| Author |
Message |
|---|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
                 
|
|
The speed of this sieve has caught us by surprise. Up to p=1405T has been inserted into the buffer. What shows on the front page is all that's left in the phase.
The sieve will pause before starting the next phase. tpsieve will be replacing ppsieve as the application. We'll soon be sieving both +1 and -1 forms in BOINC.
The reason for the pause is that all previous work must be completed and validated before switching to tpsieve. We expect to run out of work in a couple of days with only expired and resends remaining after that. The "clean up" should be completed sometime next week. Therefore, make sure your CPU's and GPU's have other work to do until then.
This is AMAZING work that's being done. Ken Brazier's ppsieve/tpsieve* truly has revolutionalized the speed of sieving. We'll be using tpsieve from this point forward. Once this next phase reaches p=1405T, we'll combine the entire n<3M range and move forward from 1405T. The PSA will turn its focus to 3M < n < 6M.
Thank you to everyone participating in this effort. :)
* based on Geoff Reynold's original work.
____________
|
|
|
 Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
                           
|
|
I suppose I have a few questions...
1) If I understand correctly, the PPS sieve BOINC subproject will soon be doing double duty, working on both the +1 and -1 forms simulataneously. Is the name of the subproject going to change?
2) Is it a new subproject altogether?
3) I guess what I'm really asking is, are we rapidly running out of hours to get a PPS Sieve badge?
4) ... and/or will there be a new badge to get?
5) You mentioned that there's a tpsieve effort going on in the PSA. While the PPS Sieve project is paused, should we point our GPUs in that direction? If so, how? The only information I found about this, if I was looking in the right place, said to PM you. So, you can consider this a public PM. :)
6) Is the ppsieve/tpsieve application usable for the other sieves? If so, would it be possible to create separate sub-projects for the CPU and GPU versions of the sieves? The reasons this would be preferable is that once there's more than one GPU application, there's no way to select what you want to run on the CPU and what you want to run on the GPU. With only one GPU project there's a workaround, but once there's two it's impossible without going through a lot of convolutions (running all the CPU apps in a VM, as an example.)
7) (I'm putting on my asbestos underwear before asking this one...) Has any thought been given to removing the CPU version of the PPS sieve? Especially once the Fermi 'no-work' problem is solved, is there any reason to run the sieve on a CPU? IMHO CPUs could be put to far better use crunching something that a video card can't do 100 times faster.
And a comment on the progress of the sieve. Is anyone really surprised? For a few months now we've had an amazing GPU application that can outrun an i7 by two orders of magnitude. Coupled with the fact that it's the only PrimeGrid GPU application (so everyone is running it, as if it was a 24/7/365 challenge) and the amount of crunching done on this has to be completely unprecedented. The only downside is the problem with getting work to the Fermi cards, which is undoubtedly keeping some of the fastest processors on the sidelines.
____________
My lucky number is 75898524288+1 |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
I suppose I have a few questions...
A few??? :D
1) Is the name of the subproject going to change?
2) Is it a new subproject altogether?
3) I guess what I'm really asking is, are we rapidly running out of hours to get a PPS Sieve badge?
4) ... and/or will there be a new badge to get?
Probably, No, No, and No. For naming concerns, consider this like the PSP (Sieve) which was eventually called the PSP/SoB (Sieve). Nothing about the project changed except the name. At most, this will become known as the PPS/RSP (Sieve) but no other changes are expected.
5) You mentioned that there's a tpsieve effort going on in the PSA. While the PPS Sieve project is paused, should we point our GPUs in that direction? If so, how? The only information I found about this, if I was looking in the right place, said to PM you. So, you can consider this a public PM. :)
The PPS/RSP (Sieve) preparation is a manual effort and may be found in the PST forum. The pause will be very short so it might be easier to just do another BOINC project during the downtime. We expect to be up and running again next week.
6) Is the ppsieve/tpsieve application usable for the other sieves?
The PPS/RSP (Sieve) is currently the only sieve that benefits from ppsieve/tpsieve. One sieve that might be in jeopardy is the 27121 manual sieve in PSA. Most likely it will be swallowed up by this effort. The 321 (Sieve) is currently the only BOINC sieve that's approaching an end within 6-12 months.
7) (I'm putting on my asbestos underwear before asking this one...) Has any thought been given to removing the CPU version of the PPS sieve? Especially once the Fermi 'no-work' problem is solved, is there any reason to run the sieve on a CPU? IMHO CPUs could be put to far better use crunching something that a video card can't do 100 times faster.
Yes, thought has been given. Same discussion a while back with 32 bit vs. 64 bit in the sieves. Same solution...inform the users and let them decide for themselves.
AND yes, the problem with Fermi's not being able to "freely" contribute without app_info is frustrating.
____________
|
|
|
|
|
|
A question regarding ATI cards: will there also be an application with OpenCL? Otherwise ATI cards will be locked out again, just when I'm appraoching my silver badge :( |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
A question regarding ATI cards: will there also be an application with OpenCL? Otherwise ATI cards will be locked out again, just when I'm appraoching my silver badge :(
Yes, tpsieve has an OpenCL build as well.
____________
|
|
|
|
|
|
Whoa! Less than 9600 WUs to go! |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
Update
The buffer is now empty: Available: Proth Prime Search (Sieve) 0
Now we must wait for all outstanding work to be completed and validated. Regular progress updates will be posted. We expect to be up and running again with tpsieve sometime next week.
Work In Progress: 45859
____________
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
|
The longest WU deadlines seem to be 100 hours (there's a lot of variation at the moment due to resends). If and when those expire, will they be resent with the same 100 hour deadline, or will it be shortened so as to expedite the clean-up?
FWIW, a GTX 280 with a maxed out 100 WU cache, assuming all are the longer 2G WUs (which is unlikely) will burn through the cache in about 16 hours. Fermis will be a lot faster, 9xxx, 8xxx, and CPUs a lot slower. Assuming that a lot of people with GPUs extended their cache sizes, or were already running at max, I would expect the majority of the WUs to come back within the next 24 hours.
A request from the peanut gallery: It's nice to be able to keep a decent buffer of work around in case of unexpected outages. (It's financially irresponsible to build dual sites and other high availability options for a project such as this, so occaissional outages are a fact of life and a necessary evil.) PG has actually been VERY stable as BOINC projects go (kudos to you guys!), but stuff happens. So a big queue is nice.
With the CUDA sieve application, that's currently impossible. Even on my dinosaur GPU (hehe, a tera-flop super-computer on a chip is a dinosaur?!?) the most I can queue up is less than 3/4 of a day. On a Fermi GPU -- and they just released the GTX 580 -- the queue length might be down to something around 3 hours! ATI is releasing a new series of GPUs this month too, I believe.
Could either A) the number of WUs per host be increased, or better yet, B) the length of the WUs be increased even more? 10G per WU would let people on moderate GPUs queue up several days worth (subject to the deadlines, of course), and those with the real speed demons could at least get close to a day's work.
Even better, since you don't have to validate WUs against wingmen, could you do what Rosetta does, and give users the option of selecting the WU length? That way people running this on, say, an 8600 (or a CPU) could use the old 1G size, while people with Fermi cards could go to 10G. Bigger WUs is better all around for the fast cards since building large queues causes CPU loading issues on both the client and the server.
EDIT: I think 10G would be ideal on my GPU, yielding a run-time of about 50 minutes. I think WUs of between about 1 hour and 24 hours are ideal, giving a good balance. So I'd be happy if you went as high as 250G!! Yes, increasing the WU size by a factor of 250 would wreck havoc on slower systems, but the converse is true, too: the small WUs (and these are really TINY on top-end hardware) aren't very efficient and cause their own set of problems.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Let us see how fast the GTX 580 drops in prices and when they are available.
MIX claims to have some, downside is they are to expensive and i have no money. ;) |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
The longest WU deadlines seem to be 100 hours (there's a lot of variation at the moment due to resends). If and when those expire, will they be resent with the same 100 hour deadline, or will it be shortened so as to expedite the clean-up?
I'll answer my own question here: Based on the resends I'm getting, the answer is the WU retains the same deadline as the original results.
____________
My lucky number is 75898524288+1 |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
will they be resent with the same 100 hour deadline, or will it be shortened so as to expedite the clean-up?
same deadline
I would expect the majority of the WUs to come back within the next 24 hours.
Yes, this is the expectation.
Could either A) the number of WUs per host be increased, or better yet, B) the length of the WUs be increased even more?
A) This can be reviewed. B) The first step was to 2G. 10G's might be too big of a jump at this time.
give users the option of selecting the WU length?
Not sure if this is possible with PG's current configuration.
____________
|
|
|
|
|
The first step was to 2G. 10G's might be too big of a jump at this time.
probably this will scrare away the last cpu-crunchers...
|
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 11,549,356
RAC: 0
                 
|
The first step was to 2G. 10G's might be too big of a jump at this time.
probably this will scrare away the last cpu-crunchers...
Another case where everyone could be made happy by separating GPU stuff into its own sub-project. CPU = 1G, GPU = 5 or 10G.
____________
Murphy (AtP)
 |
|
|
|
|
|
Damn i get CPU resends but none for GPU, BOINC again seems broken to me on this one. |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
Update
We must wait for all outstanding work to be completed and validated. Regular progress updates will be posted. We expect to be up and running again with tpsieve sometime next week.
Work In Progress
17 Nov: 141 tasks
16 Nov: 292 tasks
15 Nov: 763 tasks
14 Nov: 1616 tasks
13 Nov: 3792 tasks
12 Nov: 7797 tasks
11 Nov: 16110 tasks
10 Nov: 45859 tasks
____________
|
|
|
|
|
|
This may not be the best place to ask but I have a pc with a GeForce 9500GT installed and anther pc with an ATI Radeon HD 2400 XT. 1) how do I know it is being used and 2) are these cards that can be used? |
|
|
|
|
This may not be the best place to ask but I have a pc with a GeForce 9500GT installed. 1) how do I know it is being used and 2) is this a card that can be used?
It is listed as CUDA enabled on NVIDIA's website. Since the app "requires" only compute capabilty 1.0 it should work (if the card has enough memory) once the switch to tpsieve is complete.
If BOINC recognized the card you can see a message similiar to this (GTX 460/Linux/BOINC 6.10.58):
Do 11 Nov 2010 19:41:56 CET NVIDIA GPU 0: GeForce GTX 460 (driver version unknown, CUDA version 3020, compute capability 2.1, 1023MB, 650 GFLOPS peak)
If you don't want to wait you can download the ppsieve-cuda archive and try a command like this to check if the card works without waiting for new WUs:
Testing the "old" ppsieve:
ppsieve-cuda-x86-windows -p 42070e9 -P 42070100e6 -k 5 -K 9999 -N 2000000
Testing the "new" tpsieve
tpsieve-cuda-x86-windows -p 42070e9 -P 42070100e6 -k 5 -K 9999 -N 2000000 -M2
____________
|
|
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2165
ID: 1178
Credit: 8,777,295,508
RAC: 0
                                     
|
...anther pc with an ATI Radeon HD 2400 XT...
The 2xxx and 3xxx series ATI cards cannot run OpenCL, and therefore, cannot crunch here at PG. If a native CAL/Brook ATI application version is eventually created, you would be able to crunch with this card (assuming it has enough RAM...some 2400 series cards have only 128 mb, and thus, probably would not work well if at all).
For your 2400XT, the Collatz project offers a mathematical project where you can crunch.
____________
141941*2^4299438-1 is prime!
 |
|
|
|
|
|
the HD2XXX and 3XXX series are really fairly slow at crunching. Not much faster than a newer multicore CPU and you still get to deal with the excessive heat. I burned my 2600 up running collatz. It just doesnt have what it takes to do some really serious crunching |
|
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2165
ID: 1178
Credit: 8,777,295,508
RAC: 0
|
the HD2XXX and 3XXX series are really fairly slow at crunching. Not much faster than a newer multicore CPU and you still get to deal with the excessive heat. I burned my 2600 up running collatz. It just doesnt have what it takes to do some really serious crunching
I respectfully disagree. The 3850 and 3870 are comparable in speed to the 4650/4670 series and the 5550. However, unlike the latter two, both of these 3xxx cards can do DP calculations, and thus, can also run on projects needing DP. On Collatz, those cards are also a fast or a bit faster that 9800GTX/GTX+ and GTS 250 cards.
As for the slower 24xx/34xx or 26xx/36xx cards, I have been very successfully running cards of all four designations for months now on Collatz without one failure or any heating issues (though always in mid-tower or larger well-ventilated cases). One just needs to make sure that the cooling is adequate to successfully crunch with these (which might mean adding/changing to a different fan). Also, all of these series cards can easily be obtained for less than $25 on EBAY (under $10 in some cases), so even if they burn up it is not too big of a deal.
____________
141941*2^4299438-1 is prime!
|
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
         
|
|
Work In Progress
13 Nov: 3153 tasks in the moment of writing this posting.
____________
Best wishes. Knowledge is power. by jjwhalen
 |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
Work In Progress
1660 tasks are open in the moment of writing this posting.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
Work In Progress
861 tasks open
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
Work In Progress
382 tasks open
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
|
During the last hour I got 30 PPS Sieve resends (from totally 286 In Progress atm).
Fortunately not for the CPU.
If I'don't get more estimated return time of my last one: 17-Nov-2010 20:08 UTC.
____________
 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
|
The PrimeGrid home page has been showing 15 PPS Sieve WUs available for a little while now, but I can't seem to download them on either my GPU or even my CPU.
So, what's up with that?
____________
My lucky number is 75898524288+1 |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
The PrimeGrid home page has been showing 15 PPS Sieve WUs available for a little while now, but I can't seem to download them on either my GPU or even my CPU.
So, what's up with that?
PPS (Sieve) has been disabled for the duration of the Challenge. Afterwards, work flow will resume.
____________
|
|
|
|
|
|
I'm getting PPS LLR WU during the challenge. Someone needs to double check the server settings. |
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 11,549,356
RAC: 0
|
I'm getting PPS LLR WU during the challenge. Someone needs to double check the server settings.
No. It is the user's responsibility to select the proper sub-project in the preferences.
____________
Murphy (AtP)
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
Update
We must wait for all outstanding work to be completed and validated. Regular progress updates will be posted.
Work In Progress
19 Nov: 19 tasks
18 Nov: 63 tasks
17 Nov: 141 tasks
16 Nov: 292 tasks
15 Nov: 763 tasks
14 Nov: 1616 tasks
13 Nov: 3792 tasks
12 Nov: 7797 tasks
11 Nov: 16110 tasks
10 Nov: 45859 tasks
____________
|
|
|
|
|
|
relatively speaking...would it not be more efficient to cancel them all immediately [Canceled by server] and include them at the front of the next batch which should, by all accounts, complete them in a fraction of the time that you will spend waiting for them to decay through the node-caches? Just a thought.
____________
 |
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 11,549,356
RAC: 0
|
relatively speaking...would it not be more efficient to cancel them all immediately [Canceled by server] and include them at the front of the next batch which should, by all accounts, complete them in a fraction of the time that you will spend waiting for them to decay through the node-caches? Just a thought.
That would be highly irritating to the user who was crunching them and almost finished (especially if within the deadline)!
____________
Murphy (AtP)
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
|
And, for what it's worth, 7 of the remaining WUs were downloaded yesterday evening to a single computer (as CPU apps, even though that computer has a GTS240), and it's already returned 4 of the 7. Since that computer has no other PrimeGrid WUs, presumably it's now working on the other 3 right now (although it could be working on another project.) Those WUs should be completed and returned in a few hours, hopefully.
If you want to kill the WUs, a better plan is to prevent new results from being sent out when/if the current ones either time or get returned. That way you don't kill someone's work due to no fault of their own.
____________
My lucky number is 75898524288+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
|
FWIW, here's some of the remaining few (18 as I write this) WUs I could locate:
Of seven downloaded last night to this computer, two are still being worked on:
139376979
139150474
Another 7 were downloaded to this computer on November 15th, and they expire about 12 hours from now. Presumably they'll just time out, although they may get processed and returned by then.
138096857
137821864
138039349
138038293
138028569
138036324
138037201
So of the 18, I could find 9. Two of those should be returned within 4 hours, and 7 will likely be sent out to a new computer 12 hours from now. With luck, that new computer could return those 7 within an hour. That accounts for half of what's left. Cross your fingers!
____________
My lucky number is 75898524288+1 |
|
|
Ken_g6Volunteer developer
Send message
Joined: 4 Jul 06
Posts: 915
ID: 3110
Credit: 183,164,814
RAC: 0
             
|
|
I was wondering if a small number of WUs could have extra copies sent out to a trusted user, like Lennart, say. Then he could complete the work, and it wouldn't get sent out to anyone who didn't already have it, but those who had the work could still complete it and get credit.
____________
|
|
|
|
|
|
That would be a great solution. Perhaps something to think about for the next time something like this happens.
But let's hope we will be crunching again sometime tomorrow afternoon!
My GTX275 is hungry!
____________
 |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
Work In Progress
4 tasks open
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
|
4 tasks, now you are just taunting us |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
Work In Progress
2 tasks open
Not taunting, only reality.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
|
I interpreted John's post incorrectly.
I thought we were waiting for the sum total of the wu's issued on those days.
Oops.
In reality the return rate is much faster than expected.
Almost there.
Probably there as I post this.
____________
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
Work In Progress
2 tasks open
The good news is those two (actually 3; one is done) were sent to a computer which is processing them on a GPU.
The bad news is it's a relatively slow GPU. The first WU took about 2 hours to complete.
The great news is these are the last two WUs and they're at a computer that's actually crunching them, so the end is in sight!
____________
My lucky number is 75898524288+1 |
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1893
ID: 352
Credit: 3,142,312,174
RAC: 0
                       
|
I was wondering if a small number of WUs could have extra copies sent out to a trusted user, like Lennart, say.
Couple years ago, I suggested to Dave A. to have some *gold* host attribute (or whatever similar) for those host with very high reliability (valid results, very short turn-around with small WU cache, close 24/7 up-time). Those are suitable for clean-up older WUs, where inconsistent results etc. I would be happy to provide such hosts doing dirty job.
I guess server scheduller is far from doing this job...
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186 |
|
|
|
|
|
When these 2 WU's are finished will the project continu rightaway?
Or will there be some down time?
____________
|
|
|
samuel7Volunteer tester
Send message
Joined: 1 May 09
Posts: 89
ID: 39425
Credit: 257,416,435
RAC: 0
        
|
|
The oldest WU 139203872 is still lingering (and perhaps others). Task limits on the WU have been upped to 35 but nothing has happened in 2.5 days. |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
When these 2 WU's are finished will the project continu rightaway?
Or will there be some down time?
Take a look at http://www.primegrid.com/download/?C=M;O=D. There is no "tpsieving" app in the moment...
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
When these 2 WU's are finished will the project continu rightaway?
Or will there be some down time?
Take a look at http://www.primegrid.com/download/?C=M;O=D. There is no "tpsieving" app in the moment...
We are working on it and I think ( if nothing goes wrong ) that it will be running in ~4-6 hr.
Lennart
____________
 |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
Update
Now that all work has been completed, returned, and validated, we can transition from ppsieve to tpsieve. When Rytis is available, this transition will take place.
We will be starting the PPR3M range at p=210T and work up to 1405T. Once we reach the end, we'll combine the entire n<PPR3M range and work up from 1405T until "optimal" depths are reached for each 1M range.
Work In Progress
20 Nov: 0 tasks
19 Nov: 19 tasks
18 Nov: 63 tasks
17 Nov: 141 tasks
16 Nov: 292 tasks
15 Nov: 763 tasks
14 Nov: 1616 tasks
13 Nov: 3792 tasks
12 Nov: 7797 tasks
11 Nov: 16110 tasks
10 Nov: 45859 tasks
____________
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
The oldest WU 139203872 is still lingering (and perhaps others). Task limits on the WU have been upped to 35 but nothing has happened in 2.5 days.
OK, that is WIERD.
According to the links from the Subproject Status page, even though Range Statistics shows 0 tasks in progress, as you pointed out, the Oldest Unfinished Workunit is showing the result you mentioned, which clearly isn't completed.
____________
My lucky number is 75898524288+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
Once we reach the end, we'll combine the entire n<PPR3M range and work up from 1405T until "optimal" depths are reached for each 1M range.
At the pace we were going, how long do you think it will be until we hit the optimal depth?
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
If I have one pc with a GPU and set it to run the PPS, will BOinc send only GPU wu's or does it send both gpu and cpu tasks? |
|
|
|
|
|
no 64-bit cpu-app for windows???? |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
If I have one pc with a GPU and set it to run the PPS, will BOinc send only GPU wu's or does it send both gpu and cpu tasks?
It depends of your project-settings.
If you enable CPU and GPU you will get work for both. If you want to crunch PPS/TPSieve on GPU only and the other PG-subprojects on CPU you can use an app_info.xml from the thread App_info file.
For linux64 i will update this in the next few hours...
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
If I have one pc with a GPU and set it to run the PPS, will BOinc send only GPU wu's or does it send both gpu and cpu tasks?
BOINC's task selection mechanism is very, very poor, and completely inadequate in my opinion.
If you set your options to allow both CPU and CPU computing (as most of us will want to do), and you select PPS-Sieve, then yes, it should send both CPU and GPU tasks to your computer.
Assuming that's not what you want, there's several ways to prevent this.
1) (This is what I'm doing right now) Turn OFF the PPS-Sieve checkbox and turn ON the "Send work from any subproject if selected projects have no work" checkbox. This will work, but it does have two drawbacks: if there's ever more than one GPU sub-project here, you'll have no way of telling BOINC which sub-project you want to run on the GPU, and if the sub-project(s) you run on your CPU run dry then you'll get something else (which you might not want) on your CPU.
2) Run a VM (virtual machine) on your computer, with a second copy of BOINC. Run the GPU on the host machine's BOINC and the CPU on the VM's BOINC. Since you're running two separate copies of BOINC, you can choose completely different settings for them. The disadvantages are that setting up a VM can be intimidating if you've never done it before (it's actually very simple), the VM itself uses memory which might be an issue on your computer, and extra CPU cycles are used running two operating systems at once. Also, there may be issues with thread priorities between the host and guest systems which may result in the CUDA CPU thread being starved, thus reducing GPU performance.
3) Pray for Dave A and company to release a version of BOINC that is a lot more sophisticated in the way it handles GPUs and multiple sub-projects.
EDIT: I knew I was forgetting something. See the post before mine for option 4 using app_info.xml.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
first WU is running.. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
no 64-bit cpu-app for windows????
If they're done and that's the complete list (which is a big "if"), there's also no apps at all for Macs.
____________
My lucky number is 75898524288+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
|
Ok, I've got a bunch of new WUs. It looks like they'll take 20 minutes to run, double what the old ones took. I'm hoping that's expected (and that the credit is doubled as well.)
Thanks guys, it's nice to be crunching again!
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Getting no ATI Wus and my CPU units are all erroring out ("output file absent").
Project reset didn't help. :-/
____________
Grrrrrrrrreetings from the Lazy Cat...  |
|
|
|
|
Getting no ATI Wus and my CPU units are all erroring out ("output file absent").
Project reset didn't help. :-/
the first 100 wu was wrong. Wrong filename on output file.
Lennart
|
|
|
|
|
|
1st one reported - more than 1k factors.. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
|
Got the first WU returned. Looks like exactly double the work as before, but substantially less credit than expected.
Old WUs were returning about 270 credits per 10 minute WU. This one was 20 minutes and exactly 400 credits.
If that's intended, that's fine, but if it's not I wanted to give you a heads up about it.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Hm, still getting this:
20.11.2010 16:42:16 PrimeGrid Computation for task pps_sr2sieve_3969524_0 finished
20.11.2010 16:42:16 PrimeGrid Output file pps_sr2sieve_3969524_0_0 for task pps_sr2sieve_3969524_0 absent
20.11.2010 16:42:16 PrimeGrid Computation for task pps_sr2sieve_3969523_0 finished
20.11.2010 16:42:16 PrimeGrid Output file pps_sr2sieve_3969523_0_0 for task pps_sr2sieve_3969523_0 absent
____________
Grrrrrrrrreetings from the Lazy Cat... |
|
|
|
|
Got the first WU returned. Looks like exactly double the work as before, but substantially less credit than expected.
Old WUs were returning about 270 credits per 10 minute WU. This one was 20 minutes and exactly 400 credits.
If that's intended, that's fine, but if it's not I wanted to give you a heads up about it.
same picture over here.. |
|
|
|
|
|
Gigabyte GTX 470 SOC ( 700MHz core ):
- 6m 10s - 400 credits = 94k a day;
Old PPS was giving a 150k a day!
Huge drop, 33% less.
____________
Polish National Team |
|
|
|
|
|
I'm seeing a trend on the linux cpu app. What is option 'T', why is it invalid and why is it being called?
Stderr output
<core_client_version>6.10.58</core_client_version>
<![CDATA[
<message>
process exited with code 1 (0x1, -255)
</message>
<stderr_txt>
../../projects/www.primegrid.com/primegrid_tpsieve_1.35_i686-pc-linux-gnu: invalid option -- T
09:54:39 (10685): called boinc_finish
</stderr_txt>
]]>
Stderr output
<core_client_version>6.10.58</core_client_version>
<![CDATA[
<message>
process exited with code 1 (0x1, -255)
</message>
<stderr_txt>
../../projects/www.primegrid.com/primegrid_tpsieve_1.35_x86_64-pc-linux-gnu: invalid option -- 'T'
09:50:58 (29985): called boinc_finish
</stderr_txt>
]]> |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
I'm seeing a trend on the linux cpu app. What is option 'T', why is it invalid and why is it being called?
All CPU apps have been pulled until they are fixed. The -T flag forces use of the new tpsieve version. This will help prevent any app_info users from mistakenly using ppsieve which would only produce +1 factors. tpsieve is producing +1 AND -1 factors.
We do not yet have Mac version of any of the apps. However, that should be resolved in the next couple of days.
As for credit, adjustments will be made as necessary...which means yes, they will be increasing.
____________
|
|
|
|
|
As for credit, adjustments will be made as necessary...which means yes, they will be increasing.
600 looks quite okay - thanks!
|
|
|
|
|
|
ATI HD4670 (OpenCL) 7,463.65 sec = 2:04:23.65 / 600 credits |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
Ken's page at https://sites.google.com/site/kenscode/prime-programs lists new cpu-applications.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
ATI HD4670 (OpenCL) 7,463.65 sec = 2:04:23.65 / 600 credits
GT240(GT215) Cuda23 ~2400sec = ~40min / 600 credits
TPSieve needs now 4 times (PPS with 1G) or 2 times (PPS with 2G) longer on this card.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
no 64-bit cpu-app for windows????
If they're done and that's the complete list (which is a big "if"), there's also no apps at all for Macs.
Yes, waiting for that Mac CUDA app to appear. Hopefully soon; until then, the folks over at Collatz are happy, I'm sure.
gary |
|
|
|
|
|
The ATI-App is listed under Apps but got no work. Is there a need in using the app_info.xml or just waiting for new WUs?
____________
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
|
I'm not exactly complaining, but credit (for the same sized WUs) has gone from 400 to 600 to 1,800.
Is the 1800 intentional?
____________
My lucky number is 75898524288+1 |
|
|
|
|
The ATI-App is listed under Apps but got no work. Is there a need in using the app_info.xml or just waiting for new WUs?
Same question here.
I don't want to fiddle with the app_info.xml unless it is necessary.
Anyone able to advise?
Has all the available ati work been handed out?
____________
|
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
The ATI-App is listed under Apps but got no work. Is there a need in using the app_info.xml or just waiting for new WUs?
Maybe a problem with your AMD/ATI driver?
IIRC, the tpsieve-app for ATI needs CL. Please take a look in the thread ppsieve ATI/OpenCL testing.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
"Rebirther" wrote: The ATI-App is listed under Apps but got no work. Is there a need in using the app_info.xml or just waiting for new WUs?
Your host with the ID=120773 looks okay. 5 units got each 1800credits and no errors for PPSieve are listed...
[add]
But your CPU times seems to be a little bit high. Take a look at my host 42893.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
Ken_g6Volunteer developer
Send message
Joined: 4 Jul 06
Posts: 915
ID: 3110
Credit: 183,164,814
RAC: 0
|
|
rroonnaalldd wrote:
Your host with the ID=120773 looks okay. 5 units got each 1800credits and no errors for PPSieve are listed...
Those results are for the anonymous platform. Which means he was using an app_info.xml file for those results.
____________
|
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
Ahhh, i forgot the thread Meow?. This thread declares the differences in the higher cpu-time.
I posted my values of "avg_ncpus" and "max_ncpus" for Cuda in the thread App_info file.
They are:
<avg_ncpus>0.020000</avg_ncpus>
<max_ncpus>0.020000</max_ncpus>
[edit]
link
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
Ken, i am a little bit confused. I compared the times for calculating a tpsieve-wu on your host 111822 with a GTX460 and Rebirther's host 120773 with a HD5800?
Is OpenCL slower than Cuda or hits the GTX460 just a homerun with TPSieve?
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
Ken, i am a little bit confused. I compared the times for calculating a tpsieve-wu on your host 111822 with a GTX460 and Rebirther's host 120773 with a HD5800?
Is OpenCL slower than Cuda or hits the GTX460 just a homerun with TPSieve?
The OpenCL is slower than the GTX460 app and yes the cpu usage is taking around 5% of one core :/
____________
|
|
|
|
|
|
I got 3 out of 4 running.
1 x linux 64, NVIDIA 9600GT + Cuda23 resumed right where it left off.
1 x WinXP 32, ATI HD5770, *downgraded to* 10.10 APP 32 driver, inserted the x86 boinc client and created the app_info.xml and it 'just works'
repeated the same process for
1 x Win 7 64, ATI HD5770, *downgraded to* 10.10 APP 64 driver, inserted the x86 (#32 bit) boinc client and created the app_info.xml and it 'just works'
(#32 bit - there is no Win 64 bit client but the Win 32 bit client works just fine on a 64 WIn+64 BOINC install)
I'll have a go at the linux 64 ATI HD5770 setup later.
I just ran out of weekend. :(
lessons learned
Use the ATI 10.10 APP version driver.
It 'just works' and it's a lot simpler than fiddling with the Stream 2.2 SDK
The ATI 10.11 APP version does NOT work for these clients (not in my case anyway)
The ATI 10.11 version is NOT officially supported with Stream SDK 2.2.
The Stream SDK page shows the SDK is only tested as working with 10.9, but if you use the APP 10.9 or 10.10 you should not need the SDK anyway.
Looking forward to the linux 64 install when I can make some time free.
THANKS TO EVERYONE WHO WORKED ON THIS (esp. Ken_g6 for the OpenCL ports)
YOU ARE DOING A BRILLIANT JOB !!!
____________
|
|
|
|
|
|
AMDave, Iam running it with 10.11 driver and SDK 2.2 without any issue on win7 64bit
____________
|
|
|
|
|
AMDave, Iam running it with 10.11 driver and SDK 2.2 without any issue on win7 64bit
envious ;)
I think possibly I was not using the 10.11 APP version on that machine, it may have been the 10.11 standard edition of the driver that was last installed which would have meant that the OpenCL components would not have been there for the client to use.
If I had checked more closely I could have saved myself half an hour. ha ha
Thanks for the verification.
I'll let this setup bed in and try the upgrade to 10.11 later on.
NOTE -
I definitely did try the 10.11 APP driver ion the WinXP-32 machine and it flat-out would not work.
That is to say - the driver worked GREAT and crunch time on DNETC improved as well, but I could get no PPS(Sieve) WUs at all.
As soon as I downgraded to 10.10 it immediately started downloading work and started crunching without errors.
If anyone has ati32ati working with the 10.11 APP driver on WinXP-32, Id be interested to hear that verification.
____________
|
|
|
|
|
Getting no ATI Wus and my CPU units are all erroring out ("output file absent").
Project reset didn't help. :-/
the first 100 wu was wrong. Wrong filename on output file.
Lennart
I modified my app_info.xml today and get following for CPU-WUs
So 21 Nov 2010 14:26:52 CET PrimeGrid Output file pps_sr2sieve_4007263_0_0 for task pps_sr2sieve_4007263_0 absent
Does this sort out or should someone worry? |
|
|
|
|
Getting no ATI Wus and my CPU units are all erroring out ("output file absent").
Project reset didn't help. :-/
the first 100 wu was wrong. Wrong filename on output file.
Lennart
I modified my app_info.xml today and get following for CPU-WUs
So 21 Nov 2010 14:26:52 CET PrimeGrid Output file pps_sr2sieve_4007263_0_0 for task pps_sr2sieve_4007263_0 absent
Does this sort out or should someone worry?
The current CPU app doesn't work since it doesn't understand the dummy option -T that was introduced to be prevent the people from crunching newer WUs with the old ppsieve app.
Edit: Just tested it:
14:38:48 (3524): Can't open init data file - running in standalone mode
primegrid_tpsieve_1.35_windows_intelx86.exe: invalid option -- T
14:38:48 (3524): called boinc_finish
Edit^2: I don't know if Ken already has newer versions available for download...
____________
|
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
"John" wrote: All CPU apps have been pulled until they are fixed. The -T flag forces use of the new tpsieve version. This will help prevent any app_info users from mistakenly using ppsieve which would only produce +1 factors. tpsieve is producing +1 AND -1 factors.
We do not yet have Mac version of any of the apps. However, that should be resolved in the next couple of days.
As for credit, adjustments will be made as necessary...which means yes, they will be increasing.
http://www.primegrid.com/apps.php
There are no CPU-appz for PPSieve listed in the moment.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
|
It does not help that it is not listed, it is in the download-folder...
I downloaded it an included it into the app_info.xml |
|
|
VatoVolunteer tester
Send message
Joined: 2 Feb 08
Posts: 785
ID: 18447
Credit: 263,436,450
RAC: 0
         
|
|
I finally got my setup working with OpenCL after a lot of fiddling.
Please can someone post a pointer to a working Linux x86_64 ATI app_info.xml?
Ideally one that covers the various CPU apps?
____________
 |
|
|
|
|
I'm not exactly complaining, but credit (for the same sized WUs) has gone from 400 to 600 to 1,800.
Is the 1800 intentional?
We can use that credit now :) this will only be some more days.
We will soon reach 1.405P on this range ( 5<k<9999 2M<n<3M)
When we reach 1.405P range will be changed to 5<k<9999 1<n<3M
We will adjust credit when we change.
Lennart |
|
|
|
|
I'm not exactly complaining, but credit (for the same sized WUs) has gone from 400 to 600 to 1,800.
Is the 1800 intentional?
We can use that credit now :) this will only be some more days.
We will soon reach 1.405P on this range ( 5<k<9999 2M<n<3M)
When we reach 1.405P range will be changed to 5<k<9999 1<n<3M
We will adjust credit when we change.
Lennart
GPU POWER! :D
On cpu it would be few years. ;) Joking off course.
____________
Polish National Team |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
I finally got my setup working with OpenCL after a lot of fiddling.
Please can someone post a pointer to a working Linux x86_64 ATI app_info.xml?
Ideally one that covers the various CPU apps?
For linux64 and cuda i posted a App_info file for all PG-projects.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
|
Wish I had my GTX260 up and running. Outdated - maybe. Overclocked - you bet. Too bad I don't yet have a machine in which I can use it!
I'd crunch in CUDA on my MBP but the interface lag is too great...perhaps overnight, we'll see.
Glad this is up and running again! |
|
|
|
|
|
Just compared granted credits for my HD4670 with Collatz during the PPS Sieve pause:
Collatz about 295 cr/1000 sec runtime
PPS Sieve about 242-247 cr/1000 sec runtime
Conclusion: Even with 1800 cr/wu PPS Sieve results in less credits than Collatz for ATI cards. But as OpenCL is slower than a native library client, this drawback is justified. |
|
|
|
|
|
Amazing overall speed. last workunit I downloaded is at 788T, where did we start 3 days ago? 210T ?
Is there any prediction where optimal depth will be at? |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
Actually only 1G-units are calculated but i see more than 4000 2G-units at http://www.primegrid.com/stats_pps_sieve.php
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
|
I really like the credit at 1800 per work unit because that makes the work units worthwhile to crunch on much older and crappier cards, e.g.: the ones in MBP4,1.
____________
 |
|
|
|
|
|
I think you misunderstand the range statistics:
Shown are the Ranges 0P to 1P and 1P to 2P.
1P = 1000T = 1000000G
So what the page shows is that range 0P to 1P (up to 1000T) will be finished soon and new Workunits from range 1P to 2P (above 1000T up to 2000T) are generated. |
|
|
|
|
I really like the credit at 1800 per work unit because that makes the work units worthwhile to crunch on much older and crappier cards, e.g.: the ones in MBP4,1.
Tell me that means MacBook Pro four point one - given I went to the trouble of making my mid-2010 MacBook a CUDA-enabled optimized crunching machine, I'd love to help on this new sieve.
Then again there is no OSX app for this sieve. Yet. |
|
|
|
|
|
The mac apps have been build and will hopefully be uploaded to BOINC soon. |
|
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2165
ID: 1178
Credit: 8,777,295,508
RAC: 0
|
|
Okay...I have tried getting these to run by default on systems with HD4670's, 4650's, and 4550's across Vista 64-bit, Win 7 64-bit, and XP Media Center 32-bit using both 10.10 and 10.11 OpenCL versions. In all cases I get the error:
"No work available for the applications you have selected. Please chack your project preferences on the website."
That isn't the real issue though as all clients are configured to use the GPU. All web preferences for all venues have use both CUDA and ATI GPUs. I get the same error selecting the "work from any other project..." option or when selecting PPS Sieve directly. Am I missing something here??? I really do not want to be forced to run with an app_info file.
____________
141941*2^4299438-1 is prime!
|
|
|
|
|
I really like the credit at 1800 per work unit because that makes the work units worthwhile to crunch on much older and crappier cards, e.g.: the ones in MBP4,1.
Tell me that means MacBook Pro four point one - given I went to the trouble of making my mid-2010 MacBook a CUDA-enabled optimized crunching machine, I'd love to help on this new sieve.
Then again there is no OSX app for this sieve. Yet.
You are correct, but I run linux for crunching.
____________
|
|
|
|
|
|
Ok, cool. The GTX 260 I mentioned will be running on a machine with both Win7 and Ubuntu (on two separate HDs) once I get it set up, so there are still a lot of options open for me to contribute more to PrimeGrid (and GPU projects in general).
Iain - thanks, good to know. |
|
|
|
|
Okay...I have tried getting these to run by default on systems with HD4670's, 4650's, and 4550's across Vista 64-bit, Win 7 64-bit, and XP Media Center 32-bit using both 10.10 and 10.11 OpenCL versions. In all cases I get the error:
"No work available for the applications you have selected. Please chack your project preferences on the website."
That isn't the real issue though as all clients are configured to use the GPU. All web preferences for all venues have use both CUDA and ATI GPUs. I get the same error selecting the "work from any other project..." option or when selecting PPS Sieve directly. Am I missing something here??? I really do not want to be forced to run with an app_info file.
Same experience here the CUDA computer gets WUs by default with the use gpu setting. The ATI card systems get:
11/23/2010 4:45:21 PM PrimeGrid Message from server: No work sent
11/23/2010 4:45:21 PM PrimeGrid Message from server: _("No work available for the applications you have selected. Please check your project preferences on the web site.")
Is there a sample app_info.xml file for ATI on windows? Anyone getting the default non-app_info ATI cards to crunch?
____________
 |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
I think you misunderstand the range statistics:
Shown are the Ranges 0P to 1P and 1P to 2P.
1P = 1000T = 1000000G
So what the page shows is that range 0P to 1P (up to 1000T) will be finished soon and new Workunits from range 1P to 2P (above 1000T up to 2000T) are generated.
You are right. Now the 1P-units seems to be done and PG change to 2P-units...
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
The mac apps have been build and will hopefully be uploaded to BOINC soon.
Excellent!
Is there a recommended Mac driver version? I'm at 3.1.17, have been for a while; that's what I was running with up until the PPS sieve pause. The CUDA page in System Preferences is telling me that version 3.2.17 is available. Should I upgrade? Should I not? Does it matter? Thanks!
--Gary |
|
|
|
|
|
Due to a bug in the 3.1 driver on the Mac, the new CUDA app will require 3.2 - so you should upgrade to the latest whenever is convenient from you. BOINC won't sent WUs to hosts with older driver levels. |
|
|
|
|
|
What is with the CPU-app?
And how it the status about the Fermi-problem?
Swapped my ATI HD5770 with my old Nvidia GTX 260 with my colleague today, should be crunching this evening, so this one should not be affected by app_info.xml, but my GTX 460 in my other host is... |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
There are new CPU-appz uploaded for Mac, Linux and Windows in 32 and 64bit.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
A quick look at the applications page shows that all builds for the tpsieve app have been released excepet Mac CUDA...awaiting lib files for that. Also, still no build for Mac ATI (OpenCL).
The BOINC Fermi bug still remains. Only way to receive work is to use an app_info file. The ATI "no work" issues are probably related...again, use app_info to receive work. Sadly, no progress has been made on resolving either of these.
____________
|
|
|
|
|
|
Do i have to reset the project?
I just added the GTX 260 to my host and immediately trashed 8 CPU-WUs
EDITH says:
Same on the client with app_info.xml, downloaded the new 1.36-app
Mi 24 Nov 2010 18:40:35 CET PrimeGrid Sending scheduler request: To fetch work.
Mi 24 Nov 2010 18:40:35 CET PrimeGrid Requesting new tasks for CPU
Mi 24 Nov 2010 18:40:45 CET PrimeGrid Scheduler request completed: got 8 new tasks
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting pps_sr2sieve_4102478_0
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting task pps_sr2sieve_4102478_0 using pps_sr2sieve version 136
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting pps_sr2sieve_4102465_0
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting task pps_sr2sieve_4102465_0 using pps_sr2sieve version 136
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting pps_sr2sieve_4102461_0
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting task pps_sr2sieve_4102461_0 using pps_sr2sieve version 136
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting pps_sr2sieve_4102460_0
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting task pps_sr2sieve_4102460_0 using pps_sr2sieve version 136
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting pps_sr2sieve_4102458_0
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting task pps_sr2sieve_4102458_0 using pps_sr2sieve version 136
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting pps_sr2sieve_4102457_0
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting task pps_sr2sieve_4102457_0 using pps_sr2sieve version 136
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting pps_sr2sieve_4102455_0
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting task pps_sr2sieve_4102455_0 using pps_sr2sieve version 136
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting pps_sr2sieve_4102454_0
Mi 24 Nov 2010 18:40:47 CET PrimeGrid Starting task pps_sr2sieve_4102454_0 using pps_sr2sieve version 136
Mi 24 Nov 2010 18:40:48 CET PrimeGrid Computation for task pps_sr2sieve_4102478_0 finished
Mi 24 Nov 2010 18:40:48 CET PrimeGrid Output file pps_sr2sieve_4102478_0_0 for task pps_sr2sieve_4102478_0 absent
Mi 24 Nov 2010 18:40:49 CET PrimeGrid Computation for task pps_sr2sieve_4102465_0 finished
Mi 24 Nov 2010 18:40:49 CET PrimeGrid Output file pps_sr2sieve_4102465_0_0 for task pps_sr2sieve_4102465_0 absent
Mi 24 Nov 2010 18:40:50 CET PrimeGrid Computation for task pps_sr2sieve_4102461_0 finished
Mi 24 Nov 2010 18:40:50 CET PrimeGrid Output file pps_sr2sieve_4102461_0_0 for task pps_sr2sieve_4102461_0 absent
Mi 24 Nov 2010 18:40:51 CET PrimeGrid Computation for task pps_sr2sieve_4102460_0 finished
Mi 24 Nov 2010 18:40:51 CET PrimeGrid Output file pps_sr2sieve_4102460_0_0 for task pps_sr2sieve_4102460_0 absent
Mi 24 Nov 2010 18:40:52 CET PrimeGrid Computation for task pps_sr2sieve_4102458_0 finished
Mi 24 Nov 2010 18:40:52 CET PrimeGrid Output file pps_sr2sieve_4102458_0_0 for task pps_sr2sieve_4102458_0 absent
Mi 24 Nov 2010 18:40:53 CET PrimeGrid Computation for task pps_sr2sieve_4102457_0 finished
Mi 24 Nov 2010 18:40:53 CET PrimeGrid Output file pps_sr2sieve_4102457_0_0 for task pps_sr2sieve_4102457_0 absent
Mi 24 Nov 2010 18:40:54 CET PrimeGrid Computation for task pps_sr2sieve_4102455_0 finished
Mi 24 Nov 2010 18:40:54 CET PrimeGrid Output file pps_sr2sieve_4102455_0_0 for task pps_sr2sieve_4102455_0 absent
Mi 24 Nov 2010 18:40:55 CET PrimeGrid Computation for task pps_sr2sieve_4102454_0 finished
Mi 24 Nov 2010 18:40:55 CET PrimeGrid Output file pps_sr2sieve_4102454_0_0 for task pps_sr2sieve_4102454_0 absent
Something is afoot here.. |
|
|
pschoeferVolunteer developer
Volunteer tester
Send message
Joined: 20 Sep 05
Posts: 662
ID: 845
Credit: 2,220,370,221
RAC: 0
              
|
Do i have to reset the project?
I just added the GTX 260 to my host and immediately trashed 8 CPU-WUs
Still the same error with the app not supporting the -T switch. :|
I veryfied that the Win64 app I compiled is working with -T, but at least the Linux64 app seems to be an older version of tpsieve.
____________
  |
|
|
|
|
|
For me, it looks like x86 linux is working, but 64-bit not. |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
I posted a log from the tpsieve switches in ppsieve testing
Do we really need the -T switch ???
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
Do i have to reset the project?
I just added the GTX 260 to my host and immediately trashed 8 CPU-WUs
Still the same error with the app not supporting the -T switch. :|
I veryfied that the Win64 app I compiled is working with -T, but at least the Linux64 app seems to be an older version of tpsieve.
It's the executable from Ken's archive which contains the 0.3.10b version for 32 bit linux but the older 0.3.10a version for 64 bit linux. The 0.3.10b version is the one with added support for the (dummy) -T switch.
____________
|
|
|
|
|
|
is there a reason why the cuda wu's require 3/4 of a cpu core? could this be changed or would I need to run an app_info file? kind of a shame to have 3 out of 4 cores idle while my 4 cuda cards have all the fun. |
|
|
|
|
Do i have to reset the project?
I just added the GTX 260 to my host and immediately trashed 8 CPU-WUs
Still the same error with the app not supporting the -T switch. :|
I veryfied that the Win64 app I compiled is working with -T, but at least the Linux64 app seems to be an older version of tpsieve.
It's the executable from Ken's archive which contains the 0.3.10b version for 32 bit linux but the older 0.3.10a version for 64 bit linux. The 0.3.10b version is the one with added support for the (dummy) -T switch.
I think it is bad for the server to trash that much WUs, i see the load is between 2 and 4 now. |
|
|
|
|
|
All of my PPS Sieve CPU work units are getting errors immediately.
Linux 9.10
____________
|
|
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2165
ID: 1178
Credit: 8,777,295,508
RAC: 0
|
The BOINC Fermi bug still remains. Only way to receive work is to use an app_info file. The ATI "no work" issues are probably related...again, use app_info to receive work. Sadly, no progress has been made on resolving either of these.
Any possibility of providing app_info files for various platforms (e.g., Win64, Win32, etc.)? A Linux64 version is available in a number crunching thread, but it would be nice to have some example files placed in a centralized location given these issues.
A note on the applications page that the ATI apps cannot run via default BOINC install would also seem in order.
____________
141941*2^4299438-1 is prime!
|
|
|
|
|
|
So far, at 5 hours on a HT P4 3.0 GHz (32bit), I have a task that is at 14%. Is that a normal run time? Will it net me 1800 credits?
I might be able to get 2 of these out for the challenge. It will be close. I may have to turn HT off to get one out. |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
The BOINC Fermi bug still remains. Only way to receive work is to use an app_info file. The ATI "no work" issues are probably related...again, use app_info to receive work. Sadly, no progress has been made on resolving either of these.
Any possibility of providing app_info files for various platforms (e.g., Win64, Win32, etc.)? A Linux64 version is available in a number crunching thread, but it would be nice to have some example files placed in a centralized location given these issues.
A note on the applications page that the ATI apps cannot run via default BOINC install would also seem in order.
I suppose this is a necessity now. I was hoping there'd be a solution already. However, with the next Challenge less than a month a way, allowing as many people to join it is a priority. Therefore, I'll create a centralized location for app_info files. However, I must rely on y'all to provide the files. :)
PM me the files with details for what projects they include.
____________
|
|
|
Ken_g6Volunteer developer
Send message
Joined: 4 Jul 06
Posts: 915
ID: 3110
Credit: 183,164,814
RAC: 0
|
|
I still suspect a plan class of "ati_opencl" might be helpful. I also wonder if the same CUDA apps distributed under a "cuda_23" class could be distributed under a "cuda_fermi" class, and if that would help?
____________
|
|
|
|
|
|
Whatever will work. My computer (ID 163849) got two sieve tasks but they're CPU ones that look to need close to 10 hours. Not a problem exactly, but I CAN use my CUDA GPU instead. I didn't check though...those WUs may have been resends or something. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
Whatever will work. My computer (ID 163849) got two sieve tasks but they're CPU ones that look to need close to 10 hours. Not a problem exactly, but I CAN use my CUDA GPU instead. I didn't check though...those WUs may have been resends or something.
Even if they're resends of a WU originally sent to a CPU, they can still go to your GPU. So it's a problem (and a well known one, at that.) A WU isn't tied to a particular platform, so you could have, say, a 32-bit MAC CPU result be sent to a 64-bit LINUX GPU as a resend.
____________
My lucky number is 75898524288+1 |
|
|
|
|
So far, at 5 hours on a HT P4 3.0 GHz (32bit), I have a task that is at 14%. Is that a normal run time? Will it net me 1800 credits?
I might be able to get 2 of these out for the challenge. It will be close. I may have to turn HT off to get one out.
My P4 did one in 30 hours and did receive 1800 credits. That is really good for a P4 and I could have done two at once. |
|
|
|
|
|
It was a really long time since I looked at it, but at the time P4 (netburst family in general) were about 1/3 the clock efficiency (without HT) at sieve compared to other CPU types. I don't know if sieve code improvements have found more performance out of them since then. They were clock competitive at LLR for their age at least. |
|
|
|
|
|
Thanks Michael.
germane: I've a P4 Prescott 3.0GHz (HT) running SGS LLR and PPS LLR - takes not quite twice as long as my i5 (but close) - usually 40-ish minutes for SGS and 20 to 30 for PPS. No sieving on this, that's for sure. Odd thing about the older Pentiums, especially the ones from 2005 when HT was a big deal, is that you've got one core no matter what, and can choose (in BIOS) whether or not you want two virtual cores or not. I have it enabled, and it changes nothing performance-wise. Odd thing is that sse2 apps will run on either one or two cores and won't say something like "Running (1.95 CPUs)"...That said, it IS a Pentium chip from 2005. |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
We have reached the range 2P - 3P.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
|
What range is our goal for now?
____________
Polish National Team |
|
|
