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

Linpack Xtreme Released

There is no score, it's not a benchmark. It's a program to maximally stress the CPU with AVX instructions, which also causes the maximum power consumption and heat, and check for any miscalculations/instability. The GFLOPS number is only there for reference, not to compare different PCs among each other in a benchmark table.

Like i said, as soon as you observe bad performance in an actual CPU benchmark, then we can take a look at it. I wouldn't worry about the GFLOPS in Linpack at all. If you suspect something is wrong, i gave you several pointers what to look for (possibly some kind of limiting or throttling), and how to look for it. It could be a safety mechanism of the CPU+mobo, because if there was no limiting, you'd exceed certain power/temp parameters with your overclock.
 
unclewebb said:
What test are you running? How much memory did you originally have during your first test?
Click to expand...

I was just doing stresstest with 4gb 3 pass. had 2x8gb 3600 16-19-19-39 hynic d die in there before, right now have 4x8gb 3600-16-16-16-36 samsung b die in there.

CiTay said:
There is no score, it's not a benchmark. It's a program to maximally stress the CPU with AVX instructions, which also causes the maximum power consumption and heat, and check for any miscalculations/instability. The GFLOPS number is only there for reference, not to compare different PCs among each other in a benchmark table.

Like i said, as soon as you observe bad performance in an actual CPU benchmark, then we can take a look at it. I wouldn't worry about the GFLOPS in Linpack at all. If you suspect something is wrong, i gave you several pointers what to look for (possibly some kind of limiting or throttling), and how to look for it. It could be a safety mechanism of the CPU+mobo, because if there was no limiting, you'd exceed certain power/temp parameters with your overclock.
Click to expand...

Yah I got that from the previous post, thanks for letting me know. I was more just trying to understand the difference and why im seeing the gflops low.
 
TensorVortex said:
UPDATE:

I put 4x8gb samsung b die 3600 16-16-16-36 ram in the 10850k pc, and now linpack xtreme reports around 490Gflops.

So it's the ram but no idea why. anyway that was bothering me a lot yesterday, glad it's not my pc issue.

Now I'm going to try oc my ram see what it can do.

Thanks for all the help.



Hi sorry I been busy yesterday, my ram is 3600 16-19-19-39 Hynix.

<SNIP>
Click to expand...

It is probably your RAM, or more correctly, your timings.

Post a screenshot of AsRock Timing Configurator, please.
 
Hey

Does anyone know why LinpackXtreme hangs when my monitor enters power saving mode? If I change "Turn off monitor after" to "never" it runs fine, but when screen turns off, Linpack just seems to hang/freeze and do nothing. I can still use PC and don't see any WHEA errors.

Yes I can run Linpack without it enabled, but it is concerning me that it hangs Linpack, is this a sign instability? I've tested non OC settings and same issue.

I am using onboard graphics as awaiting 3080 still.

Update: I tested starting PC and letting the PC turn monitor off, then started Linpack and it doesn't run the test, just sits there? I have to restart to run it.

Tx
 
Last edited:
Yachtie2020 said:
Hey

Does anyone know why LinpackXtreme hangs when my monitor enters power saving mode? If I change "Turn off monitor after" to "never" it runs fine, but when screen turns off, Linpack just seems to hang/freeze and do nothing. I can still use PC and don't see any WHEA errors.

Yes I can run Linpack without it enabled, but it is concerning me that it hangs Linpack, is this a sign instability? I've tested non OC settings and same issue.

I am using onboard graphics as awaiting 3080 still.

Update: I tested starting PC and letting the PC turn monitor off, then started Linpack and it doesn't run the test, just sits there? I have to restart to run it.

Tx
Click to expand...
Display driver timeout, bug, or instability. Try the bootable version.
 
it works with Ryzen 5 3600
 
My 8700k@4.7ghz all core work for years at 1.18-1.20v(load voltage in prime95 etc), was at offset + 0.005mv llc2 on asus z370.Tested with linpack, crash in 3-4 minutes. Then I put 4.4ghz all core, 1.18-1.20v load voltages, and it pass hours of linpack.
But in those years I never seen an freeze, shutdown, blue screen.
I begin to test with linpack because I bought some micron e die chips.
But what I don't get is huge vccio vccsa set by xmp for 3200 ram and the same problem on 10900k with 64gb 3600, 1.3-1.4v.
 

Attachments

  • 426AAFCE-1C63-4B59-9673-E24EEB1242A4.jpeg
    426AAFCE-1C63-4B59-9673-E24EEB1242A4.jpeg
    2.6 MB · Views: 202
  • E0A2694C-0F35-4809-9C67-CFBE0467BD9A.jpeg
    E0A2694C-0F35-4809-9C67-CFBE0467BD9A.jpeg
    3.8 MB · Views: 196
reeven said:
But what I don't get is huge vccio vccsa set by xmp for 3200 ram and the same problem on 10900k with 64gb 3600, 1.3-1.4v.
Click to expand...

That seems to be the norm. Enabling the XMP profile makes the board raise certain voltages, namely VCCSA, VCCIO, and for Rocket Lake, also VCCIO 2. They can be set crazy high sometimes, which is not necessary at all though, unless you go way above DDR4-3600.

The board makers like to use a very high safety margin for those voltages, which just isn't needed. At DDR4-3600 and especially below, you can most likely set the same low standard voltages as it sets without XMP at DDR4-2133.

Do the following: Disable XMP, reboot, and if it uses DDR4-2133 frequency now, note the SA and IO voltages in the BIOS. Now set them manually to the values that are active. Here are my standard values with Rocket Lake for a lower RAM frequency, which i manually enforced for DDR4-3200:

BIOS Voltages SA IO.png


By the way, you can press F12 to save a clean screenshot like this to a FAT32 USB Stick in the BIOS.

Note that i - for the time being - use a kit of older Kingston HyperX DDR4-3200 that only needs 1.2V DRAM Voltage to reach that frequency, albeit at looser timings than normal. I will soon use this as my main system and put my DDR4-3600 in there which will require the more common 1.35V for its XMP frequency.

Now it should still run like a charm, without elevated SA and IO voltages. You can do some RAM stress-testing now.

On my Z390 board, i'm even running at DDR4-3600 with stock SA and IO voltages. It just doesn't need more for stability. This is a good way to save a couple Watts.
 
reeven said:
My 8700k@4.7ghz all core work for years at 1.18-1.20v(load voltage in prime95 etc), was at offset + 0.005mv llc2 on asus z370.Tested with linpack, crash in 3-4 minutes. Then I put 4.4ghz all core, 1.18-1.20v load voltages, and it pass hours of linpack.
But in those years I never seen an freeze, shutdown, blue screen.
I begin to test with linpack because I bought some micron e die chips.
But what I don't get is huge vccio vccsa set by xmp for 3200 ram and the same problem on 10900k with 64gb 3600, 1.3-1.4v.
Click to expand...
Common occurrence.
If it crashes in linpack or p95, it's simply not completely stable. This might or might not show in BSODs and the sort. For me, such borderline instability showed in application/games occassional crashes. Let's say 2 times/week some more CPU intensive title like Assassin's Creed would crash to desktop.
Once I got my 8700K fully stable, all crashes went away. Used to be a saw line in the View Reliability History thing in Windows 10. Now it's a line that sticks to the top without any "red" i.e. crashes.
 
În linpack, 64gb corsair 3600c18 m-die hynix not stable at xmp 1 or 2 with 10900k with 1.4 vcio vccsa.
Those other 32gb 3200c16 micron e-die double sided /8700k seems stable at xmp but with 2t.
And my oem 32gb c-die Samsung goes stable even at 3200c16 1t with 1.21.
I miss old micron days like bh5 ddr1, or ddr3 micron that went 6-6-6 or 7-7-7 1600. Those just work.

Edit: with 8700k 1.0 vcssa vccio auto at 2400mhz. Same for 10900k.
Those aren't stable in linpack at all at xmp speed. Not stable even at 1.15, 1.20, 1.25. Those are bios voltages, în hwinfo is higher.
 
reeven said:
În linpack, 64gb corsair 3600c18 m-die hynix not stable at xmp 1 or 2 with 10900k with 1.4 vcio vccsa.
Those other 32gb 3200c16 micron e-die double sided /8700k seems stable at xmp but with 2t.
And my oem 32gb c-die Samsung goes stable even at 3200c16 1t with 1.21.
I miss old micron days like bh5 ddr1, or ddr3 micron that went 6-6-6 or 7-7-7 1600. Those just work.

Edit: with 8700k 1.0 vcssa vccio auto at 2400mhz. Same for 10900k.
Those aren't stable in linpack at all at xmp speed. Not stable even at 1.15, 1.20, 1.25. Those are bios voltages, în hwinfo is higher.
Click to expand...
RAM stability is a pain.
In my case, with some SpekTek chips, the RAM was unstable because of the high XMP DDR voltage. I literally had to reduce voltage, against all previous experience, from 1.35V to 1.20 for the XMP at 3GHz to work. For a 3600 MHz RAM OC to be stable, anything from 1.25 to 1.32V would do. Around 1.34V instability starts to creep up.
As for the VCCSA and VCCIO voltages, the motherboard would set them at huge values for XMP or OCs, like 1.33 and 1.25V. For my CPU and RAM it was useless. I am now keeping it around 1.15 and 1.10 and I could probably reduce them further (3500MHz OC on RAM), but I got bored a while ago with the tweaking.
Took me a year of running Blend and Linpack on and off over night to figure out this weird stuff. Sometimes I'd have workers stop after 10 hrs as the PC was close to stability. Much more annoying to stabilize the RAM compared to CPU OCs.
 
I find the culprit. It was 1t vs 2t. I put it 1t, but my micron hate this. 2t pass linpack and karhu over 20.000%, memtest hci over 1500%.
I have now 1.1 vccsa vccio manual set, along with 2t on ram on 8700. I can't acces pc with 10900k till next week to test it as well.
I always use 1t on any ram...so this 2t is weird for micron e-die.
 
Very odd using this software linpack xtreme v1.1.5 on my Ryzen 7 3700X all passed but linx v0.9.5 for AMD has failed (not enough memory) did try that on clean windows 10 pro with no other app running. Still the same error not enough memory (is that linx v0.9.5 don't work for Zen2 CPU?)
Memory test passed.jpg


Untitled.jpg
 
Both LinX and Linpack Xtreme are basically interfaces for Linpack. I know back in the day when i used LinX, it was quickly outdated, and some forum members provided updated Linpack binaries etc... but it was all unofficial. So in the end i just switched to Linpack Xtreme and never looked back.

They are doing one and the same thing, just with a different UI. So if Linpack Xtreme works and confirms the stability, there's no need to use LinX in addition. That "not enough memory" is probably just a bug of LinX.
 
CiTay said:
Both LinX and Linpack Xtreme are basically interfaces for Linpack. I know back in the day when i used LinX, it was quickly outdated, and some forum members provided updated Linpack binaries etc... but it was all unofficial. So in the end i just switched to Linpack Xtreme and never looked back.

They are doing one and the same thing, just with a different UI. So if Linpack Xtreme works and confirms the stability, there's no need to use LinX in addition. That "not enough memory" is probably just a bug of LinX.
Click to expand...
That's good to know. :)
 
Optiplex 9020 USFF, Delided CPU
Stock Fan+Heatpipe
All system max at 134.4W by Watt Meter :)
 

Attachments

  • 4.png
    4.png
    455.9 KB · Views: 288
Last edited:
linpackX115 linx0911 4-29-21.JPG
 
Have done some benching for my 5950x with different core amounts in Linpack Xtreme_x64 1.15 downloaded from here

  • 5950x @ custom water+TechN block with latest CTR running.
  • SMT enabled for all runs
  • 4x8GB (gskill 3600 cl16) running flat 14-14-14 timings running on T1 setup-time (no GDM)
  • Minimum required vdimm = 1.54v (maxtemp below 40degress even when gaming with a 3090)
1623534993609.png


Benchmark 3GB numbers with 6 cores
  • 6 cores at 4.5ghz static OC = 266 Gflops
  • 6 cores at PBO CO -30 = 281 Gflops
  • 6 cores maxed (static 4.85ghz) = 286 Gflops

Benchmark 8GB extended numbers with 6 cores
  • 6 cores at PBO CO -30 = 283 Gflops
  • 6 cores maxed (static 4.85ghz) = 287 Gflops
1623535160802.png

Benchmark 3GB numbers with 16 cores
  • 16 cores at 4.5ghz static OC = 641 Gflops
  • 16 cores maxed (static ~4750/4600mhz) = 654 Gflops

Benchmark 8GB extended numbers with 16 cores
  • 16 cores at 4.5ghz static OC = 668 Gflops
  • 16 cores maxed (static ~4750/4600mhz) = 680 Gflops
1623535222756.png

Have numbers for my cpu at 8 and 12 corecount also if anyone are interested in those

*edit*

Aha, i see!
I'm was using "benchmark" while you guys is using "stresstest" :)


Static OC @ 4675/4500mhz, 1.31v set = ~1.24 vcore get
Around 320watt usage (maxtemp = 93)

Stresstest 10GB = ~700 Gflops
1623541764898.png
 
Last edited:
linextreme.png
 
it can possible to know the root password for distro of linpak?
 
You must log in or register to reply here.
Back
Top