09-04-2009, 05:58 PM

copper addict:D

nice design, looks a tad heavy with .25" thick stainless used?

am looking forward to some actual pics.

rubidium

09-05-2009, 08:39 AM

I took a few shots of the completed reservoir support. Here's a look at the front, showing one of the 3 display windows and the "RUBIDIUM" logo milled into the stainless:

http://farm3.static.flickr.com/2657/3887955448_460b7409f2_b.jpg

The logo letters were patiently cut into the steel, under computer control, with a 0.039” (1 mm) tungsten carbide end mill turning at 20,000 RPM. I wet-sanded the entire exterior of the support with 220 grit wet-or-dry held in a rubber sanding block, using frequent flushings of both the steel and the paper with soapy water per coolmiester's guidance.

Another shot of the front left-hand corner:

http://farm3.static.flickr.com/2675/3887953564_4f80376cdf_b.jpg

This shot shows the mounting holes for one of the Aquatubes and the matching oval cutout in the top of the support, as well as the tubing exit holes in the bottom of the support:

http://farm3.static.flickr.com/2538/3887161699_5171fb0472_b.jpg

More of the top:

http://farm3.static.flickr.com/2619/3887959132_46b2b46c55_b.jpg

http://farm3.static.flickr.com/2671/3887158571_3bebe3854d_b.jpg

More of the bottom:

http://farm4.static.flickr.com/3506/3887956484_d95d0f824f_b.jpg

Finally, a shot showing the details of one of the bottom holes through which my 3/4"OD x 1/2"ID R-3400 tubing will pass. The groove will hold an o-ring to prevent chafing of the tubing. A bit more deburring and sanding is needed in these areas to finish up:

http://farm4.static.flickr.com/3437/3887952684_72a4b19bd2_b.jpg

The considerably bigger challenge lies ahead: The flowmeter displays. :)

rubidium

mcoffey

09-05-2009, 09:03 AM

Great concept, design and execution Rubi. Incredible custom work.

andyc

p0Pe

09-05-2009, 09:04 AM

DAYM that looks good dude! but wont it be a bitch to to tighten the screws that hold the resevoirs in place?
really really nice finish of the holder though! great idea ! need more pics:D

Septim

09-05-2009, 09:11 AM

Beautiful work.

Good Sanding.

Snyxxx

09-05-2009, 09:27 AM

Wow! Very nice. How do you get the parallel grain finish like that? Sanding?

rubidium

09-05-2009, 09:29 AM

DAYM that looks good dude! but wont it be a bitch to to tighten the screws that hold the resevoirs in place?
really really nice finish of the holder though! great idea ! need more pics:D

Ah, but that's all part of the master plan. :D The reservoirs get mounted and their hex head cap screws get inserted and tightened through the 3 cutouts for the displays. Tried it and it's no problem at all. Then the fittings get installed into the bottoms of the reservoirs, the R-3400 tubes attached, and everything properly tightened - again through the display cutouts. Tried that too and all is fine. The display cutouts are an integral part of the "logistics" of assembly. What finally goes into the display windows is the subject for a future post - and more work on my part.

Worthy

09-05-2009, 09:39 AM

Damn. That looks incredible.
I knew you were holding out on us. ;)

rubidium

09-05-2009, 09:47 AM

Wow! Very nice. How do you get the parallel grain finish like that? Sanding?

Here's how the finish was achieved. First off, the stock material that was machined was 2" x 4" x 0.25" wall thickness rectangular tubing. Being manufactured as an extrusion, there are minor "lines" running along the length of the stock from the extrusion die. I tried to start with a "relatively clean" 14" segment cut from the 3-foot long piece of stock that I bought. After machining, I placed the support onto a pile of wet towels in a sink with running warm water. Using 220 grit 3M "Wetordry Tri-M-ite" paper affixed to a 3"x5" rubber sanding block, I dilligently wet sanded each of the 4 sides of the support - going back and forth along the 14" length and frequently rinsing the sandpaper and the support to prevent residue from caking up and causing non-uniform streaks in the finish. In the end, I mounted a fresh piece of the Wetordry in the sanding block and made one-directional strokes in the support from one end to the other, being careful not to twist the sanding block along the way. Also, taking a cue from coolmiester in this (http://www.realredraider.com/vbulletin/showthread.php?t=705&highlight=soap) thread, I used liquid dish soap and warm water in the final stroking, which made a big difference in the final finish.

rubidium

09-05-2009, 10:12 AM

I'm undecided on how to treat those milled-out logo letters. Part of me just wants to leave it as-is. However, I did have the passing thought of preparing a slurry of copper dust in clear lacquer and injecting that in with a needle/syringe. Decisions, decisions.

Worthy will appreciate that my original inspiration for the Aquatubes lined up along a beam was the timeless image of the wooden water tower atop a Manhattan building.

THP Dan

09-05-2009, 10:51 AM

i would say fill the letters with a casting acrylic, would give a nice flush effect what you can shine led's through if you want.

cheapskate

09-05-2009, 10:55 AM

Pure sweetness.:D
for the logo, there's always the tried and true trick of a red plexi backing.

Edit: Oops. Yeah, like that.^

rubidium

09-05-2009, 11:43 AM

i would say fill the letters with a casting acrylic, would give a nice flush effect what you can shine led's through if you want.

You're a mind-reader, mate. Wait for a bit and you'll see what I mean. :D

THP Dan

09-05-2009, 01:14 PM

You're a mind-reader, mate. Wait for a bit and you'll see what I mean. :D

Can't wait :D

rubidium

09-05-2009, 02:02 PM

Here's a brassboard of the display circuit for the flowmeter. It's seen here reading the RPM of a standard fan (off to the side of the photo):

http://farm4.static.flickr.com/3535/3890523954_77f88a24c8_b.jpg

The large 28-pin integrated circuit is the Maxim MAX1496 (http://datasheets.maxim-ic.com/en/ds/MAX1447-MAX1498.pdf), which is a sigma-delta analog-to-digital converter (ADC) with on-board drivers for an external 3-1/2 digit, 7-segment, LED display. It runs about $12US. It's very versatile and can directly be used to accurately measure and display voltages. I'm using it to measure the voltage out of the smaller 14-pin integrated circuit to its right: the NTE995 (http://www.nteinc.com/specs/900to999/pdf/nte995.pdf) frequency-to-voltage converter. It runs about $3.75US. I have the NTE995 configured in this circuit to measure the frequency of the pulsating signal present on a standard RPM line of a fan, and output a voltage level which is proportional to that frequency for conversion and display by the ADC. The constant of proportionality is dictated by the value of an external resistor, for which I'm using a potentiometer (the small blue package with the set screw) at present. This circuit has been tested and proven to work with the Aquacomputer flowmeters that I said I was going to use. A simple choice of the external resistor value will scale the pulse rate of the flowmeter to gal/min for display - all for under $18US.

Now, in keeping with my stainless steel/copper theme, I’ve added a small, but challenging to machine, twist to the display. I’m going to do away with the 4-digit, 7-segment, LED panel used in the brassboard to test the circuit and replace it with an intricately-milled copper plate to be inserted into one of the display windows of the stainless steel beam. The plate will have a thru-slot for each of the 28 segments in the display, as seen in these renders:

http://farm3.static.flickr.com/2483/3889735061_e9edb8eb91_b.jpg

http://farm4.static.flickr.com/3441/3889736289_4f372cb993_b.jpg

http://farm4.static.flickr.com/3519/3890531718_d5bc96c18d_b.jpg

The copper plate is 0.25" thick and has overall dimensions of 3” x 1.5”, and the rounded-rectangular portion that will protrude through the cutout in the stainless steel beam measures 2.5” x 1.25”. These slots will be slowly cut through the plate, under computer control, with a 0.039” (1 mm) tungsten carbide end mill at 20,000 RPM spindle speed. Each of the segment slots will be straddled by a corresponding circular depression part-way milled into the back side of the plate to retain a 3mm red LED to back-illuminate the slot, as seen here:

http://farm4.static.flickr.com/3441/3889737599_a742118e50_b.jpg

and here:

http://farm3.static.flickr.com/2582/3890530452_1f439321a6_b.jpg

Each segment slot will be filled with translucent red acrylic (flush to the front surface) to provide uniform illumination from its corresponding LED. (This is why I told THP Dan earlier that he was reading my mind.:rofl:) I’m still experimenting with the details of how to do that, but basically I will take some existing translucent red acrylic sheet that I have and generate shavings with a wood plane. I’ll prepare a “syrup” having suitable viscosity by dissolving those shavings in methylene chloride, and inject the syrup into each LED-backed slot with a needle-tipped syringe. I plan to fill each slot to a level just slightly above the front surface of the plate, and then sand and polish the acrylic (once it hardens) and the front copper surface together. The 28 LED’s will be connected in a common-cathode configuration to be compatible with the metering circuit I designed above.

The 3 other holes between digits, by the way, are to support variable decimal point placement controlled by the Maxim chip, and will be given the same treatment with the acrylic. Likewise, I reserved 3 other holes in the panel, to the right of the 4 digits, to possibly serve other indicator functions that I might like to have later. The result should be something like this:

http://farm3.static.flickr.com/2543/3890532814_d2fe3c8a0a_b.jpg

I haven't even ordered the copper yet, so it will be some time before I have results. Also, I'm debating between laying out and fabricating nice PC boards for the circuit, or going the "Radio Shack route" with "perf board." I also still have to design a spring-loaded mount for the copper plate (and circuit) to retain it neatly from within the stainless steel rectangular tube without drilling holes through the front of the tube for external fasteners. I want to have as clean a look as possible. Once again, for a lot of this type of activity, McMaster-Carr is my friend. So there's a lot of work ahead. :coke:

rubidium

cheapskate

09-05-2009, 06:08 PM

I'd use the message box McMaster has to get them to wrap the copper in something. The packing dept. is getting lazy as hell.

Odin Eidolon

09-06-2009, 03:31 AM

someone said...

...PERFECTION?

p0Pe

09-06-2009, 03:54 AM

someone said fucking awsome:D

Oli4v

09-06-2009, 05:31 AM

Dude, you're doing awesome jobs againg and again :D

cheapskate

09-06-2009, 12:33 PM

When can we expect the next set of machinist's porn?

rubidium

09-06-2009, 01:45 PM

When can we expect the next set of machinist's porn?

In about a couple of weeks.:D

p0Pe

09-06-2009, 02:23 PM

In about a couple of weeks.:D

:( way to long time dude!!!

cyriene

09-06-2009, 05:15 PM

Great work rubidium! Someday I hope to attain a fraction of the skill you possess!

ballzd33p

09-08-2009, 12:50 PM

will you marry me? (that way half of it is mine)
i put out

p0Pe

09-08-2009, 01:01 PM

hey rubi:D you should give your quad system a go in this thread!

http://www.realredraider.com/vbulletin/showthread.php?t=12439

Greedseed

09-08-2009, 01:55 PM

realy nice project man :D

rubidium

09-08-2009, 05:46 PM

will you marry me? (that way half of it is mine)
i put out

:rofl::rofl::rofl:

rubidium

09-08-2009, 07:30 PM

hey rubi:D you should give your quad system a go in this thread!

http://www.realredraider.com/vbulletin/showthread.php?t=12439

Problem with cinebench R10 is that I'm running this as a Linux rig. Windows won't see more than 2 sockets, unless you buy the uber-expensive Server 2008 enterprise. I'm not so keen on Windows that I want to drop several $K on it. What a ripoff, especially when I tested Server 2008 during a free demo period early on, and found that it couldn't hold up for more than a few days without hanging. Ubuntu-64 is doing just fine for crunching and in supporting the scientific apps that I built the rig for in the first place, there's no limit to how many cores or sockets you have, it hasn't hung since I installed it in March, it's a lot leaner, and it's free! :)

n0t

09-09-2009, 09:27 AM

Rubidium,
This is fAmazing!!!!!!!
Way to Go!!!!
We want MOAR!!!!!!!!!!!!!!!

rubidium

09-19-2009, 04:07 PM

Progress continues on those copper displays, but I'm busy as hell at work writing a couple of proposals, so I haven't gotten to the point I hoped to be at by now. If all goes well, I'll hope to have some raw machined pieces to show here in a few days. Meanwhile, I got some electronic parts in and I'm completing the circuits that will read the pulsed signals from the Aquacomputer flow meters, convert to gal/min values and drive 4-digit, 7-segment, displays that will be implemented in LED-backed copper plates. Here are the nearly-completed circuit cards, made using simple perforated "Vector Circboard":

http://farm4.static.flickr.com/3511/3934584071_6915a33638_b.jpg

http://farm3.static.flickr.com/2651/3935364192_48edf1c617_b.jpg

http://farm3.static.flickr.com/2440/3935365134_e7bb1e3dd3_b.jpg

I still need to mount 3-pin fan headers (They'll go to the right of the blue potentiometer). These headers will connect directly to the Aquacomputer flow meters.

Odin Eidolon

09-20-2009, 03:56 AM

uber clean work!

rubidium

09-21-2009, 01:43 PM

After a very challenging machining effort this weekend, requiring considerable diligence and patience, I finally completed the 4-digit, 7-segment, copper display panels that I showed renders of earlier. These are not yet finished articles in terms of LED mounting, filling of the segments with translucent acrylic, and final polishing of the front surface, but I wanted to show them at this point in their development.

I started with 0.25” copper-101 plate and drilled a 3mm diameter and 0.125” deep hole for each LED on the rear surfaces. The LED's fit like a glove - just a very slight interference fit that retains them in their respective holes nicely:

http://farm3.static.flickr.com/2514/3942074378_da4d4d7059_b.jpg

Flipping the plate and meticulously realigning it in the machine vice with a dial indicator, each 0.039” wide and 0.170” long segment was then carefully milled through the entire 0.25” thickness of the plate using a 1mm end mill operating at 20000 RPM spindle speed. In the image above, you can see each thru-slot from the rear, and how it aligns with it's respective LED hold. Lots of coolant flow and a slow feed paid off, and I’m very pleased with the end result:

http://farm3.static.flickr.com/2509/3941292951_78d7c1091f_b.jpg

http://farm4.static.flickr.com/3418/3941291559_9fbd479fe5_b.jpg

http://farm4.static.flickr.com/3473/3941288577_72d8ce0664_b.jpg

http://farm3.static.flickr.com/2464/3941290109_5596273e30_b.jpg

A lot of work remains in finishing the electronics, mounting LED’s, and filling each segment with red casting acrylic. After I complete all of that, the front surfaces of the copper panels, together with the acrylic filling the segments, will together get the final polish to a mirror finish. :)

rubidium

Floyd

09-21-2009, 01:57 PM

Wonderful. I cant wait to see those in action!

Oli4v

09-21-2009, 02:14 PM

this is nice.

p0Pe

09-21-2009, 02:18 PM

i was litteretly jumping on the chair as i saw thoose pics! astuneshing work dude! (i might not know how to spell the word, but i damn right know what it means!)

really, really cant wait for more!

Baleful

09-21-2009, 03:02 PM

Damn! Awesome looking stuff man, can't wait to see the finished product.

funkysnair

09-21-2009, 03:07 PM

them copper display panels look sweet as m8.....

nice job there-keep up the good work ;)

cyriene

09-21-2009, 05:22 PM

This is pure badass

Greedseed

09-21-2009, 05:56 PM

sweet and perfect man =D wone see that beast up and running :P

rubidium

09-21-2009, 06:15 PM

This is how the LED's placed into the rear holes of the copper blocks are to be wired:

http://farm3.static.flickr.com/2592/3942921362_7d0396ff02_o.jpg

http://farm3.static.flickr.com/2457/3942922182_6356f0c00d_o.jpg

and then connections to the circuit card that I photographed earlier will be made like so:

http://farm3.static.flickr.com/2595/3942922994_fcd11debf8_o.jpg

This is going to be very tight wiring, as those LED's are asses-to-elbows.;)
rubidium

p0Pe

09-21-2009, 06:19 PM

gonna be good dude! we are currently learning about operation amplifiers in school. and next semestre, im hoping to build some sort of fan controller, and then further on something more complicated:D

love the ideas you are bringing out there!

cheapskate

09-22-2009, 02:23 PM

Yay! schematic and parts list in one shot. You rule!
With that bit size, you must of been working insanely slow. What was the IPM set to?

Snyxxx

09-22-2009, 03:47 PM

Yout attention to detail is impressive. Really good stuff here.

Nvidiator

09-22-2009, 05:25 PM

I think I am witnessing the creation of "Skynet".

Fucking awesome build dude:rockout::rockon:

Worthy

09-22-2009, 06:26 PM

Awesome.

n0t

09-22-2009, 07:09 PM

:htk::youtheman::rave:

MOAR!!!!!
PICs,PICs,PICs,PICs,PICs,PICs,PICs,PICs,PICs,PICs, PICs,!!!!!!

Worthy

09-22-2009, 08:26 PM

The machine work and finish are excellent.

Crow

09-22-2009, 08:29 PM

I would pay money for one of those...

Snyxxx

10-01-2009, 10:54 AM

Did you use your home brew copper preservative on the copper in your photos?

rubidium

10-01-2009, 04:28 PM

Did you use your home brew copper preservative on the copper in your photos?

No, not yet. The segments on these panels will have to be filled with casting acrylic and the front surfaces re-polished before I can do that.

just a noob

10-06-2009, 05:40 PM

how much longer until we get some updates? :coke:

Greedseed

10-06-2009, 06:01 PM

yeah we wont update's :D:D

just a noob

12-11-2009, 09:25 PM

any updates yet? :coke:

Kuro

12-11-2009, 10:41 PM

mm like 2 months and no updates

rubidium

12-12-2009, 10:16 AM

Sorry for the extended run of silence on this project, but I've been swamped between extensive business travel, writing two major proposals at work, completely remodeling a bathroom at home, and a half-dozen other things that haven't left me with much spare time. When it rains - it pours. The travel has subsided, the proposals are complete, and the Mrs. is finally happy with the shithouse (and nothin' leaks) so my neck is just now getting free of all those albatrosses.

I've only been spot-checking here lately at the forum to occasionally throw out a comment/suggestion or two here and there. Next, I'm off on a much-needed vacation for two weeks. I'll get back on track with this project come the new year. :crack:

THP Dan

12-12-2009, 10:30 AM

Good to hear it :)
Im looking forward to the new year now :p Cant wait...

RedRaider

12-12-2009, 12:38 PM

Good to hear from you.

Now get your ass back on the forum:)

KaosII

12-20-2009, 08:31 PM

I'm glad to see you got it going...:up:

I will probably always be leery of FUFUFUFUFUFUFUFU, no plans to ever buy their products. But that's just me...

It seems they are trying real hard to redeem their history of past f*ck-ups only to to keep repeating them. At least they are try to stay on the copper bandwagon, but still have a love for mixing metals.

cheapskate

12-21-2009, 11:49 AM

Awww.... I though there was an update.:( -And here I just learned not to get tricked by someone bumping Langer's thread!

Kuro

12-21-2009, 04:28 PM

:rofl: I thought the same too

anyway have a nice holiday mate :up:

Septim

02-04-2010, 04:46 AM

tagged!

thread bump for new members and old.

this project never ceases to amaze me.

btw, we want some updates, after the new years and the 2 weeks vacation for yourself... some nice bad ass pictures would tide our appetites...

coolmiester

02-04-2010, 05:25 AM

One of the most insane builds on any forum bar none, and definitely worth going back to page 1 to remind ourselves just how awesome this build is.

Good call on the bump Septim

Here's hoping for an update in the not to distant future :crack:

just a noob

02-06-2010, 09:13 PM

hey rubidium, if you don't mind me asking, could to take a picture(sometime) of the screen on that colorfontz display(or link it)?

rubidium

02-07-2010, 08:24 AM

hey rubidium, if you don't mind me asking, could to take a picture(sometime) of the screen on that colorfontz display(or link it)?

See post #25 here (http://www.realredraider.com/vbulletin/showthread.php?t=169&highlight=crystalfontz) for now. I'll take a few more shots shortly.

just a noob

02-07-2010, 08:30 AM

thanks rubidium, I'm just trying to decide whether I want a colorfontz, or an aquero

rubidium

02-07-2010, 09:00 AM

The CFA-635 on my Erinyes rig is down right now, since I'm experimenting with Ubuntu. But I have a 635 running CrystalControl2 under XP-64 on my older rig. Here are some fresh shots from there.

Memory utilization:
http://farm3.static.flickr.com/2798/4337011447_b00246b5b6_b.jpg

HDD utilization (idle):
http://farm5.static.flickr.com/4008/4337012443_abaabbfbca_b.jpg

HDD utilization (copying a file from one folder to another on the same drive):
http://farm3.static.flickr.com/2732/4337013321_c2476fac44_b.jpg

And - my favorite - live BOINC stats:
http://farm5.static.flickr.com/4064/4337758742_ed3c4efa0b_b.jpg

BTW, the 4 LED's on the left are set up to be live indicators of CPU utilization. As this is a 4-core rig, each LED corresponds to a core. Green=0% through Yellow, Orange, up to Red=100%. Since I'm crunching, all 4 LED's are pinned at red.

Edit: The actual appearance of the display is much more refined than the images indicate. It's tough to get the exposure right and avoid one or more discrete light sources from "flaring up" in a case like this.