Anyone use LinPack for stresstesting?

[n7]

Get it here: http://www.intel.com/cd/softwa...asmo-na/eng/266857.htm

I saw some posts on XS mentioning this as the ultimate RAM test, so i decided to run it this morning on my supposedly stable system...
Ended with a BSOD.

It completely thrashes the CPU as well...temps seem to go higher than P95.

I don't really know how the bench works, but at some points it uses ALL of my 8 GB!
Pic of task manager (running runme_xeon64.bat): http://img204.imagevenue.com/i..._testing_122_546lo.jpg

I don't know how long it goes on for, but the system is basically unuseable during due to all RAM being used.

 

[Sentry2]

I'll have to check it out. Thanks.

 

[Andvari]

LOL, whoa...

 

[BlueAcolyte]

Wow. Is that safe?

 

[n7]

So ran a bit of P95 & Linkpack, & it turns out that temps spike up to ~7-8C higher than P95 :Q
It's not a consistant load on every core, so they don't stay the same temps for more than a few seconds, but it's certainly way hotter than P95!

 

[ch33zw1z]

Possible to run Linpack on XP Pro?

 

[n7]

Originally posted by: ch33zw1z
Possible to run Linpack on XP Pro?

Yeah, you'll just want to run the 32-bit .bat instead of the 64-bit one like i have been.

http://img177.imagevenue.com/i...k_folder_122_462lo.jpg

Run the runme_xeon32.bat (or runme_xeon64.bat for x64 systems).

I've been reading more on this test.

It's a lot more of a beating than P95 is, others' results have confirmed mine.

So if you want something that's significantly hotter on the CPU in portions, not to mention uses RAM far harder, this is the stresstest to run...just don't expect to use the PC during...

 

[Idontcare]

This is pretty neat, seems to be similiar idea as Prime95's "Blend" stress testing at first assessment - tons and tons of matrix math.

Prime95 might be able to reach these same temps if they implemented a stress routine using FFT Lengths <8K. I am sure it can be done by setting a manual test regiment but who wants to deal with doing that? This Intel package may very well be the newest "power bug" stress test.

Can't wait to try it out on my system. Thanks for posting the link for us.

 

[ch33zw1z]

Runs great, I left it on while I was not around and it was finished by the time I got back. Wish it just kept running!

 

[Idontcare]

I haven't ran it yet, am still working on getting a stable OC for Prime95 ATM.

So what does it tell you when it is done? I am very reward driven, so what's the reward for completion? Does a window popup exclaiming "Complete, you are an OC'ing champion!"?

 

[ch33zw1z]

Originally posted by: Idontcare
I haven't ran it yet, am still working on getting a stable OC for Prime95 ATM.

So what does it tell you when it is done? I am very reward driven, so what's the reward for completion? Does a window popup exclaiming "Complete, you are an OC'ing champion!"?

Based on what I've read, Review=Passing. lol

 

[zeroburrito]

lol @ all these torture tests. if it runs all the games you play without crashing thats called stable. or you can do these torture tests and play the same games at a lower speed. no games/apps use anywhere near the resources these programs use. you're robbing yourself of ultimate power.

 

[ch33zw1z]

Originally posted by: zeroburrito
lol @ all these torture tests. if it runs all the games you play without crashing thats called stable. or you can do these torture tests and play the same games at a lower speed. no games/apps use anywhere near the resources these programs use. you're robbing yourself of ultimate power.

lol @ people who test their oc with games. Fuck gaming.

 

[n7]

Originally posted by: zeroburrito
lol @ all these torture tests. if it runs all the games you play without crashing thats called stable. or you can do these torture tests and play the same games at a lower speed. no games/apps use anywhere near the resources these programs use. you're robbing yourself of ultimate power.

I get your point of view, but overall, i disagree.

Yes, it's true that most games won't stress things as much as program designed to, but the reality is, random BDODs or crashing is actually quite common if you don't really have a stable system.

Sure, 99% of the time, things are fine, but there's nothing worse than being in a crazy battle in a MP game online & then locking up or crashing because your OC wasn't quite stable.

I often get annoyed with attaining absolute stability myself, so i'll get lazy & only run P95 for an hour, or won't finish Memtest, & then i suffer the consequences.

So while i can understand your perspective, my own personal experiences with not using stresstesting apps & assuming stability have never worked out well, & i think most would agree.

 

[Idontcare]

Originally posted by: zeroburrito
lol @ all these torture tests. if it runs all the games you play without crashing thats called stable. or you can do these torture tests and play the same games at a lower speed. no games/apps use anywhere near the resources these programs use. you're robbing yourself of ultimate power.

Oh my, that is a rather self-centric perspective on why people stress test their rigs.

Do you really believe that your experiences and utilization of your computer is representative of what everyone else does with their computers?

What if I happen to be part of GIMPS and I "prime" all the time with my rigs? Wouldn't I be best served in stress testing my rig with Prime95?

Or what if I use my rig for HPC type tasks, computational chemistry and the like? Wouldn't I be best served stress testing my rig with Linpack?

Had you wrote "lol @ all these torture tests...insofar as I use my computer these things are not beneficial" then that would be a different, and alltogether more agreeable statement.

 

[VirtualLarry]

Intel(R) LINPACK data

Current date/time: Sat Jan 19 09:23:36 2008

CPU frequency: 3.102 GHz
Number of CPUs: 2
Number of threads: 2
Parameters are set to:

Number of tests : 9
Number of equations to solve (problem size) : 15000 14000 13000 12000 11000 10000 8000 6000 1000
Leading dimension of array : 15000 14008 13000 12008 11000 10008 8008 6008 1000
Number of trials to run : 1 2 2 2 2 2 2 3 4
Data alignment value (in Kbytes) : 4 4 4 4 4 4 4 4 4

Maximum memory requested that can be used = 1800304096, at the size = 15000
============= Timing linear equation system solver =================

Size LDA Align. Time(s) GFlops Residual Residual(norm)
Error: info returned = 3277176
15000 15000 4 332.974 6.7586 2.196558e-010 3.459615e-002
Error: info returned = 3277176
14000 14008 4 204.010 8.9688 1.886800e-010 3.407360e-002

Hmm, why did I get errors? It looks like it didn't even run.

E2200 freshly OCed to 3.1Ghz, it wouldn't get to 3.2. Seems somewhat prime stable so far.

 

[Idontcare]

Originally posted by: VirtualLarry
Hmm, why did I get errors? It looks like it didn't even run.

E2200 freshly OCed to 3.1Ghz, it wouldn't get to 3.2. Seems somewhat prime stable so far.

What does it do if you run it at stock on your computer? I'd check for myself but I am still priming for stability atm.

 

[VirtualLarry]

How long does this benchmark take to run? It's been 15 minutes so far and its still going.

I rebooted at 2.2Ghz, went to run the benchmark, and it didn't start the first time. Temps were still at idle. Now it's running, and temps are peaking at 50C. Prime95 runs at 60C, well, at 3.1Ghz. I dunno at 2.2Ghz. The whole system is running rather slow while this benchmark runs.

This is at stock:

Intel(R) LINPACK data

Current date/time: Sat Jan 19 09:59:02 2008

CPU frequency: 2.195 GHz
Number of CPUs: 2
Number of threads: 2
Parameters are set to:

Number of tests : 9
Number of equations to solve (problem size) : 15000 14000 13000 12000 11000 10000 8000 6000 1000
Leading dimension of array : 15000 14008 13000 12008 11000 10008 8008 6008 1000
Number of trials to run : 1 2 2 2 2 2 2 3 4
Data alignment value (in Kbytes) : 4 4 4 4 4 4 4 4 4

Maximum memory requested that can be used = 1800304096, at the size = 15000
============= Timing linear equation system solver =================

Size LDA Align. Time(s) GFlops Residual Residual(norm)
Error: info returned = 3277176
15000 15000 4 485.191 4.6383 2.196558e-010 3.459615e-002
Error: info returned = 3277176
14000 14008 4 405.053 4.5172 1.886800e-010 3.407360e-002
Error: info returned = 3277176
14000 14008 4 409.814 4.4648 1.886800e-010 3.407360e-002
Error: info returned = 3277176
13000 13000 4 331.926 4.4137 1.530559e-010 3.202580e-002
Error: info returned = 3277176
13000 13000 4 327.745 4.4700 1.530559e-010 3.202580e-002
Error: info returned = 3277176
12000 12008 4 256.934 4.4848 1.331741e-010 3.268739e-002
Error: info returned = 3277176
12000 12008 4 254.774 4.5228 1.331741e-010 3.268739e-002

Still getting errors. I dont get it.

 

[PCTC2]

I think it pushes temps above Prime because of it's random CPU usage. This is exactly what vDroop was specified for. During 100% load, the CPU draws more amperage at the given voltage (thus more wattage at load). But during a cycle of cpu usage, when it drops from 100% to 0%, the amperage needed drops but the motherboard can't react quick enough and supplies the increased wattage, and since it doesn't need the amps, voltage spikes. vDroop was designed to limit the spikes to below maximum rated voltages. Now, LINPACK, with it's random spikes of usage, would cause the processor to repeated experience these spikes in voltage, causing more heat than the constant load of Prime95, thus the higher temperatures.

BTW, nice find n7

EDIT: Oh yeah, prime95 25.2 only pushes my system to peak at 55C on the highest core. LINPACK pushes my system peak to 62C on the highest core.

 

[VirtualLarry]

I have the opposite. Prime95 25.5 small fft stress my CPU temp more than Linpack does.

Btw, does any else's system error out like mine does?

 

[Cheex]

I will have to check this out. It is a must.

Will post back here with my 'results'.

 

[PCTC2]

Originally posted by: VirtualLarry
I have the opposite. Prime95 25.5 small fft stress my CPU temp more than Linpack does.

Btw, does any else's system error out like mine does?

Sorry. it just crashed my computer.

Mine is Prime95 v25.2 stable and passes a day of memtest but not LINPACK stable....

 

[CurseTheSky]

At stock speed, the temperatures from LINPACK were the same as TAT. Seems like a pretty good stability test, though. TAT is much quicker at heating things up, but, of course, there's no output data other than temps.

 

[CTho9305]

Highly optimized applications / math libraries often produce very high power consumption because the people designing them work to get every last bit of performance out of the CPU - and this often means eliminating stalls from the pipeline by carefully ordering instructions and tweaking the algorithm's block sizes (where possible) to fit within the L1 cache (eliminating more stalls). When the pipeline isn't stalled, it's consuming more power, so the chip gets hotter.

 

[Cheex]

Originally posted by: CTho9305
Highly optimized applications / math libraries often produce very high power consumption because the people designing them work to get every last bit of performance out of the CPU - and this often means eliminating stalls from the pipeline by carefully ordering instructions and tweaking the algorithm's block sizes (where possible) to fit within the L1 cache (eliminating more stalls). When the pipeline isn't stalled, it's consuming more power, so the chip gets hotter.

Great information...:thumbsup:

Still haven't had a chance to try it yet though...