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

[Demo]

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?

I get something like this



On the lower left side, so i guess i have to use that far viewer ?

[Isira]

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

The number of primes dictates the amount of memory used, but its not fine grained as the Linpack configuration file since you are limited to only 4 choices. For a quick stability tool, SuperPrime fits the bill.

Select the max number of primes that doesn't cause paging and use it to "check stability".

[Demo]

The number of primes dictates the amount of memory used, but its not fine grained as the Linpack configuration file since you are limited to only 4 choices. For a quick stability tool, SuperPrime fits the bill.

Select the max number of primes that doesn't cause paging and use it to "check stability".

Thats the problem, the max size is too little.. even on 1gb machines.

[Isira]

Take a look at Commit Charge in the Performance tab of the Task Manager and you will see how much memory is being used. 144M is definitely too much for a 1GB machine, really meant for a 2GB machine unless you want paging to occur.

[Demo]

I tried running the linpak64.exe directly w/o relying on the .bat file and i noticed the CPU usage is much more stable at 100%
could be an incorrect string on the batch file ?

Im still inputting the same values of
1
2600
2600
4

[graysky]

I tried running the linpak64.exe directly w/o relying on the .bat file and i noticed the CPU usage is much more stable at 100%
could be an incorrect string on the batch file ?

Im still inputting the same values of
1
2600
2600
4

My batch file simply has the exe call the config file named, "lininput_xeon64" and inserts date/time stamps.

What is your full commandline that gave you a more stable result?

[Demo]

Well running the .exe gives me a real time monitor of the progress, i think ill stick with the .exe

[graysky]

@demo - real time progress is cool, but can you write a log as well?

[003]

I would just like to say this is the best stress test ever. I was experiencing occasional freezes in UT3, where the game would just lock up and I'd have to reset the system. Prime95 and OCCT were 100% stable and error free for hours on end, and so was ATItool artifact scanning. I thought I'd give this a try, and not only does it get the CPU close to 6c hotter than prime95 (using small FTT), but it identified errors with it set to only 5 repetitions! Increased vcore some more and now it's stable as well as UT3!

[MartinuZ]

It seems like this program (linpak 64) killed my PCI Network card.

After 16 minutes of testing my hottest cores were at 73 C, compared to 62-63 with p95 with 4 threads.

After 17 minutes i heard wierd sounds from the computer and my network stopped working. I am now running on an other network card.

CPU frequency: 3.294 GHz
Number of CPUs: 4
Number of threads: 4
Parameters are set to:

Number of tests : 1
Number of equations to solve (problem size) : 14500
Leading dimension of array : 14500
Number of trials to run : 128
Data alignment value (in Kbytes) : 4

Maximum memory requested that can be used = 1682294096, at the size = 14500
============= Timing linear equation system solver =================

Size LDA Align. Time(s) GFlops Residual Residual(norm)
14500 14500 4 53.596 37.9286 2.404240e-010 4.050385e-002
14500 14500 4 53.534 37.9726 2.404240e-010 4.050385e-002
14500 14500 4 53.785 37.7957 2.404240e-010 4.050385e-002
14500 14500 4 53.857 37.7453 2.404240e-010 4.050385e-002
14500 14500 4 54.063 37.6010 2.404240e-010 4.050385e-002
14500 14500 4 54.111 37.5681 2.404240e-010 4.050385e-002
14500 14500 4 54.152 37.5397 2.404240e-010 4.050385e-002
14500 14500 4 54.289 37.4449 2.404240e-010 4.050385e-002
14500 14500 4 54.406 37.3641 2.404240e-010 4.050385e-002
14500 14500 4 54.493 37.3044 2.404240e-010 4.050385e-002
14500 14500 4 54.626 37.2141 2.404240e-010 4.050385e-002
14500 14500 4 54.742 37.1350 2.404240e-010 4.050385e-002
14500 14500 4 54.790 37.1021 2.404240e-010 4.050385e-002
14500 14500 4 55.007 36.9557 2.404240e-010 4.050385e-002
14500 14500 4 56.042 36.2734 2.404240e-010 4.050385e-002

Seems like my cpu is Linpak 64 stable but my NIC was not =(

Anyone got an idea on why it burned my NIC?

[graysky]

dunno what to say about that dude... what is your PCI frequency set to if the BIOS has an option for it? If you don't manually set it, there's a chance it'll increase proportional to your FSB on the o/c but I honestly don't see how running linpack would make a difference.

[mikeyakame]

I've hit over 80c sporadically in Linpack 64bit, provided the main CPU temp doesnt ever get that high you should be fine. Highest I've seen CPU is at 60c or so with Linpack, but most cores are between 70-80c. Once memory/cpu/fsb clocks are all aligned and in phase Linpack temps go through the roof! It holds at 100% cpu for over 40sec on 15000x15008 dimensions! Most I let Linpack run CPU temp is 65 constantly before I kill the process. All depends on the chip.

@graysky, you mentioned you didn't quite get the differences between core 0/1 and 2/3 in linpack. The differences in load are due to out of skew CPU/DRAM clock signals on either side of the MCH. If data clock signal from Channel A arrives at MCH before Channel B, then it is passed to which even particular die is using the smallest time slice. CPU clock signals are still based on same AGTL+ ref voltage though even for each die, so while 1 die is handling the data which arrived from first memory channel, the other die is sitting there idle until it arrives, how ever many clocks late. So thats the reason for the unstable CPU usage. Mine runs nearly at 100% for every test, it usually takes a few seconds for the PLL to stabilize the clock signals, but the bouncing is the PLL not able to keep them in phase for the duration of the test.

[StevensDE]

Hi,

here is my linpack 64 bit score:

Average 66.6707 GFlops with a Dell Precision T7400. (all default settings)

[mikeyakame]

Damn nice! My kentsfield peaks at 42GFlops average at 1940mhz FSB! Thats impressive for a notebook CPU!

Batch file I use to launch Linpack from inside Windows Powershell cause I got sick of doing cd and typing the whole input line. Modify to your needs, I launch it from C:\Users\MyUser

< lp.bat >
-----------------------------------------------------------------------------------------

@ECHO OFF
cd AppData\Roaming\Intel.Linpack.Benchmark
.\linpack_xeon64.exe .\lininput_xeon64

-----------------------------------------------------------------------------------------

< lininput_xeon64 >

-----------------------------------------------------------------------------------------

Intel(R) LINPACK data file
User Specified Datafile
1 # number of tests
14000 # problem sizes ( x = problem_sz; 8 * x^2 = mem_usage; )
14008 # leading dimensions ( x = problem_sz, y = lead_dim; y > x && y % 8 == 0 && y % 16 == 8; )
5 # times to run a test
4 # alignment values ( in kBytes )

-----------------------------------------------------------------------------------------

[SmaKKed]

NIce looking app.
The problem with most software testing is it is coded by man and i can guarantee there will be mistake somewhere that can cause error, so why bother? Until some one can code 100% with no errors and no all bugs on all known hardware then we will have true stability, but that will never happen. A combination of Man made product and man coding is always going to give issues.

[Demo]

NIce looking app.
The problem with most software testing is it is coded by man and i can guarantee there will be mistake somewhere that can cause error, so why bother? Until some one can code 100% with no errors and no all bugs on all known hardware then we will have true stability, but that will never happen. A combination of Man made product and man coding is always going to give issues.

I dont quite understand this.

[graysky]

@demo - the only version of linpack currently offered at the intel url is 10.1 which now requires a dll to run! I haven't been able to locate it after an hour or so on google... you? libiomp5md.dll

[burebista]

May I ask why don't you use LinX or IntelBurnTest?