Post Your Linpack Score

[Regeneration]

Both run in 50% utilization. Overclocked 6800K is likely to be faster than stock Ryzen.

 

[Wavetrex]

Both run in 50% utilization. Overclocked 6800K is likely to be faster than stock Ryzen.

Ugh.
But why?

It should use all the logical threads available, like any other software, even if that means sharing cores via SMT.
LinX (with the old build) uses all 12 threads on my Intel and shoots utilization to 100% and temperatures to the center of the sun, like it should.

 

[Regeneration]

Ugh.
But why?

It should use all the logical threads available, like any other software, even if that means sharing cores via SMT.
LinX (with the old build) uses all 12 threads on my Intel and shoots utilization to 100% and temperatures to the center of the sun, like it should.

Linpack Xtreme has two operating modes: benchmark and stress test.

Benchmark runs only on true cores to provide accurate performance rating (cross-platform, cross-vendor, etc).

Stress test runs on all threads to push your CPU to its limits.

 

[Wavetrex]

Linpack Xtreme has two operating modes: benchmark and stress test.

Benchmark runs only on true cores to provide accurate performance rating (cross-platform, cross-vendor, etc).

Stress test runs on all threads to push your CPU to its limits.

I see.
Well, it doesn't feel accurate at all, in all multi-threaded tasks that I've thrown to my two computers (video encoding, compiling, 3D rendering, data analysis), the 8-core Ryzen 7-1700 @ 3.6 Ghz beats the older Intel i7-6800K @ 4.2 Ghz soundly.
Yet, in Linpack my Intel CPU wins by a huge margin.

Looking at the results of others as well, this entire test seems extremely skewed towards Intel, so the entire "cross-vendor" thing feels like a big lie.

 

[Mussels]

I think thats linpack in general, its never been AMD friendly

 

[Regeneration]

I see.
Well, it doesn't feel accurate at all, in all multi-threaded tasks that I've thrown to my two computers (video encoding, compiling, 3D rendering, data analysis), the 8-core Ryzen 7-1700 @ 3.6 Ghz beats the older Intel i7-6800K @ 4.2 Ghz soundly.
Yet, in Linpack my Intel CPU wins by a huge margin.

Looking at the results of others as well, this entire test seems extremely skewed towards Intel, so the entire "cross-vendor" thing feels like a big lie.

First of all, you'll have to run the standard benchmark.

And second, unlike other apps, Linpack uses AVX instructions and a lot of RAM.

Memory bandwidth is a factor too.

 

[Nutria]

FX-6100@3.0GHz

 

[Arctucas]

v0.9 with new binaries.

 

[er557]

where did you get new binaries and linpack_xeon64?
cheers

 

[Arctucas]

where did you get new binaries and linpack_xeon64?
cheers

Intel website.

 

[damric]

164

 

[nico_80)]

Here's my ryzen 7 3700x.

 

[Conar-XP]

i5 9600kf Stock.

 

[jasko]

i5 4690K@4.3GHz with a strangely high score

 

[LiquidTrance]

Highest frequency I could run my 9900K and still be stable in Linpack Xtreme 1.1.3. The heat is too much for anything higher. Could totally use some custom watercooling right about now. Interested to see what others are getting on their 9900K on linpack xtreme version 1.1.3 as there is barely any data out there for that specific chip/linx version. Please test/share when you can.

 

[Arctucas]

~45°F air through radiator.

 

[LiquidTrance]

~45°F air through radiator.

View attachment 169179

Thank you for sharing this mate. Ok so it looks like the gflops range is about the same for 9900k frequencies as one of the older versions of linx, i think it was either 0.9.6 or 0.9.7., 9900K 5g was getting 550 gflops on one of those. I should clock to 5g and hook up an air conditioner to see if I can get 5g to pass lol. Can do up to 4.85g HT on and 4.9g HT off with regular air. Tried 5g the other day with regular air and temps exceeded 100c quickly/got errors/mismatched residuals, didn't really wanna add any more volts after that to see if it was lack of voltage or heat because 195 amps was already getting pulled.

 

[Arctucas]

Thank you for sharing this mate. Ok so it looks like the gflops range is about the same for 9900k frequencies as one of the older versions of linx, i think it was either 0.9.6 or 0.9.7., 9900K 5g was getting 550 gflops on one of those. I should clock to 5g and hook up an air conditioner to see if I can get 5g to pass lol. Can do up to 4.85g HT on and 4.9g HT off with regular air. Tried 5g the other day with regular air and temps exceeded 100c quickly/got errors/mismatched residuals, didn't really wanna add any more volts after that to see if it was lack of voltage or heat because 195 amps was already getting pulled.

You did not mention what cooler you are using.

What VCore are you using? VCCSA and VCCIO also contribute a little heat. Your AIDA64 memory bench looks good. Is that 32GB of RAM?

 

[LiquidTrance]

You did not mention what cooler you are using.

What VCore are you using? VCCSA and VCCIO also contribute a little heat. Your AIDA64 memory bench looks good. Is that 32GB of RAM?

I'm on a 360mm aio, 4x8gb ram. Not sure how to gauge vcore, i'm on a powersaving preset for acdc loadlines, normal vcore llc(least aggressive vcore LLC in bios), dynamic voltage mode with a +80mv offset. Okay so I just got done completing this. Same settings, different ram OC(latency is worse because motherboard sucks at training rtls/iols at 4000mhz and up). Memory benchmark taken right after the test complete, read/write/copy can look better if I take it right after booting in windows(62/63/61). Regular ambient. Gflops all over the place, my guess is something went off in the background. Based on how my other frequencies scored, I think 5g will give somewhere between 550-555 gflops, maybe 557ish at best.

Ok so I tried for 5ghz/4.6ghzCache again with ram at c15-4133 for the even 35.0ns latency. Cache error with +80mv offset 5 seconds into first loop, then tried again and failed on first loop with +90mv offset. Temps maxed at 106c. vcore was something like 1.28v during the loop. I think 4.9ghz hyperthreading turned on is my limit on regular ambient for linpack xtreme 1.1.3 with my cooling solution or unless I bust out an air conditioner.

 

[Arctucas]

I'm on a 360mm aio, 4x8gb ram. Not sure how to gauge vcore, i'm on a powersaving preset for acdc loadlines, normal vcore llc(least aggressive vcore LLC in bios), dynamic voltage mode with a +80mv offset. Okay so I just got done completing this. Same settings, different ram OC(latency is worse because motherboard sucks at training rtls/iols at 4000mhz and up). Memory benchmark taken right after the test complete, read/write/copy can look better if I take it right after booting in windows(62/63/61). Regular ambient. Gflops all over the place, my guess is something went off in the background. Based on how my other frequencies scored, I think 5g will give somewhere between 550-555 gflops, maybe 557ish at best.
View attachment 169194

Ok so I tried for 5ghz/4.6ghzCache again with ram at c15-4133 for the even 35.0ns latency. Cache error with +80mv offset 5 seconds into first loop, then tried again and failed on first loop with +90mv offset. Temps maxed at 106c. vcore was something like 1.28v during the loop. I think 4.9ghz hyperthreading turned on is my limit on regular ambient for linpack xtreme 1.1.3 with my cooling solution or unless I bust out an air conditioner.

Might I suggest running HWiNFO, as I did, while LinPack X is running?

Then the Voltages, Temperatures, etc. Min/Max can be seen.

 

[LiquidTrance]

Might I suggest running HWiNFO, as I did, while LinPack X is running?

Then the Voltages, Temperatures, etc. Min/Max can be seen.

I run hwinfo64 for a 3 loop intro just to see temps + min/max voltage. at 4.9ghz, it was around 1.256v load voltage I think. At 4.85ghz it was 1.21-1.22v load voltage, i think average was 1.21v load. I don't run hwinfo during the 30 loop because I don't want polling to mess with gflops and/or residuals though i don't think it affects residuals, it might affect gflop consistency. Temps during 4.85v are 90c-95c and temps during 4.9ghz were 95c-97c, maybe topping out at 100c.

Btw, that's a sick 5.266ghz cinebench r20 score you got ^^

 

[Arctucas]

I run hwinfo64 for a 3 loop intro just to see temps + min/max voltage. at 4.9ghz, it was around 1.256v load voltage I think. At 4.85ghz it was 1.21-1.22v load voltage, i think average was 1.21v load. I don't run hwinfo during the 30 loop because I don't want polling to mess with gflops and/or residuals though i don't think it affects residuals, it might affect gflop consistency. Temps during 4.85v are 90c-95c and temps during 4.9ghz were 95c-97c, maybe topping out at 100c.

Btw, that's a sick 5.266ghz cinebench r20 score you got ^^

OK. Thought it might help identify where you could make some adjustment to lower temps.

Which CBR20 score was that? I have done several.

 

[LiquidTrance]

OK. Thought it might help identify where you could make some adjustment to lower temps.

Which CBR20 score was that? I have done several.

The score on the front page of the share your cinebench thread^^. In regards to temps, I think the only thing I could do is lower frequency/volts because i'm already on the lowest form of llc.

 

[Arctucas]

The score on the front page of the share your cinebench thread^^. In regards to temps, I think the only thing I could do is lower frequency/volts because i'm already on the lowest form of llc.

The score on the front page of the share your cinebench thread^^. In regards to temps, I think the only thing I could do is lower frequency/volts because i'm already on the lowest form of llc.

OK. Has not been updated for a while.

This was my best score. https://www.techpowerup.com/forums/threads/post-your-cinebench-r20-score.213237/post-4221302.

You have a 360 AIO, correct? That seems as if it might do a better job of cooling than what you seem to have. Perhaps there is an issue with the cooler?

 

[LiquidTrance]

OK. Has not been updated for a while.

This was my best score. https://www.techpowerup.com/forums/threads/post-your-cinebench-r20-score.213237/post-4221302.

You have a 360 AIO, correct? That seems as if it might do a better job of cooling than what you seem to have. Perhaps there is an issue with the cooler?

Cooler is working ok, pump/fans operational at the speeds/rpms i set them to. I think the chip is a guzzler, wants 1.3v load voltage for 5g in occtlarge avx2 to be stable, about 1.35v for 5.1g. A delid would probably help a bit with temps