LinX - A simple Linpack interface

[my87csx481]

Thanks a lot for the interface man....I like it a lot and it helped me to finally (I think) get my system stable....thanks for the tips on the errors too.

[Praz]

Don't really know if a fix can be done as you probably don't have the operating system. I was lucky enough to have a Windows 7 disk sent to me. Running LinX there are 2 issues. The memory defaults to 1024Mb and the number of runs to 3. Selecting a different value in the drop-down lists will not change these two values.

[Demo]

Hey thanks for this, i love the interface!

[Demo]

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.

For users with 4GB or higher ram, this or memtest is the only option since orthos and prime is still limited to 32bit memory addressing.

[Dua|ist]

Demo, thanks. I'm glad you like LinX.

Don't really know if a fix can be done as you probably don't have the operating system. I was lucky enough to have a Windows 7 disk sent to me. Running LinX there are 2 issues. The memory defaults to 1024Mb and the number of runs to 3. Selecting a different value in the drop-down lists will not change these two values.

Thanks for that. Once I finally managed (or so I hope ) to fix most Vista-related issues there's yet another new OS. It's so irritating when something works flawlessly in XP but doesn't in Vista. Or Windows 7.
I'm currently making another version which will have a good old ini-file to store settings in, so editing it might be a temporary workaround for you.

Sorry for a probably stupid question, but have you tried keying the numbers in instead of selecting from drop-down menu? What happens then? This might help me to find and fix that bug. Thanks in advance.

[Pyr0]

Just tried LinX in Windows 7. Keying in the numbers for mem ammount and number of loops works
also tried running it as an admin and tried compatibility mode for vista rtm, but neither method fixed the drop down selection

[Zucker2k]

Awesome work here Dua|list

[bowman]

ooh, linpack for both Intel and AMD! Can't believe I haven't seen this before, great!

edit - it's been running for almost 4 minutes now and there's nothing in the display yet. Is it just because of the slow CPU or is stuff supposed to show up by now? It does load at 100% so it certainly is working.. On something.. wherever.. :p

edit2 - i realised it's because i selected 'max' in the memory section. I don't have the patience for that, ha..

[Dua|ist]

Just tried LinX in Windows 7. Keying in the numbers for mem ammount and number of loops works
also tried running it as an admin and tried compatibility mode for vista rtm, but neither method fixed the drop down selection

Thanks, this sounds reassuring. At least this way it works. Once I get Win7, I'll try to fix it.

Zucker2k, you're welcome.

[Dua|ist]

ooh, linpack for both Intel and AMD! Can't believe I haven't seen this before, great!

edit - it's been running for almost 4 minutes now and there's nothing in the display yet. Is it just because of the slow CPU or is stuff supposed to show up by now? It does load at 100% so it certainly is working.. On something.. wherever.. :p

edit2 - i realised it's because i selected 'max' in the memory section. I don't have the patience for that, ha..

High problem sizes take up to a couple of minutes to calculate on my rig in sig. But 10000 and above is usually good enough for stability testing.

P.S. An AMD user in Intel thread, so unusual

[bowman]

High problem sizes take up to a couple of minutes to calculate on my rig in sig. But 10000 and above is usually good enough for stability testing.

P.S. An AMD user in Intel thread, so unusual

Well I have AMD but I'm not hysterical Besides, the Bloomfield 920 is on its way so I wander in all the areas nowadays. Want to compare the Bloomfield at 4GHz, wondering what sort of GFLOPS that one will do in Linpack

[Dua|ist]

Well I have AMD but I'm not hysterical Besides, the Bloomfield 920 is on its way so I wander in all the areas nowadays. Want to compare the Bloomfield at 4GHz, wondering what sort of GFLOPS that one will do in Linpack

The testing of an ES 965 by DeDaL has shown that Bloomfield performance in Linpack (or Linpack performance on Bloomfield?) was very inconsistent due to OS (or Linpack) switching the load from one core to another, including the virtual SMT cores. This is hopefully fixed in new Linpack 10.1, which I'll include in the newer version of LinX. Should you encounter something like that or just want to tell us about Bloomfield performance, you can share your results here.

[Praz]

Just tried LinX in Windows 7. Keying in the numbers for mem ammount and number of loops works

Same here. Manually inputting the values works fine.

[Demo]

Demo, thanks. I'm glad you like LinX.

Like it ? Who says i like it ???

I LOVE IT !!!

BTW, Whats the recommended problem size if i want to test the CPU only for stability ?

[bowman]

The testing of an ES 965 by DeDaL has shown that Bloomfield performance in Linpack (or Linpack performance on Bloomfield?) was very inconsistent due to OS (or Linpack) switching the load from one core to another, including the virtual SMT cores. This is hopefully fixed in new Linpack 10.1, which I'll include in the newer version of LinX. Should you encounter something like that or just want to tell us about Bloomfield performance, you can share your results here.

I got 24.8GFLOPS which is slightly more than a Q6600, thought something was weird. I guess that explains it!

[Dua|ist]

Demo: Linpack does test only CPU anyway. It stresses memory and NB only a little. And the effectiveness of CPU testing depends only on Problem Size/Amount of memory. The more the better error finding capability, higher temps and higher flops. I'd start with something like 5-7K for initial OC (to avoid fatal errors like hangs, BSODs, etc) and for actual stability testing anything above 10000 (~1GB) is good. But the more the better.

bowman: Maybe it does. Check this thread out in a couple of days, I'll probably upload a new version by that time with new Linpack 10.1, which was released by Intel shortly after Bloomfields, so it should fix this issue I guess.

[Demo]

Oh i thought it was a combined memory and CPU test, so anyway ill use max mem always.

[Dua|ist]

Oh i thought it was a combined memory and CPU test, so anyway ill use max mem always.

Linpack isn't that bad at memory testing, it will find most memory errors, it'll just need way more time for that when compared to Orthos/Prime and especially Memtest.

[avercros]

Great Job, Nice.

Works perfect for me, OC stable weeeeee!!

Thanks.

[slim142]

Dualist, do you think you can give us an ETA for LinX 1.0?

[Dua|ist]

Dualist, do you think you can give us an ETA for LinX 1.0?

Is there something in the 0.5.x series that doesn't suit you? Except the version number?
Seriously, I don't know. There are still many ideas to implement and there's a lot to be fixed, made more fool-proof, stable, etc.
But sooner or later... we'll eventually come to 1.0, just a matter of time.

[Dua|ist]

Here it is finally: LinX 0.5.2
List of changes:
- Linpack executables updated to version 10.1 (for Intel CPUs only; for AMD users there's still Linpack 10.0 with no CPU restrictions, just rename and replace the linpack_xeon32.exe_amd and linpack_xeon64.exe_amd). Happy Corei7 owners might want to try this one and report their performance
- added Speedfan support. LinX can now import such data as CPU temperature, vcore, fan speed and +12V voltage from Speedfan. Fill in the values in the Speedfan.ini file in LinX' folder for this to work (see readme for more details)
- some interface enhancements/fixes
- consumed memory calculation is slightly improved
- LinX now uses ini-files (LinX.ini and Speedfan.ini) to store its settings in. Those who like tweaking things may now override default values for the amount of memory to leave for OS in Max mem mode and maximum Problem size for Linpack32
- The Leading Dimensions value is now adjusted for maximum performance instead of being equal to Problem Size value. This can be disabled via the UseOptimizedLeadingDimensions=0 key in the LinX.ini



P.S. Ever tried disabling a page file in Vista with 2GB of RAM? It's a bad idea, trust me. Spent over 2 hours trying to determine why LinX no more works with high Problem Sizes because of that.

[bowman]

Still not Nehalem/HT compatible.



Also temps are 60-65C whereas actual 100% stress would bring it up to nearly 90C (Prime95 and WCG do this)

[Dua|ist]

Still not Nehalem/HT compatible.



Also temps are 60-65C whereas actual 100% stress would bring it up to nearly 90C (Prime95 and WCG do this)

Then it means we are to expect another Linpack update soon. I wonder what would happen if you disable SMT?

[bowman]

This: