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

Sponsored by:

Hosted at and sponsored by Rackspace.

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
1) Message boards : Proth Prime Search : Faulty batch PPS_Sr2 W.U.s (Message 50557)
Posted 2121 days ago by Profile Beyond
All mine are still failing:

PrimeGrid 02-26-12 13:58 Aborting task pps_sr2sieve_44697242_0: exceeded elapsed time limit 478.06 (307529.96G/643.29G)
PrimeGrid 02-26-12 13:58 Computation for task pps_sr2sieve_44697242_0 finished
PrimeGrid 02-26-12 13:58 Output file pps_sr2sieve_44697242_0_0 for task pps_sr2sieve_44697242_0 absent

Switched to another project for now...
2) Message boards : Number crunching : PPS (Sieve) Cobblestone Recalibration (Message 41890)
Posted 2252 days ago by Profile Beyond
..."with no apparent justification let's award credit higher than anyone else to attract users" philosophy.

If this ever happens, then it will be an excellent tribute to the software developers of tpsieve...Geoff Reynolds and Ken Brazier. The code advancement was phenomenal (as quantified by the increased credit generation over the previous software). Their outstanding work revolutionized sieving, and they should be duly recognized. :)

The amount of sieving that was completed after their improvements was unimaginable over a year ago. However, since the introduction of tpsieve and its port to the GPU, decades of work have been completed in less than a year (thanks to all the volunteer crunchers). And the benefits extend beyond PrimeGrid as other prime search projects are now using these incredibly sieved files.

The prime community is forever indebted to Geoff and Ken...and to all the crunchers who made use of their software. :)

Well said!

Well said indeed. Over 15 years ago I got my BComp so I am not a total novice in programming. I thought I would have a go at modifying the TRP sieve for ATI, after an hour reading the AMD introduction to that form of programming I gave up and I still do not claim to know a thing about it.

I take my hat off to our programmers, they are top notch and we are lucky to have them.

Hmm, below is my post from which the snipped above was culled:

Hi John,
Basing credits on real world benchmarks is laudable. There has increasingly been a perhaps unfortunate effect. It seems many of the larger NVidia based producers are moving to a project that's banned by some teams and that apparently bases it's credits on the "with no apparent justification let's award credit higher than anyone else to attract users" philosophy. Not much to do about it but at least be aware of the issue...

Note that I was referring to another project that is basing it's credits on seemingly little other than a desire to pull users from elsewhere. I think PrimeGrid is taking the high road (like Collatz and MW) by basing credits on the amount of data being crunched. I like credits as much as anyone but think they should have some real world basis.
3) Message boards : Number crunching : More ATI/OpenCL tasks? (Message 41261)
Posted 2268 days ago by Profile Beyond
As any 4770, 4830,4850,4870, 4870x2, 5830 5850 and 5870 have them, you see the oddity here even low models had DP.
when the 5xxx series entered the arena they took DP of the low/mid range cards

As you can see by your listing above, AMD kept DP on the whole 58xx line just as on the 48xx line. The 4770 was a special case as it was their testbed for the 40nm process and thus contained their full spectrum of technology. They decided not to include DP in the later 5770 in order to make the chip smaller and more affordable.
4) Message boards : Number crunching : PPS (Sieve) Cobblestone Recalibration (Message 41094)
Posted 2271 days ago by Profile Beyond
Credit decrease has resumed. Please see the first post for an updated time table.

Hi John,

Basing credits on real world benchmarks is laudable. There has increasingly been a perhaps unfortunate effect. It seems many of the larger NVidia based producers are moving to a project that's banned by some teams and that apparently bases it's credits on the "with no apparent justification let's award credit higher than anyone else to attract users" philosophy. Not much to do about it but at least be aware of the issue...
5) Message boards : Number crunching : Computation errors on ATI (Message 40272)
Posted 2287 days ago by Profile Beyond

I think you need an app_info for this but then you can too calculate 3 or more units in parallel.
You need only enough GPU-RAM and "we need more power" inside the GPU...

Not sure if there is any advantage to this. It would clutter the tasks screen and not achieve any more work anyway.

It's an advantage only at MilkyWay AFAIK (and not a huge deal at that).
6) Message boards : Number crunching : Computation errors on ATI (Message 40178)
Posted 2291 days ago by Profile Beyond
I'm not a programmer either, but I thought that tasks optimised for GPUs (or, rather, the combination of the two) were infinitely faster than a task running on a CPU, even if it's got multiple cores. GPUs specialise in FPU calculations as far as I know, and can do many more in parallel.

I see. I shall check.....
CPU does 2.5 billion floating point operations per core per second. GPU does oh my god..... 1000 billion. 1000>>>>>>2.5 and also greater than 8x2.5=20
50 times more on a GPU than the whole CPU? I'm sure the CPU should do better than this. Aren't flops (good word that) used a lot in general tasks?

Consider that GPUs have up to 3200 cores versus up to 12 for a CPU. If a task can be highly parallelized for the GPU it can be run very quickly.
7) Message boards : Number crunching : Computation errors on ATI (Message 40150)
Posted 2292 days ago by Profile Beyond
You could think of a GPU as a Cray-1, except that it's much larger, much faster, and physically shrunk down to the size of a cigarette carton.
The only thing lacking is that a GPU isn't nearly as comfortable when you sit on it.

Especially when running :)
8) Message boards : Number crunching : Computation errors on ATI (Message 40074)
Posted 2293 days ago by Profile Beyond
You need to install the latedst driver and the SDK.

I'm using v10.12 w/ocl on all 16 of my ATI cards (HD4770 - HD5870) on various projects and they all run perfectly. I've tried all the newer versions but keep going back to v10.12.
9) Message boards : Number crunching : PPS(S) with Nvida 280.26 (Message 40072)
Posted 2293 days ago by Profile Beyond
My OC'd GTX 460 got slower upon going from 26099 to 27533 (representing about half a year of not bothering to update), so I'm somewhat sure that updating again to 28026 will bring it back to at least normal speed (~19m/WU)...

Perhaps a good idea to try it before posting the result ;)
10) Message boards : Number crunching : PPS(S) with Nvida 280.26 (Message 39475)
Posted 2305 days ago by Profile Beyond
Update: Using 258.69 (beta), PPS(S) WUs completes in the range of 65-66mins.
I'm at a lost.. terencewee* Sicituradastra.

Why 258.69 beta instead of 258.96?


Next 10 posts
[Return to PrimeGrid main page]
Copyright © 2005 - 2017 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 1.43, 1.16, 1.22
Generated 17 Dec 2017 | 23:15:58 UTC