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

Linpack Xtreme Released

I haven't touched any other voltages. We're talking voltage required to keep it stable under Linpack Xtreme though, which is a pretty demanding test... it produces a magnificent amount of heat and requires some pretty high voltages to stay stable. The 2500k would be a lot cooler than the 2600k though due to the absence of HT...
 
Maybe this rest is too much..too unrealistic of a load. There's a reason why people dont/shouldnt run Furmark on gpus.....this feels quite similar.
 
EarthDog said:
Maybe this rest is too much..too unrealistic of a load. There's a reason why people dont/shouldnt run Furmark on gpus.....this feels quite similar.
Click to expand...
nope, stress the shit out of it. Furmark exposed shitty hardware designs that could have been left alone otherwise, and forced GPU makers to design thermal and wattage limits into cards - it wouldnt have taken long for a game or program (think openCL) to expose those flaws and kill hardware.

If your hardware cant handle 100% load, you've bought faulty goods or really f*cked up your overclock.
 
Furmark is a power virus. Both amd and nvidia suggest not using it. Furmark doesnt even test the same clockspeeds you set. Games or programs would not have done that, sorry.

This is also overkill. Unless you're a prosumer, this kind of 'stability' isnt needed. P95, also overkill, is plenty. I stick with aida64 and it has worked fine for MY uses for years. This program leaves way too much meat on the bone if its holding sandy bridge to mid 4 ghz...
 
Furmark just run an infinite loop of 3D text in OpenGL or Vulkan. It doesn't use VRAM much.

Furmark.jpg

If your hardware can't take it, something is wrong with the design.

PUBG puts more stress on the GPU than Furmark. Uses a lot of VRAM and reloading textures all the time.
 
It may test more ram, pubg, but it isnt more stressful than furmark for the GPU and temperature/power wise. Modern hardware can take it...because both amd and nvidia throttle when using the application. For example, if my boost clocks normally hold at 2000 mhz for gaming, when running furmark, I'm back to stock base boost bin in the 1800s (will vary by card a bit). This, obviously, defeats the purpose of testing. Again, it's in the nvidia press deck for generations they suggest not using it.

But, this about linpack being overkill. To each their own but this is at least a step too far for my needs. :)
 
If your hardware works fine in linpack on stock clocks, Linux or windows, it should equally pass it fine when overclocked. period. If it doesn't, something is somewhere wrong with the overclock. It is not overkill if it exposes hardware weakness.
 
er557 said:
If your hardware works fine in linpack on stock clocks, Linux or windows, it should equally pass it fine when overclocked. period. If it doesn't, something is somewhere wrong with the overclock. It is not overkill if it exposes hardware weakness.
Click to expand...


^ this is a good way of wording what i wrote earlier

If your hardwares VRMs cant handle 100% load at stock clocks, its faulty
if it cant handle it OCd, your OC is bad
 
But what if it can handle things you normally do on a PC? It leaves a lot of meat on the bone (too much). I see a point in overkill testing, but this seems to be in a whole new level.

What I do see a use for this test is what I call blip stress testing....running it for 5 mins before I raise mutli/bclk/or lower voltage. At least it can identify problems faster. But yeah, I dont need this kind of stability on my daily rig when I know aida64 works for me and my uses.
 
Who cares if the test stresses things more than your usual load levels?

Its like saying your car does 60KM/h in normal use, so you dont care if it falls apart at 80KM/h - so you're fine with the manufacturer never testing it for it for safety.
Logically speaking, i want to know if i suddenly need it to go harder, it always can.

I feel like we're de-railing the thread a little bit now, however.
 
My point is that it's well above an already overkill stress test (p95) and this leaves too much meat on the bone being 'more' stressful. To me, for my uses (which I'd call a power user) this goes in the bin along with Furmark. This is more akin to testing your car at rev limiter/fuel cutoff than testing in normal situations (redline - which is what P95 and others do already). It is well pas the 'go harder' stage. P95, and others, do a solid job as it stands of 'testing it for when one 'suddenly' needs to go harder'.

er557 said:
If your hardware works fine in linpack on stock clocks, Linux or windows, it should equally pass it fine when overclocked. period. If it doesn't, something is somewhere wrong with the overclock. It is not overkill if it exposes hardware weakness.
Click to expand...
Is it really a weakness though when it is a situation you will never, I repeat, NEVER run into? What good is testing it like that when it isn't remotely realistic? Other programs are already overkill... this is a whoooooole 'nother level of stress testing.



Apologies... I figured discussing the application (its merits and demerits) the thread is about was on topic...
 
Last edited:
I'm curious guys, what's the differences between this and the LinX I've used previously? Tougher testing or extra tests for newer instruction sets? :)
 
Regeneration said:
Both. LinX is no longer maintained by the developer.
Click to expand...

Thanks :)

I've had some funny experiences with stress testing software.. It seems that sometimes even stress testing for 16 hours, put it into WCG, it would crash inside of 30 seconds.. It was rather frustrating :(
 
phill said:
I've had some funny experiences with stress testing software.. It seems that sometimes even stress testing for 16 hours, put it into WCG, it would crash inside of 30 seconds.. It was rather frustrating :(
Click to expand...

LLC, dynamic voltage and C-states (if enabled) are likely to blame for this.

It is wise to test stability of low and medium load (limit Linpack to 25 or 50 percent of available threads).
 
This was back in the day when I was using Orthos stress testing software and even Prime 95 I believe with Core 2 Duo CPUs :) I think I went to WCG for stress testing as it just worked and soon showed an error but over time of learning pro benching and such, it's taught me a lot of little things that I have picked up :)
 
Do these non-matching Residual values indicate there is an instability in the overclock?

244241c271.png
 
hutt132 said:
Do these non-matching Residual values indicate there is an instability in the overclock?
Click to expand...

Different residual values are normal on modern Intel CPUs.

Abnormal on AMD and legacy Intel CPUs (Sandy Bridge and older).
 
Nice app. Gonna replace LinX and IBT with it I guess.
It makes my 8700K draw the same amount of power as the last p95 smallFFTs /w AVX, around 130W (uncapped power limits and undervolted but stock multis, i.e. 4.3GHz for this specific case). I like that I can test lots of RAM too however, I made the mistake of buying some cheaper RAM and fought to stabilize XMP for basically an entire year. Who would have thought that all I had to do was actually decrease voltages, not increase them. My 3000MHz CL16 RAM runs at 1.2V VDIMM and 1.15 SA, 1.10 IO. Tried for ages to get it stable with higher voltages and always failed (p95 Blend workers would crash very late, sometimes after 11 hrs, and some games would also crash, typically with some 0xc0000005 (Access Violation) error.
Happily it's all in the past, but damn it was a struggle.
 
Hello to everybody, I'm new member and I'm trying to undervolt my Dell XPS 9570 i7 8750h, 32 GB RAM using Throttlestop.
Even though I read a lot of tutorials about it, I still have some doubts.
I started to do tests from - 50mV with TS Bench built in, the machine didn't crash but i was able to get 0 errors at 1024M and 12 threads only around an undervolt of -113.3 mV for core and cache... if I try to undervolt less, i start to get a lot of errors and around -140 mV I get tons of errors despite no BSODs with Prime95 (for 12 hours) or Aida64.
After asking for some help, I discovered the great Linpack Xtreme to test the stability, but I'm not sure how to set it up...
I tried to set it using 30 GB of Ram (experimental) (problem size: 62897), with the maximum of trials (also if i could set something like 3) and all of cores.
Unfortunaly, most of times, my system gets freezes or Bsods before to finish the first trial, also with a voltage offset of -113mV or -105mV...
It seems i can't undervolt anything.
This looks like very strange... because with other stress test software like Prime95, I can continue the test also for 12 hours (I tried with -140mV) without any errors or Bsods.
I read that all the people who have this machine (with i7 8750h) are able to undervolt it also with -140mV ore more... at least with -120mV.
Is there anything I do wrong? How should I set Linpack Extreme for my machine to be sure it is stable?
Or maybe isn't my laptop so lucky or has any issues? I don't know if an undervolt which is around -70 or - 80 mV still makes sense for this pc/cpu model.
I'm quite disappointed.
Thank you for your help.
cleardot.gif
 
Last edited:
Dont worry about this application. It's an unrealistic load. P95 or similar applications are fine. If you are stable using your machine after testing with those others, than its stable for you....which is the important part. Nkt trying to pass every stress test around, including this overkill app.

Regarding undervolting, every cpu is different as to how far, if any, it can undervolt.
 
Solid said:
How should I set Linpack Extreme for my machine to be sure it is stable?
Or maybe isn't my laptop so lucky or has any issues? I don't know if an undervolt which is around -70 or - 80 mV still makes sense for this pc/cpu model.
cleardot.gif
Click to expand...

Undervolt is like overclock. A matter of luck.

The right way to run Linpack Xtreme: choose stress test, 10GB is sufficient enough, enter 10000 times and use all available threads.

For maximum stability, keep it running for 8-12 hours and ensure cooling is sufficient (20c below Tjunction, usually 80c).

EarthDog said:
Dont worry about this application. It's an unrealistic load. P95 or similar applications are fine.
Click to expand...

Both Linpack and Prime95 solve mathematical equations. Linpack just solves a larger equation and distributes the load between cores in a more balanced manner. On modern Intel CPUs, Linpack uses the latest AVX-512 instruction set to speed up the calculation.
 
Last edited:
80C is this application? That's like 70-75C using something that isnt overkill. The amount of meat left on the bone here by using this app AND saying a max of 80C leaves enough meat on the bone for a Xmas meal.
 
Thank you to both, the machine goes in thermal throttling during the test, and throttlestop shows that some alarms are active, like PL1, PL2, VR Current and EDP OTHER. The Dell XPS is a hot pc.. so, the system provides to lower the temp around 84°C when it is throttling.
I know that every chip is different, but I've never heard of an i7 8750H which is unstable at -100mV
 
You must log in or register to reply here.
Back
Top