New Prime95

[prime95]

Prime95 version 26.2 is available for beta testing. See http://www.mersenneforum.org/showthread.php?t=13887

It wouldn't hurt to have an SoB user or two give it a try.

[imwithid]

I was wondering if it was okay to use with SoB. I'll start phasing it in as soon as prior tests have completed, in case there are any errors in starting part way.

[Frodo42]

Trying it out now on my i7 64bit linux, so far seems to be working without problems

[Frodo42]

hmmm seems that it is only starting 4 workers where is normally 25.11 started 8 workers on an i7 ... is this expected?
It says something about running helper threads on a cpu ... haven't been checking if the 4 workers are able to work double as fast or more than the 8 workers yet.

[prime95]

Your best configuration is likely to be 4 worker threads with no multi-threading. Use Test/Worker-windows to change turn off the multi-threading.

Time it to see which is faster.

[Frodo42]

Yup seems you are right 4 workers on 26.2 is faster, than with 8 workers on 25.11 per iteration time was in the range 0.084-0.091 and with 26.2 and 4 workers the range was 0.034-0.040

Only problem is now to get the 4 jobs that was underway on worker 5-8 moved to worker 1-4 so that once they are done with their current tests they will continue with the ones that was underway on 5-8.
It should work if I in worktodo.txt just copy the lines from worker 5-8 over below the current lines on worker 1-4 right?

[prime95]

Yes. A quick edit of worktodo.txt will fix the problem

[Joe O]

Prime95 version 26.2 is available for beta testing. See http://www.mersenneforum.org/showthread.php?t=13887

It wouldn't hurt to have an SoB user or two give it a try.

Prime95 version 26.3 is now available for beta testing. See http://www.mersenneforum.org/showthread.php?t=14055

[Joe O]

Has anyone besides Frodo42 tried this?
Does anyone have timings before and after available?
It should be possible to simply:
1) stop your client
2) Back up your folder, just in case
3) replace your old executable with the new executable (executable only, not your prime.txt)
4) start your client
5) make sure you resumed where you left off and are still connected to the SB server and not to mersenne.org instead.

[umccullough]

Guess I better get on the bandwagon and update the Haiku port(s) :)

[imwithid]

Apparently, Prime95 26.4 (http://mersenneforum.org/showthread.php?t=14198) has been out for about a few days now. So far, this one seems to have addressed the minor bugs from the last release without any problems.

[imwithid]

This version has been out for over two weeks now and addresses many bug fixes from the previous releases (26.3 and its three or four builds). How difficult (or advisable) would it be to use this version?

[Joe O]

This version also has had a few builds in the two weeks it has been out. If you want to try it, and are not using a Mac, then all you have to do is drop in the correct executable into your current directory. Just replace the executable, not the prime.txt file.

[imwithid]

Thanks, I thought that there may be, perhaps, some added settings that would conflict (with prime.txt?) if I simply dropped the prime95.exe or mprime.

If the speed and stability improvements are as they have been purported to be, then this is a worthwhile update to implement.

[umccullough]

<sigh>

Guess I'll have to update the Haiku port(s) again...

[imwithid]

Wow, I just upgraded one of my machines with Prime95 26.4 and what a boost! From .077 to .055 on a Core2 4300 @ 1.8Ghz (relatively clean install). That's almost a 30% boost! I may be paranoid to want to finish my workers before upgrading but such an increase in performance is too tempting. This is far better than the 3-5% average improvement in the 25.x series of updates. I've got a Quad Core AMD Phenom that I'm curious how well it will perform after these updates (perhaps none, but worth a try).

[tqft]

The current download from seventeenorbust.com for linux appears to be for 263 dated 18 Oct 2010 - should I wait for 264 or upgrade now?

[Joe O]

I haven't sent Louis the updated packages to post. Do you need the package or are you already running Prime95? If you are already running Prime95 then just download the new executable. If you are just starting, then download the package, and replace the executable before startiing. If you are just starting then please make sure that your userid is correct in the prime.txt file so that you will get credit for your work.

[Cowering]

I was setting up prime95 for a friend. His machine is a Pentium-D (true dual core) OCed to 3.2GHZ. It runs perfectly, but... worker thread #1 is reported as a 'Pentium 4-type3' in the prime95 output window. worker thread #2 is reported as a 'Core2-type3'. Thread 2 is running iterations about 20% faster than thread 1, when both cores are just idling. Is this a detection error? Can I somehow force thread 1 into Core2 mode for some extra speed?

Go TeamRetro!

[prime95]

worker thread #1 is reported as a 'Pentium 4-type3' in the prime95 output window. worker thread #2 is reported as a 'Core2-type3'.

This is probably normal. On a Pentium 4, some FFT lengths are actually faster using FFTs optimized for the Core 2. Prime95 knows this and selects the right one accordingly. Admittedly, the "Core2 type 3" wording is poor/misleading.

I'll bet worker #1 and worker #2 are using different FFT lengths. If you can tell me which FFT lengths the two workers are using and if this is 32-bit or 64-bit prime95, then I can tell you if this is a bug or a feature.

I also own a Pentium D which I benchmarked to pick the fastest FFT implementation for each FFT length.