Info PrimeGrid Challenges 2025

Page 9 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

mmonnin03

Senior member
Nov 7, 2006
317
266
136
Errors from the client would not make the score go down as the server would never see them as possible valid work. The 35k would need to be something completed as OK (Potential points during the competition) by the client but checked as No Good (Score reduced) by the server or comparing between main/check tasks. An Invalid so those CPU tasks.
 

Skillz

Golden Member
Feb 14, 2014
1,131
1,151
136
Errors from the client would not make the score go down as the server would never see them as possible valid work. The 35k would need to be something completed as OK (Potential points during the competition) by the client but checked as No Good (Score reduced) by the server or comparing between main/check tasks. An Invalid so those CPU tasks.

1750240734883.png

That's what I said.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,094
16,014
136
That pic doesn't show the whole picture.

That box has:
14 CPU errors
19 GPU errors
20 CPU invalids

This leads me to believe it's most likely RAM related or possibly the SSD/disk is going bad. The latter part I'm guessing because some of the errors are related to this error
"Error: gproof_102.ckpt: bad file (crc32)"

Could this be the result of a drive failing and not writing the data correctly? It's a stretch.

Which is why with both CPU and GPU tasks being effected I'm leaning towards something wrong with the system memory causing errors.

However, I do agree with Tony that even an under volt or power limit can cause errors just as much as an overclock.

Bottom line is. Something is up with that system and it's effecting more than one thing.

edit
Upon further investigation it seems the only valid CPU tasks were returned late on the 16th and on the 17th. No CPU tasks before that were valid results. This leads me back to a possible heat problem. If the temps got lower on those days, but was higher on the days before. Then that could be the problem.

Edit2
Also seems the CPU tasks that got marked invalid took more time than those CPU tasks that did validate. Which could have been the result of heavy CPU throttling. Again pointing towards heat.

I'm on my phone trying to look this up. Someone, like Mark, should put all that data in a spreadsheet to sort through the data better by separating the errors, invalids and valids for CPU then do it again for the GPU in that box.
Well, the first thing is to check BIOS for that and the 9950x (also a problem, 7 of 40) The 9950x temp was at 95c (stock) and I set it to 80c. The SOC in bios was red at 1.35 and I set it down to 1.2. Now on the 7950x3d, it had already rebooted, so I know there is issue there. I wanted to check SOC and max temp in bios, but its the damn ASUS ROG bios and I can't find those 2 to change. There are screens that have SOC in the name, but is it memory or CPU ? and what name might they go by. Somebody that knows the ASUS ROG bios could be of a lot of help here. !
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,094
16,014
136
OK, thought I found it and it was 1.39, changed it to 1.2 and it would not boot, so I set it to 1.3 on the 7950x3d
 

StefanR5R

Elite Member
Dec 10, 2016
6,551
10,291
136
Speaking generally without having looked at the results tables, there can be calculation errors which either the application or the client detect themselves, and calculation errors which only the validator ( = server with help of another host) detects. The latter type of errors is the one which cause point deductions in PrimeGrid challenges. Yet *if* there is interest in improving the host to compute correctly, it is obviously worthwhile to look at *all* of the types of errors to get a better picture of what went on. :-)
 
  • Like
Reactions: Ken g6 and Skillz

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,094
16,014
136
I know this is about the August PG race, but before I forget, I will be out of down for it.
 

Ken g6

Programming Moderator, Elite Member
Moderator
Dec 11, 1999
16,621
4,539
75
This one's fast and at a good time for me so I'm aiming for half-day stats.

Day 0.5 stats:

Rank___Credits____Username
5______16806751___markfw
7______14457768___crashtech
11_____8102172___10esseeTony
19_____5478265____cellarnoise2
20_____5069922____Orange Kid
29_____3889366____mmonnin
95_____747955_____Ken_g6
110____626665_____johnnevermind
111____618579_____w a h

Rank__Credits____Team
1_____57564234___Antarctic Crunchers
2_____55797443___TeAm AnandTech
3_____36273796___SETI.Germany
4_____21084245___Czech National Team
5_____19475131___Romania
 

Ken g6

Programming Moderator, Elite Member
Moderator
Dec 11, 1999
16,621
4,539
75
Day 1 stats:

Rank___Credits____Username
5______31830539___markfw
6______29364309___crashtech
11_____16964428___10esseeTony
12_____16923998___cellarnoise2
21_____9804275____Orange Kid
25_____8498386____mmonnin
81_____2158962____[TA]Skillz
101____1536340____w a h
105____1463566____Ken_g6
112____1273545____biodoc
117____1196728____johnnevermind
133____1014793____Icecold
308____28301______kiska
376____4043_______zzuupp

Rank__Credits____Team
1_____122058170___TeAm AnandTech
2_____107450811___Antarctic Crunchers
3_____73643245___SETI.Germany
4_____49150751___Czech National Team
 

Ken g6

Programming Moderator, Elite Member
Moderator
Dec 11, 1999
16,621
4,539
75
Day 1.5 stats:

Rank___Credits____Username
5______45976996___markfw
7______44869214___crashtech
11_____25143417___cellarnoise2
12_____24803805___10esseeTony
20_____14146457___Orange Kid
25_____12783966___mmonnin
28_____11979409___Icecold
62_____4669665____[TA]Skillz
72_____4014699____biodoc
100____2361112____w a h
111____2114489____Ken_g6
124____1774877____johnnevermind
142____1406964____ChelseaOilman
250____226408_____kiska
367____16172______zzuupp

Rank__Credits____Team
1_____196271478___TeAm AnandTech
2_____149041152___Antarctic Crunchers
3_____106739243___SETI.Germany
4_____73881782___Czech National Team
 
  • Like
Reactions: crashtech