| Author |
Message |
|---|
|
|
|
Hi folks, after spending many hours I got this version also compiled successfully in windows. This one is the 64bit version.
SR5 tests looking good with around 50% speed improvement. I dont know other pfgw projects yet but I think its also working with that.
Stay tuned for download. Iam testing one more unit to know the time improvement.
SR5: before/after--->50-60min/30min |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1256
ID: 12001
Credit: 18,565,548
RAC: 0
 
|
Hi folks, after spending many hours I got this version also compiled successfully in windows. This one is the 64bit version.
SR5 tests looking good with around 50% speed improvement. I dont know other pfgw projects yet but I think its also working with that.
Stay tuned for download. Iam testing one more unit to know the time improvement.
SR5: before/after--->50-60min/30min
I maintain pfgw and I want to make sure that gwnum v27.x has any kinks worked out before I release a new version with that library. |
|
|
|
|
Hi folks, after spending many hours I got this version also compiled successfully in windows. This one is the 64bit version.
SR5 tests looking good with around 50% speed improvement. I dont know other pfgw projects yet but I think its also working with that.
Stay tuned for download. Iam testing one more unit to know the time improvement.
SR5: before/after--->50-60min/30min
I maintain pfgw and I want to make sure that gwnum v27.x has any kinks worked out before I release a new version with that library.
Iam testing it and have some little issues with it. One total crash of client, dont know why, and sometimes a pfgw.log file, creating from pfgw.exe or client? Performance boost is now 30-50%.
Yes, I have planned to contact you. I only replaced the older gwnum files + one change in code because of an error. If you want my source files or only the program, send me a PM. |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1256
ID: 12001
Credit: 18,565,548
RAC: 0
|
I am testing it and have some little issues with it. One total crash of client, dont know why, and sometimes a pfgw.log file, creating from pfgw.exe or client? Performance boost is now 30-50%.
Yes, I have planned to contact you. I only replaced the older gwnum files + one change in code because of an error. If you want my source files or only the program, send me a PM.
You can send me the code change that you made.
As for the crash in the client, if it wasn't the known issue in the client that I fixed in 5.0.4, then it would be helpful if you could provide additional information on it. I would also like to know if you could reproduce it. |
|
|
|
|
|
With 4 WUs running the time is increasing much higher than expected from 2300-2600s. No more problems seen so far. |
|
|
|
|
|
BETA-TEST!
Hi all,
just looking for beta testers with a test client (better 4 cores / tasks / clients running), for speed improvement of SR5, 2xSR5, 2xppselow:
What you need:
- prpnet client 4.3.7, 5.0.2, 5.0.4
- running SR5 and ppselow (pfgw for the prime test)
- a compatible AVX cpu (Sandybridge/Bulldozer)
- win7 sp1 or better
- 64bit only!
Please note to observate the client most of the time during rare crashes.
What can you expected
- app crash
- client crash after app crash (mostly with 5.0.4 client)
- lost work
Pls report crash dumps in this thread (with what client version was running) and speed improvement (before/after).
Thx!
Download-link
-Reb |
|
|
|
|
BETA-TEST!
Hi all,
just looking for beta testers with a test client (better 4 cores / tasks / clients running), for speed improvement of SR5, 2xSR5, 2xppselow:
What you need:
- prpnet client 4.3.7, 5.0.2, 5.0.4
- running SR5 and ppselow (pfgw for the prime test)
- a compatible AVX cpu (Sandybridge/Bulldozer)
- win7 sp1 or better
- 64bit only!
Please note to observate the client most of the time during rare crashes.
What can you expected
- app crash
- client crash after app crash (mostly with 5.0.4 client)
- lost work
Pls report crash dumps in this thread (with what client version was running) and speed improvement (before/after).
Thx!
Download-link
-Reb
OK will try to help out with exactly your specs..
Q: on SR5 task.. is your pfgw the only .exe to uncomment ??
parabol
____________
I'm a prime millionaire !
9*2^3497442+1 |
|
|
|
|
OK will try to help out with exactly your specs..
Q: on SR5 task.. is your pfgw the only .exe to uncomment ??
parabol
this line must be activated:
pfgwexe=pfgw64.exe
Thx for helping!
-Reb |
|
|
|
|
|
OK rig setup as follows:
2 CPU-Cores running PPSElow with 'llravx'
2 CPU-Cores running SR5 with your 'pfgw'
1 GPU-Core running GFN-work with 'GeneferCuda'
Client 5.0.4
i7-2600K OC to 3600 (temp 60C)
GTX-560 setting 925/1850/1715 (running 99% GPU , temp 64C) driver 285.62
OS: win7/64 prof SP1
Will let it run now for 48h+
Any more info you need ??
Regards .. parabol
____________
I'm a prime millionaire !
9*2^3497442+1 |
|
|
|
|
OK rig setup as follows:
2 CPU-Cores running PPSElow with 'llravx'
2 CPU-Cores running SR5 with your 'pfgw'
1 GPU-Core running GFN-work with 'GeneferCuda'
Client 5.0.4
i7-2600K OC to 3600 (temp 60C)
GTX-560 setting 925/1850/1715 (running 99% GPU , temp 64C) driver 285.62
OS: win7/64 prof SP1
Will let it run now for 48h+
Any more info you need ??
Regards .. parabol
Runtimes of SR5 and client version (+ earlier values with standard app)
Thx.
-Reb |
|
|
|
|
OK rig setup as follows:
2 CPU-Cores running PPSElow with 'llravx'
2 CPU-Cores running SR5 with your 'pfgw'
1 GPU-Core running GFN-work with 'GeneferCuda'
Client 5.0.4
i7-2600K OC to 3600 (temp 60C)
GTX-560 setting 925/1850/1715 (running 99% GPU , temp 64C) driver 285.62
OS: win7/64 prof SP1
Will let it run now for 48h+
Any more info you need ??
Regards .. parabol
Runtimes of SR5 and client version (+ earlier values with standard app)
Thx.
-Reb
-Client version see my post: 5.0.4
Your 'pfgw.exe' seem to be very fast with SR5 !!!
Let me run it for a time and I will report further.
PPSElow with 'llravx' is terribly fast !!
Greetings .. parabol
____________
I'm a prime millionaire !
9*2^3497442+1 |
|
|
|
|
OK rig setup as follows:
2 CPU-Cores running PPSElow with 'llravx'
2 CPU-Cores running SR5 with your 'pfgw'
1 GPU-Core running GFN-work with 'GeneferCuda'
Client 5.0.4
i7-2600K OC to 3600 (temp 60C)
GTX-560 setting 925/1850/1715 (running 99% GPU , temp 64C) driver 285.62
OS: win7/64 prof SP1
Will let it run now for 48h+
Any more info you need ??
Regards .. parabol
Runtimes of SR5 and client version (+ earlier values with standard app)
Thx.
-Reb
-Client version see my post: 5.0.4
Your 'pfgw.exe' seem to be very fast with SR5 !!!
Let me run it for a time and I will report further.
PPSElow with 'llravx' is terribly fast !!
Greetings .. parabol
Thx, I also tested pfgw with ppselow but looks the same timings or slower.
With 2 SR5 only you can get a 45-50% faster calculation :) |
|
|
|
|
|
Thx, I also tested pfgw with ppselow but looks the same timings or slower. (from rebirther)
Agree .. about 10% slower
parabol
____________
I'm a prime millionaire !
9*2^3497442+1 |
|
|
|
|
|
If the app crashed on your test client 5.0.4:
- stop the client
- overwrite pfgw64 with the originally one
- start the client and let the workunit end
After that I only need tests with 4.3.7 client.
Thx
-Reb |
|
|
|
|
If the app crashed on your test client 5.0.4:
- stop the client
- overwrite pfgw64 with the originally one
- start the client and let the workunit end
After that I only need tests with 4.3.7 client.
Thx
-Reb
Well the app (SR5) actually did not 'crash'
It stopped with a ^C command which I did not give myself.
Greedtings
parabol
____________
I'm a prime millionaire !
9*2^3497442+1 |
|
|
|
|
If the app crashed on your test client 5.0.4:
- stop the client
- overwrite pfgw64 with the originally one
- start the client and let the workunit end
After that I only need tests with 4.3.7 client.
Thx
-Reb
Well the app (SR5) actually did not 'crash'
It stopped with a ^C command which I did not give myself.
Greedtings
parabol
My last 5.0.4 test client crashed again but 4.3.7 works fine ^^ |
|
|
|
|
|
Please stop testing until I can find the issue with compiler! |
|
|
|
|
|
I have tested a compiled version from rogue in same size. It crashed also but took some hours.
Pls testing further on your systems... |
|
|
|
|
|
Hello beta-testers,
we have discovered a problem/bug in gwnum code. After some local tests, all SR5 numbers around larger than 260k starting with crash.
Some of gfn-prime-testings with ppselow are also affected.
Please run out your current work and stop!
Thx
-Reb
|
|
|
 rroonnaalldd Volunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
                 
|
|
Could be the same problem what we see with the llr386(devsrc)_AVX-build on hosts without AVX...
____________
Best wishes. Knowledge is power. by jjwhalen
 |
|
|
|
|
|
pfgw3.6.0-AVX-32bit
pfgw3.6.0-AVX-64bit
Successful tests of SR5 >260k now without crashing.
Due bandwidth issues only run 2 of 4 cores to get full speed (around 45%)
The new early beta of gwnum 27.3 should also fixed the Bulldozer problem.
Pls test and report! |
|
|
|
|
|
Both versions of pfgw 3.6.0 fail on an AMD FX-4100 Bulldozer. |
|
|
|
|
|
The new one is a little bit faster:
2 x SR5 = 1500-1700s vs. 1800-2100s
more than 2 decrease speed due cpu cache limits (this not affects the llravx) |
|
|
|
|
|
Passed some more tests vs. non avx (gwnum 27.3):
2 x SR5 = ~56% speedup
3 x SR5 = ~47% speedup
4 x SR5 = ~39% speedup
The ivy bridge has 8MB cache, so it can be more profitable than 6MB with Sandy Bridge (bottleneck). It would be great to get all 4 cores running with double output.
If someone has a Sandy Bridge-E (10MB cache) post some results here without HT on. |
|
|
|
|
|
wow: pfgw64 avx is faster than llravx on ppselow port (32sec vs 34sec).
pfgw64 (non avx): 55 sec in the same port.
trying sr5 now. |
|
|
|
|
wow: pfgw64 avx is faster than llravx on ppselow port (32sec vs 34sec).
pfgw64 (non avx): 55 sec in the same port.
trying sr5 now.
Nope, the pfgw64avx is always slower than llravx. |
|
|
|
|
wow: pfgw64 avx is faster than llravx on ppselow port (32sec vs 34sec).
pfgw64 (non avx): 55 sec in the same port.
trying sr5 now.
Nope, the pfgw64avx is always slower than llravx.
Not on my host:
[2012-02-20 16:12:14 GST] Server: PPSElow, Candidate: 9869*2^323399+1 Program: pfgw64.exe Residue: 018A2463F053E94B Time: 32 seconds
[2012-02-20 16:12:07 GST] Server: PPSElow, Candidate: 9069*2^323399+1 Program: llravx.exe Residue: 30C13A6A690FB9FC Time: 35 seconds
Note: llravx is 32bit version. |
|
|
|
|
If someone has a Sandy Bridge-E (10MB cache) post some results here without HT on.
You still need a test run? I've got a 3960X running with 8 out of 12 on GFN Fermet Sieving, but I can put it on hold for a couple of hours to run 4 cores without HT (need to keep the other two going due to other Project committments I have made)
I've run PPSElow before with llravx, let me know changes you need to that setup for a test, and if you still need one doing, I'll get it done.
Regards
Zy |
|
|
|
|
If someone has a Sandy Bridge-E (10MB cache) post some results here without HT on.
You still need a test run? I've got a 3960X running with 8 out of 12 on GFN Fermet Sieving, but I can put it on hold for a couple of hours to run 4 cores without HT (need to keep the other two going due to other Project committments I have made)
I've run PPSElow before with llravx, let me know changes you need to that setup for a test, and if you still need one doing, I'll get it done.
Regards
Zy
Nice! Do it fast, the port is empty soon! But need a full core run to see if the speed is slowing down or not ;) |
|
|
|
|
|
PPSELow or SR5 - also need the app - where from? Putting the Sieve on hold and rebooting whilst waiting for reply, back in a couple
Regards
Zy |
|
|
|
|
PPSELow or SR5 - also need the app - where from? Putting the Sieve on hold and rebooting whilst waiting for reply, back in a couple
Regards
Zy
You can get all files from the link below. |
|
|
|
|
|
Running 6 cores on PPSElow - tell me if you need different
Regards
Zy |
|
|
|
|
Running 6 cores on PPSElow - tell me if you need different
Regards
Zy
Need a run on SR5 only! |
|
|
|
|
|
opps - on its way rofl
Regards
Zy |
|
|
|
|
|
Phew, close .... managed to get work on four cores, other two failed to, looks like port out of work - but first four cores running on port 7171
Running 5.0.4, 3960X @4.4Ghz on four cores
Regards
Zy |
|
|
|
|
|
Its running hot - PCH went to 60 degrees
Fiirst results:
1475.1299 + 0.0173s
1476.0073s+0.0173s
1284.1522s+0.0174s
1284.4345s+0.0173s
If they all got what was set for then will have 10 per core on four cores, want all run for the test ? If so how you want results reported, as above ??
Regards
Zy |
|
|
|
|
Its running hot - PCH went to 60 degrees
Fiirst results:
1475.1299 + 0.0173s
1476.0073s+0.0173s
1284.1522s+0.0174s
1284.4345s+0.0173s
If they all got what was set for then will have 10 per core on four cores, want all run for the test ? If so how you want results reported, as above ??
Regards
Zy
This is a 6 core processor right? If so run 6 x SR5 and report. I have run my i-2500k@4Ghz, runtimes could be the same with higher clock but my temps are @70°C. |
|
|
|
|
|
Cant run on 6, no work, only 4 cores managed to grab any.
PCH went back down to 38 (GPU apps finishing off so running cooler), cpu running at 55 (has cpu water cooler), with four cores running
Regards
Zy |
|
|
|
|
Cant run on 6, no work, only 4 cores managed to grab any.
PCH went back down to 38 (GPU apps finishing off so running cooler), cpu running at 55 (has cpu water cooler), with four cores running
Regards
Zy
Yes, we must wait for fill up. |
|
|
|
|
|
Okie Doke - I'll finish up what I have rather than abandon, and keep an eye open, as soon as I see a refill, I'll come back and do a full run. I'll keep the sieve suspended for now and run GPU apps whilst waiting for refill.
Regards
Zy |
|
|
|
|
I have run my i-2500k@4Ghz,
When the refill comes in, want me to run @4Ghz as well ?
EDIT:
Refill's in, setting up again, need to finsh off some GPU WUs, will be running in around 15 mins or so circa 2055hrs UTC
EDIT2:
All six cores running, 10WUs on each core
Regards
Zy |
|
|
|
|
|
12 results in .... how do you want them reported ... or are you grabbing from the server?
Regards
Zy |
|
|
|
|
|
There's around 50 done now, so I'm stopping the run and assuming you are grabbing from the server.
I'm restarting the sieve and GPU app, if you want more done post |
|
|
|
|
12 results in .... how do you want them reported ... or are you grabbing from the server?
Regards
Zy
Report only the time on all six cores. There is enough work now :) |
|
|
|
|
|
Updated to gwnum 27.4 lib:
Download |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1256
ID: 12001
Credit: 18,565,548
RAC: 0
|
Updated to gwnum 27.4 lib:
Download
It appears that some of the AMD Bulldozer issues have been resolved, I will post new builds today with gwnum v27.4. If you don't mind, could you please remove your builds to eliminate confusion. |
|
|
|
|
Updated to gwnum 27.4 lib:
Download
It appears that some of the AMD Bulldozer issues have been resolved, I will post new builds today with gwnum v27.4. If you don't mind, could you please remove your builds to eliminate confusion.
There are some people having problems with Bulldozer and llr so pfgw should also not working. If someone can confirm this pls report here! |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1256
ID: 12001
Credit: 18,565,548
RAC: 0
|
Updated to gwnum 27.4 lib:
Download
It appears that some of the AMD Bulldozer issues have been resolved, I will post new builds today with gwnum v27.4. If you don't mind, could you please remove your builds to eliminate confusion.
There are some people having problems with Bulldozer and llr so pfgw should also not working.
It might work in some circumstances. George had told me days ago that it would fix the issues. In any case, this is an official release which will make some happy. |
|
|
|
|
|
On an AMD FX-4100 Bulldozer using Windows 7 SP1, both 32 bit and 64 bit pfgw 3.6.2 fail with "error 1002 initializing FFT code". |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1256
ID: 12001
Credit: 18,565,548
RAC: 0
|
On an AMD FX-4100 Bulldozer using Windows 7 SP1, both 32 bit and 64 bit pfgw 3.6.2 fail with "error 1002 initializing FFT code".
George is aware that some issues still exist with gwnum v27.4. He doesn't have a box, but is asking someone with one to help. |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1256
ID: 12001
Credit: 18,565,548
RAC: 0
|
|
George has told me that gwnum v27.5 is official. I have built pfgw 3.6.3 and posted on sourceforge. Right now only the Windows version is there, but the Mac and Linux builds should follow later today. |
|
|
|
|
|
Is it safe to run pfgw 3.6.3 on PRPNet?
____________
My Top 5000 Primes:
110059!+1 is prime! (FPS)
14493618614235*2^666666-1 is prime! (SGS) |
|
|
|
|
Is it safe to run pfgw 3.6.3 on PRPNet?
Yes, this is the official version. |
|
|
|
|
|
Wow thanks for updating it PPESLOW is going by so fast! 40-50 seconds per work unit :) |
|
|
