LinX - A simple Linpack interface

[safan80]

That's it, my noobness and no Nehalems around to test on. I think I know why this happens, but can't reproduce that on Atom.
I should have somehow said that HT support is... sort of beta. But since the version number is starting with 0 anyway, I hope you can forgive me that.
Here you go, LinX 0.5.4.2, hopefully fixed.

no noobness here... we know it's beta. I like to beta test

btw it works!

I have one request though: could you put the version number in the titlebar in your next release?

[Dua|ist]

I have one request though: could you put the version number in the titlebar in your next release?

This is not an easy task. But I'll do it.

BTW, Happy New Year everybody!

[slim142]

Dualist
I dont know if you would like this or not, but a .txt file with the changelog of the program since it first came out would be great.

[Dua|ist]

Dualist
I dont know if you would like this or not, but a .txt file with the changelog of the program since it first came out would be great.

Yep, this should have been done from the very beginning. Luckily, I have a full changelog in russian. Looks like I'll have a nice time translating it.

[KURTZ]

hi folks, is there a link for this superb tool that is continuously updated (for the new versions) so i can link this onto my board?

[Dua|ist]

Forget those 2 instances. There IS a way to force Linpack use more threads than number of physical cores in a system from within a single process. If only I discovered this earlier...

Intel suggests using only physical cores probably because of P4 Hyperthreading which does perform worse in Linpack (not only there ) when turned on. Why is Intel so secretive anyway? There is no mention about threading control in Linpack's documentation. Google is a real life-saver.

[Dua|ist]

hi folks, is there a link for this superb tool that is continuously updated (for the new versions) so i can link this onto my board?

Currently there is no permanent link. I'm not sure if Fileden supports it. I think if I always keep the same filename and update the archive, link should remain the same. We'll see how it goes since next version.

[RJARRRPCGP]

Now more serious errors:
- a BSOD during testing (at the very first runs) is often caused by too low vcore
- if the system hangs and remains hung it is almost 100% not a CPU but memory or possibly NB issue
- if the system reboots (with no hang prior to reboot and no BSOD) - a CPU issue, but not vcore related (insufficient PLL or Vtt I guess)
- if it hangs for a short while and then BSODs - once again NB or memory problem (but might be wrong Vtt / GTL setting as well)
- if it hangs and then just reboots - wrong Vtt (too low or too high) or GTL settings

P.S. Despite being probably the most stressful stability checking program for CPU Linpack is not that good at finding memory errors, so checking memory with tools like Memtest, OCCT & Prime Blend will be a better idea.

P.P.S. Please correct me if I'm wrong.

At least with earlier systems, especially socket A, the Vcore being too low causes a BSOD with "STOP: 0x0000009C MACHINE_CHECK_EXCEPTION" or "0x0000000A IRQL_NOT_LESS_OR_EQUAL" and sometimes "0x00000050 PAGE_FAULT_IN_NON_PAGED_AREA" (0x00000050 is the least common of the three, because "PAGE_FAULT_IN_NON_PAGED_AREA" is 99 percent RAM-related)

[KURTZ]

I think if I always keep the same filename and update the archive, link should remain the same. We'll see how it goes since next version.

thanks mate, this is real usefully for us

[vega22]

Dualist, the cpu detection is not correctly working in xp x64 sp2. The 2 instances HT Cpu buttom is grayed. I can turn this buttom active editing myself the cpu.cfg, but the program goes wrong anyway for testing HT...

[Dua|ist]

Dualist, the cpu detection is not correctly working in xp x64 sp2. The 2 instances HT Cpu buttom is grayed. I can turn this buttom active editing myself the cpu.cfg, but the program goes wrong anyway for testing HT...

Yes, I'm aware of that. This is because of the fact that WMI section containing the number of physical cores is present only in Vista and XP SP3. And LinX needs to know the number of physical cores to determine whether CPU has Hyperthreading or not.
Good news is that I've finally found a way to override the number of threads Linpack creates, so there's is gonna be no need to run two Linpack processes and thus no need to query WMI at all. By default number of threads (and cores stressed) will already be set to maximum (you can lower it if you want via Number of threads field though). Since next version everything should be OK, please wait a little.

RJARRRPCGP, thanks, the IRQL_NOT_LESS_OR_EQUAL BSOD is the one I see most when there's insufficient vcore. As for the MACHINE_CHECK_EXCEPTION BSOD - it seems weird to me. Every time I've seen it, it was very consistent, while usually BSODs differ slightly from time to time (in case of CPU instability). That led me to think it was some sort of MoBo problem maybe. But I may be wrong.

[davidk21770]

RJARRRPCGP, thanks, the IRQL_NOT_LESS_OR_EQUAL BSOD is the one I see most when there's insufficient vcore. As for the MACHINE_CHECK_EXCEPTION BSOD - it seems weird to me. Every time I've seen it, it was very consistent, while usually BSODs differ slightly from time to time (in case of CPU instability). That led me to think it was some sort of MoBo problem maybe. But I may be wrong.

I used to see IRQL_NOT_LESS_OR_EQUAL a lot when I was testing overclocks on my old AMD dual. It was always memory instability. I haven't seen it with the Intel quad that I have now.

[vega22]

Yes, I'm aware of that. This is because of the fact that WMI section containing the number of physical cores is present only in Vista and XP SP3. And LinX needs to know the number of physical cores to determine whether CPU has Hyperthreading or not.
Good news is that I've finally found a way to override the number of threads Linpack creates, so there's is gonna be no need to run two Linpack processes and thus no need to query WMI at all. By default number of threads (and cores stressed) will already be set to maximum (you can lower it if you want via Number of threads field though). Since next version everything should be OK, please wait a little.

thank you very much, master

[safan80]

here's a LinX 5.4.2 run that finished nicely


a link to the loaded version is here
http://www.xtremesystems.org/forums/...postcount=1197

[ReverendMaynard]

Not too shabby. Really like the interface and how if you select over the available amount of ram, it auto selects it right down to the last mb.
Good work

[RJARRRPCGP]

RJARRRPCGP, thanks, the IRQL_NOT_LESS_OR_EQUAL BSOD is the one I see most when there's insufficient vcore. As for the MACHINE_CHECK_EXCEPTION BSOD - it seems weird to me. Every time I've seen it, it was very consistent, while usually BSODs differ slightly from time to time (in case of CPU instability). That led me to think it was some sort of MoBo problem maybe. But I may be wrong.

My T-Bird 900 at 1050 Mhz would have "IRQL_NOT_LESS_OR_EQUAL" (and I couldn't get it stable no matter what! at least on the Soyo SY-K7VTA-B and Chaintech CT7AJA2E boards, which I had at the time, in 2002) (I first OC'ed that processor in 2001)

I was a n00b in 2001 and 2002, anyways, because I hoped it would be stable, because I could run a Sega Genesis game with Gens with the intro looping for hours without crashing LOL!

while my T'bred 2000+ AIUCB at 2055 Mhz would give me "MACHINE_CHECK_EXCEPTION" when playing GoldenEye 007 with the 1964 0.9.9 Nintendo 64 emulator or Project64 Nintendo 64 emulator LOL! (at 1.82V and the Asus A7V8X-X, which it was one, wouldn't let me go higher than 1.82V!) So I was forced to lower the core frequencies for both systems.

[jcool]

Maynard can U repeat that with 11GB ram usage? I think not

[Dua|ist]

Here you go, only one Linpack process for all cores/threads: LinX 0.5.5
Tested only on P4 and Atom, so please, i7 owners, if you're not yet tired beta-testing, try this one and check, whether all cores are loaded or not.

KURTZ and everybody who wants to link to LinX (sounds funny ), I think (almost sure) this link will remain the same for newer versions.

No WMI calls now, should detect everything correct on all OSes.
Quote from changelog (yes, there's even a changelog now! ):

LinX 0.5.5
- improved threading control. 2 Linpack processes are created no more to load all "cores" on HT processors
- added tray icon option. When enabled, LinX will be minimized to tray
- small interface fixes

[safan80]

LinX 0.5.5
- improved threading control. 2 Linpack processes are created no more to load all "cores" on HT processors
- added tray icon option. When enabled, LinX will be minimized to tray
- small interface fixes

it's not loading my system all the way. cpu is only at 26% and memory is at 54%

here's a loaded picture of 5.4.2 if you need that
http://www.xtremesystems.org/forums/...postcount=1252

[stasio]

Nice release and link(for my system)

[Dua|ist]

it's not loading my system all the way. cpu is only at 26% and memory is at 54%

here's a loaded picture of 5.4.2 if you need that
http://www.xtremesystems.org/forums/...postcount=1252

Damn. Number of threads is set to 8 and still no go?

[ReverendMaynard]

Maynard can U repeat that with 11GB ram usage? I think not

I'm on a 32b OS, using 3gb. Should I care if I can run 12gb? no.

Nice clocks though... aren't they?

[safan80]

Damn. Number of threads is set to 8 and still no go?

with threads set to 8(it defaults to 4 didn't see that lol) it maxes the cpu usage out. I have to use a larger problem size to get close to the same memory usage though.

[KURTZ]

Here you go, only one Linpack process for all cores/threads: LinX 0.5.5
Tested only on P4 and Atom, so please, i7 owners, if you're not yet tired beta-testing, try this one and check, whether all cores are loaded or not.

KURTZ and everybody who wants to link to LinX (sounds funny ), I think (almost sure) this link will remain the same for newer versions.

No WMI calls now, should detect everything correct on all OSes.
Quote from changelog (yes, there's even a changelog now! ):

LinX 0.5.5
- improved threading control. 2 Linpack processes are created no more to load all "cores" on HT processors
- added tray icon option. When enabled, LinX will be minimized to tray
- small interface fixes

sounds good mate thanks

[rge]

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