• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Linpack Xtreme Released

JKRsega said:
I'm a n00b! Erm, where would I see that information?
Click to expand...
On Intel/AMD website, just Google your CPU name. If its Ryzen 5 3600, then Tjmax is 95c, so your should aim for maximum of 85-90c/185f-194f.

You can use software like AIDA64 / HWMonitor / HWinfo / Ryzen Master to monitor the CPU temperature.
 
I've ran it about 10 times now and my computer hasn't turned into a ball of flames, so all good and it got to 83 degrees with no crashes. Stable, unlike me, lol.

So I did the stress test and got scared when my cpu got to 93C so quit the test. Need better cooling.
 
Are there going to be any updates, @Regeneration ? It's been a long time since last version. The linpack libraries must have been updated a few times since then?
 
JKRsega said:
With how demanding your program is Mr Regeneration, I'm a bit scared to run it! It's not going to damage anything is it?
Click to expand...
If it damages your system, then your system was setup incredibly badly.... like overclocking to 400W and unplugging your AIO's pump with all safety features disabled.
 
Mussels said:
If it damages your system, then your system was setup incredibly badly.... like overclocking to 400W and unplugging your AIO's pump with all safety features disabled.
Click to expand...
Hasn't damaged mine, but it got too hot for my liking, so I CTRL and break'd it.
 
Regeneration said:
I didn't like the MKL 2020-2021 binaries and that's why I didn't update. The recent ones are much better, so maybe in the near future.
Click to expand...
Oh? Could you elaborate? I'm curious.
 
Hi,
i was trying Linpack Extreme because i want to find a Stress Test with "scores" while it is running. Also i need a benchmark that works on Windows and Linux.

On Ubuntu 22.04, Kernel 5.19rc8 i don't get the benchmark to perform right. I missing 20% of performance. Also, the system get not near that hot like on windows.

Is there something on the linux version to take care of? The permission, normal, sudo, root does not make any difference. Disabling all mitigations gets me around 10 Gflop/s more.

My System: i7-8086k, 16GB DDR4 4000, Maximus XI Apex, RTX 3080.

Thanks for your benchmark!
 

Attachments

  • Screenshot_Windows.png
    Screenshot_Windows.png
    166.7 KB · Views: 205
  • Screenshot_Ubuntu.png
    Screenshot_Ubuntu.png
    300.6 KB · Views: 194
24/7 -882.48
Екранна снимка (301).png
 
I can also play that game :laugh:

24/7 rock stable settings

907 average gflops
1672055810036.png


875 averag gflops when using the benchmark-part (3200 problem size)
1672055944106.png
 
Not much interest in this thread it seems, but i have completed a new round of memory stresstests for the 7950X3D
Follow stresstest all done in a single run: (no disabling of L2 prefetchers to artificially increase the performance numbers in certain benchmarks)
  • 3 hours karhu @ 314.69MB/s
  • 1 hour y-cruncher all tests enabled
  • 1 hour testmem5 1usmus
  • 30 cycles Linpack Extreme 14GB stresstest
54.9ns-19.png
 
It's a great stress test, but it's quite uneven between architectures (and has false positive errors on some) so it dropped out of fashion

still a useful tool to have
 
Maybe it just needs an update... when was the last one? Two years ago?
 
If it doesnt stress the extra FADD pipeline, just FMA will not crash ryzen 7000 series. Actual flops per cycle of zen 4 core is 48. 32 from fma, 16 from fadd. Aida64 gpgpu says my ryzen is 2100 gflops but actually it is 3150 gflops if fadd is used.
 
Looks I'm going to need an upgrade soon..

Joining the 500 club a couple of years ago was cool, joining the 1000 club would be sweet :)

Screenshot 2023-05-21 152716.png
 
Ryzen 9 7900 (its 5.3 GHz stable on all other kind of work) [theoretical max = 3150gflops, theoretical FMA max = 2100 gflops, real-world algorithm = 560 gflops]
This is at 142 Watts total power of CPU or 117 Watts without iGPU, etc.
VmL8X9[1].jpg


I don't know why but picking extreme test (30GB one) increases GFLOPS value to ~600 and temperature sees 77 Celcius max with 5.0 GHz boost(not 4.8 as it says in there):

0VP2nI[1].jpg

If GCC v12 was not used for this project, probably it doesn't even use the AVX512 of this CPU, let alone the extra FADD pipeline.
 
Last edited:
Is it game over for the IMC on my Ryzen 7 5800X? Linpack has a problem of randomly failing with more than around 16 GB tested. If I set it to the highest problem size. And it also failed at only 2133!

It never failed with the current 16 GB kit I have in there right now. FFS, I have a 2x8GB kit at 2666 running fine. But the 32 GB of TridentZ Neo ends up resulting in a fail even at 2133!

Gives me a hard time, as if I have all 4 slots filled up, but never have more than 2 slots filled up! Is this a known issue with using more than somewhere around 16 GB?

Mussels said:
It's a great stress test, but it's quite uneven between architectures (and has false positive errors on some) so it dropped out of fashion

still a useful tool to have
Click to expand...
Linpack?
 
Last edited:
RJARRRPCGP said:
Is it game over for the IMC on my Ryzen 7 5800X? Linpack has a problem of randomly failing with more than around 16 GB tested. If I set it to the highest problem size. And it also failed at only 2133!

It never failed with the current 16 GB kit I have in there right now. FFS, I have a 2x8GB kit at 2666 running fine. But the 32 GB of TridentZ Neo ends up resulting in a fail even at 2133!

Gives me a hard time, as if I have all 4 slots filled up, but never have more than 2 slots filled up! Is this a known issue with using more than somewhere around 16 GB?
Click to expand...

Not that I know of. Memory allocation errors happen in the beginning and not just randomly, unless there is some background service that consume a lot of memory.

Try running other multithreaded memory stress test software like HCI MemTest, MemTest64, MemTest86 to be sure.
 
You must log in or register to reply here.
Back
Top