LinX - A simple Linpack interface

[vodka3dg]

hi! just got a new rig today.

was wondering how much memory and loops i had to set linx to manage a stable OC.

in my previous rig (EP45-UD3R + Q9550) i used 20 passes and 3.5 as stability criteria; would the same apply to my core i7 rig?

thanks in advance!

[Red Maw]

Hey, I tried to use this version of linx for stability testing and I think there's a bug in it. No matter what I do it encounters an error just after 3 minutes of running while two other IBT programs can complete the task successfully. I'm using win7 pro x64 btw.

[donmarkoni]

hi! just got a new rig today.

was wondering how much memory and loops i had to set linx to manage a stable OC.

in my previous rig (EP45-UD3R + Q9550) i used 20 passes and 3.5 as stability criteria; would the same apply to my core i7 rig?

thanks in advance!

More memory - the better. 20 passes is more then enough IMO. I personally use just 5 passes with 5GB of RAM. Then I use my rig for what it's meant (folding, gaming...) and see if its stable.

Hey, I tried to use this version of linx for stability testing and I think there's a bug in it. No matter what I do it encounters an error just after 3 minutes of running while two other IBT programs can complete the task successfully. I'm using win7 pro x64 btw.

Usual answer is: unstable OC (no offence meant). I find LinX most stressing of all stressing application as many OC'ers will concur. Of course it is not the only needed for stability testing (prime, hyper pi, etc...)

[jstanthr]

I downloaded linx and installed it, set it as described and when i hit start the counter starts counting, but none of the cores are loaded?

[jstanthr]

ok, the zipped version that is just extracted runs just fine, but the setup one doesn't it will load the app but doesnt load the cores. but anyway, my problem is
once it stops at an error, how do you find out what the error was? mine stopped upon error after 52m 16s on run #14 i was averaging 52.3 Gflops on a i7 920
running at 3.9, im trying my dambdest to get over 4.0

[Red Maw]

Usual answer is: unstable OC (no offence meant). I find LinX most stressing of all stressing application as many OC'ers will concur. Of course it is not the only needed for stability testing (prime, hyper pi, etc...)

No offense taken, but my point was that no matter what this version of linpack fails at ~3min (iirc) independent of OC. My system has no trouble passing either of the other two linpack versions I have, just this one.

[jstanthr]

here is the link or w/e to the SS of the installer version of linx running but not loading any cores at all. the .rar version works fine for me tho.

[donmarkoni]

No offense taken, but my point was that no matter what this version of linpack fails at ~3min (iirc) independent of OC. My system has no trouble passing either of the other two linpack versions I have, just this one.

Well, LinX is just a GUI and uses Intel's linpack modules, but it could be a bug showing only with your specific HW/SW used...
BTW, try raising VTT a notch, usually works for me when I get error after 1st loop.

[GigaByte]

Using version 0.6.4 I think theres a bug when another app is running while testing, such as WMP or browsing the net. When watching a movie while testing it tends to fail within 1 hour regardless of settings, when just running LinX alone it went for over 170 runs all memory (4850 MB) compared to less than 20 runs when watching a movie or browsing the net with the exact same settings, temps, and memory use.

[Delirious]

Using version 0.6.4 I think theres a bug when another app is running while testing, such as WMP or browsing the net. When watching a movie while testing it tends to fail within 1 hour regardless of settings, when just running LinX alone it went for over 170 runs all memory (4850 MB) compared to less than 20 runs when watching a movie or browsing the net with the exact same settings, temps, and memory use.

Are you able to duplicate this using Intel Burn Test?

[GigaByte]

Are you able to duplicate this using Intel Burn Test?

Yes, when both apps were in their earlier stages I didn't have that issue. Even happened once at stock, IBT seems to finish several seconds faster than LinX at the same memory useage, don't really know why.

[Mastakilla]

I downloaded linx and installed it, set it as described and when i hit start the counter starts counting, but none of the cores are loaded?

same problem here...

I'll try the zip now

edit:
also the zipped version doesn't do anything

[Mastakilla]

no one knows how to fix this?

[Red Maw]

Linx doesn't work for everyone, not sure why. I use the linpack version/gui by AgentGod since it has worked every time on all my systems, just search on XS or google and you should find it.

[Lsdmeasap]

This needs an update to show correct Gflops for 980X

[fonze98]

for the 980x It is possible to edit the program with the new Linpack from intell and get correct ~Gflops at a 25000 problem size but for me anyway trying to run at 37000 problem size it will still give lower temps and way lower Gflops for me running the 6.4 when I ran at 25000 problem size I was getting ~50Gflops andtemps of 60C and at problem size of 37000 I was getting 80-83Gflops and 70C temps after replacing the linpack with the intell version it was the exact oppisite at a 25000 problem size I would get 80-82Glops and 70Ctemps and at 37000 problem size I get 50Gflops and 60C temps so still not sure intell has got the linpack correct for 980X series

[Mastakilla]

I just succeeded in running IntelBurnTest!

The trick was to execute it with admin rights.
Not sure if I did that with Linx.

In noticed that in between each run, it pauses for a moment (few secs) and the CPU cools down
Is this normal?

edit:
just tried Linx out again... It works if you run it as administrator... Seems like an easy fix for the next version to automate this...

Also Linx pauzes and cools down between each run btw...

[eternal_fantasy]

For Ci7 980X here is how to update your LinX:

http://www.evga.com/forums/tm.aspx?m=333894

Hello Everyone,

So yesterday I joined the ranks of all the folks that have shiny new Gulftown 6 core CPU's. One of the first things I noticed when trying to overclock is that LinX was behaving just downright weird. The GFlops were way off and it was just hit or miss getting it to work correctly at times. So this prompted me to search for a solution rather than to wait for a new version of LinX.

So one thing that's a known is that LinX is really just a GUI frontend for LinPack, so that's where I started to look. And to my surprise I found a newer version of LinPack on Intel's site. So this post will tell you how to use this to make LinX work the way it should on your Gulftown CPU.

1. Make sure you have the latest version of LinX (0.6.4 as of this posting)
2. Go to software.intel.com/en-us/articles/intel-math-kernel-library-linpack-download/ and download the windows package for LinPack
3. Open the zip file and throw the folder inside on your desktop for ease of access
4. Go into that folder, then the benchmarks folder, then the linpack folder
5. In that folder you will see linpack_xeon32.exe and linpack_xeon64.exe. These are the files that we need.
6. Now open up another explorer window with the folder that you have LinX installed in.
7. Take the new linpack exe's and drop them in the LinX folder, allowing it to overwrite the old exe's.
8. Now run LinX and observe that it runs properly and also displays the proper GFlops that we should be seeing with these CPU's

Word of warning though, this version of Linpack loads your CPU correctly once again so expect it to run your CPU hotter than Prime95 does, so all of you should keep an eye on your temps while using this.

[Lsdmeasap]

It's little bit better, but it's still not correct is it?

New Linpack added



Compare with SiSandra Gflops (Clicky) 75 vs 130




Of course I realize they are different programs, but shouldn't Glfops be a universal thing or at least very close?

Thanks for the tip by the way!

[donmarkoni]

It's little bit better, but it's still not correct is it?

New Linpack added

[ img ]

Compare with SiSandra Gflops (Clicky) 75 vs 130

[ img ]


Of course I realize they are different programs, but shouldn't Glfops be a universal thing or at least very close?

Thanks for the tip by the way!

IMO, should be right. I never took Sandra seriously.
Comparing yours 74GFLOPS @ Core 4200 / UnCore 3360 / QPI 3696 / RAM 1680
against mine 54GFLOPS @ Core 4200 / UnCore 3600 / QPI 3600 / RAM 1600
looks right. With 50% more Cores you could have 81GFLOPS, but the difference comes from slower UnCore speed & higher RAM timings.

[tool_462]

First loop or two is always quite a bit lower than the rest too, I would suspect you would be closer to donmarkoni's 81Gflop estimate.

[Lsdmeasap]

IMO, should be right. I never took Sandra seriously.
Comparing yours 74GFLOPS @ Core 4200 / UnCore 3360 / QPI 3696 / RAM 1680
against mine 54GFLOPS @ Core 4200 / UnCore 3600 / QPI 3600 / RAM 1600
looks right. With 50% more Cores you could have 81GFLOPS, but the difference comes from slower UnCore speed & higher RAM timings.

So Sandra is wrong? Shouldn't Glfops all be the same, I guess maybe not since we just showed how different linpack versions change them.

I suppose it is normal then, I just read online that 980X should be 100 or so at default speeds I think, maybe overclocked was meant? Either way though that would still mean LinX is still off a bit

[justin.kerr]

also remember that with hyper threading and Linx, you will get more G flops with out HT, but Sandra the opposite, if my lil memory is working this morning.

[Lsdmeasap]

Ya, I did notice that about LinX, but didn't know Sandra was the opposite.

Wonder why the differ so much?

[AlderaaN]

Hello Dua|ist,

Thank you very much for this software, most useful!

Would like to seek your advice please regarding the following setup:
I'm currently testing a Core i7-860 CPU (4 Physical + 4 "Virtual"/Hyper-Threading Cores) with:
- 4GBs of RAM
- Operating system Windows 7 32-bit.

This is how I currently test my system's available memory with your software:
1. Run a session of LinX, set it to 32-bit, 4 Cores and test 1700MB of RAM.
2. Run a second session of LinX from the same directory as I ran the first one, set it to 32-bit, 4 Cores and test the remaining available memory (minus 20-40MBs)

Question 1 :
Should I change the Cores setting for each session to just 2, so each session will actually use 2 Physical + 2 Virtual cores?

Question 2:
Should I run both LinX sessions from the same directory, or create a separate directory with its own settings for each session?

Thank you very much in advance for your answer!

Best regards,
Samuel