Linpack Xtreme Released

[EarthDog]

How do you know?

 

[Regeneration]

The latest hardware always ship with a brand new bugs.

Unexplained throttling can be result of a UEFI bug. It happened before.

 

[EarthDog]

Maybe.

But he didn't give any info. We were told performance dropped.. but not sure why. Clocks remain the same through testing?
Esit: he says they did. But verify with other software.

I like to have intel xtu up while running stress tests as it will show throttling reasons.

 

[er557]

ran the 1.1.0 ver on windows 64, result is 630gflops, but it states it detects 1 cpu, 18 cores, 36 threads, while I have 2 such cpus. So is this result reliable?

 

[Regeneration]

ran the 1.1.0 ver on windows 64, result is 630gflops, but it states it detects 1 cpu, 18 cores, 36 threads, while I have 2 such cpus. So is this result reliable?

If you ran it on benchmark mode, it is accurate and perfectly fine (36 threads for 36 cores).

 

[hat]

Keep up the good work @Regeneration this tool already told me my rig won't be stable at anything over 4GHz (at this voltage) and did so pretty quickly. Most people will say if you can pass Linpack, you can pass anything, though there are strange cases where that's not 100% true... but it's true enough for me to run Linpack first, then I'll run Prime95 for a long ass time.

 

[WoodyUK]

Anyone using v1.0.0 with a modern Intel CPU should update to v1.1.0.

v1.0.0 throws false positive hardware faliures.

If only I had know this four days ago...

 

[Omegacain]

Maybe.

But he didn't give any info. We were told performance dropped.. but not sure why. Clocks remain the same through testing?
Esit: he says they did. But verify with other software.

I like to have intel xtu up while running stress tests as it will show throttling reasons.

NOT THROTTLING

 

[hat]

That's weird. Why is it detecting 10 threads? You should have 16.

 

[Solaris17]

That's weird. Why is it detecting 10 threads? You should have 16.

thats 1.0.0. Maybe detection is improved in 1.1.0?

 

[Omegacain]

1.1.0 just crashes when i try to open it

 

[Regeneration]

v1.1.1 is now available.

Users with non-modern Intel CPUs can use the /residualcheck switch to activate checking of residual checksum.

 

[Jstn7477]

Thank you for making this... did have the problem with 1.0.0 and false hardware failure warnings, but 1.1.0 seems to run fine now on my 5820K/X99 system. Both my 5820K and FX-8150 ended up being unstable and I likely fixed the FX-8150 with a 0.5x multi drop, but my 5820K has always been troublesome and still needs work (had to raise voltages to 1.23/1.2 for core/ring at 40x/30x and RAM is currently at 1600MHz to rule it out, and finally got a clean 100 runs this morning).

I used to use IBT but kind of drifted away due to crazy temps with non-soldered Intel chips and AVX, but will definitely be testing all of my overclocked machines with whatever the latest version is at that time and making sure they all survive at least a night of this.

 

[hat]

v1.1.1 is now available.

Users with non-modern Intel CPUs can use the /residualcheck switch to activate checking of residual checksum.

How old is "non- modern"?

 

[Omegacain]

I downloaded the new version... so what are the steps that you use for testing a 9900k ... act like I never use it before... thanks

 

[Regeneration]

How old is "non-modern"?

Sandy Bridge and older, when the residual checksum doesn't change after several runs.

I downloaded the new version... so what are the steps that you use for testing a 9900k ... act like I never use it before... thanks

Stress test > 10GB of RAM > try both all available threads and then custom 8 threads, see which one is more stressful (temperature and power output) and run it over-night.

 

[baryluk]

It appears some of my various complains were addressed partially for Linux version I was testing.

• tar files doesn't look to be corrupted anymore.
• tar.gz file now unpacks to sub-directory. Nice!
• Directory has a version number in the name. linpack-xtreme-1.0.0. Nice.
• Files are executable. Nice. However for unknown reason, all files are executable, instead of only the exeuctables. I.e. readme.txt, license.txt, settings, defaulti686, defaultx86_64, shouldn't really be executable.

Still various issues remain:

• No checksums (SHA-256 preferably) anywhere to be found. Neither in this thread or on homepage.
• No version in the filename of the downloaded tar.gz file. linpack-xtreme-1.0.0.tar.gz would be awesome.
• read -p "Bleble..." at the end of supplied scripts is weird for Linux. I guess this is because you expect people to double click it, which will launch a terminal emulator, that might auto close when the benchmark/test is finished? Still a bit weird, for general use.
• Scripts now autodetect AMD vs Intel CPUs, however it will probably break on other manufacturers like Via or Hygon.
• The tool outputs the version to the terminal now on Linux, but it doesn't tell you which version (Intel / Amd) it is running. It seams like not a problem, but it is, because when copying, pasting and sharing results (terminal output) on Internet, it is preferably to have this information included by default.
  • That is because on few of my Intel CPUs, the AMD (generic gcc compiled one) version is actually faster than Intel one!
• The current Linux version is v1.0.0, which is out of sync with Windows.
• There is no changelog included in the archive.
• Other issues remain.

 

[er557]

 

[Vlada011]

Guys how you set Hyperthreading On?
He detect 6 cores, 12 threads but run only 6 core, 6 threads.

 

[EarthDog]

It's a setting in the bios. Typically found in the cpu features.

 

[Mussels]

If you mean in this program, its off for benchmark, on for stress test

 

[Vlada011]

If you mean in this program, its off for benchmark, on for stress test

Thanks.

 

[hat]

Still using this program to find what my 2600k can do. I can hit 4.6 but it takes some pretty insane voltage and temps are in the 90s... trying 4.4 with more sane voltage.

 

[Mussels]

Still using this program to find what my 2600k can do. I can hit 4.6 but it takes some pretty insane voltage and temps are in the 90s... trying 4.4 with more sane voltage.

my 2500k did 5.1Ghz on a mere H80, and it stayed under 80C
Wish i could remember the settings, still got the chip with no board in the shed - try lower SA/DRAM voltages and see if it helps lower CPU temps?

 

[EarthDog]

Two different chips.... but 2600k should be at least a 4.8ghz chip... even when cooled by a potato.