Stress Test System

[xmanrigger]

Hey guys, I will be unloading my Sabertooth build I just finished. I have about a week to tweak everything.

Although this machine has run stable with everything I threw at it so far, I want to be sure as it is going to a client. I have the CPU at 3.8GHZ, the memory is around 1866mhz, and a slight oc on the GTX580s.

Without frying anything, I want to stress everything over a few hrs to ensure stability for the client. Is there a single app to do this, or do I need to use a couple? Also, what apps are recommended and run time? I wouldnt mind running Furmark on the GPUs, but worried they might cook.

I have only sent 1 client machine out overclocked and it has been fine to this point. This build however is considerably more expensive so I want a good piece of mind it is stable.

Thanks, 'rigger

[cpufrost]

OCCT 4. is nice and new...

http://majorgeeks.com/OCCT_d5612.html

Beware if using the PSU test though. It runs (essentially furmark) and Linpack AT THE SAME TIME! It will literally rape your power supply!

The stress test is pretty good though. Set it to 8 hours and let it fly.

[xmanrigger]

OCCT 4. is nice and new...

http://majorgeeks.com/OCCT_d5612.html

Beware if using the PSU test though. It runs (essentially furmark) and Linpack AT THE SAME TIME! It will literally rape your power supply!

The stress test is pretty good though. Set it to 8 hours and let it fly.

Thanks, will give it a go.

[Snyxxx]

Agreed with OCCT.

Also, I am still partial to Prime 95, blend test. I go about 12 hours and call it good if no errors.

Further, I boot MEMTEST and run a few passes to just be sure.

[xmanrigger]

OCCT 4. is nice and new...

http://majorgeeks.com/OCCT_d5612.html

Beware if using the PSU test though. It runs (essentially furmark) and Linpack AT THE SAME TIME! It will literally rape your power supply!

The stress test is pretty good though. Set it to 8 hours and let it fly.

OK, I got it cooking away with the CPU: OCCT test running and set for 8hrs. What is the difference between the CPU operations with CPU: OCCT and CPU:LINPACK tests? As far as I can tell, LINPACK also stresses the sys memory. Am I correct?

Also, can I, or would I even want to run both OCCT and LINPACK at same time?

Lastly, is there any issue running either CPU test and the GPU test in tandem?

Thanks, 'rigger

[cpufrost]

GPU stress is Furmark, CPU stress is either Linpack or OCCT's blend of stress tests. I find that using IBT or LinX more convenient than using OCCT Linpack.

If you run both GPU and CPU stress test simultaneously that is the PSU test. Nothing you do with your system (normally) will put as much load on the PSU as that test.

[B NEGATIVE]

You want 10 passes of LinX for CPU,12 hrs on P95 blend for Memory and CPU and GPU stress using furmark or kombustor.

[cpufrost]

Honestly I don't trust Prime on Nehalem/Gulftown CPUs any more. You actually want to use small FFT if you use it.

[B NEGATIVE]

sml fft doesnt stress the memory,although your running an intel. AMD clocking involves CPU NB oc'ing which is tested with blend.
Prime95 is still to my mind the best for 24/7 high OC crunching/folding machines.

[cpufrost]

My last AMD system was an FX60 (which to this day is crunching!). Yes I did test with Prime back then. I used Prime until I found it just was not catching errata in extreme overclocks with Nehalem processors in early 2009.

Of course if you can crunch/fold 24/7 for weeks on end without crashing or getting computational errors chances are the system should be adequately stable. :)

[p0Pe]

Coredamage is also great for cpu testing. Have not found anything that raped the cpu more.

[cpufrost]

Any details?
Does it check for errata or do you just wait for the machine to crash? :D

[p0Pe]

Just wait for the machine to crash:) If it does not crash, its stable lol

[xmanrigger]

Ok, so I have decided on final clock speeds for the Sabertooth system and I am now going to give it the final go-around before delivery.

I am using OCCT as suggested. My final clock settings are as follows:
CPU - i7 950: 3.65GHZ / 3.8GHZ-Turbo. 1.3v (Left Turbo function enabled)
Memory - Dominator GT 1866mhz: 1907mhz. 1.65v
VGA - EVGA GTX580 X2-SLI: 797mhz / 2025mhz (Yes, not much of an OC, but hey, no voltage increase-1.0v, and he is getting SuperClocked speed at reference cost.)
PSU: AX850

I presently have it at these speeds running the PSU test, set to 8hrs, essentially stressing entire system. Is this a bad idea?

I want this thing rock solid before delivery.

Thanks, 'rigger

EDIT AND UPDATE: 2 hrs: highest CPU core = 70C, GPU(s) = 44/47C, water temp = 39C (23C ambiant air). Temps good ? bad? See below.

WC = 1X TC TA120.3 / 1X MCP350 with XSPC rez-top / 1X HK Rev3.0 / 2X HK GPU-X3 in parallel / 1/2" X 3/4" tubing

[cpufrost]

39C water temp seems high to me but your cpu/gpu temps otherwise are fine.
If you are running the PSU test also record your PSU exhaust air temp.

[xmanrigger]

After 2+ hrs, I am not liking the water temp at 39C and holding. I am thinking a second 140.2 rad maybe required in the loop, or run a second loop with the 140.2 for the CPU alone.

Suggestions? A second pump in paralell or bigger pump with existing system maybe? I would have originaly incorporated a MCP655, but already had the 350.

[cpufrost]

Adding a pump isn't going to lower your water temperatures...as a matter of fact the additional dump will raise it. (albeit slightly) Pumps in parallel definitely are not recommended.

The only way to lower that water temp (aside from cooling down your room) is to increase radiator capacity. If the pressure drop from the additional plumbing and radiator(s) causes the flow to drop too much another pump may be warranted (but in most cases un-necessary).

Sixteen degree delta is a bit excessive though. ;)

[xmanrigger]

Adding a pump isn't going to lower your water temperatures...as a matter of fact the additional dump will raise it. (albeit slightly) Pumps in parallel definitely are not recommended.

The only way to lower that water temp (aside from cooling down your room) is to increase radiator capacity. If the pressure drop from the additional plumbing and radiator(s) causes the flow to drop too much another pump may be warranted (but in most cases un-necessary).

Sixteen degree delta is a bit excessive though. ;)

The only reason I mentioned dual pumps or a bigger pump, it that I originally bought the MCP350 on the premise that I was using a Swiftech 120.3, different routing, and the cpu with only a single GPU block. I have a few 90s in the plumbing also.

And ya, the delta seems high as I mentioned above, but temps otherwise 'normal'.

[cpufrost]

Where are you measuring water temp?

[xmanrigger]

Well, it passed the 8HR PSU test. All indications are that this bitch is very stable. I now have a good piece of mind selling this.The guy buying the machine will never realize loads even close to that.

As cpufrost suggested and I noticed, the delta was quite high, at about 16deg. That said, the system temps were satisfactory in my opinion for the clock speeds, loads, and duration. Ambiant air was about 23-24C, and after about 15 minutes of full load, water temp was at 39C and held there.

This acceptable this time, but in the future and as I did with my previous water build, bigger pump and more rad. This build was originally only going to be 1X GTX480 w/ BP Black Freezer, so it would have been great. But I finished it with 2X GTX580s. Also as I said above, the dude likely wont even see half the potential of this bitch, so cooling will be sufficient....this time.
http://img64.imageshack.us/img64/3975/3814190779720258hrscrop.jpg (http://imageshack.us/photo/my-images/64/3814190779720258hrscrop.jpg/)

If somebody can decipher these, can you tell me how the PSU faired?
http://img856.imageshack.us/img856/5536/2011081500h2712v.th.png (http://imageshack.us/photo/my-images/856/2011081500h2712v.png/)
http://img845.imageshack.us/img845/3741/2011081500h275v.th.png (http://imageshack.us/photo/my-images/845/2011081500h275v.png/)
http://img850.imageshack.us/img850/6627/2011081500h2733v.th.png (http://imageshack.us/photo/my-images/850/2011081500h2733v.png/)

Thanks, 'rigger

[xmanrigger]

Where are you measuring water temp?

Between the pump o/p and rad i/p. I guess I should likely move it to the rad o/p?

[p0Pe]

Between the pump o/p and rad i/p. I guess I should likely move it to the rad o/p?

You wont see much of a change. And where you have it there is the best position, as it shows the temp where the temps #should# be highest in the loop.

The difference between radiator input and output is only a few degrees, i would guess:D

[xmanrigger]

You wont see much of a change. And where you have it there is the best position, as it shows the temp where the temps #should# be highest in the loop.

The difference between radiator input and output is only a few degrees, i would guess:D

Thanks. Makes perfect sense. Will leave as is. The probe is were it is at because I had to put a Q-Block there for the drain valve, so why not plug a hole with the probe.

[cpufrost]

The more difference you see the more of a flow hindrance you may have. In ideal conditions temperature of the water should be nearly the same in the entire loop regardless where it's taken. Again actual conditions are always never ideal, but your delta is rather large. If water temp everywhere is nearly the same with a delta that high you're definitely limited by your rad(s).