1)
Message boards :
Number crunching :
Prime Meridian Day Challenge
(Message 157642)
Posted 211 days ago by  Mr. Cool
I run a few short benchmarks which directly measure LLR2 performance with a given workunit. Those benchmarks are fully reproducible and enable me to find the optimum or near optimum config (for throughput optimum, or power efficiency optimum, or whatever) in a short timeframe and with high confidence.
How do you run the benchmark? Could you please explain?
|
2)
Message boards :
Number crunching :
Sub-project "life" expectancy
(Message 116084)
Posted 1903 days ago by Mr. Cool
Michael, with the discovery of 3 GC primes, is there much life left in GCW Sieve?
Thanks in advance.
|
3)
Message boards :
Number crunching :
PrimeGrid is Moving!!!
(Message 109873)
Posted 2098 days ago by Mr. Cool
If there's a way I can donate directly, I'd be happy to throw some coins your way.
I concur.
|
4)
Message boards :
Generalized Fermat Prime Search :
EFF offering $150,000 prize for 1st 100M digit prime
(Message 109455)
Posted 2111 days ago by Mr. Cool
Hi everyone.
Will quantum computers ever help us find larger and larger prime numbers? Or are they only useful at factoring numbers?
Thanks
|
5)
Message boards :
Proth Prime Search :
ATI sieve app turned off; Help needed.
(Message 78948)
Posted 3206 days ago by Mr. Cool
Try leaving a full core for the sieve? Lots of OpenCL apps benefit from that.
You're also on an i7. Try turning off HT and leaving a full core for the sieve?
It's definitely going to work, but it's not a good idea for me. I'm running PSP Sieve on the CPU and it's better with HT on. If I can't get the GPU PPS Sieve to 100% load with all CPU cores occupied, then I'll leave it be.
Thanks anyway
|
6)
Message boards :
Proth Prime Search :
ATI sieve app turned off; Help needed.
(Message 78946)
Posted 3206 days ago by Mr. Cool
I'm ruining the sieve on HD5870, but get a load lower 80%. Is there a way to increase the GPU load?
Oh! I might understand. Basically, even when vectorized, the sieve uses about four integer units at a time. Older VLIW-architecture cards like yours have five integer units available. 4/5 = 80%. :) And no, there's no good way to get to 100% if that's the issue. GCN cards really are better at this task.
If the issue is CPU limitation, which I don't think it would be on a 3770k and a 5870, I'm working on that. I finally found the intrinsic I need. Now I need to add it and see if I can compile for 64-bit with VS 2008 Express.
Thanks for your previous answer. The problem is actually in CPU load. I've assigned the highest priority to the sieve task and the load went up from 60% to 84%. If I stop all CPU tasks all together then the load goes up to 100%.
I'm using the latest Catalyst driver and I wonder if I should try another driver version.
|
7)
Message boards :
Proth Prime Search :
ATI sieve app turned off; Help needed.
(Message 78929)
Posted 3207 days ago by Mr. Cool
I'm ruining the sieve on HD5870, but get a load lower 80%. Is there a way to increase the GPU load?
|
8)
Message boards :
Number crunching :
Credit Milestones
(Message 78893)
Posted 3209 days ago by Mr. Cool
Took me 4.5 years but I finally hit 500M....
Congratulations! It's a big achievement!
|
9)
Message boards :
Project Staging Area :
Call for wwwwcl beta testers (OpenCL)
(Message 78862)
Posted 3211 days ago by Mr. Cool
Is there anyone succesfully running WallSunSun on HD5870 in Win7 64bit?
I'm having troubles running my set up. I'm using latest Catalyst 14.4.
|
10)
Message boards :
Sieving :
DC on sieving using CPU is Only waste of TIME.
(Message 78758)
Posted 3215 days ago by Mr. Cool
Since we have a way to calculate density of factors at a specific P (or we can have a look at recent figures), then we can expect a certain number of factors per a number of tasks. Say we expect one factor to be returned per 100 tasks, then if a host doesn't return a factor after 300 WUs, we DC those tasks to see if there's a problem there.
Also at some depth of a sieve when factors become so scarce so there's no point to DC, we can stop DCing and continue sieving, only double-checking suspicious hosts occasionally.
I was also thinking about working out a point of optimal sieve depth. Since we know all factors returned for a range, is it possible to estimate time saved by sieving? Say we choose an i7-4770 as a default CPU and estimate the time it would take to LLR all factors found for a range and compare the time with the period it would take to sieve the range.
For instance, we sieve TRP 'n' between 10m and 50m, but not all factors are found at a higher 'n'. We could get most factors in a lower 'n' value and that could make a difference.
I think this way we could better estimate the point where to stop sieving.
I think that GPU sieving depth is a little bit more difficult to understand. First, it's difficult to compare GPU to CPU time spent. We could compare watts used?
Second, if at some point we decide that we should stop GPU sieve and don't want to sieve a higher n-range, then we can just loose all our GPU power to other projects. Not all of our members will move their video cards to GFN. I'm saying that GPU power is "free" for PrimeGrid and as long as we have a good paying [cobblestones] sieve like PPS, we can freely use GPUs. But once we stop sieving, we will loose them. So we shouldn't stop sieving at any point as it's free.
These are my 2 cents. Thanks.
|
