PrimeGrid
Please visit donation page to help the project cover running costs for this month
1) Message boards : Number crunching : Yuri's Night Challenge (Message 150095)
Posted 59 days ago by Profile Chooka
Thanks for organising this focus on WW.
2) Message boards : Number crunching : Yuri's Night Challenge (Message 150078)
Posted 60 days ago by Profile Chooka
44% 1st rate rate with a Radeon VII - Ouch!
It's not a fan of WW :)
3) Message boards : Number crunching : Yuri's Night Challenge (Message 149990)
Posted 63 days ago by Profile Chooka
Dare I say... I feel this is the worst, GPU subproject for Radeon cards lol.
Terrible.

GLHF
4) Message boards : Wieferich and Wall-Sun-Sun Prime Search : Slow speed on RX 5700XT (Message 149841)
Posted 70 days ago by Profile Chooka
Just over 1000 seconds for a Radeon VII :( It gets belted by NVIDIA here.
1660Ti is quicker.

Oh well.
5) Message boards : Number crunching : Sier"pi"nski's Birthday Challenge (Message 149825)
Posted 71 days ago by Profile Chooka

Hi Nick,

Yes it was you I was referring to! :D
Your PC's with AVX-512 are really great for PG. I assume that's why you bought them? They are WAY faster than my 3950X


Guilty! I bought my 2nd computer - a 9960X - entirely for primegrid.
Initially I wasn't able to use AVX512 and comparing computer throughput, it ran not much better than the 8 core 9900K.
I remember that to get best throughput, it was better to use around 8 to 10 cores.
That all changed when AVX512 was introduced into LLR.

I initially chose the 9960X because I was considering a 9920X - I thought maybe the 9940X, but settled on the 9960X because of the extra cache.
But it left me wondering if I should have gone all out and got the 9980XE. So I got one.
It underperformed the 9960X, core for core by 15% and was hotter.

Then I got really silly and bought another 9960X confirming that it is (was) a better choice.
I say was because there are 2 or 3 projects that now test it's ability to run 2 tasks in cache at once, whereas the 9980XE should have an advantage with those 2 or 3 projects over the 9960X.
If I was to buy another intel, I would buy another 9960X (or newer AVX512 16 core if it exists). I am not going to buy any more hardware - I have already gone too far.
My 4th computer is ready to go for a real graphics card, but I am not going to go there.


Hahaha there you go! :)

I hear you on the hardware front. Enough (is never enough) but it will have to do when it comes to crunching pc's.

It will be interesting to see if the rumour is true about Zen 4 supporting AVX-512.
Could be a worthy upgrade. Especially if it's on 5nm. (code name GENOA)

Start saving. :D

https://www.techpowerup.com/279129/amd-zen-4-microarchitecture-to-support-avx-512#:~:text=The%20next%2Dgeneration%20%22Zen%204,web%20on%20the%20ChipHell%20forums.
6) Message boards : Number crunching : Sier"pi"nski's Birthday Challenge (Message 149745)
Posted 75 days ago by Profile Chooka
Interesting!
Thanks for sharing Scott.

I really noticed during the challenge just how much quicker an Intel PC with AVX-512 was compared to say a Ryzen 3950X.

A 9960X (16 cores) did SoB w/u's in around... 30-33K sec per w/u vs a 3950X (16 cores) which was around 70 - 80K per w/u.

I think I'm the only person with 9960Xs.
What is interesting is that some tasks ran much slower.
All slow tasks had FFT of 3000K (except for one task with FFT of 3200K)
All tasks with an FFT of 3360K were fast (33K sec)
A fast task with FFT of 3000K was 29K sec.

Even my 9980XE was nearly as fast as a 9960X for fast tasks and it has shown that it can nearly keep up - so now I know that it slacks off. Up until running this challenge, in nearly every project, the 9960Xs were faster than the 9980XE by about 10-15% and run about 10 degrees C cooler. Go figure?

I know an AMD 5950X that was doing 2 x 8 thread tasks in 54K sec which was with FFT of 3000K
And 65K sec for FFT of 3456K - these are faster than my 9960Xs by about 7% for the FFT of 3000K

I learnt on the last day that I could download many tasks and quickly work out which ones were going to run as well as possible - aborting the rest.


Hi Nick,

Yes it was you I was referring to! :D
Your PC's with AVX-512 are really great for PG. I assume that's why you bought them? They are WAY faster than my 3950X
7) Message boards : Number crunching : Sier"pi"nski's Birthday Challenge (Message 149665)
Posted 80 days ago by Profile Chooka
Interesting!
Thanks for sharing Scott.

I really noticed during the challenge just how much quicker an Intel PC with AVX-512 was compared to say a Ryzen 3950X.

A 9960X (16 cores) did SoB w/u's in around... 30-33K sec per w/u vs a 3950X (16 cores) which was around 70 - 80K per w/u.
8) Message boards : Number crunching : Sier"pi"nski's Birthday Challenge (Message 149659)
Posted 80 days ago by Profile Chooka
Thanks for organising the challenge Michael.
Its amazing to see how well one can do when not up against a lot of people renting servers to crunch with ie - the TdP) :D (or perhaps those users just couldn't be bothered with SoB)

Till the WW challenge! See you then.
9) Message boards : Number crunching : Sier"pi"nski's Birthday Challenge (Message 149299)
Posted 95 days ago by Profile Chooka
Sorry Freestman... could you please break down what I'm looking at with your results?

Lets take the 6700K for example.
Running all 8 threads, it takes 5.48 days to complete 8 tasks.
Running 4 cores, it takes on average around 4 days to complete.
I'm struggling to work out if it's better to run 8T or 4C based on your figures sorry. If I'm reading it right, you'll finish 8 tasks in 5.48 days vs 4 tasks in around 4 days. Is that right?

I didn't think running all threads was ever better than having H/T off? Either way, you'll only complete 8 tasks within the 10 day period.

Did I get that right?
10) Message boards : Number crunching : a variety of CPU generations, overwelmed by all the subprojects in a good way (Message 149269)
Posted 96 days ago by Profile Chooka
On the 3700x it's still faster to run anything above ~2M fft with 8 threads.

3950x which scales near enough to 2x the 3700x:
Timings for 2880K all-complex FFT length (16 cores, 2 workers): Throughput: 1132.42 iter/sec.
Timings for 2880K all-complex FFT length (16 cores, 4 workers): Throughput: 503.76 iter/sec.


And with my 3950X (for the upcoming SoB challenge for anyone interested) -

Timings for 3200K FFT length (16 cores, 1 worker): Throughput: 464.14 iter/sec.
Timings for 3200K FFT length (16 cores, 2 workers): Throughput: 602.66 iter/sec.
Timings for 3200K FFT length (16 cores, 4 workers): Throughput: 289.66 iter/sec.



Next 10 posts
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2021 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 3.10, 3.35, 3.49
Generated 14 Jun 2021 | 4:48:59 UTC