LinX - A simple Linpack interface

[KGB7]

@KGB7
does increasing the volts solved the BSOD?

It did not. I was getting a random BSOD while surfing the web and moving files from one drive to another.

I eventually found that the BSODs were caused by Voltage Offset setting it self. By switching back to manual Voltage setting, the BSODs were gone.

[xpower]

thanks...

[m4gg0t]

Any updates?

[stasio]

New Intel Linpack (11.0.5.009)

http://registrationcenter.intel.com/...11.0.5.009.zip

[SBB]

I updated my LinX to the latest linpack, and if I just place linpack_xeon32.exe and linpack_xeon64.exe in the folder it complains that a DLL is missing (libiomp5md.dll). If I place this DLL in the folder linpack errors out as follows, this is at stock speeds btw:

[xpower]

WTH same error
the digital signature for linpacks 11.0.1.005 in not valid but no error
but as ver 11.0.5.009 licensed is valid but with this error

[kiwi]

Same issue.

You can still run Linpack from the command line if you wanna heat your Haswell

Download Linpack from Intel:
http://software.intel.com/en-us/arti...npack-download

Unpack and go to folder (this is for version 11.0.5.009):
w_lpk_p_11.0.5.009\linpack_11.0.5\benchmarks\linpa ck

Edit lininput_xeon64 file, for example, for 4GB RAM the problem size is 23118 (you can check Linx which problem size you wish to run):

Sample Intel(R) Optimized LINPACK Benchmark data file (lininput_xeon64)
Intel(R) Optimized LINPACK Benchmark data
1                        # number of tests
23118 # problem sizes
23118 # leading dimensions
30  # times to run a test
4  # alignment values (in KBytes)

Run runme_xeon64.bat

Results will be in win_xeon64.txt file

That's it

[orky87]

Here's the same LinX v0.6.4 from Page 1 with the latest AVX Linpack v11.0.5 integrated .

https://mega.co.nz/#!MFhX2IxI!VKJ0UB...S9xnOBtYhtIqPU

[xpower]

@ orky87

[SonDa5]

@ orky87

I got the same problem.

[munim]

I got the same problem.

Me too.

[munim]

I tried several different version combinations of Linx and the latest intel linpack library but I kept getting out of memory errors, even with something absurd like 32 megabytes selected. Finally a combination worked.

I used this linx version from post 838 in this thread: http://www.mediafire.com/?7w8xpkd233i85f8

And then I copy pasted the xeon64.exe ONLY from the latest linpack library (as at time of writing) into the folder that contains linx.exe

for reference, with approx. 12 GB selected for memory, I'm getting ~200 gflops on my 4770K with HT disabled and 4.2 GHz. With HT it drops down to ~165.

[Barr3l Rid3r]

LinX with Intel Linpack v11.0.1.005 for Intel and AMD CPU.

For AMD users who might face with the Non-Loading issue with the latest Linpack v11.0.1.005,
I have come up with a patch to resolve this problem.

LinX 32-bit


LinX 32-bit log


LinX 64-bit


LinX 64-bit log


For the AMD patch I changed the hex byte 75H (JNE) to EBH (JMP), ie I forced the code
to jump straight to the Linpack Math library routine regardless of the CPU type whether
it's Intel or AMD. Nice, right?

Some hash values and short DisAssembly explanation to the original and patched files,
32 and 64 bit.

linpack_xeon64.exe (Intel CPU only)
CRC-32: ee84f31b
---MD5: ea2c1ed5d9ea12a3df2e47ee7666ab12
-SHA-1: a4f9effd7b23afc85b8b3858ff44c9764c580f8c

00046C 85 C0 Test EAX, EAX......................; Not Zero = Intel, Zero = AMD
00046E 75 2B JNE Some Memory Address.....; Intel, so go to Math Routine

linpack_xeon64.exe (Patched, Both Intel and AMD)
CRC-32: 3f167792
---MD5: 4032d811603594a0785c7bb1dd67bf7b
-SHA-1: 8a7f62581846bd33be45161f09738f348e07882c

00046C 85 C0 Test EAX, EAX......................; Not Zero = Intel, Zero = AMD
00046E EB 2B JMP Some Memory Address.....; Don't care just go to Math Routine

linpack_xeon32.exe (Intel CPU only)
CRC-32: c184babb
---MD5: ab203fd622c162782933591e273cc880
-SHA-1: 6a0f78d647ff3fc60a8972939326b9048af283d4

001038 85 C0 Test EAX, EAX......................; Not Zero = Intel, Zero = AMD
00103A 75 2B JNE Some Memory Address.....; Intel, so go to Math Routine

linpack_xeon32.exe (Patched, Both Intel and AMD)
CRC-32: e84cd5c4
---MD5: 60882e2d5024364fab26910daa815b4b
-SHA-1: 5a6a91e46f6245fd97a438c90e87ae59480aaa6b

001038 85 C0 Test EAX, EAX......................; Not Zero = Intel, Zero = AMD
00103A EB 2B JMP Some Memory Address.....; Don't care just go to Math Routine

The short DisAssembly explanation might be wrong or inaccurate but I think you get
the general idea.

Credit to AgentGOD (IBT creator) who I think to come up with this brilliant idea and
be the first person to do this patch years ago. ;-)

For AMD users who are creative enough you can use this latest Math library with IBT v2.54
too by copying and renaming the appropriate files. Don't forget to copy the Intel OMP
Runtime Library (libiomp5md.dll), 32 and 64 bit version to the appropriate folder also.

Download links :
Mediafire :

http://www.mediafire.com/download.php?kna2w379ggxr7hz

Datafilehost :

http://www.datafilehost.com/download-d5b49d25.html

Thanks a lot! What software did you use to edit the exe?

[Dua|ist]

Hey everyone. It's been a while.
I was told that LinX is having trouble coping with latest Linpacks, so I tried to fix it.
I also uploaded the source to GitHub in case I decide to take another leave. It's ugly as hell tho.
Anyhow, here you go, v0.6.5. Linpack binaries themselves haven't been patched for AMD compatibility, but I see that this problem has already been solved here.

Updated 1st post as it appears that it actually works not only on my Windows VM.

[slim142]

Hey everyone. It's been a while.
I was told that LinX is having trouble coping with latest Linpacks, so I tried to fix it.
I also uploaded the source to GitHub in case I decide to take another leave. It's ugly as hell tho.
Anyhow, here you go, v0.6.5. Linpack binaries themselves haven't been patched for AMD compatibility, but I see that this problem has already been solved here.

Updated 1st post as it appears that it actually works not only on my Windows VM.

Hey Dualist

So does your 0.6.5 comes with the latest linpacks update?

[SimpleTECH]

Hey Dualist

So does your 0.6.5 comes with the latest linpacks update?

It does..

[stasio]

Yea,same hash as Linpack ver. 11.0.5.009.....

[xpower]

thanks for update Dualist.

[slim142]

Thank you and Dualist for this update.

Definitely handy! Hope to have more when the haswell upgrade becomes available!

[Barr3l Rid3r]

Hey everyone. It's been a while.
I was told that LinX is having trouble coping with latest Linpacks, so I tried to fix it.
I also uploaded the source to GitHub in case I decide to take another leave. It's ugly as hell tho.
Anyhow, here you go, v0.6.5. Linpack binaries themselves haven't been patched for AMD compatibility, but I see that this problem has already been solved here.

Updated 1st post as it appears that it actually works not only on my Windows VM.

Can you patch it for AMD use?

[stasio]

New Intel Linpack 11.1.0.002

http://registrationcenter.intel.com/...11.1.0.002.zip

[stasio]

LinX 0.6.5-11.1.0

http://www.mediafire.com/?x1gp435o55h61ef

- with latest Intel Linpack binaries (ver. 11.1.0.002)

[xpower]

thanks for sharing stasio.....

[chrissnake]

Thank you Stasio

[Barr3l Rid3r]

can someone patch this lates version LinX 0.6.5-11.1.0 to be used in AMD systems?