PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
11) Message boards : Number crunching : Better multi-threading (Message 132377)
Posted 144 days ago by tng*Project donor
I'll get them done by the 2d.
12) Message boards : Number crunching : Better multi-threading (Message 132282)
Posted 148 days ago by tng*Project donor
I wish that error didn't exist. It's really quite useless.


+1 -- it had me hyperventilating a bit until I figured it out.
13) Message boards : Number crunching : Sub-project "life" expectancy (Message 130854)
Posted 199 days ago by tng*Project donor
Clever way to deal with this.



There's been a slight change in plans.

PPS-MEGA is currently searching the PPS k range (k<1200). We're going to move it to the PPSE k range (1200<k<10000) and drop n back down to about 3.322M so that the numbers are once again just above 1 million digits. With 7 times as many k's to search, n will grow a lot slower. This will keep the size of the candidates lower, making it easier to find mega primes.

This changes, somewhat, what will happen with PPS when it reaches the beginning of PPS-MEGA. Instead of merging with PPS-MEGA, it will jump over PPS-MEGA and continue searching the low k range (k<1200) upwards, to at least n=9M or n=12M.

PPSE will eventually catch up to the beginning of PPS-MEGA (a decade or three from now), and will then merge with PPS-MEGA since both will be searching the same k range.

If you're completely confused by this, here's the tl;dr:

* PPS-MEGA numbers will soon become smaller than they are currently. This will make it easier to find mega primes. (They will drop from about 1.1 million digits to 1.0 million digits.)

* In about 2 years, when PPS catches up to PPS-MEGA, it will skip over PPS-MEGA. The size of the numbers will jump from 1 million digits to about 1.1 million digits.

* Many years from now, PPSE will catch up to the beginning of PPS-MEGA, and the two will become one.

14) Message boards : Seventeen or Bust : 67607*2^n+1 (Message 129252)
Posted 254 days ago by tng*Project donor
I notice that the range statistics for Seventeen or Bust show that there is only 1 workunit in progress for 67607*2^n+1. Possibly meaningless, but interesting.
15) Message boards : Number crunching : Badges III (Message 129109)
Posted 261 days ago by tng*Project donor
16) Message boards : Number crunching : Badges III (Message 128617)
Posted 284 days ago by tng*Project donor
17) Message boards : Sieving : Invalid Command queue error? (Message 128462)
Posted 293 days ago by tng*Project donor
It appears that this may have been due to a physical misconfiguration. When I installed the second card, the system had two PCI power cables available, which were bundled together and had 6+2 connectors. The card required one 8-pin connector. I plugged in the 6-pin connector from one cable and the 2-pin connector from the other.

Since I fixed this, no errors.

There's still an issue with this system. The reason I am running 3 sieving reservations on each card is that I can't get the expected performance without doing that. Running one or two sieving tasks per card, I get just under 170P/day for each task. Running three per card, performance is slightly less per task, which puts it in line with expectations.

Still trying fo figure that out.
18) Message boards : Number crunching : Badges III (Message 128330)
Posted 298 days ago by tng*Project donor
19) Message boards : Sieving : Invalid Command queue error? (Message 128276)
Posted 300 days ago by tng*Project donor
I think you should reduce your clock and memory speeds. Those errors look like what comes up and happens when the card is pushed to it's limits. Reduce it a bit and see if the errors go away.

I learned, from Azmodes, if you get errors like that reduce clocks and memory till they go away. You can adversely increase clocks till those errors occur then reduce till they subside. We learned for sieving high clock speed is best obviously but low memory is also fine and doesn't affect sieving performance.


Will try that if needed, but I havn't overclocked these cards, and one is definitely stock clocks (the other came with the system. Also, the errors are occurring on both cards. Will at least reinstall drivers first.

It really seems to act like a memory leak -- I stopped and restarted all sieving tasks last night, and no problems so far (maybe 20-22 hours).
20) Message boards : Sieving : Invalid Command queue error? (Message 128264)
Posted 301 days ago by tng*Project donor
Update:

Just checked and had the same error on three instances and a different one on one instance:

Error clGetEventInfo on device 1 (131.7P/day) Found 6465 ETA 42h22m
Error code -5, message: Out of resources
GPU Error -5 @ 2365: Out of resources

Ran out of GPU RAM, maybe?


Next 10 posts
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2020 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 3.77, 2.77, 2.21
Generated 18 Jan 2020 | 8:55:35 UTC