PrimeGrid's 2022 Challenge Series

How many threads per unit are you running? I would've thought the E cores would still do ok, if it is scheduling intelligently.
 
Ok, that should be decent then. Just trying to make sure you weren't overwhelming the memory.
 
Elapsed​
CPU Time​
Average​
Computer​
Clock​
Cores​
Clock*Cores​
Mem Chan​
Threads/Core​
MT Threads​
Avg​
Avg​
# Running​
hr/WU​
WU/day​
PPD​
Utilization​
%​
2P E5-2695V3 (L)​
2.7​
28​
75.6​
8​
1​
14​
9084​
108330​
2​
2.52​
19.02​
114135​
11.93​
85%​
2P E5-2695V3 (L)**​
2.7​
28​
75.6​
8​
1​
7​
15894​
105887​
4​
4.42​
21.74​
130464​
6.66​
95%​
2P Epyc 7601 (L)​
2.7​
64​
172.8​
16​
1​
16​
16960​
234382​
4​
4.71​
20.38​
122264​
13.82​
86%​
2P Epyc 7601 (L)**​
2.7​
64​
172.8​
16​
1​
8​
25659​
194229​
8​
7.13​
26.94​
161627​
7.57​
95%​
2P Epyc 7601 (L)​
2.7​
64​
172.8​
16​
1​
4​
46840​
183854​
16​
13.01​
29.51​
177079​
3.93​
98%​
Ryzen 5900X (L)**​
4.1​
12​
49.2​
2​
1​
6​
12225​
70810​
2​
3.40​
14.13​
84810​
5.79​
97%​
Ryzen 5900X (L)​
4.1​
12​
49.2​
2​
1​
3​
24555​
72849​
4​
6.82​
14.07​
84447​
2.97​
99%​
Ryzen 5900X (L)​
4.1​
12​
49.2​
2​
1​
1​
101588​
101513​
12​
28.22​
10.21​
61236​
1.00​
100%​
E5-2683V4 (L)​
2​
16​
32​
4​
1​
16​
12248​
168199​
1​
3.40​
7.05​
42325​
13.73​
86%​
E5-2683V4 (L)​
2​
16​
32​
4​
1​
8​
21308​
161094​
2​
5.92​
8.11​
48658​
7.56​
95%​
E5-2683V4 (L)**​
2​
16​
32​
4​
1​
4​
37336​
146775​
4​
10.37​
9.26​
55539​
3.93​
98%​
Xeon Phi 7250 (L)​
1.3​
68​
88.4​
8 (MCD)​
1​
17​
26577​
259245​
4​
7.38​
13.00​
78022​
9.75​
57%​
Xeon Phi 7250 (L)**​
1.3​
68​
88.4​
8 (MCD)​
1​
4​
58450​
223799​
17​
16.24​
25.13​
150775​
3.83​
96%​
i9 10980XE (W)​
2.8​
18​
50.4​
4​
1​
9​
7755​
62610​
2​
2.15​
22.28​
133694​
8.07​
90%​
i9 10980XE (W)**​
2.8​
18​
50.4​
4​
1​
6​
11175​
63078​
3​
3.10​
23.19​
139168​
5.64​
94%​
i9 10980XE (W)​
2.8​
18​
50.4​
4​
1​
4​
15782​
61342​
4​
4.38​
21.90​
131390​
3.89​
97%​
i9 10980XE (W)​
2.8​
18​
50.4​
4​
1​
3​
33806​
99490​
6​
9.39​
15.33​
92007​
2.94​
98%​
i9 10980XE (W)​
2.8​
18​
50.4​
4​
1​
1​
150337​
150093​
18​
41.76​
10.34​
62069​
1.00​
100%​
2P E5-2698V4 (L)​
1.9​
40​
76​
8​
1​
20​
11734​
180791​
2​
3.26​
14.73​
88359​
15.41​
77%​
2P E5-2698V4 (L)​
1.9​
40​
76​
8​
1​
10​
17383​
159685​
4​
4.83​
19.88​
119289​
9.19​
92%​
2P E5-2698V4 (L)**​
1.9​
40​
76​
8​
1​
5​
31472​
153034​
8​
8.74​
21.96​
131774​
4.86​
97%​
Ryzen 3950x (L)​
3.7​
16​
59.2​
2​
1​
16​
9151​
132073​
1​
2.54​
9.44​
56650​
14.43​
90%​
Ryzen 3950x (L)**​
3.7​
16​
59.2​
2​
1​
8​
15680​
120292​
2​
4.36​
11.02​
66122​
7.67​
96%​
Ryzen 3950x (L)​
3.7​
16​
59.2​
2​
1​
4​
33018​
129713​
4​
9.17​
10.47​
62802​
3.93​
98%​
Ryzen 3700X (W)**​
3.5​
8​
28​
2​
1​
8​
9404​
10919​
1​
2.61​
9.19​
55125​
1.16​
15%​
Ryzen 3700X (W)​
3.5​
8​
28​
2​
1​
4​
18995​
18849​
2​
5.28​
9.10​
54583​
0.99​
25%​
Ryzen 5800X3D (W)**​
4.3​
8​
34.4​
2​
1​
8​
8093​
13545​
1​
2.25​
10.68​
64055​
1.67​
21%​
Ryzen 5800X3D (W)​
4.3​
8​
34.4​
2​
1​
4​
15779​
17457​
2​
4.38​
10.95​
65708​
1.11​
28%​
Ryzen 5800X3D (W)​
4.3​
8​
34.4​
2​
1​
1​
62124​
55987​
8​
17.26​
11.13​
66757​
0.90​
90%​

Not done yet - but as we are running out of time I figured I should post. I'll update later tonight with the remaining data. I put "**" next to what I plan to run.

Edit: Updates Completed!
 
Last edited:
It was perfect for me - just headed in to work a few minutes late. I know PG says they try to move them around throughout the day to help folks in different timezones/schedules.
 
These are definitely weird - but I think that makes them fun! I think the Xeon Phi may be the most efficient CPU I have. How often can you say that - a failed product from several years ago does great. Cache and memory limitations also made these a blast to tune - each architecture and platform was different. Not just a question of how many 3xxx or 4xxx GPUs you can fit in your house before it burns down, lol.
 
Well done. Team finished 8th in this TRP-LLR subproject cpu challenge.

1665506124412.png


Overall team ranking. Our team performance is pretty consistent thus far by finishing either 7th or 8th place in the last 7 races. The last two races will be a 10 day challenge and it is going to be brutal.
1665505911368.png



Individual ranking.
1665505989265.png
 
Next challenge starts tomorrow.

Any advice on setup?

Multithreading off?

max cores per wu?

These look like they will be time consuming.
 
Yep - definitely use multithreading. I would start with what you used for the last challenge and test out higher core counts to see if they scale. I just started testing unfortunately, so I can only confirm that they are LONG. LOL
 
I was able to run two tasks just to see if one project was faster than the other. The results were extremely close and the difference is probably FAH in the background using a virtual thread or me browsing the internet.

CPU: AMD 5900X (6 threads/task)

Cullen -
Run Time: 42,291
CPU Time: 245,282
Credit: 29,125
Points Per Second Run Time: 0.689 p/s or 59,529 ppd

Woodall -
Run Time: 39,346
CPU Time: 228,072
Credit: 27,198
Points Per Second Run Time: 0.691 p/s or 59,702 ppd
 
I was able to run two tasks just to see if one project was faster than the other. The results were extremely close and the difference is probably FAH in the background using a virtual thread or me browsing the internet.

CPU: AMD 5900X (6 threads/task)

Cullen -
Run Time: 42,291
CPU Time: 245,282
Credit: 29,125
Points Per Second Run Time: 0.689 p/s or 59,529 ppd

Woodall -
Run Time: 39,346
CPU Time: 228,072
Credit: 27,198
Points Per Second Run Time: 0.691 p/s or 59,702 ppd
If you have 7950x, it will run much faster, at least according to this post in TAAT forum. 11 hrs for 5950x versus 7 hrs for 7950x. My guess is that it could be due to AVX-512 and larger L3 cache memory. I believe the ETA run time is based on 8 cores per task.
 
What is the power limit on the 5950x? If stock the 7950x can pull a ton more power, which could be a bigger factor.
 
Nice 'trudging' so far wareyore! Certainly putting up impressive numbers!

I'll try to post my table this evening. These units are slow and with two projects (randomly assigned) it takes a little while to build a dataset.
 
Ok, I just fired up i7-6900K which I think is probably better than 3950X in LLR. Keep the 3950X on TN-Grid. 8 cores per task and I set the cpu affinity to see if it helps a bit. Hopefully the 20MB L3 cache should be sufficient for this task and if not the 4 channel memory should at least help a bit. Est run time is about 10 hours.

Edit. Task finished in 10.05 hours for Cul-LLR subproject. I'll try Woo-LLR tomorrow.
1667407937801.png
 
Last edited:
Finally posting my performance numbers. First Woodall:
Computer​
Clock​
Cores​
Clock*Cores​
Mem Chan​
Threads/Core​
MT Threads​
Avg​
Avg​
# Running​
hr/WU​
WU/day​
PPD​
2P E5-2695V3 (L)**​
2.7​
28​
75.6​
8​
1​
14​
25491​
326641​
2​
7.08​
6.78​
184724​
2P E5-2695V3 (L)​
2.7​
28​
75.6​
8​
1​
7​
50935​
345415​
4​
14.15​
6.79​
184894​
2P Epyc 7601 (L)​
2.7​
64​
172.8​
16​
1​
16​
4​
0.00​
#DIV/0!​
#DIV/0!​
2P Epyc 7601 (L)**​
2.7​
64​
172.8​
16​
1​
8​
74116​
572537​
8​
20.59​
9.33​
254131​
Ryzen 5900X (L)**​
4.1​
12​
49.2​
2​
1​
12​
21044​
239982​
1​
5.85​
4.11​
111880​
Ryzen 5900X (L)​
4.1​
12​
49.2​
2​
1​
6​
43506​
256232​
2​
12.09​
3.97​
108233​
E5-2683V4 (L)​
2​
16​
32​
4​
1​
16​
1​
0.00​
#DIV/0!​
#DIV/0!​
E5-2683V4 (L)**​
2​
16​
32​
4​
1​
8​
60743​
468757​
2​
16.87​
2.84​
77520​
Xeon Phi 7250 (L)​
1.3​
68​
88.4​
8 (MCD)​
1​
17​
4​
0.00​
#DIV/0!​
#DIV/0!​
Xeon Phi 7250 (L)**​
1.3​
68​
88.4​
8 (MCD)​
1​
8​
104245​
747953​
8​
28.96​
6.63​
180682​
i9 10980XE (W)**​
2.8​
18​
50.4​
4​
1​
9​
26102​
213811​
2​
7.25​
6.62​
180400​
i9 10980XE (W)​
2.8​
18​
50.4​
4​
1​
6​
52093​
297442​
3​
14.47​
4.98​
135588​
2P E5-2698V4 (L)​
1.9​
40​
76​
8​
1​
20​
28800​
494923​
2​
8.00​
6.00​
163500​
2P E5-2698V4 (L)**​
1.9​
40​
76​
8​
1​
10​
50004​
475368​
4​
13.89​
6.91​
188337​
Ryzen 3950x (L)**​
3.7​
16​
59.2​
2​
1​
16​
26925​
407896​
1​
7.48​
3.21​
87443​
Ryzen 3950x (L)​
3.7​
16​
59.2​
2​
1​
8​
52042​
405342​
2​
14.46​
3.32​
90481​
Ryzen 3700X (W)**​
3.5​
8​
28​
2​
1​
8​
31009​
19192​
1​
8.61​
2.79​
75926​
Ryzen 3800X (W)**​
4.4​
8​
35.2​
2​
1​
8​
28124​
219650​
1​
7.81​
3.07​
83715​
Ryzen 5800X3D (W)**​
4.3​
8​
34.4​
2​
1​
8​
22558​
24977​
1​
6.27​
3.83​
104371​
 
And now Cullen:
Computer​
Clock​
Cores​
Clock*Cores​
Mem Chan​
Threads/Core​
MT Threads​
Avg​
Avg​
# Running​
hr/WU​
WU/day​
PPD​
2P E5-2695V3 (L)**​
2.7​
28​
75.6​
8​
1​
14​
27368​
350324​
2​
7.60​
6.31​
184367​
2P E5-2695V3 (L)​
2.7​
28​
75.6​
8​
1​
7​
57461​
391490​
4​
15.96​
6.01​
175624​
2P Epyc 7601 (L)​
2.7​
64​
172.8​
16​
1​
16​
4​
0.00​
#DIV/0!​
#DIV/0!​
2P Epyc 7601 (L)**​
2.7​
64​
172.8​
16​
1​
8​
83269​
645271​
8​
23.13​
8.30​
242384​
Ryzen 5900X (L)**​
4.1​
12​
49.2​
2​
1​
12​
21905​
250401​
1​
6.08​
3.94​
115174​
Ryzen 5900X (L)​
4.1​
12​
49.2​
2​
1​
6​
44362​
260916​
2​
12.32​
3.90​
113741​
E5-2683V4 (L)​
2​
16​
32​
4​
1​
16​
1​
0.00​
#DIV/0!​
#DIV/0!​
E5-2683V4 (L)**​
2​
16​
32​
4​
1​
8​
64657​
501962​
2​
17.96​
2.67​
78039​
Xeon Phi 7250 (L)​
1.3​
68​
88.4​
8 (MCD)​
1​
17​
4​
0.00​
#DIV/0!​
#DIV/0!​
Xeon Phi 7250 (L)**​
1.3​
68​
88.4​
8 (MCD)​
1​
8​
127045​
917331​
8​
35.29​
5.44​
158865​
i9 10980XE (W)**​
2.8​
18​
50.4​
4​
1​
9​
26542​
227103​
2​
7.37​
6.51​
190105​
i9 10980XE (W)​
2.8​
18​
50.4​
4​
1​
6​
55129​
322193​
3​
15.31​
4.70​
137290​
2P E5-2698V4 (L)​
1.9​
40​
76​
8​
1​
20​
30593​
525002​
2​
8.50​
5.65​
164932​
2P E5-2698V4 (L)**​
1.9​
40​
76​
8​
1​
10​
53119​
505299​
4​
14.76​
6.51​
189979​
Ryzen 3950x (L)**​
3.7​
16​
59.2​
2​
1​
16​
28609​
432448​
1​
7.95​
3.02​
88185​
Ryzen 3950x (L)​
3.7​
16​
59.2​
2​
1​
8​
53816​
419440​
2​
14.95​
3.21​
93759​
Ryzen 3700X (W)**​
3.5​
8​
28​
2​
1​
8​
32986​
40897​
1​
9.16​
2.62​
76483​
Ryzen 5800X3D (W)**​
4.3​
8​
34.4​
2​
1​
8​
24789​
32925​
1​
6.89​
3.49​
101774​
 
At least your results confirm that the i7-6900K would run faster than the 3950X for both Cullen and Woodall. The avx2 implementation in 3950x is lagging behind 6-year old cpu.

I only ran two tasks on 6900K (3.7GHz, 8 cores per task, 50% SMT, 4 channel memory at 2400MHz)
Cullen (2400K FFT length): 36,193s vs 53,816s in 3950x
Woodall (2304K FFT length: 31,765s vs 52,042s in 3950x
 
  • Like
Reactions: EXT64
like this
My 5900 is averaging just under 11.5 hours for Woodall running 6 threads per task. My 3600 is averaging around 12 hours for Woodall at 6 threads per task. I am not getting too many Cullen tasks.
 
  • Like
Reactions: EXT64
like this
I hate traveling during these challenges. Two systems dropped offline and no way to bring them back until I get home.
 
Coming up on the home stretch! I lost a 2P system yesterday. I won't get home until after the end :( .
 
I hate traveling during these challenges. Two systems dropped offline and no way to bring them back until I get home.
Dropped offline like restarted or also no internet? If just restarted can setup something like chrome remote desktop. Can even use on phone. Can also consider using smart plugs if you need to remotely power cycle a rig that is offline. I have used a combo of both for a rig i keep in storage room
 
Dropped offline like restarted or also no internet? If just restarted can setup something like chrome remote desktop. Can even use on phone. Can also consider using smart plugs if you need to remotely power cycle a rig that is offline. I have used a combo of both for a rig i keep in storage room
I remote in already. They are showing offline. Not sure if they have power or there is a network issue or they are sitting at a bios screen or what. No eyes/hands there to check. They are on different circuits so the issue is specific to the box.
 
Sorry for the late post. We finished 7th as a team. A very consistent performing team!

As usual, yours truly is still last (ahem, yes I'm still a consistent performer too ;) )

Last PG challenge is mostly GPU though you can still run cpu for GFN-21. Do You Feel Lucky?

1668274039606.png


1668273875543.png
 
Last challenge of 2022 starts tomorrow evening (US time).

Definitely a good time to replace the space heaters with GPU's.
 
Yep - and any of you rocking Intel Arc or Apple M1/M2, note PG just added official support for both!
 
Nice bonus on the DYFL tasks. My 3080 took around 18 hours but it granted almost 1M points. I only had it at 64% power. I pushed it up to 80% to see what kind of gain I can get.

3070 is around 29 1/2 hours at 62%. I will bump that one to 80% as well to see what the time shortens by.

Update:
3080 @80% power dropped roughly 1 1/2 hours off the time down to 16 1/2 hrs from 18 hrs.
3070 @80% power dropped almost 3 hours off the time down to around 26 1/2 hrs from 29 1/2 hrs.

I don't run them at 100% due to where they are and the extra heat annoys me working from home in the same room if it is not real cold outside. Next week the temperature drops here in the DFW area.
 
Last edited:
Day two finished. Still early for this long challenge but it looks like it'll be a good cap for the year. I decided last year to participate in all the PG challenges this year and see if I could make top 20. If things don't break down before the end of this series I may make top 10.
 
Back
Top