LinX - A simple Linpack interface

[RJARRRPCGP]

Yep, but now it tells me that is not enough memory. The bright side is that it doesn't remain memory resident.

Attachment 85269

Dua|ist, I believe to get around the 2 GB per-process limit is to have 2 Linpack processes! Sorry.

[Dua|ist]

It loaded mine 99.9%, all 8 threads, but there is some memory bug. Refuses to run sometimes with stopped unexpectedly, error not enough memory. If I try to run with 256 mb, very small, it tells me not enough memory and kicks it up to 1748. It will only let me run at 1748, no less, and half the time when I pick 1748, it tells me not enough memory, though I have 2748 free. (I know there is 1748 limit with xp 32 bit).

But once that bug gets fixed, pretty sweet being able to load all 8 cores with one instance at max memory

At least it works on all cores. Here's the fixed version : LinX 0.5.5.1

Dua|ist, I believe to get around the 2 GB per-process limit is to have 2 Linpack processes! Sorry.

Question is, whether we need to get around that limit at all? Most systems with 4GB of RAM and more have a 64-bit OS, and x64 version of Linpack has no such limit. There are of course 32-bit OSes with 4 GB of RAM and ~3.5 GB of useable memory, but that 2 processes idea was just a temporary fix to load all cores when HT is on. Now that it can be done in a much better way via Linpack itself I don't see any reason to keep this option. The current ~1750 MB limit for 32-bit Linpack is enough to adequately stress-test any CPU, while RAM testing is not Linpack's goal anyway. If you still want to test all your memory, just run 2 instances of LinX.

[rge]

Dualist, Linx 0.5.5.1 worked perfect..that was a fast fix

Here is pic with max memory 1748 on XP 32 bit, loading 99.9% avg all cores.

I also tried it at 2 notches of vcore below stable at 4.2Ghz, where prime reboots around the 2 to 3 hour mark, linx rebooted at 4th run on 1748 mem, then 3rd run on second try, so linx is back in business with 1 instance.

Unfortunately on core i7's, I cant take screen pics of prime errors or linx errors as almost all OC failures on my core i7 are reboots or crashes.

[ReverendMaynard]

Dualist, Linx 0.5.5.1 worked perfect..that was a fast fix

Here is pic with max memory 1748 on XP 32 bit, loading 99.9% avg all cores.

I also tried it at 2 notches of vcore below stable at 4.2Ghz, where prime reboots around the 2 to 3 hour mark, linx rebooted at 4th run on 1748 mem, then 3rd run on second try, so linx is back in business with 1 instance.

Unfortunately on core i7's, I cant take screen pics of prime errors or linx errors as almost all OC failures on my core i7 are reboots or crashes.

I can only load the next lowest settings that you have rge before I get the "not enough memory" message. I find that after I get that message, I have to close and re-open linx. I did get it to run at around 2.3gb of ram and all cores loaded for a 20 run.

Where did you get that i7 core load app? Is it part of realtemp?

[rge]

loadtester is unclewebbs program (author or realtemp), but not part of realtemp, he sent it to me to beta test load programming on realtemp.

[nusihb]

Great program!
Heats up my CPU to 78+C, that's at 4.1ghz/1.4v water cooled... 'only' 55+C at stock speeds, though.

I seem to have a problem to get the 'residuals' to match up...
I'm running i7 965EE on a Asus P6T deluxe, triple channel 3x2gb G.skill, Vista 64.
It behaves the same way regardless if I O/C or not - fails at 2nd run when I select 'all mem'.
In fact, anything over 1.8gb fails (in 64bit setting).

Anyone else with a P6T, 6gb & Vista64 that could try?

[Pyr0]

i seem to be having the same problem here with the new version.
stock or oc'd it always fails the 2nd loop with all mem selected

[SetteUltras]

Yeah me too. Settings 64bit/all mem/optimal leading dimensions checked/number of threads 8/data allingment 4/memory to leave for os 5mb on vista64/6gb ram always fail on second loop.

[burningrave101]

i seem to be having the same problem here with the new version.
stock or oc'd it always fails the 2nd loop with all mem selected

Same here. Gigabyte EX58-Extreme and Vista x64 with all memory selected fails on 2nd loop.

[tomb18]

Clicked on one of the links for this program and there is a virus! The HTML_DLOADER.NQZ virus appeared and was detected by trend micro before the software was loaded. The actual link is http://rs524.rapidshare.com/files/160909159/LinX.7z and I think it was on page 5.
Is thisa actually a virus or a false alarm?

[Dua|ist]

nusihb, Pyr0, SetteUltras, burningrave101,
seems like it is only a 64-bit Linpack issue, when forced to run in 8 threads. I heard of that problem before but I don't think I can do anything about it. Reverting back to 2 Linpack processes doesn't seem a good idea to me. For now I can only suggest you to disable the Stop on error option and see whether all other results (starting from the 2nd one) are the same. If they are - it is definitely that Linpack glitch I heard about (the 1st result differing from the rest). I don't think I can do anything about it till the newer version of Linpack will possibly fix this issue. But if all the forthcoming results are same, I can just add a feature to skip the very 1st test.

tomb18, there are no viruses in LinX, must be a false positive. The exe is packed though - this is the only explanation I have that might have caused the detection of a virus.

[msgclb]

Up until today LinX has only stopped once with an error. When it has a problem it usually crashes Windows. Here's the second time that I've gotten an error:



This is the first time that I've used LinX 0.5.5.1. I didn't see if it loaded my 965.

Note: After posting this message I started LinX again and it did load my 965 100%.

[Pyr0]

nusihb, Pyr0, SetteUltras, burningrave101,
seems like it is only a 64-bit Linpack issue, when forced to run in 8 threads. I heard of that problem before but I don't think I can do anything about it. Reverting back to 2 Linpack processes doesn't seem a good idea to me. For now I can only suggest you to disable the Stop on error option and see whether all other results (starting from the 2nd one) are the same. If they are - it is definitely that Linpack glitch I heard about (the 1st result differing from the rest). I don't think I can do anything about it till the newer version of Linpack will possibly fix this issue. But if all the forthcoming results are same, I can just add a feature to skip the very 1st test.

Hi Dua|ist,

yes, thats exactly whats happening. i just disabled the Stop on error option and apart from the 1st set, the rest all match

[Dua|ist]

Up until today LinX has only stopped once with an error. When it has a problem it usually crashes Windows. Here's the second time that I've gotten an error:



This is the first time that I've used LinX 0.5.5.1. I didn't see if it loaded my 965.

Note: After posting this message I started LinX again and it did load my 965 100%.

LinX 0.5.5 runs Linpack in 8 threads now on Core i7, this must be a reason. Your case is definitely not the worst (see posts above).
This might be some non-CPU problem as Residual values differ much from each other. Could be RAM or NB.

Pyr0, thank you. That is what I was afraid of. There must be some way to fix it without any workarounds in LinX. I'll see what I can do. For now you can consider it stable if all results after the 1st one match.

[slim142]

Dualist

Is the lastest version 0.5.5.1 safe to use on Core 2 cpus? No problems reported regarding memory issues, thread issues or residual not matching?
I wish I could test this and not bother you asking you this, but my system is on the brink of melting so I cannot really test it (not even 1 min).

[Dua|ist]

Dualist

Is the lastest version 0.5.5.1 safe to use on Core 2 cpus? No problems reported regarding memory issues, thread issues or residual not matching?
I wish I could test this and not bother you asking you this, but my system is on the brink of melting so I cannot really test it (not even 1 min).

Yes, for Core 2 everything remains as usual. No problems reported. The oddities affect only Core i7 systems (and only x64 Linpack as it seems). But thanks for your willingness to help.

[nusihb]

None of the residues match for me.
But, it seems to work OK if I turn of the HT... so maybe it's the x64 linpack that's not ready for i7 yet.

Nevertheless, it's a great & easy way to determine your O/C limit on i7 with this tool.
Mine runs for hours at 4.1, but just raising it to 4.2 and it reboots after 2-3 'laps'.

[burningrave101]

None of the residues match for me.
But, it seems to work OK if I turn of the HT... so maybe it's the x64 linpack that's not ready for i7 yet.

Nevertheless, it's a great & easy way to determine your O/C limit on i7 with this tool.
Mine runs for hours at 4.1, but just raising it to 4.2 and it reboots after 2-3 'laps'.

Yeah I'm having the same problem. I'm running Vista x64 on my i7 920 and I can't get the residuals to match up for nothing. Like two in a row will match now and then but that's it. I even backed my overclock back down to 3.6Ghz with 1.4V vCore and adjusted most all of the voltages up high and back down to see if that helps but it doesn't seem to be. It keeps running though and doesn't BSOD or anything. The residuals are just all different and that's not including the first run since it's of course failing due to me setting all memory on Vista x64.

[Pyr0]

i just tested 10 loops at stock speeds with HT and memory set to all.
all of the residuals and norms matched except #3

[jcool]

I am using v 0.5.5 on i7 with HT and in x64 mode, all residuals and norms match or it is unstable (freeze/BSOD). Finally got BCLK 213 20 runs linpack stable, what a PITA that was.. my CPU sucks

[Pyr0]

could you select all memory and run it again please?

[jcool]

No time now.. but I did run 5 passes with 3Gb ram, and it worked. VTT requierements for stability are insane past 210 BCLK... I can run 211 stable at around 1,46V but I need close to 1,6 to get 213 stable.
Good night guys!

[burningrave101]

Ok, I'm running stock settings here with all the voltages in the BIOS set to auto and Turbo mode even disabled and the residuals are NOT matching up. So unless my hardware is bad and can't run stock speeds there is something up. And like I said this is under Vista x64 with HT enabled on an i7 920 and 3x2GB of memory. I did notice that several of you other guys are only running 3-4 gigs of memory.

I've been running Prime95 Blend and Small FFTs and I'm not getting any errors there. I also ran Windows Memtest to try and test my memory but I don't know how effective it is because even with four instances it only loads the cores 50% due to HyperThreading.

[burebista]

Sorry for OT but burningrave101 use latest beta RealTemp, it's heavily modified for Nehalem.

[Dua|ist]

Changed some of Linpack settings, this most likely won't fix anything, but still worth a try I think. Anyone experiencing differing Residuals with Core i7 and 64-bit Linpack is invited to try this one. (But don't hope too much)

Also corrected (lowered) the upper limit for 32-bit Linpack for Vista and Win7. Thanks to SetteUltras for that.

Link: LinX 0.5.5.2. Hopefully I didn't screw anything up this time.