PrimeGrid's 2022 Challenge Series

[pututu]

I know right? I don't even have all boxes contributing....this should be a cakewalk for the rigs he has...lol

No way in PG. I downsized my xeon rigs over the past year or so. WFeather has one of my haswell rigs. Where is he???

My Amd rigs are currently in use to race with motqalden in MLC.

 

[motqalden]

No way in PG. I downsized my xeon rigs over the past year or so. WFeather has one of my haswell rigs. Where is he???

My Amd rigs are currently in use to race with motqalden in MLC.

Well you are gonna lose so...

 

[pututu]

Day 3.5 result. 1.5 days to the finishing line.

 

[pututu]

Final result. We finished a respectable 8th place with 5 participants.

Next PG challenge is a sub-project GFN-19, best run with gpus.

 

[EXT64]

Nice work team!

It is unfortunate they had so few competitions this spring when it is cooler, yet so many coming up in the summer. Oh well. And the next is GFN-19, which I am not very familiar with so I will need to do some CPU/GPU performance testing on. I imagine at that large of a size and only single threaded, CPUs may be slow and painful compared to GPUs, but we'll see.

 

[wareyore]

Nice work team!

It is unfortunate they had so few competitions this spring when it is cooler, yet so many coming up in the summer. Oh well. And the next is GFN-19, which I am not very familiar with so I will need to do some CPU/GPU performance testing on. I imagine at that large of a size and only single threaded, CPUs may be slow and painful compared to GPUs, but we'll see.

This one will be GPU dominant. We last ran these on a PG Challenge at the end of 2020. GFN-18, 19 and 20.

On the 19's my GTX1060 3G gets a task completed in ~2.5 hours. 3090 in under 30 min.

So, anyone that has a GPU with some spare cycles can help a lot. Run a single gaming rig overnight while sleeping. It'll be a little cooler in the northern hemisphere.

 

[EXT64]

Information on the upcoming PG challenge: https://www.primegrid.com/forum_thread.php?id=9915#155479

Unfortunately no mention of that new GFN CPU core, which is quite disappointing. I was hoping that it would be released in time for this.

 

[EXT64]

Or maybe it is ... now I see a post about multithreading. I'm confused, so I guess I will have to start testing to see what is going on.

 

[Gilthanis]

For those not wanting to click to get the details

25-30 May​

18:00:00

GFN-19

Geek Pride Day Challenge

5 days

Here is some information on multi-threading.
The allocated data size is 10 MB.
On processors with two memory channels, the throughput is generally optimal if the data size of all tasks is lower than the L3 cache size. For example, i7-11700K: L3 = 16 MB => 1 task; i7-12700K: L3 = 25 MB => 1 or 2 tasks; Ryzen 5950X: L3 = 2 x 32 MB => 6 tasks or less.
If the L3 cache size is below 10 MB, hyper-threading may increase the speed of genefer because two threads per core hide the memory latency.
Processors with four memory channels can be more efficient if the data size is greater than the L3 cache size.
On Ryzen processors with two CCX, it is recommanded that the affinity of genefer threads is set (see http://www.primegrid.com/forum_thread.php?id=9905).

 

[EXT64]

Wow, ok. Props to Primegrid. The MT CPU core is out and working. Not only that, it follows the core count set on the website (like it does for LLR). Nice work PG! It looks like I have a lot of testing to do before the competition next week!

 

[EXT64]

For those interested, here are some performance numbers for the upcoming GFN-19 competition:

​

​

​

​

​

​

​

Elapsed Time​

​

​

CPU Time​

​

​

​

​

​

Computer​

Clock​

Cores​

Threads/Core​

MT Threads​

Avg​

Min​

Max​

Avg​

Min​

Max​

# Running​

hr/WU​

WU/day​

PPD​

2P Epyc 7601 (L)*​

2.7​

64​

1​

4​

40336​

38936​

44873​

153302​

149273​

165829​

16​

11.20​

34.27​

267151​

2P Epyc 7601 (L)​

2.7​

64​

1​

8​

20664​

19827​

21631​

159824​

149273​

165829​

8​

5.74​

33.45​

260739​

Xeon Phi 7250 (L)*​

1.3​

68​

1​

4​

59450​

58299​

60077​

237501​

232905​

239996​

17​

16.51​

24.71​

192587​

Xeon Phi 7250 (L)​

1.3​

68​

2​

8​

60216​

59764​

60486​

480855​

477220​

482998​

17​

16.73​

24.39​

190137​

i9 10980XE (W)​

2.8​

18​

1​

6​

12105​

12089​

12115​

60981​

60924​

61040​

3​

3.36​

21.41​

166912​

i9 10980XE (W)​

2.8​

18​

1​

9​

9176​

9175​

9178​

65675​

65675​

65676​

2​

2.55​

18.83​

146793​

i9 10980XE (W)​

2.8​

18​

1​

4​

15589​

15566​

15604​

59168​

59102​

59223​

4​

4.33​

22.17​

172811​

i9 10980XE (W)*​

2.8​

18​

2​

6​

23231​

23205​

23242​

121283​

121101​

121401​

6​

6.45​

22.32​

173946​

2P E5-2695V3 (L)​

2.7​

28​

1​

7​

13964​

13707​

14254​

97660​

95854​

99681​

4​

3.88​

24.75​

192921​

2P E5-2698V4 (L)*​

1.9​

40​

1​

5​

27356​

26597​

28189​

136204​

132387​

140361​

8​

7.60​

25.27​

196955​

2P E5-2698V4 (L)​

1.9​

40​

1​

10​

14307​

13690​

14952​

142637​

136433​

149114​

4​

3.97​

24.16​

188296​

E5-2683V4 (L)*​

2​

16​

1​

4​

32064​

31983​

32124​

128162​

127852​

128396​

4​

8.91​

10.78​

84018​

E5-2683V4 (L)​

2​

16​

1​

8​

16200​

16199​

16201​

129489​

129484​

129494​

2​

4.50​

10.67​

83147​

Ryzen 3950x (L)*​

3.7​

16​

1​

4​

18863​

18702​

18978​

75324​

74681​

75805​

4​

5.24​

18.32​

142817​

Ryzen 5900x (L)*​

4.1​

12​

1​

3​

16225​

16106​

16434​

48635​

48273​

49279​

4​

4.51​

21.30​

166037​

Ryzen 5900X (L)​

4.1​

12​

1​

6​

9148​

9145​

9150​

54825​

54805​

54845​

2​

2.54​

18.89​

147243​

Ryzen 5800X3D (W)​

4.3​

8​

1​

4​

11776​

11685​

11868​

41956​

41948​

41946​

2​

3.27​

14.67​

114383​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

2x Radeon VII (W)​

~125W​

2​

1​

1​

3570​

3548​

3593​

621​

612​

630​

2​

0.99​

48.40​

377304​

Radeon VII (W)​

~125W​

1​

1​

1​

3566​

3566​

3566​

301​

301​

301​

1​

0.99​

24.23​

188864​

Tesla P100 (L)​

~150W​

1​

1​

1​

3797​

3797​

3797​

3793​

3793​

3793​

1​

1.05​

22.75​

177374​

RTX 3070 (W)​

~160W​

1​

1​

1​

2662​

2648​

2678​

66​

47​

92​

1​

0.74​

32.46​

253001​

GTX 1080Ti (W)​

~170W​

1​

1​

1​

4565​

4565​

4565​

36​

36​

36​

1​

1.27​

18.93​

147533​

Titan Xp (W)​

~175W​

1​

1​

1​

4273​

4273​

4273​

544​

544​

544​

1​

1.19​

20.22​

157615​

GTX 1070 (W)​

~120W​

1​

1​

1​

6900​

6900​

6900​

21​

21​

21​

1​

1.92​

12.52​

97607​

 

[EXT64]

On some CPUs, the impact of going to higher thread counts has only a minor negative impact (such as the Epyc and the 2683V3). So although it is likely a slight throughput decrease, I will likely run the higher thread counts to improve my chance of being 'first' (just in case we find any primes this year, I think that they are fairly rare though).

 

[Gilthanis]

I got a slow start but have my GPU's asking for work. Trying to stay cool though...Summer...grumble grumble

 

[wareyore]

Looks like we managed 7th on the Geek Pride Day Challenge. Which puts us in 7th overall at the end of three challenges.

 

[EXT64]

Great job team! Another good and consistent showing. Don't worry, for the next competition the weather will be even hotter!

 

[EXT64]

The next competition is coming soon (June 17) - yikes!

 

[wareyore]

Heard someone has lots of memory idling.

 

[EXT64]

​	​	​	​	​	Elapsed Time​	CPU Time​	​	​	​	​
Computer​	Clock​	Cores​	Threads/Core​	MT Threads​	Avg​	Avg​	# Running​	hr/WU​	WU/day​	PPD​
2P Epyc 7601 (L)​	2.7​	64​	1​	4​	33453​	127364​	16​	9.29​	41.3​	247,942​
Xeon Phi 7250 (L)​	1.3​	68​	1​	4​	47198​	167262​	17​	13.11​	31.1​	186,720​
i9 10980XE (W)​	2.8​	18​	1​	4​	11044​	42153​	4​	3.07​	31.3​	187,758​
2P E5-2695V3 (L)​	2.7​	28​	1​	7​	11234​	69746​	4​	3.12​	30.8​	184,583​
2P E5-2698V4 (L)​	1.9​	40​	1​	5​	21049​	96899​	8​	5.85​	32.8​	197,026​
E5-2683V4 (L)​	2​	16​	1​	4​	24927​	94504​	4​	6.92​	13.9​	83,187​
Ryzen 3950x (L)​	3.7​	16​	1​	4​	18547​	68800​	4​	5.15​	18.6​	111,802​
Ryzen 5900X (L)​	4.1​	12​	1​	6​	9373​	51696​	2​	2.60​	18.4​	110,616​
Ryzen 5800X3D (W)​	4.3​	8​	1​	4​	8765​	33489​	2​	2.43​	19.7​	118,289​

Here are my (semi optimal-ish) performance numbers for the upcoming competition (10.5 hours to start!). Somewhat interesting how several of my very different configuration computers get nearly identical performance. Not certain why the 3950x is doing abysmal here - due to the large WU size it may be memory bandwidth starved.

 

[EXT64]

Competition has started! It is supposed to be near 98F (36.5C) here today. YAY!

 

[EXT64]

Made a few updates to the table as the sample size increases (I've skipped the ones with small time increases, as that is expected as the numbers being tested get longer). Interestingly the 5800X3D now soundly beats both multi-chiplet systems, indicating that power or memory (or cache?) is the more limiting factor here. Anyone have a 5800X for comparison?

 

[pututu]

Nope but I only ran i7-6900K broadwell with 4 threads (AVX2) completed in about 12,087 seconds with default turbo clock of 3.7GHz.

Intel is still the king in LLR sub-projects.

 

[EXT64]

I realized in the table I copied the points/unit from a prior sub-project. Corrected now.

Points seem to vary a decent bit. Using 6K as a rough estimate for now.

 

[wareyore]

Any suggestions on settings for the upcoming challenge? Similar to the previous challenge?

 

[EXT64]

These are smaller, faster units. So in general I would trend towards smaller thread counts (~4t). I'll try to start testing shortly, but that is where I plan to start.

 

[wareyore]

These are smaller, faster units. So in general I would trend towards smaller thread counts (~4t). I'll try to start testing shortly, but that is where I plan to start.

That's what I tested at last weekend. 4, 5 and 8. 4 seemed the best.

 

[EXT64]

I just tested 4 vs 2 and 2 won by a good margin. So you may want to test 2 (or even 1, on faster systems).

 

[EXT64]

Looks like 1T is the best performing on my 10980xe (4T > 100k, 2T > 140K, 1T > 170K). I'll fire up the rest of my systems tomorrow to get a more complete picture.

 

[wareyore]

Starting today - 3PM Pacific - 6PM Eastern.

 

[EXT64]

Yep, 3 hours to go! Still wrapping up testing, but looking like generally very low multithreading or single threading is best here.

 

[EXT64]

And we're off!

 

[EXT64]

For those interested, here are some performance numbers:

​

​	​	​	​	​	​	Elapsed Time​	CPU Time​
Computer​	Clock​	Cores​	Clock*Cores​	Threads/Core​	MT Threads​	Avg​	Avg​	# Running​	hr/WU​	WU/day​	PPD​
2P E5-2695V3 (L)​	2.7​	28​	75.6​	1​	1​	4624​	4617​	28​	1.28​	523.18​	143875​
2P Epyc 7601 (L)​	2.7​	64​	172.8​	1​	2​	4046​	7959​	32​	1.12​	683.34​	187919​
2P Epyc 7601 (L)​	2.7​	64​	172.8​	2​	2​	8447​	16619​	64​	2.35​	654.62​	180021​
2P Epyc 7601 (L)*​	2.7​	64​	172.8​	1​	1​	7270​	7267​	64​	2.02​	760.61​	209166​
Ryzen 5900X (L)​	4.1​	12​	49.2​	1​	1​	2486​	2482​	12​	0.69​	417.06​	114690​
E5-2683V4 (L)​	2​	16​	32​	1​	1​	6199​	6196​	16​	1.72​	223.00​	61326​
Xeon Phi 7250 (L)​	1.3​	68​	88.4​	2​	2​	10339​	20044​	68​	2.87​	568.26​	156270​
i9 10980XE (W)​	2.8​	18​	50.4​	1​	4​	976​	3610​	4​	0.27​	354.10​	97377​
i9 10980XE (W)​	2.8​	18​	50.4​	1​	2​	1524​	2952​	9​	0.42​	510.24​	140315​
i9 10980XE (W)*​	2.8​	18​	50.4​	1​	1​	2522​	2517​	18​	0.70​	616.65​	169580​
2P E5-2698V4 (L)​	1.9​	40​	76​	1​	2​	5441​	10446​	20​	1.51​	317.59​	87337​
2P E5-2698V4 (L)​	1.9​	40​	76​	2​	2​	7376​	14272​	40​	2.05​	468.55​	128850​
2P E5-2698V4 (L)*​	1.9​	40​	76​	1​	1​	6257​	6251​	40​	1.74​	552.34​	151894​
Ryzen 3950x (L)​	3.7​	16​	59.2​	1​	1​	2862​	2856​	16​	0.80​	483.02​	132830​
Ryzen 5800X3D (W)​	4.3​	8​	34.4​	1​	1​	2770​	2764​	8​	0.77​	249.53​	68621​
Ryzen 3800X (W)​	4.4​	8​	34.4​	1​	1​	2775​	2768​	8​	0.77​	249.08​	68497​

A bit late for this year I know, but at least it is all set for next time!

 

[Gilthanis]

I'm getting a late start due to my daughter's birthday. But moving things over. Probably will just be a drop in the bucket.

 

[EXT64]

It's blazing hot here too (95F + humidity), so I'm glad this competition is short.

 

[Gilthanis]

It is also one of my smaller badges at PG...so kinda wanted to run this one...haha

 

[wareyore]

Got one:

 

[pututu]

Got one:
View attachment 494573

Gratz.

 

[MN Scout]

Have fun y'all. I'd consider running, but my apartment has very questionable wiring. Currently it is dedicated to the little AC.

PM me in October, please. Then I'll help.

 

[EXT64]

That never stopped Gilthanis - at least it better not have!

Also interesting note - my browser wants to auto-correct Gilthanis to Afghanistan. Fascinating.

 

[pututu]

Have fun y'all. I'd consider running, but my apartment has very questionable wiring. Currently it is dedicated to the little AC.

PM me in October, please. Then I'll help.

I'm having fun for finishing last in the last 5 challenges of this year. Thought I might pass Gilthanis but took my feet off the pedal early this morning
Next challenge with be a 3-day gpu. Sub-project is WW.

 

[EXT64]

I need your key to push you past him next time ... so you can stop using that as an excuse