LinX - A simple Linpack interface

[Dua|ist]

I always get low Gflops with HT enabled, I thought that was a Linpack general problem and "normal"?

Uhm, actually not. At least it isn't for Core i7 with HT on and Linpack being run in all 8 threads. Running Linpack on a Core i7 in only 4 threads, however, would indeed result in lower GFlops (this is normal).

LinX is supposed to detect the change in # of logical processors (when turning HT on/off) and set the number of Linpack threads to the actual value.

As for the i9 of yours, it looks really weird. Maybe it'll be possible to do some Linpack tweaking to see if anything changes. Once I implement this, you'll be the first to know.

In the IntelBurnTest thread, it mentions doing 16 threads on i7s, is that recommended here also?

Sure, if you have HT on, 16 threads is way to go. As already mentioned, LinX should default to 16 threads on an i7 PC on the first run.

So finished one batch, quit the app, reopened, still working. I noticed when it's not working, it doesn't save changes you make to it like the settings, but it does remember the changes when it's working right, so it seems like it can't write to it's config file. I wonder if turning off UAC fixed something?

What you said about the settings not being saved is interesting indeed. Never heard of such an issue and never experienced it myself. I can't imagine what could be preventing it from saving a dozen of string into an ini-file… Thanks for the observation, I've gotta check it.

UAC is turned off by me, but never heard of LinX having issues with it. Any chance you can turn it on and see if the bug you mentioned will re-appear?

[eva2000]

0.6.4 looks okay here on win7 ultimate 64bit rtm with core i7 920 and asus p6t6 ws revo

[Bitgod]

Sure, if you have HT on, 16 threads is way to go. As already mentioned, LinX should default to 16 threads on an i7 PC on the first run.


What you said about the settings not being saved is interesting indeed. Never heard of such an issue and never experienced it myself. I can't imagine what could be preventing it from saving a dozen of string into an ini-file… Thanks for the observation, I've gotta check it.

UAC is turned off by me, but never heard of LinX having issues with it. Any chance you can turn it on and see if the bug you mentioned will re-appear?

Huh, it defaults to 8 threads on mine, I had to manually set it to 16, but maybe still a UAC issue...

So I tested some more, tried it first tonight with UAC still off and it's still working. So turned UAC back to default, restart, test, it's NOT working. When it doesn't work, like I mentioned it doesn't save any changes you make and the stop button doesn't work either, I have to hit the close button and the app closes. At least it doesn't need to be force quit.

Anywho, restarted again to test it, still not working with UAC on. Turned UAC back off, restart, and it now works again.

Not using any 3rd party security SW at the moment, haven't gotten around to installing an anti-virus and frankly I'm enjoying the quick response of a new system. I dread slowing it down with some AV, heh. But I'll have to do that some day.

[jcool]

Uhm, actually not. At least it isn't for Core i7 with HT on and Linpack being run in all 8 threads. Running Linpack on a Core i7 in only 4 threads, however, would indeed result in lower GFlops (this is normal).

Really? I thought I even read about Linpack taking a perf hit with HT enabled.. because it's already making such efficient use of the CPU HT is counter-productive.

I am positive I get lower Gflopws with HT enabled (and Linpack running on 8 threads) vs. HT off and running 4 threads. This is with all Linpack based apps, from LinX to IBT right down to the original unmodified Intel linpack.

[Dua|ist]

Huh, it defaults to 8 threads on mine, I had to manually set it to 16, but maybe still a UAC issue...

So I tested some more, tried it first tonight with UAC still off and it's still working. So turned UAC back to default, restart, test, it's NOT working. When it doesn't work, like I mentioned it doesn't save any changes you make and the stop button doesn't work either, I have to hit the close button and the app closes. At least it doesn't need to be force quit.

Anywho, restarted again to test it, still not working with UAC on. Turned UAC back off, restart, and it now works again.

Not using any 3rd party security SW at the moment, haven't gotten around to installing an anti-virus and frankly I'm enjoying the quick response of a new system. I dread slowing it down with some AV, heh. But I'll have to do that some day.

I just realized I said a nonsense in my previous post about those 16 threads (I've gotta sleep more). Sorry. 8 threads should be actually fine, since there are only 8 logical processors. Not sure whether increasing the # of threads beyond the # of logical CPUs improves error finding, but GFlops-wise it usually produces a bit lower results than with the number of threads equal to the number of logical CPUs (8 for an i7).

Now what you said about UAC is interesting. Will try to fix this, thanks a lot.
OT: I don't like AVs too. Haven't been using any for half a year now, still doing fine.

Really? I thought I even read about Linpack taking a perf hit with HT enabled.. because it's already making such efficient use of the CPU HT is counter-productive.

I am positive I get lower Gflopws with HT enabled (and Linpack running on 8 threads) vs. HT off and running 4 threads. This is with all Linpack based apps, from LinX to IBT right down to the original unmodified Intel linpack.

Yes, Linpack did take a performance hit back when i7s only entered the scene, but this was due to the fact that Linpack by default (I still don't get why Intel suggests this) runs in 4 threads on an i7 CPU, no matter HT on or off. While this is OK with HT off (4 cores, 4 threads), it is apparently not with it on (8 cores visible to Windows, but only 4 threads, causing perhaps a Windows thread manager to switch the threads between all 8 visible cores without regard to their nature, so that two threads may sometimes run on two logical processors from a single real core thus causing a performance hit). But then a way has been found to make it possible to override the number of threads Linpack creates and run it in 8 threads or even more on 8 logical cores on an i7.

I never had a chance to run Linpack on Core i7 but I did run it on an old P4 Prescott and on Atom 270. On P4 with HT on running Linpack in 2 threads vs 1 thread as Intel suggests was counter-productive indeed (only a tad but still), while on Atom it produced a nice boost in GFlops (like 0.51 vs 0.37 or something like that). So I figure HT of our days isn't as bad as it was back in P4 times. Correct me if I'm wrong.

[mskvorc1]

Hi,all!

I was running my E8500 @ 3,825 GHz @ 1.304V. Recently I tride Prime95 and it crashed after 20-minutes (before I could run it 24 hours). So I decided to lower my speed and my current speed is 3,375 GHz (7,5x450) @ 1.225V (default voltage). I read a lot about Linx, and manage to run it on this settings with maximum meomry usage, two threads and for 80 times (for around cca. 2 hours). Is this enough to test my stability or should I run Prime95 for 24 hours?

[aka_GK]

mskvorc1, I can't directly answer your question, but I reached 3,6GHz (9x400) at stock voltage fully Linx stable with ease on E8500 E0

[felix_w]

Anyone has a hint for what may cause LinX to stop by error at same time every run (3m 7s) ? Could it just be a coincidence ( 11 runs ) , or just a temp problem ?

EDIT : Nevermind...ram was crappy...

[dalekdukesboy]

linx .6.4 will NOT work for me...I tried with windows vista 32 bit and now windows 7 64 bit and it fails within the first run every time...and once I close the program...it never runs again, it turns on etc, but I start to test and the cores don't load up at all and the timer runs but nothing happens...ALSO with windows vista and xp I got it to load the cpu about 98.5-99.7 percent load...both were 32 bit systems, now I have windows 7 with 64 bits and it load the cpu 100 percent the whole test period except when it unloads between runs...has anyone else seen that difference in either windows 7 vs xp/vista or is it the 64 bit operating system or what??? I know it's testing it better technically but 99 percent is enough...100 is really pushing my temps a good several celcius higher and making stability significantly harder and no program will load this thing anywhere near 100 percent with almost 100 percent memory as well...is there any way to adjust how much it actually loads the cpu in linx's settings?

[vazel]

So I was wondering why I got 7-17Gflops sometimes when I run it and why it wouldn't heat up my CPU as much as it should then I figured out Linx was using too much RAM. Seems paging file activity significantly hurts Linx performance. I set Linx to 3GB and now I'm getting 21Gflops and temps were in the high 80s. So if you're having similar issues try leaving 1GB to the OS.

Edit: nvm still happens even when limiting it to 3GB. I don't know what's going on.

[DavyBoy]

I noticed earlier today, when I use 1t command rate, I get much lower Gflops than I do when using 2t.

41.5642 when using 1t, and around 52.0000 when using 2t, my chip is o/ced to 4GHz atm.

Is this normal?

I thought I should have gotten higher Gflops using 1t?

[autox]

I have a i7 DO clocked to 4ghz using win7 x64.
Here are a few problems/questions I have.

1) I get variable Gflop when I run Linx. It would always be consistant during the entire run, but would differ from one session to the next. The funny thing is that when th Gflop is lower, my temps don't heat up as much. Before you say that my CPU is throttling, i have i7 turbo loaded and noticed no drop in the multiplier. This also occured when I was on a q9550.

2) I get lower Gflop (by about 9), when I have HT on vs. Off. Any idea why?

[sukebe]

Hey also i find something strange, i think it was mentioned somewhere before, lowering VTT volts results in higher Gflops...

I'm on i7 920 and evga E760, testing at 4.2ghz i get around 53 Gflops whit vtt 1.27v, and down to 48 Gflops just by raising vtt to 1.3v...

edit. nvm it reported 54 on second try at 1,3v, and seems it's not stable at all for me yet...

[jcool]

I think we should simply disregard the program's Gflops rating. It has never displayed reliable or reproducable results for me either. But it is a damn good stability tester, which is the program's intended use, after all. Not YAB (yet another benchmark)

[autox]

I think we should simply disregard the program's Gflops rating. It has never displayed reliable or reproducable results for me either. But it is a damn good stability tester, which is the program's intended use, after all. Not YAB (yet another benchmark)

well, I don't disregard the Gflop ratings because (on same overclock settings) when I get higher Gflop, the temps are also higher (higher stress). If I had a run where the gflop was lower, the overclock may have passed that session because it wasn't stressed as much (kind of like using smaller amount of memory).

[autox]

Hey also i find something strange, i think it was mentioned somewhere before, lowering VTT volts results in higher Gflops...

I'm on i7 920 and evga E760, testing at 4.2ghz i get around 53 Gflops whit vtt 1.27v, and down to 48 Gflops just by raising vtt to 1.3v...

edit. nvm it reported 54 on second try at 1,3v, and seems it's not stable at all for me yet...

This is what I'm talking about. The amount of stress (Gflop) is veriable from one session to the next.

[jcool]

Now that's weird.. hmmm
Dualist?

[sniggle]

2) I get lower Gflop (by about 9), when I have HT on vs. Off. Any idea why?

From the official release notes:

Known Limitations
1. Intel(R) Optimized LINPACK Benchmark is threaded to effectively use multiple processors.
Therefore, in MP systems, best performance will be obtained with hyperthreading turned off.
This insures that the operating system assigns threads to physical processors only.

Everyone seems to get approximately 20% worse performance with HT enabled in Linpack.

I have HT disabled always because:

1) Various benchmarks show that in general HT only helps encoding/rendering applications and hinders performance in many other things.
2) When all 8 logical CPU's are maxed it really feels like Windows is slower to respond in general than when HT is disabled and all 4 physical CPU's are maxed.
3) My CPU runs cooler and I get a higher overclock

[sukebe]

This is what I'm talking about. The amount of stress (Gflop) is veriable from one session to the next.

That's right, when it produces higher Gflop numbers it's putting more stress therefore shows instabilities quicker... in my case the gflops can vary from 48 to 54 in different sessions with identical settings, which means it can take about 30 minutes more to get an error when gflops are at the lowest (compared to the highest)...

@sniggle
Interesting stuff, we've also seen many single-threaded benchmarks perform better with HT off. What i wonder now is if games can perform better without HT, considering that it's probably most demanding apps on i7 rigs nowadays

[sniggle]

@sniggle
Interesting stuff, we've also seen many single-threaded benchmarks perform better with HT off. What i wonder now is if games can perform better without HT, considering that it's probably most demanding apps on i7 rigs nowadays

From what I've seen games generally do better with it off. There was a very comprehensive benchmark somewhere but I don't have the link, but here's something I found with a quick search: http://www.bit-tech.net/hardware/cpu...-cpu-review/10

Bottom line is that unless you're running a video crunching box you're probably better off with HT off.

[phsinc1]

with Q9550, testing the 1024MB @ 4GHz gives me normally 50GFLops all the session but sometimes in other sessions it would drop to 30GFLops, I keep closing the program and reopening till the first test finish with normal GFLops and I leave it to continue the other runs. Also when producing 30GFLops, temps were lower by 20c compared to 50GFLOps and the CPU usage was 100% either with the 30 or 50 GFLops.
happens on Windows 7 x64 and Windows Vista x64

[dejanh]

with Q9550, testing the 1024MB @ 4GHz gives me normally 50GFLops all the session but sometimes in other sessions it would drop to 30GFLops, I keep closing the program and reopening till the first test finish with normal GFLops and I leave it to continue the other runs. Also when producing 30GFLops, temps were lower by 20c compared to 50GFLOps and the CPU usage was 100% either with the 30 or 50 GFLops.
happens on Windows 7 x64 and Windows Vista x64

That's because something has crashed. Programs have bugs, that's what you are seeing.

[vazel]

I think we should simply disregard the program's Gflops rating. It has never displayed reliable or reproducable results for me either. But it is a damn good stability tester, which is the program's intended use, after all. Not YAB (yet another benchmark)

But as others have reported when you get lower gflops temps aren't as high which means it's not stressing your cpu as much. When I get 21gflops my oc'ed cpu heats up to the 80s but when I get 15-18gflops temps are in the 70s.

[Dua|ist]

Yes, folks, sorry for not being here and not doing anything about these GFlops issues. I am aware of them. And I too can't understand why it would work all right in some cases and show less GFlops and lower temps while keeping CPU usage at the same 100% in other cases.

Once I'll have free time, I'll get back to trying to fix these problems as well as looking for ways to make it i9 compatible.

For now (and for at least half a month) University is all I'm concerned about.

[felix_w]

University is all I'm concerned about.

Hey mate,your studies is your priority.... Good luck..