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

[Shampoo]

Strange.
I ran the Xeon32bit version of the program with the default settings and got much higher temperatures than with the provided changes by the OP.

Anyways, seems like my system is safe and stable.

Now if only I could figure out why my system hangs when playing videos for a long period of time.

[Cronos]

Up.

[KTE]

I've done all the testing, just need to find the time to look over it again and set all details straight on the other system. For a day XS servers have been failing to load up the site to us.

One thing I'll tell though; Linpack 32-bit is no where near the same power/temps as Intel TAT if run properly, it is only just higher than P95 small FFT if left for a few hours. TAT with a Q6600 is considerably higher load/temp than Linpack 32b so you need to edit your statements in OP out to reflect facts better.
The vdroop I had with TAT nothing came close to that.

Linpack is also not a consistent load but fluctuating, so your temps and power is high/low constantly and consistent load/temp isn't as high as max observed during testing whereas for all other stability tests, it is pretty much and so rise.

Although the calculation error sensitivity is a pretty good test for stability on it's own as an addition to the normal routines.

[KiD0M4N]

I have wrapped a nice UI around this one guys, give this a try: http://www.superprime.org/

[Cronos]

One thing I'll tell though; Linpack 32-bit is no where near the same power/temps as Intel TAT if run properly, it is only just higher than P95 small FFT if left for a few hours. TAT with a Q6600 is considerably higher load/temp than Linpack 32b so you need to edit your statements in OP out to reflect facts better.
The vdroop I had with TAT nothing came close to that.

Linpack is also not a consistent load but fluctuating, so your temps and power is high/low constantly and consistent load/temp isn't as high as max observed during testing whereas for all other stability tests, it is pretty much and so rise.

Please show some comparative screenshots of TAT vs. Linpack 32b together with Linpack input and output files.
I have not done any extensive testing with 32b version ( i have not used any 32 bit OS for over 2 years already )

[KTE]

Don't worry, I will do. I'm busy yet.
But I have a measured SMPS and power guaging device here, they show me the load power along with the PF and Eff. best when running such tests. Especially since I have it's test report through an ATE by the work engineer.

I don't have 64-bit Windows XP because I don't need it at home. Only use it at work and the tests I ran last on x64 were with my work system since my boss had queried (it runs stock). Have a Vista Ultimate x64 DVD but no need to install or use it since I run the 32b version and even that, it being Vista, hardly. It will obviously produce higher results if an application is RAM intensive but for sub-2GB, still not much difference.

[Tony]

I have wrapped a nice UI around this one guys, give this a try: http://www.superprime.org/

Any chance of adding the amount of ram you want testing?

[Kolapso]

Linpack looks nice, and it´s always nice with some new tools.

[KTE]

I have wrapped a nice UI around this one guys, give this a try: http://www.superprime.org/

That's a class, quick and simple prime, performance and stability tester. More load than Linpack 32b at 144M inside the 1st minute.

I just gave it a quick try;
Heck, 73, 73, 69, 69 C and 292-296W 85% consistent load, 298W max observed load with the 144M bench whereas Linpack 32b using 1.6GB/2GB with 10 trials managed 72, 71, 68, 67 C and 269-293W 70% consistent load, 298W max observed load. Impressive.

It also never just ran on but failed with instability right there and then saying "error". It'll give low temp readings if it finds an error in a core because it won't be able to load it and so the calculation will be slower but it'll tell that as "error" before end of calc. It's also very good for memory stability testing unlike Linpack, although Memtest/Windows Memory Diagnostic/RightMark Memory Stability Tester is the much better way.

Someone get Charles or Intel's Francois to post a 144M run on the Skulltrail 4.4GHz they're backing seen as no way they've ran any proper stability test incl. linpack.

EDIT2 above - reason: ran then both again and realized one of my system fans was spinning a little slower on the Superprime test. Updated both for current balanced results to make it fair.

[NickS]

Linpack is nuts! My CPU shot up to 60C and topped off around 65/65 on both cores. Orthos on the other hand tops out at around 57/57, small FFT's.

[KTE]

Please show some comparative screenshots of TAT vs. Linpack 32b together with Linpack input and output files.

Here you go, took around 10 minutes each.

Linpack 32b input and output files are like this:



Running:



Didn't get the max momentary temps captured and they didn't appear again, so I just ran that batch file again at 450x8=3600 and these are the max temps I had seen before too.



As a comparison, TAT after 5 minutes:



And SuperPrime posted above after a minute (at lower volts):




More Testing Details

My setup has trouble cooling with those volts, hence why I don't run those MHz usually but I did it just for this comparison.

Air (Zalman 9700 - no other system fan), ambient heatsink 23C

BIOS: 1.40V
Idle: 1.392V 129W AC, 32C-31C-26C-26C

Prime 95 25.6 Small FFT
Load Vcore: 1.360V (1.354-358V DMM)
Power: 278-289W consistent
Temps: 69C-69C-65C-65C consistent

Prime 95 25.6 Large FFT
Load Vcore: 1.360V (1.354-358V DMM)
Power: 280-282W consistent
Temps: 63C-63C-60C-60C consistent

Linpack 32-bit
Load Vcore: 1.360V (1.354-358V DMM)
Power: 269-300W momentary (285-292W consistent)
Temps: 72C-72C-69C-69C momentary (66-69C/66-68C/63-65C/63-65C consistent)

Intel TAT
Load Vcore: 1.344V (1.339-1.332V DMM)
Power: 312-315W consistent
Temps: 78C-78C-72C-72C consistent

That app, SuperPrime above at the same volts and ambient as Linpack 32b has a higher load thus resulting temps.

[Bobsama]

I'm going for it shortly. I'll be running the 64-bit version...

E2140 @ 3.0GHz (375*8), 1.3575V
Scythe Infinity with stock fan
Abit IP35-E Rev1.00 (flaky)
Transcend JetRAM 1x2GB DDR2 800 (Memtest86 stable at 1.9V, 900MHz 5-5-5-16 for many hours)


I just got relatively stable with my HD3850 @ 770/1998.

[Bobsama]

I'm getting a bunch of errors--can someone post a proper config file for a dual-core with 2GB of RAM (Vista)?

[Cronos]

Read the first post, you will find there all information you need, as well as an example of input file.

[Charles Wirth]

Thanks Cronos for the link and instructions. Vista Ultimate 64 Skulltrail

Intel(R) LINPACK data

Current date/time: Sun Feb 17 14:56:16 2008

CPU frequency: 3.591 GHz
Number of CPUs: 8
Number of threads: 8
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 : 1
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 63.366 84.1796 4.455000e-010 3.943651e-002

Performance Summary (GFlops)

Size LDA Align. Average Maximal
20000 20000 4 84.1796 84.1796

End of tests
Sun 02/17/2008
02:57 PM

[Charles Wirth]

Linpack really loads the cores up, I cannot get 4Ghz air done unless I subcool more with AC.

Im also running cas 3

[jcool]

Damn this is evil.. I just caught my Q6600 at 78C using linpack x64, problemsize 20000 (running XP x64, too many other processes running to use 21000).
Normally it's just 60C or so when folding...
Speed reading is wrong, it's actually 440x8=3520... my current 24/7 folding setting. Oh and it isn't finished but I need to leave now

Intel(R) LINPACK data

Current date/time: Mon Feb 18 00:44:47 2008

CPU frequency: 3.960 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 : 200
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 4 113.761 46.8889 4.455000e-010 3.943651e-002
20000 20000 4 114.600 46.5458 4.455000e-010 3.943651e-002
20000 20000 4 113.134 47.1487 4.455000e-010 3.943651e-002

[Cronos]

My Linpack 64b temps for Q6600 @3600/1.4V are 72-72-66-66, 25C ambient.

[CyberDruid]

Thanks for the new test SW Cronos. It's always good to have new ways to compare and test the rigs.

[Cronos]

Linpack really loads the cores up, I cannot get 4Ghz air done unless I subcool more with AC.

Im also running cas 3

I thought you have them watercooled?
4GHz with quadcores on air is overkill even for 45nm.

By the way, you scored 84Gigaflops. At 3.6GHz, your theoretical peak value is 3.6*4*8=115Gigaflops. Your efficiency is 84/115=73%.

You can improve somewhat by using large problem size. But you most probably memory-limited here, although Linpack makes very efficient use of CPU cache (Hence such a large efficiency. Most normal programs can hardly reach 10-20%, benches 30-40%)

With my Q6600, i can reach almost 88% efficiency in Linpack 64b.

[Charles Wirth]

25K it hits the hard drive, 20k seems to be the best with bloated Vista. Im running all sorts of stuff.

I setup again, changed up a little.

[laragirl83]

So what if my seventh row has 002, 001 and 000 as well?
it means it failed...?

[jcool]

Some DUal CPU scores from my E5420's (snappy little beasts they are):

Intel(R) LINPACK data

Current date/time: Tue Apr 29 21:12:07 2008

CPU frequency: 3.225 GHz
Number of CPUs: 8
Number of threads: 8
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 73.852 72.2269 4.455000e-010 3.943651e-002
20000 20000 4 73.795 72.2827 4.455000e-010 3.943651e-002
20000 20000 4 73.785 72.2926 4.455000e-010 3.943651e-002
20000 20000 4 73.752 72.3257 4.455000e-010 3.943651e-002
20000 20000 4 73.758 72.3192 4.455000e-010 3.943651e-002
20000 20000 4 73.832 72.2472 4.455000e-010 3.943651e-002
20000 20000 4 73.823 72.2561 4.455000e-010 3.943651e-002
20000 20000 4 73.773 72.3044 4.455000e-010 3.943651e-002
20000 20000 4 73.774 72.3041 4.455000e-010 3.943651e-002
20000 20000 4 73.767 72.3106 3.635155e-010 3.217909e-002
20000 20000 4 73.774 72.3040 4.455000e-010 3.943651e-002
20000 20000 4 73.753 72.3247 4.455000e-010 3.943651e-002
20000 20000 4 73.827 72.2515 4.455000e-010 3.943651e-002

and so on ^^

Win XP x64, 2x E5420@7,5*430, Asus Z7S WS, 4GB Ram (no QuadChannel yet)

SmallFFT drains 295W while Linpack consumes up to 340W - go figure

[Jeremy617]

Some DUal CPU scores from my E5420's (snappy little beasts they are):


and so on ^^

Win XP x64, 2x E5420@7,5*430, Asus Z7S WS, 4GB Ram (no QuadChannel yet)

SmallFFT drains 295W while Linpack consumes up to 340W - go figure

You failed...

20000 20000 4 73.774 72.3041 4.455000e-010 3.943651e-002
20000 20000 4 73.767 72.3106 3.635155e-010 3.217909e-002

[jcool]

Oh really? Hm, maybe more MCH or Vdimm. Prime found nothing though, possibly the NB was overheating with linpack, as it's getting very hot.