P-1 factoring anyone? - Page 67

James Heinrich · 2011-10-23, 22:20

Quote:

Originally Posted by bcp19

I thought P-1 was only supposed to use 90% of available memory

Prime95 will only let you allocate up to 90% of your system memory, but it will use 100% of what you allocate to it where possible (it might not be possible to use every last MB since it's allocated in discrete chunks depending how many relative primes are being processed).

Mr. P-1 · 2011-10-23, 22:35

Quote:

Originally Posted by bcp19

I thought P-1 was only supposed to use 90% of available memory, which would kinda explain why at 112MB available it uses 92M, but doesn't explain the use of 112MB at 113MB available...

Earlier versions of the client would never use all the memory you allowed it, but I haven't seen this behaviour for a long time now. Even if it did, it should decide to do stage 2 with a certain amount of memory avaiable, then be unable to do it with that amount of memory.

Mr. P-1 · 2011-10-23, 22:43

Quote:

Originally Posted by James Heinrich

Prime95 will only let you allocate up to 90% of your system memory, but it will use 100% of what you allocate to it where possible (it might not be possible to use every last MB since it's allocated in discrete chunks depending how many relative primes are being processed).

You can put any amount of memory, even > 100%, into local.txt and have it "use" that when calculating P-1 bounds. Of course, you must ensure that it never actually goes to stage 2 in this condition. I do this on my secondary machine, which only has 256MB of physical memory. I use LowMemWhileRunning=prime95 to stop it from going to stage 2.

Every once in a while I shift the accumulated stage-2-ready assignments to my main box.

nucleon · 2011-10-24, 12:16

Any progress on P-1 on GPUs?

P-1 loves ram - most top end cards have >=1GB
P-1 loves fast ram - somewhat modest GTX560Ti - 52.61GB/s (beats definitely not modest SNB-E quad channel theoretical max of 51.2GB/s)

Anything on the drawing board?

-- Craig

Jwb52z · 2011-10-24, 15:33

Is there any way to get a more accurate credit for the P-1s? I am asking because I only do 1 P-1 at a time and have a handful always queued up to start. For some reason, even though each P-1 only takes a bit over a day to compute, the credit given to me is several days worth as if it's crediting the time the P-1 tests are sitting there unprocessed. I feel like it's an unforseen cheat that isn't on purpose by the user.

James Heinrich · 2011-10-24, 15:52

The credit given is based on the difficulty of the assignment alone, it doesn't matter how long you've had the assignment. GHz-days is what credit is given in, and that's how many days it would take to complete the assignment on a 1.0GHz single-core Intel Core² CPU. Your CPU is likely more powerful and can complete a (for example) 4GHz-days assignment in one day.

Jwb52z · 2011-10-24, 21:41

Quote:

Originally Posted by James Heinrich

The credit given is based on the difficulty of the assignment alone, it doesn't matter how long you've had the assignment. GHz-days is what credit is given in, and that's how many days it would take to complete the assignment on a 1.0GHz single-core Intel Core² CPU. Your CPU is likely more powerful and can complete a (for example) 4GHz-days assignment in one day.

Thank you for explaining it. Just FYI personally, I have an i7 Q720 1.6 GHz quad core.

Christenson · 2011-10-24, 21:53

Quote:

Originally Posted by James Heinrich

The credit given is based on the difficulty of the assignment alone, it doesn't matter how long you've had the assignment. GHz-days is what credit is given in, and that's how many days it would take to complete the assignment on a 1.0GHz single-core Intel Core² CPU. Your CPU is likely more powerful and can complete a (for example) 4GHz-days assignment in one day.

James:
Mr P-1 seems to have found a problem where he gets a different amount of credit depending on manual or automatic submission.
E

James Heinrich · 2011-10-24, 22:10

Quote:

Originally Posted by Christenson

Mr P-1 seems to have found a problem where he gets a different amount of credit depending on manual or automatic submission.

Yes, there are apparently some issues there, and I'll try and have a look as to why that's happening. But that's a bug in the manual results submission code somewhere, and it'll be several weeks before I can try and track that down.

Dubslow · 2011-10-24, 22:12

Didn't P95 say that was due to a lack of FFT size information? Or is this something else I missed?

kladner · 2011-10-25, 02:34

To whom it may concern (Mr. P-1,etc.):

I am currently running 3 1090T cores on P-1. 2 cores are feeding mfaktc on a GTX 460. (But that belongs in another thread.) 1 core is cleaning up the LL assignments which I moved from the other cores. In Win7-64 I have 8GB RAM, so I can handle giving fully adequate amounts to 4 cores once the LL's finish.

I have been taking whatever PrimeNet gives me in the P-1 department, without changing the depth.

I would gladly accept assignments from other sources, if it would be helpful.

2011-10-25, 02:34	#737
kladner "Kieren" Jul 2011 In My Own Galaxy! 2·3·1,693 Posts	CPU cycles available for P-1 To whom it may concern (Mr. P-1,etc.): I am currently running 3 1090T cores on P-1. 2 cores are feeding mfaktc on a GTX 460. (But that belongs in another thread.) 1 core is cleaning up the LL assignments which I moved from the other cores. In Win7-64 I have 8GB RAM, so I can handle giving fully adequate amounts to 4 cores once the LL's finish. I have been taking whatever PrimeNet gives me in the P-1 department, without changing the depth. I would gladly accept assignments from other sources, if it would be helpful.

2011-10-24, 12:16	#730
nucleon Mar 2003 Melbourne 5·103 Posts	Any progress on P-1 on GPUs? P-1 loves ram - most top end cards have >=1GB P-1 loves fast ram - somewhat modest GTX560Ti - 52.61GB/s (beats definitely not modest SNB-E quad channel theoretical max of 51.2GB/s) Anything on the drawing board? -- Craig

2011-10-24, 15:33	#731
Jwb52z Sep 2002 2⁴·3·17 Posts	Is there any way to get a more accurate credit for the P-1s? I am asking because I only do 1 P-1 at a time and have a handful always queued up to start. For some reason, even though each P-1 only takes a bit over a day to compute, the credit given to me is several days worth as if it's crediting the time the P-1 tests are sitting there unprocessed. I feel like it's an unforseen cheat that isn't on purpose by the user.

2011-10-24, 15:52	#732
James Heinrich "James Heinrich" May 2004 ex-Northern Ontario 3711₁₀ Posts	The credit given is based on the difficulty of the assignment alone, it doesn't matter how long you've had the assignment. GHz-days is what credit is given in, and that's how many days it would take to complete the assignment on a 1.0GHz single-core Intel Core² CPU. Your CPU is likely more powerful and can complete a (for example) 4GHz-days assignment in one day.

2011-10-24, 22:12	#736
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	Didn't P95 say that was due to a lack of FFT size information? Or is this something else I missed?