Advanced search

Message boards : Graphics cards (GPUs) : New application in preparation

Author Message
Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 20198 - Posted: 20 Jan 2011 | 19:13:54 UTC

The changes will be mostly on usability for GPUGRID users. Planned changes (we cannot guarantee to have them all, but we will try):

1) Project preference for running the app at higher priority than normal (higher speed)

2) Project preference for using 1 core to control the GPU (equivalent to current SWAN_SYNC=0)

3) Having two applications, acemd_long and acemd. acemd_long will have longer workunits, you can opt out of it in project preferences. Longer WUs will give relatively more credits per hour.

4) All new WU will be shipped with the same options of the current DHFR1000, so they will be generally faster.

5) Higher GPU usage in general (closer to 99%)

Planned release in mid February.

GDF

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20201 - Posted: 20 Jan 2011 | 21:06:42 UTC - in response to Message 20198.

That sounds really nice - thumbs up!

MrS
____________
Scanning for our furry friends since Jan 2002

Tom Philippart
Send message
Joined: 12 Feb 09
Posts: 57
Credit: 23,376,686
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 20223 - Posted: 21 Jan 2011 | 22:14:27 UTC

great to hear!

MarkJ
Volunteer moderator
Volunteer tester
Send message
Joined: 24 Dec 08
Posts: 738
Credit: 200,909,904
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20224 - Posted: 22 Jan 2011 | 6:03:42 UTC - in response to Message 20198.

The changes will be mostly on usability for GPUGRID users. Planned changes (we cannot guarantee to have them all, but we will try):


I presume this will be a cuda 3.2 app? Meaning we'll need to make sure we have 260.89 or later drivers?
____________
BOINC blog

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 20240 - Posted: 25 Jan 2011 | 15:58:32 UTC - in response to Message 20224.

No. we will stay to cuda3.1.

gdf

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 20259 - Posted: 27 Jan 2011 | 21:58:05 UTC - in response to Message 20240.

Points 3),4),5) are now ready to go.

gdf

Profile [AF>Libristes>Jip] Elgran...
Avatar
Send message
Joined: 16 Jul 08
Posts: 45
Credit: 78,618,001
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20260 - Posted: 28 Jan 2011 | 10:56:17 UTC - in response to Message 20259.

Points 3),4),5) are now ready to go.

gdf

Great
____________

Profile Stoneageman
Avatar
Send message
Joined: 25 May 09
Posts: 224
Credit: 34,057,224,498
RAC: 231
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20262 - Posted: 28 Jan 2011 | 13:40:56 UTC - in response to Message 20260.

I see the 'long' option under GPUgrid preferences now. When will the long tasks be released?

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20264 - Posted: 28 Jan 2011 | 14:50:31 UTC - in response to Message 20262.

I guess they will start feeding them into the system when they make them. I completed an et today, so I expect they have tested to the extent that they are setup correctly and ready to go with large tasks.

It's worth noting that the present adc5-IBUCH tasks take almost twice as long as the KASHIF_HIVPR and more than twice as long as the GIANNI_DHFR1000 tasks, so I expect when new tasks are produced they will include the new and faster routines.

I have already set my GTX470's to accept long tasks, and set my GT240's to only pick up short tasks, so when the long tasks are released I won't have to do anything.

roundup
Send message
Joined: 11 May 10
Posts: 57
Credit: 1,425,368,693
RAC: 8,158,385
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20271 - Posted: 29 Jan 2011 | 8:53:19 UTC

Great to hear.
All of those planned changes make sense.
I am looking forward to the improved GPU usage and set my 460 and 470s to accept long tasks.

Rabinovitch
Avatar
Send message
Joined: 25 Aug 08
Posts: 143
Credit: 64,937,578
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 20272 - Posted: 29 Jan 2011 | 10:02:23 UTC

So what about GPUs like GTS250 or close to this? Not all of volunteers have GTX4xx, 5xx-class cards... But is their hardware completely useless for the project?
If only "short" WUs (at least) could run normally on non-top GPUs!.....
____________
From Siberia with love!

cristipurdel
Send message
Joined: 31 Mar 10
Posts: 45
Credit: 103,429,292
RAC: 0
Level
Cys
Scientific publications
watwatwatwat
Message 20281 - Posted: 29 Jan 2011 | 13:53:49 UTC

On my FX580 (32 shaders), adc5-IBUCH takes around 100h.
Will the time decrease when Point 2) will be available?

Hypernova
Send message
Joined: 16 Nov 10
Posts: 22
Credit: 24,712,746
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwat
Message 20289 - Posted: 30 Jan 2011 | 19:31:06 UTC - in response to Message 20264.


I have already set my GTX470's to accept long tasks, and set my GT240's to only pick up short tasks, so when the long tasks are released I won't have to do anything.


skgiven, how do you do that. I mean adapt the type of GPU to the length of the task. It would make sense indeed to switch long tasks to the most powerful GPUs, and the short ones to the others.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20291 - Posted: 31 Jan 2011 | 7:30:57 UTC - in response to Message 20289.
Last modified: 31 Jan 2011 | 8:37:37 UTC

In Your account, under GPUGRID preferences create two (or up to 4) profiles (default, home, work and school). Set one to crunch short tasks and the other to crunch long tasks. Add each of your systems to one of the profiles. You can do that by going to Your computers, clicking Details and changing from default.

It has not yet been announced that there are short and long tasks.

cristipurdel
Send message
Joined: 31 Mar 10
Posts: 45
Credit: 103,429,292
RAC: 0
Level
Cys
Scientific publications
watwatwatwat
Message 20333 - Posted: 2 Feb 2011 | 6:53:31 UTC - in response to Message 20281.

On my FX580 (32 shaders), adc5-IBUCH takes around 100h.
Will the time decrease when Point 2) will be available?

Actually I'm doing 67h for 141-GIANNI (with the SWAN_SYNC option working apparently). Any chance for smaller WUs or code improvements?

Profile SMTB1963
Avatar
Send message
Joined: 27 Jun 10
Posts: 38
Credit: 524,420,921
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 20342 - Posted: 4 Feb 2011 | 15:15:16 UTC - in response to Message 20198.

Well, something's definitely changed for the better - at least for my GTX275s. I hadn't run them since the beginning of January; they produced around 35-40K ppd each at that time. Started 'em back up the other day, and they now seem to be getting 50K+ ppd each. Utilization is higher on most (but not all) WUs for that machine. Too early to tell if the 570 has seen any increase yet.

I didn't make any changes to my preferences, but when I looked at them yesterday, the "long" option was already checked...is this by design? I would've opted in anyways, but shouldn't one have to proactively select this option?

In any case, great to see the improvements! Now if only something could be done about GTX460 performance... ;-)

Andrew3000
Send message
Joined: 1 Feb 10
Posts: 24
Credit: 1,220,848
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwatwat
Message 20358 - Posted: 7 Feb 2011 | 7:11:30 UTC - in response to Message 20342.
Last modified: 7 Feb 2011 | 7:11:49 UTC

I have a gainward gtx 460 768MB and the gpu usage is just horrible sometimes: 44%.
and it usually stays at around 60%.
the gpu usage problem for gtx 460 is really urgent and important because a lot of people will buy gtx 460 video cards. which would mean A LOT of power wasted.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20360 - Posted: 7 Feb 2011 | 9:19:15 UTC - in response to Message 20358.

Performance changes with different tasks, but this should generally improve soon*.

All CC2.1 cards, including the GTX460, have a somewhat incomplete ability at GPUGrid due to shader access problems, mostly down to the ACEMD application, I think. This is why I advise people to get CC2.0 cards if possible. Perhaps the remaining third of the shaders will be usable in a future ACEMD edition. It would be a very nice boost to the project.

It’s down to the project managers to say what is urgent for GPUGrid (an ATI app, new faster routines*, long tasks, server or website maintenance, support, presentations, publications...), and they can’t do much in some areas (tools and drivers) other than wait. Unfortunately this is the situation with CC2.1 cards.

All we crunchers can do is report problems, make suggestions and optimize our systems as best we can; using swan_sync=0, freeing up a CPU core for every GPU, configuring your preferences, reporting tasks immediately and using a fast operating system - due to the WDM overhead Vista and Win7 are slower than XP and Linux, though this may not be apparent from looking at GPU Utilization, but rather through task performance.


Good luck,

Andrew3000
Send message
Joined: 1 Feb 10
Posts: 24
Credit: 1,220,848
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwatwat
Message 20393 - Posted: 10 Feb 2011 | 8:27:50 UTC - in response to Message 20360.

what i know is that someone messed up big time(if it is nvidia then the gpu grid people should communicate the problem so that nvidia can fix it or give a solution how to go around the bug). and that this problem with the gpu usage needs to be fixed fast.

what i don't know is why games are able to use 100% of my card and gpu grid can't.

Andrew3000
Send message
Joined: 1 Feb 10
Posts: 24
Credit: 1,220,848
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwatwat
Message 20396 - Posted: 10 Feb 2011 | 18:16:01 UTC - in response to Message 20393.

on other thought... some of the games run with 60% gpu usage while having 40fps and that's odd.. so maybe this problem affects the games too and nvidia will solve it later.
another odd thing is that if i overclock from 650mhz to 850mhz the gtx 460 768mb when the game has 60%gpu the fps does not increase.

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 20437 - Posted: 14 Feb 2011 | 17:38:19 UTC - in response to Message 20396.

Did anybody of you tested if with higher priority was still beneficial to assign a full cpu core to the application (SWAN_SYNC)?

Are there any collateral effects in increasing the priority to high-priority (like slow response of the desktop)?

thanks,
gdf

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2343
Credit: 16,201,255,749
RAC: 7,520
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20441 - Posted: 14 Feb 2011 | 18:41:57 UTC - in response to Message 20437.

Did anybody of you tested if with higher priority was still beneficial to assign a full cpu core to the application (SWAN_SYNC)?

I tried increasing ACEMD's priority before I knew about the SWAN_SYNC environmental setting. Boosting ACEMD's priority in itself increased minimally the GPU usage and decreased the running time. (It was a previous version of ACEMD) Then, I searched the forum for performance increasing tips, and I learned about the SWAN_SYNC setting. I haven't turned it off since then. Using SWAN_SYNC=1 and increasing ACEMD's priority at the same time helps ACEMD to maintain the GPU usage, regardless of other tasks or applications I'm using. If I don't boost ACEMD's priority, the GPU usage drops when I start other applications (and it jumps back to normal very soon).

Are there any collateral effects in increasing the priority to high-priority (like slow response of the desktop)?

I've increased acemd's priority to "above normal". Everything seems to be fine, except watching movies (especially HD movies).

Profile nenym
Send message
Joined: 31 Mar 09
Posts: 137
Credit: 1,308,230,581
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20444 - Posted: 14 Feb 2011 | 19:02:09 UTC - in response to Message 20437.
Last modified: 14 Feb 2011 | 19:04:43 UTC

Are there any collateral effects in increasing the priority to high-priority (like slow response of the desktop)?
If I set ACEMD to normal, desktop had time to time slow response - up to 2s, and PT i had to set to high priority. If I set ACEMD to high priority, desktop was sluggish. (GTX560Ti factory OC to 900MHz, XP 64bit, Swan_Sync=0 set, one CPU core free out of Boinc, GIANNI_DHFR task - GPU utilization 99%).

Werkstatt
Send message
Joined: 23 May 09
Posts: 121
Credit: 321,525,386
RAC: 177,358
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20445 - Posted: 14 Feb 2011 | 19:20:10 UTC - in response to Message 20437.
Last modified: 14 Feb 2011 | 19:21:55 UTC

Did anybody of you tested if with higher priority was still beneficial to assign a full cpu core to the application (SWAN_SYNC)?

Are there any collateral effects in increasing the priority to high-priority (like slow response of the desktop)?

thanks,
gdf

I did some experiments with the app_info.
I kept one cpu free with the setting
<avg_ncpus>1</avg_ncpus>
<max_ncpus>1</max_ncpus>
This had the best effect.

Setting the priority to 'above normal' helped when the original setting of 0.24 was applied (~2% faster) but had nearly no effect with one cpu for the app.
When GPU-usage is below 80% the highest increase of speed can be achieved when running two apps together with a setting of <count>0.5</count> and the <avg_ncpus> also to 0.5
This setting brought an overall increase of more than 10%, but is senseless for wu's with gpu-usage of 90% or better.
I would not prefer the higher priority because I use my system for daily work, where I would otherwise see a performance decrease.
Edit: I forgot: win7x64 E8400 6GB 6.12.13

Kirby54925
Send message
Joined: 21 Jan 11
Posts: 31
Credit: 70,061,988
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 20446 - Posted: 14 Feb 2011 | 19:34:38 UTC

When I increased the priority of the GPUGrid task, my computer's regular performance wasn't affected. I could still do day-to-day things such as going online, watching YouTube, and playing music with no performance hits. However, when I checked how much CPU time was being used by GPUGrid, it's at 0%! To make matters worse, it seemed like it wasn't doing any work at all! This was when I set the number of processors to be used to 100%. Rosetta@Home then proceeds to use all four cores. What's interesting, though, is that even though Rosetta@Home has the lowest priority compared to GPUGrid, it still seems to hog nearly 100% of the CPU time in the fourth core.

Another thing to note is that I removed SWAN_SYNC, as it doesn't appear to do anything. There is no mention of SWAN synchronization in any of the stderr outputs I can see in my completed tasks. For the record, I'm using Linux Mint 10 (which is based on Ubuntu 10.10) with a Core i5-750 processor and a GTX 570 GPU.

Could you look into this GDF? I really want to crunch for both Rosetta@Home and GPUGrid at the same time.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20447 - Posted: 14 Feb 2011 | 22:04:50 UTC - in response to Message 20446.

Kirby54925, set your CPU usage to 75%. You have a GTX570, it needs to be able to use some of the CPU. If you let Rosetta hog all the CPU cores you will knobble the GTX570 performance. I would expect the difference between CPU Time and RunTime to also drop somewhat for your Rosetta tasks, so you would not quite be losing a full CPU core anyway.

GDF, I pushed the Priority to High on all four GT240's in the one system when not using swan_sync. It only increased GPU Utilizaion on one GPU by about 4% - none of the others changed. No problems with the system (also running CPU tasks at the time). I have not tried it when swan_sync is being used, but I see little point. Tasks were running between 65% and 90% utilization. The increase from using swan_sync now appears to be less than 10% for GT240’s running the 6.12app on Vista x64, while running CPU tasks.

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 20452 - Posted: 15 Feb 2011 | 19:28:44 UTC - in response to Message 20447.

So, we can focus on SWAN_SYNC alone and let people use it via project preferences?

gdf

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20453 - Posted: 15 Feb 2011 | 20:47:09 UTC - in response to Message 20452.

A preference for swan_sync would be useful, especially if you could do it for Linux.

It is already possible to reduce the number of CPU cores to use (for CPU projects) from Computing preferences (and Boinc Manager):
On multiprocessors, use at most - 100.0 % of the processors.

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 20457 - Posted: 16 Feb 2011 | 8:53:54 UTC - in response to Message 20453.

ok

Profile nenym
Send message
Joined: 31 Mar 09
Posts: 137
Credit: 1,308,230,581
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20459 - Posted: 16 Feb 2011 | 19:09:22 UTC
Last modified: 16 Feb 2011 | 19:19:42 UTC

Swan_Sync XOR one free core out of Boinc makes the same for GPU load (Win XP). By my point of view is more useful to use Swan_Sync than free core for extraordinary cases - e.g. GPUGRID is down, GPU is running backup project with low CPU utilization (i have set PG) and machine is not baby-sited. When I leave machine for longer than buffer of GPUGRID tasks has got, I switch Boinc from 75% CPU cores to 100% CPU cores. So I switch in the evening 75% to 100% and in the morning from 100% to 75%. Why? Swan_Sync AND one free core makes GPU load from +2% to 5% higher depending on tasks except GIANNI_DHFR as that kind of task makes GPU load 96% - 98% using only one of these features.

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 20475 - Posted: 19 Feb 2011 | 19:01:36 UTC - in response to Message 20459.

In the future most of the tasks will reach 96-98% load.

gdf

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20489 - Posted: 21 Feb 2011 | 9:49:05 UTC - in response to Message 20396.

another odd thing is that if i overclock from 650mhz to 850mhz the gtx 460 768mb when the game has 60%gpu the fps does not increase.


Sounds like you're running into a cpu or other limit.

MrS
____________
Scanning for our furry friends since Jan 2002

Profile robertmiles
Send message
Joined: 16 Apr 09
Posts: 503
Credit: 727,920,933
RAC: 155,858
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20508 - Posted: 25 Feb 2011 | 4:40:52 UTC - in response to Message 20360.

Performance changes with different tasks, but this should generally improve soon*.

All CC2.1 cards, including the GTX460, have a somewhat incomplete ability at GPUGrid due to shader access problems, mostly down to the ACEMD application, I think. This is why I advise people to get CC2.0 cards if possible. Perhaps the remaining third of the shaders will be usable in a future ACEMD edition. It would be a very nice boost to the project.

It’s down to the project managers to say what is urgent for GPUGrid (an ATI app, new faster routines*, long tasks, server or website maintenance, support, presentations, publications...), and they can’t do much in some areas (tools and drivers) other than wait. Unfortunately this is the situation with CC2.1 cards.

All we crunchers can do is report problems, make suggestions and optimize our systems as best we can; using swan_sync=0, freeing up a CPU core for every GPU, configuring your preferences, reporting tasks immediately and using a fast operating system - due to the WDM overhead Vista and Win7 are slower than XP and Linux, though this may not be apparent from looking at GPU Utilization, but rather through task performance.


Good luck,


What's the best CC2.0 card you can suggest that uses no more power than a GT240? It looks like I'll be getting another computer soon, but with a rather strong limit on the amount of power it can use.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20511 - Posted: 25 Feb 2011 | 15:48:32 UTC - in response to Message 20508.

There is none and this is a problem for the project.
The GT440 would perform no better than a GT240. Stay away from the GT430 and below (if you can); as far as I know these are cc2.1. Ditto for the GTS450, GTX460 and GTX560.
There is a very big gap between the GT240 and the GTX465 (the lowest powered CC2.0 Feri). The only CC2.0 cards are the GTX 465, 470, 570 and 580. These are all 200W+

Profile robertmiles
Send message
Joined: 16 Apr 09
Posts: 503
Credit: 727,920,933
RAC: 155,858
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20512 - Posted: 25 Feb 2011 | 16:04:44 UTC - in response to Message 20511.
Last modified: 25 Feb 2011 | 16:05:56 UTC

How long do you expect the CC2.1 problem to last, or do you expect it to be permanent even for new versions of the application?

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20514 - Posted: 25 Feb 2011 | 18:31:45 UTC - in response to Message 20512.

I am fairly sure the CC2.1 problem is here for the immediate to midterm future (0-6 months); I don’t know of any mid-range CC2.0 cards being developed and software changes take a long time and may never resolve the situation. Beyond 6 months there are some possible changes (program/kernel or app) that might improve CC2.1 performance but there is no guarantee. It would be wrong to get a GPU thinking that improvement is just around the corner.

Kirby54925
Send message
Joined: 21 Jan 11
Posts: 31
Credit: 70,061,988
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 20551 - Posted: 28 Feb 2011 | 23:21:57 UTC

Looks like the long tasks have finally arrived!

Profile Stoneageman
Avatar
Send message
Joined: 25 May 09
Posts: 224
Credit: 34,057,224,498
RAC: 231
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20555 - Posted: 1 Mar 2011 | 8:53:46 UTC

Yes they have, but so far all my completed ones are 'pending validation'. However on the server status page, it shows none are waiting for validation?

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 20556 - Posted: 1 Mar 2011 | 9:21:34 UTC - in response to Message 20555.

It should have been fixed now by Toni.

gdf

Kirby54925
Send message
Joined: 21 Jan 11
Posts: 31
Credit: 70,061,988
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 20558 - Posted: 1 Mar 2011 | 9:38:02 UTC
Last modified: 1 Mar 2011 | 9:44:22 UTC

Sweet! Thanks for the fix!

As for you Stoneageman, how's your electric bill? It must be really high to be running all those computers! (Not to mention that getting all those GTX 580s costs a pretty penny...)

EDIT: GDF, if I select both the ACEMD2 and ACEMDLONG tasks in my GPUGrid preferences, does the ACEMDLONG task have a higher priority?

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20566 - Posted: 1 Mar 2011 | 12:54:34 UTC - in response to Message 20558.

Tony’s long tasks seem to be running well and giving plenty of credit :) On my GTX470’s they complete in under 8h (XP x86, i7-920).

Run time 29089, Claimed credit 27186.875, Granted credit 40780.3125 task
Expected (24/7) daily credit from one (ref/stock) GTX470 = (86400/29089)*40780=120120

Compare that to this different Tony task, on the same GTX470:
Run time 10254, Granted credit 9283
Expected daily credit from one GTX470 = (86400/10254)*9283=78218

Just over 50% more credit.

Kirby54925
Send message
Joined: 21 Jan 11
Posts: 31
Credit: 70,061,988
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 20567 - Posted: 1 Mar 2011 | 13:01:51 UTC

Happy to report that they finish in about 6.5 hours on a GTX 570. And yes, they give quite a bit of credit. I compared the credit/sec for these new workunits compared to the highest-yielding standard workunit I finished. The long workunits yield about 1.7 credits/sec compared to the standard workunits, which only get around 0.8 credit/sec. That said, the queue for the long workunits is running low.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20569 - Posted: 1 Mar 2011 | 13:27:39 UTC - in response to Message 20567.

Their popularity seems to have risen sharply since yesterday; 200 in progress when I checked. Now there are over 1000 in progress, about 1/3 of the tasks.

Kirby54925
Send message
Joined: 21 Jan 11
Posts: 31
Credit: 70,061,988
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 20570 - Posted: 1 Mar 2011 | 13:35:05 UTC - in response to Message 20569.

But out of that thousand, there will probably be hundreds that will fail to complete the workunits. That should keep the queue from completely emptying out. However, there really should be a buffer of about a couple hundred workunits for both acemdlong and acemd2 at all times.

Toni
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 9 Dec 08
Posts: 1006
Credit: 5,068,599
RAC: 0
Level
Ser
Scientific publications
watwatwatwat
Message 20572 - Posted: 1 Mar 2011 | 13:39:22 UTC - in response to Message 20570.

I am still tuning the acemdlong tasks, se we don't have a steady supply of them yet (there will be soon). Glad to see that they worked well. :-)

Profile Stoneageman
Avatar
Send message
Joined: 25 May 09
Posts: 224
Credit: 34,057,224,498
RAC: 231
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20573 - Posted: 1 Mar 2011 | 13:43:34 UTC

Hi Toni. Just make sure after 'tuning', that if they error, they do it at the start : )

Kirby54925
Send message
Joined: 21 Jan 11
Posts: 31
Credit: 70,061,988
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 20574 - Posted: 1 Mar 2011 | 13:43:52 UTC - in response to Message 20572.

Awesome! Keep up the good work!

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 20575 - Posted: 1 Mar 2011 | 15:03:01 UTC - in response to Message 20574.

the higher credits should compensate for the fact that the probability to fail is higher due to longer running time.
If you have a stable machine, then it's all good.

gdf

ftpd
Send message
Joined: 6 Jun 08
Posts: 152
Credit: 328,250,382
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20580 - Posted: 1 Mar 2011 | 17:37:35 UTC
Last modified: 1 Mar 2011 | 17:38:34 UTC

Downloaded four long wu's. All cancelled after some time.

GTX295!

What cards are the best for these long wu's? Fermi?
____________
Ton (ftpd) Netherlands

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20583 - Posted: 1 Mar 2011 | 21:09:15 UTC - in response to Message 20580.

Long run tasks run on the 6.13app (cuda31). They should work for Fermi cards and CC1.3 cards such as a GTX295. I expect they are least likely to fail on the top Fermi cards.

For now, I would suggest you chose to only run the normal tasks on your GTX295 system and you might want to restart it. This might be a temporary issue so try again in a week or so.

Good luck,

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2343
Credit: 16,201,255,749
RAC: 7,520
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20591 - Posted: 2 Mar 2011 | 1:57:23 UTC

For these long tasks, running time is about 5h40m (20400s) on a GTX 480 (at 800 Mhz) and 5h10m (18600s) on a GTX 580 (at 850 Mhz). GPU utilization is 72% and 68%, respectively. CPU is a C2Q9650 at 4.0 GHz, SWAN_SYNC=0 applied.

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 20594 - Posted: 2 Mar 2011 | 8:24:52 UTC - in response to Message 20591.

Utilization will increase in new jobs, this was an old one made longer as we needed further results.

gdf

ben
Send message
Joined: 27 Oct 08
Posts: 19
Credit: 22,658,253
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwat
Message 20599 - Posted: 3 Mar 2011 | 11:41:17 UTC - in response to Message 20594.

when will we be getting the 99% utilization tasks again? march 4?

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20601 - Posted: 3 Mar 2011 | 13:12:27 UTC - in response to Message 20600.

We are introducing the possibility for users to select between shorter (acemd2, approx 4 hours on GTX275 cards) and longer (acemdlong, 8 to 12 hours on GTX480) tasks. Simply go to your account preference to select which applications you want to crunch for. Long task are of course credited higher, but GPUs with less cores cannot simply compute them in time. 31 Jan 2011

With respect to the short tasks, I am seeing tasks as short as 6h on a GT240 and as long as 20h on the same system, or 28h on another GT240 (Vista, without swan_sync=0). I guess different projects require different runtimes, and perhaps you were testing, but if possible closer runtimes would be much preferred.

Thanks,

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 20607 - Posted: 4 Mar 2011 | 8:47:52 UTC - in response to Message 20601.


We have released internally guidelines for short and long runs. Most likely a factor 2 difference is unavoidable between different workunits in the same queue.

gdf

Profile Fred J. Verster
Send message
Joined: 1 Apr 09
Posts: 58
Credit: 35,833,978
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20611 - Posted: 4 Mar 2011 | 15:45:41 UTC - in response to Message 20607.
Last modified: 4 Mar 2011 | 15:48:55 UTC


We have released internally guidelines for short and long runs. Most likely a factor 2 difference is unavoidable between different workunits in the same queue.

gdf


I noticed 1 long run WU, when I checked, I found this ,
WU , but isn't a
long runner?
They have Long-run in the name of the WU, IMHO.
Is the difference in CPU time, indicating, using SWAN_SYNC?
I use a X9650 @3510MHz CPU.
____________

Knight Who Says Ni N!

Profile Fred J. Verster
Send message
Joined: 1 Apr 09
Posts: 58
Credit: 35,833,978
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20614 - Posted: 4 Mar 2011 | 20:12:55 UTC - in response to Message 20611.
Last modified: 4 Mar 2011 | 20:15:13 UTC

The WU a long run, 8-12 hour on the fastest card is at 55,95% after
4 hours, so I'm curious how much time is really needed, in this case/rig?
(X9650@3510MHz; GPU=GTX480; XP64;BOINC 6.10.58 64BIT)
Running smooth, load is better, as before.
And the latest EINSTEIN CUDA tasks, which also have been optimized, have a better GPU load as before.

Running 3 SETI MB WU's at a time, gives a 99% load and very little CPU
footprint or time to load and unload, also report and upload.
____________

Knight Who Says Ni N!

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 20619 - Posted: 5 Mar 2011 | 9:33:53 UTC - in response to Message 20614.

Hi Fred, I thought these KASHIF_HIVPR_wo tasks were quite long too.

I see you missed out on the full credit bonus. Perhaps lowering your cache would help, and/or man-managing which projects you run on a day-day or week-week basis.

Not using swan_sync would explain some of the slowness (16.6% slower). On such a powerful card using swan_sync & freeing a CPU core becomes more important.

While your GPU is clocked at 1.4GHz the other card clocks in at 1.45GHz. That said, you might have a faster CPU and are using 64bit XP.

It also looks like your task also restarted about 5 times. Each time can lose up to 5min. Changing your processor usage, Switch between applications everyto a high number such as 999 should prevent this, or using Leave Applications in Memory (LAIM), if you don't already have it selected.

Your task run, 18,302.34 1,226.58 10,347.84 12,934.80 ACEMD2: GPU molecular dynamics v6.13 (cuda31)
Other cruncher, 15,688.01 15,616.50 10,347.84 15,521.76 ACEMD2: GPU molecular dynamics v6.13 (cuda31)

The task on two systems

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 21182 - Posted: 13 May 2011 | 10:42:14 UTC - in response to Message 20619.

So, we would like to update the applications to the newest version.
1) Some little speed improvement
2) Only cuda3.1 version
3) Some bug fixes
4) If we can the read of SWAN_SYNC as a boinc parameter.

g

Profile Carlesa25
Avatar
Send message
Joined: 13 Nov 10
Posts: 328
Credit: 72,619,453
RAC: 251
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21184 - Posted: 13 May 2011 | 16:35:38 UTC - in response to Message 21182.

Hi, Linux and Windows and are using Cuda 4 on Nvidia drivers. Greetings.

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 21187 - Posted: 14 May 2011 | 10:57:55 UTC - in response to Message 21184.

We will add cuda4 app when ready of course. Always keeping only two versions.
gdf

Post to thread

Message boards : Graphics cards (GPUs) : New application in preparation

//