Lets put your CPU into REAL test! Intel Linpack 64bit

[Falkentyne]

It saddens me to see how XS is becoming XT(XtremeTesting).

All you guys that think a benching rig should be primed or whatever for hours on end can rip me all you want. I just hate to see the forum go this direction.

Maybe there should be a section at XS for all those who what to show how many hours of prime or whatever there rig can do. Instead of spi or 3D benching... it could be whoever can prime or do Linpack the longest has the most stable computer.

I can't help but agree.
How can someone say a system isn't stable if only a non real world application like this linpack is the only thing that can bring it down?
This is a bit silly...I mean.why not take it to the next level ? Turn on the heater and heat up your room to 40 degrees Celsius (104 F) and then try benching, all while in a closed case !! That extra 15 C sure can't hurt, now can it? You want super stability, lets push things to the limit of physical AND hardware endurance!

Next thing people will be trying to run linpack with passive cooling on 100 degree F days to see if their system can REALLY deal with the worst you can offer it.

I hope I'm beginning to make a point here....

NOW LINPAK IS GOOD for fun testing, just to see what you have to do to bring the system down, and its an interesting bench, but it should NOT be used to determine if a system is completely stable or not, unless EVERYTHING ELSE (heatsinks/cooling, ambient temps, etc) is STANDARDIZED so that everyone uses the exact same setup, which..is...well....impossible....

[Falkentyne]

this thing made my cpu 2c hotter than 8K FFT prime. not bad at all

Except if I recall correctly, Intel TAT (which won't work on a QX9650, in fact it actually completely messes up the CPU speed (gets throttled to 300 mhz until you uninstall)), made the temps on my old X6800 around 10C higher than 8k Prime, on a Ultra 120 (not True). 2C doesn't seem like an average TAT gain unless you're on some extreme cooling...
I remember TAT giving temps similar to that old BurnBX/BurnP6 program....

[road-runner]

Maybe set up to run your rig in a oven at 250F and time how long it lives...

[jcool]

If you don't wanna run it, just don't for ****'s sake
We heard enough already about people who disapprove in this thread. It may not be important for your benching or gaming rigs to be stable in every situation, but it certainly is crucial for a workstation-class system running under full, continuous load. All day and every day, come hell or high water.

By the way, I fixed the "error" in linpack with 1,30V MCH setting. Strange that prime blend survived 10h at 1,25V but what the heck.

[overclockerd]

i hope my cpu is gonna survive from this test:P

[Demo]

Here are my thoughts on the program after understanding how to play around with it;

It is indeed better that prime/occt/memtest86+/dual orthos/superPi/wPrime , its only weakness that it wont hold 100% cpu usage, it fluctuates around 98~100% and drops to 25% at times ( quadcore )
This program can really heat up the CPU and introduce further vdrop when loaded and increased PWM/NB/RAM temps ALOT! Think of this program as a burst-stress-testing, but still cannot replace prime 95 v25.6

Prime95 v25.6 was the king of heating up and rocking down my system, but now this rewrote the books in terms of stability testing on my end, even though i have no application remotely close to what prime does it still want my entire rig 100% stable on any given program.

My prime 95 v25.6 72hrs stable settings are as follows
400*8
1.335v bios volts ( 1.30x windows idle & 1.284 prime loaded small ftt )

with Linpak my load volts are 1.230 ( a 0.075 vdrop w/c is out of spec already )

And im getting a BSOD w/ an error of "Page fault in non paged area" or something along those lines, im not 100% sure of what voltage to increase so i upped the vcore by 2 notches in bios 1.375 and the vDIMM at 1.95 & vMCH at 1 notch of 1.29

EDIT:
Can anyone correct the settings i made, i seem to get errors about incorrect parameters but the test will run though

Sample Intel(R) LINPACK data file (lininput_xeon64)
Intel(R) LINPACK data
12 # number of tests
26000 # problem sizes
26000 # leading dimensions
5 # times to run a test
4 # alignment values (in KBytes)

I get his error

Warning: incorrect parameter 'Number of equations to solve (problem size)' (1),
must be not less than (1),
set to default value (1).
Warning: incorrect parameter 'Number of equations to solve (problem size)' (1),
must be not less than (1),
set to default value (1).
Warning: incorrect parameter 'Leading dimension of array' (1),
must be not less than (1),
set to default value (1).
Warning: incorrect parameter 'Leading dimension of array' (1),

[Demo]

To run the test, run runme_xeon(32,64).bat. The best way is to run it from Far manager (google Far manager and download it, its free) or similar program.
The results will be in file with the name lin_xeon(32,64).txt. The progress can be conveniently monitored in another Far instance with F3 viewer.

Maybe im doing something wrong in this part, I downloaded Far manager and ran the 64bit batch file using the far manage and i get an error saying "Could not find executable PKZIPC" but the 64bit batch file runs properly outside far manager..

[Isira]

12 # number of tests
This should be 1.

5 # times to run a test
Should definitely be raised. Just set it to a high number like 100 so it can run for a while. You can kill the program anytime and check the log files.

My CPU was 1.47V Prime95 stable, but required 1.52V to be Linpack stable.

As far as raising voltage, I would first loosen your RAM timings so you can focus on stabilizing your CPU. Once you find your stable voltage for the CPU, you can work on VDIMM/VMCH for tightening memory.

Before running Linpack, I like to use SuperPrime (Uses Linpack for its engine) as a quick stability check.

[Demo]

12 # number of tests
This should be 1.

5 # times to run a test
Should definitely be raised. Just set it to a high number like 100 so it can run for a while. You can kill the program anytime and check the log files.

My CPU was 1.47V Prime95 stable, but required 1.52V to be Linpack stable.

As far as raising voltage, I would first loosen your RAM timings so you can focus on stabilizing your CPU. Once you find your stable voltage for the CPU, you can work on VDIMM/VMCH for tightening memory.

Before running Linpack, I like to use SuperPrime (Uses Linpack for its engine) as a quick stability check.

Thanks for the input, The memory timings are already very loose, its 5-5-5-15

[ZOMGVTEK]

LOL!

I was a hour prime stable, and my system BSOD'd after 5 seconds of this.... CPU was touching 80+C!

[jonny_ftm]

Running your system (CPU and memory especially, but also NB if you overvolt it too much) at extreme temperatures will only cause irreversible damage yielding to overclock potential diminishing.

Furthermore, it won't help you to say your overclock is stable or not since if you run your system out of temp specs, the temperatures would be the limiting factor, not your inadequate settings.

I won't say Linpack is great or crap, every benchmark is a plus for people looking into ultimate stability in addition to an xtreme benchmark exciting expierience.

Personally, I don't want to run my CPU at +70°C or anything above 65°C on cores. For people running their system at the limits of their cooling capabilities and fearing Linpack overheating damaging issues, the best solution is to profit from cold nights/days, and run your session at open window so that your CPU won't go above the 60-65°C. If your benchmarking session is stable, than for daily PC use, just monitor your temps and assume 65°C as a limit. If you notice, daily, some application exceeding your fixed limit you're left to either improve your cooling or lower your overclock. If it is unstable, at least you know it is not a heat issue and you should tweak your settings. Of course, keep at least temperatures between 55-60°C, as lower temperatures, especially in the 40°C zone could need less vcore and mislead your reslts

But definately, running any benchmark at +70°C for the quest of stability is a silly autosuicide nonsense behavior: Control the temp, even by cold weather and get your system stable for those extreme benchmarks, that's how it should be done.

[Demo]

@jonny_ftm

Good point, on a 65nm i think intel rates it at 72c, so 70c flat should be good enough.. the 45nm though are rated lower at 65c if i remember correctly.. its pretty bad for people who live in hot places..

[jonny_ftm]

@jonny_ftm

Good point, on a 65nm i think intel rates it at 72c, so 70c flat should be good enough.. the 45nm though are rated lower at 65c if i remember correctly.. its pretty bad for people who live in hot places..

Benching 18h a 65nm CPU at 70°C or even above 65°C is a bad idea, since, compared to intel specs, that temperature is added to excessive vcore usually on overclocked systems. Really no need to kill your silicone while checking the stability. Use any benchmark tool, but keep temperature ideally in the 50-60 during bench session. At least, you're not cooking silicone that way

[Demo]

During afternoons i get about 62~64c temps, during the evening is a health 54~58, all this priming small ftt.. with linpak in moves up to 68c tops

[jonny_ftm]

During afternoons i get about 62~64c temps, during the evening is a health 54~58, all this priming small ftt.. with linpak in moves up to 68c tops

Great example. In such a case, it is really useless to continue killing the sytem with linpack at 68°C. One should better wait for a cold night, open the window, and run linpack to maintain a T°C around 55-60°C even if I'd recommand 50-55°C. If linpack is unstable, than the OC settings need tuning. Remember that in most daily usage of a PC, no one will reach linpack temperatures unless some very hot summer nights coming soon. And we're supposed to solve the heating issue rather than the nonsense reflex of increasing vcore to stabnilize an overheating system.

[rge]

I think Linpack32 (cant speak for linpack 64) works well for its intended purpose, which is heat stressing. But I still have yet to find a use for it for stability testing. At 4.1ghz, I am very unstable at 1.31v vcore. Running 10ffts on orthos, or prime 95, or cpu occt if I run it 10x in a row, all 10 times will fail in 30 seconds to 5 minutes, most being 1-2 minutes. But I can run Linpack for 30 minutes repeatedly without error. Although it did make an error at 60 trials, about 40 minutes into it.

Though Linpack32 will eventually find an error, it is inefficient at finding unstable overclocks in my setup. On the other hand, since it does increase temps 5C above prime, it might tell you if you are still stable at 5C higher temps, if that is important for those without A/C control in summer.

Linpack 64 may be better, maybe someone with 64 bit could post some pics of it compared with other cpu testing.

[Demo]

Well on my end, increasing the voltage 1 notch more from 1.335 to 1.355v only increases the heat by 2c more.

[jonny_ftm]

I think Linpack32 (cant speak for linpack 64) works well for its intended purpose, which is heat stressing. But I still have yet to find a use for it for stability testing. At 4.1ghz, I am very unstable at 1.31v vcore. Running 10ffts on orthos, or prime 95, or cpu occt if I run it 10x in a row, all 10 times will fail in 30 seconds to 5 minutes, most being 1-2 minutes. But I can run Linpack for 30 minutes repeatedly without error. Although it did make an error at 60 trials, about 40 minutes into it...

I already confirmed Linpack32 ineffectivness compared to OCCT or Prime95 in a previous post, despite the overheating. I just didn't bother retesting to give sceptics some screenshots
I also can't talk about Linpack64, maybe once I migrate to Vista64 I will give it a new comparisions tests and post my findings

[Demo]

What linpak64 lacks is a constant 100% CPU usage.

[Demo]

@rge

How do you get a real time monitor of the progress ?

[rge]

By real time monitor, do you mean the cmd screen as shown?

I just double click on linpack32, type in the variables, and then stretch the cmd screen size, and hit print screen just before it finishes. Or did you mean something else?

[graysky]

Dunno what to make of this.... my system is p95 stable for >30 h under these settings. I guess this benchmark confirmed the stability since the numbers are identical:

Intel(R) LINPACK data

Current date/time: Wed May 21 15:00:58 2008

CPU frequency:    3.400 GHz
Number of CPUs: 4
Number of threads: 4
Parameters are set to:

Number of tests                             : 1
Number of equations to solve (problem size) : 20000
Leading dimension of array                  : 20000
Number of trials to run                     : 100  
Data alignment value (in Kbytes)            : 4    

Maximum memory requested that can be used = 3200404096, at the size = 20000
============= Timing linear equation system solver =================

Size   LDA    Align. Time(s)    GFlops   Residual      Residual(norm)
20000  20000  4      117.053    45.5704  4.455000e-010 3.943651e-002
20000  20000  4      116.505    45.7846  4.455000e-010 3.943651e-002
20000  20000  4      116.076    45.9537  4.455000e-010 3.943651e-002
20000  20000  4      116.418    45.8190  4.455000e-010 3.943651e-002
20000  20000  4      116.994    45.5934  4.455000e-010 3.943651e-002
20000  20000  4      115.492    46.1864  4.455000e-010 3.943651e-002
20000  20000  4      117.306    45.4718  4.455000e-010 3.943651e-002
20000  20000  4      115.621    46.1346  4.455000e-010 3.943651e-002
20000  20000  4      117.541    45.3809  4.455000e-010 3.943651e-002
20000  20000  4      118.747    44.9200  4.455000e-010 3.943651e-002
20000  20000  4      117.721    45.3116  4.455000e-010 3.943651e-002

It fcking scorched my cores. I ran HDMon which logs highest temps. I hit 69 °C during the tests; p95 by contrast didn't break 63 °C. If I open up my case, the 69 °C has dropped to 63 °C with this thing but damn!

Something else I found striking (perhaps a major knock against this util as a stability app) is that despite a 100 % core load on all 4 cores as-per task manager, I'm seeming some pretty radical differences in temps from cores 0/1 to cores 2/3 which I do not see when running p95. I think this means that p95 does a more effective job spreading the load out evenly which may have stability implications.

Example:
Temps as read by HWMon for linpack: 63,63,56,56
Temps as read by HWMon for p95 large FFT's: 52,52,51,50
Temps as read by HWMon idle: 35,35,36,35

Again, the magnitude of the numbers don't bother me as much as the fact that linpack gives that massive 7 °C difference vs. the small 2 °C diff on p95. Yeah, I know about issues affecting temp spreads such as HS mounting, uneven pressure on the mount via the screws, uneven surfaces, TIM irregularities, etc. but other stability tests such as p95, OCCT, etc. do not show this major difference on my machine indicating to me that whatever is causing this is not hardware related.

What I would really love is a benchmark that can quickly (<15 min) tell me if my vcores are high enough for a given set of conditions. Perhaps this might do that if one adjusts the work size to generate a smaller problem size such that the cycle time would be like 30-40 sec? It might be as-good or superior to p95 for verifying not thermal stability, but the "stability" of mathematical precision. As you know, p95 has the round off number checking errors which we all use to verify "stability" of a given set of conditions. What sucks is that as you approach a "stable" system, the errors from p95 often take 10s of hours to manifest themselves.

I'm not interested in doing the Pepsi challenge with this software in this regard, but maybe one of you out there is?

Question 1: How quickly do errors as indicated by differences in the residual (norm) values take to propagate? In other words, if it takes p95 doing large FFT's 6 hours to fail on a core, how long does it take this software given a reasonable work size (my machine did a 20k size in about 3 min)?

Question 2: Has anyone run this thing for 12+ hours? If so was it stable and if not, how many hours did it take to give an error, again characterized by differences in the residual (norm) values?

[Isira]

Dunno what to make of this.... my system is p95 stable for >30 h under these settings. I guess this benchmark confirmed the stability since the numbers are identical:

You have to run it for many more iterations. Set it to 100 iterations and just let it run overnight. I had errors somewhere between 15-20 iterations before.

[Isira]

What I would really love is a benchmark that can quickly (<15 min) tell me if my vcores are high enough for a given set of conditions.

SuperPrime http://www.superprime.org/
It uses Linpack to compute primes. Set it to "check stability" and let run. 144m primes on an E2180 takes about 30 mins. If this passes, then run the Intel Linpack overnight to ensure no errors crop up about 30 iterations down the road since Superprime only runs about 16 iterations.

[Demo]

Superprime has no sort of memory testing like linpak, id stick to linpak.