Prime95 vs. Linpack Xtreme

Nov 3, 2007
28
3
91
#1

The video above demonstrates the superiority of Linpack Xtreme versus the latest version of Prime95. The overclocked PC passed nearly 2 hours of Prime95's small FFTs torture test and yet completely crashed within less than a minute and a half (77 seconds to be precise) with the bootable version of Linpack Xtreme integrated on Porteus Linux.

Computer specifications: Intel Xeon W3680 @ 4.17 GHz (144x29), ASUS P6T (vanilla), 3x 4GB G.Skill Ares F3 @ 2304 MHz (10-12-12-31 CR1), and Nvidia GeForce GTX 970.
 
Last edited:

JDG1980

Golden Member
Jul 18, 2013
1,522
0
136
#2
The Xeon W3680 is an ancient Westmere CPU. Do we know if Linpack Xtreme works as well as Prime95 on modern processors (Coffee Lake or Ryzen)? I do know that current versions of Prime95 use AVX instructions, which can push recent Intel CPUs harder than non-AVX workloads. However, it's still possible that Linpack Xtreme is exercising a wider variety of instructions and might catch some overclocking glitches that Prime95 won't. But if you're worried about stability, the Linux kernel compilation might be an even better choice. Testing for maximum power/temp and testing for stability are two different things, though they are related.
 
Nov 3, 2007
28
3
91
#3
Same outcome for AVX CPUs.

I use Westmere for my experiments, extremely cheap on eBay, with a lot of room for extreme overclocking and tweaking.
 

Charlie22911

Senior member
Mar 19, 2005
519
26
116
#4
In my experience the various Linpack benchmarks have done a far better job at finding instability than Prime. I’ve used LinX with similar results for some time now (windows tool, not to be confused with Linux). In fact I don’t consider my overclocks stable unless it can pass 8 hours of LinX with no AVX offset; absolute stability is a must IMO.
 

ehume

Golden Member
Nov 6, 2009
1,279
3
91
#5
In my experience the various Linpack benchmarks have done a far better job at finding instability than Prime. I’ve used LinX with similar results for some time now (windows tool, not to be confused with Linux). In fact I don’t consider my overclocks stable unless it can pass 8 hours of LinX with no AVX offset; absolute stability is a must IMO.
LinX is simply a GUI for Linpck, so when you use LinX you are using Linpack. Before I retired I used LinX 0.6.5 -- Linpack with AVX2. That ran hotter than Linpack with AVX (e.g. OCCT).
 

tamz_msc

Platinum Member
Jan 5, 2017
2,148
86
106
#6
One can just run the latest version of Intel's MKL benchmarks containing Linpack from here.
 
Jun 30, 2004
13,409
38
126
#7
In my experience the various Linpack benchmarks have done a far better job at finding instability than Prime. I’ve used LinX with similar results for some time now (windows tool, not to be confused with Linux). In fact I don’t consider my overclocks stable unless it can pass 8 hours of LinX with no AVX offset; absolute stability is a must IMO.
From a reference point of an i7-6700K and earlier K processors, I would think that 8 hours of LinX is unnecessary punishment. There IS a Prime95 test that takes about an hour to show any instability if it exists, or close to it. After that, if LinX gave 35 iterations without error and a narrow variance in the GFLOPS results among those 35 iterations, I would think you'd have total stability. Caveat -- you can best evaluate the GFLOPS variation by affinitizing LinX. That is, if you have a chip featuring HT, you would affinitize so that only every other "virtual processor" is used.
 
Apr 27, 2000
10,494
346
126
#8
Prime95 needs to be tuned to the CPU for maximum heat.

For example, on my old A10-7700k, I had to set it to Custom, in-place FFTs, and fix the min and max FFT sizes to . . . 3/4 the L2 cache size per module, or something like that?
 
Nov 3, 2007
28
3
91
#9
A user on 3DCenter forums posted a comparison of Linpack Xtreme vs. other stress tests on i7 8700K.



And that's with the Windows version, which is a bit inferior to the Linux version.
 

wantkitteh

Junior Member
Nov 24, 2018
1
0
6
#10
Sorry to necro the thread a little, but the OP may be mistaken about the actual cause of the instability exhibited in the video. Running the Small FFTs test in Prime95 places zero load on system memory while Linpack hammers it pretty hard - I would speculate that the system CPU stability was absolutely fine in both tests but marginal memory stability was the factor that caused the Linpack test to crash so quickly.
 
Apr 27, 2000
10,494
346
126
#11
That's why you run Blend on Prime95 if memory stability is in question.
 
Nov 3, 2007
28
3
91
#12
Memory stability wasn't the reason Linpack crashed after a minute.

Linux pushes modern hardware even further when stress tested. The scheduler is designed for MP.

MemTest64 and HCI MemTest are better to determinte memory stability than Prime95 blend.
 
Apr 27, 2000
10,494
346
126
#13
MemTest64 and HCI MemTest are better to determinte memory stability than Prime95 blend.
I find that MemTest64 gives a better picture of memory stability based on which test fails and how severely. But if you just want to crash a machine with unstable memory, Linpack; Prime95 Blend; and y-cruncher can all get the job done. In some cases, so can SuperPi 1.5 XS .
 

Similar threads



ASK THE COMMUNITY