Advanced search

Message boards : Number crunching : NATHAN_KID WUs

Author Message
Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30168 - Posted: 21 May 2013 | 19:05:59 UTC
Last modified: 21 May 2013 | 19:06:56 UTC

Saw some new NATHAN WUs coming through a couple hours ago and said OH BOY because the last NATHANs ran so well on all my cards. So I dumped the other projects from my GTX 460/768MB GPUs and grabbed the new WUs. Not sure if it's that they won't work properly on < 1GB or what, but they are SLOW and the projected time is well over 24 hours on all 4 GPUs. What happened? The previous NATHANs ran in 10-11 hours on the GTX 460/768MB. Ouch :-(

Profile dskagcommunity
Avatar
Send message
Joined: 28 Apr 11
Posts: 456
Credit: 817,865,789
RAC: 0
Level
Glu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30186 - Posted: 22 May 2013 | 6:32:16 UTC

Hm i will need around 14 hours on 570s with min. 96% gpu load. So these are really long units ^^
____________
DSKAG Austria Research Team: http://www.research.dskag.at



Crunching for my deceased Dog who had "good" Braincancer..

Profile Mumak
Avatar
Send message
Joined: 7 Dec 12
Posts: 92
Credit: 225,897,225
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 30189 - Posted: 22 May 2013 | 8:58:36 UTC

I'm currently running one such on 660 Ti - 75% after 10 hours.

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 30197 - Posted: 22 May 2013 | 12:00:36 UTC - in response to Message 30189.
Last modified: 22 May 2013 | 12:02:42 UTC

Long runs (8-12 hours on fastest card)

Presently, the fastest single GPU card is the GTX680.
On XP the GTX680 takes the following times for different Long WU's:

SDOERR_2HDQd ~25,700sec (7.1h) 135,000
NOELIA_klebe ~27,500sec (7.6h) 127,000 (yes, there is a slight credit disparity, but you could argue that the SDOERR WU's are the ones that overpay)
NATHAN_KIDc22 ~36,500sec (10.1h) 167,550 (smack in the middle of the 8-10h estimate)

NATHAN_dhfr36 ~13,600sec (3.8h) 70,800 {these are the old NATHAN WU's and would normally have been in the short queue if there wasn't a shortage of WU's}
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Profile Mumak
Avatar
Send message
Joined: 7 Dec 12
Posts: 92
Credit: 225,897,225
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 30200 - Posted: 22 May 2013 | 14:01:43 UTC
Last modified: 22 May 2013 | 14:58:20 UTC

My results on 660 Ti (Boost @ 1228 MHz):

NATHAN_dhfr36: 19,700 s
SDOERR_2HDQd: ~36,700 s
NOELIA_klebe_run2: ~39,300 s
NATHAN_KIDc22: ~48,100 s

klepel
Send message
Joined: 23 Dec 09
Posts: 189
Credit: 3,552,598,778
RAC: 91,768
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30204 - Posted: 22 May 2013 | 14:50:29 UTC

My first result on EVGA GTX 670 SC @ AMD 8150 (One Core reserved for the GPU):
http://www.gpugrid.net/result.php?resultid=6892235
I13R6-NATHAN_KIDc22_2-0-8-RND0204_0
Time GPU: 52042.51
Time CPU: 51079.20
It seems to me quite long in comparison to other times posted.

On the other hand the new task on the same machine:
http://www.gpugrid.net/result.php?resultid=6892718
I60R3-NATHAN_KIDc22_2-0-8-RND5457_0 (Advanced 36% of total task, 5 h 7 m)
On EVGA OC Scanner X: The GPU POwer drops to ca. 96% TDP after restart, as I found it this morning at ca. 64% TDP.

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30206 - Posted: 22 May 2013 | 15:02:37 UTC - in response to Message 30197.
Last modified: 22 May 2013 | 15:03:35 UTC

Presently, the fastest single GPU card is the GTX680.
On XP the GTX680 takes the following times for different Long WU's:
NATHAN_KIDc22 ~36,500sec (10.1h) 167,550 (smack in the middle of the 8-10h estimate)

Realistically, how many real world people are running the fastest GPU on the fastest but now unsupported (arguably obsolete) operating system? Only a few crunchers specializing in this one project. When they finally get the titan running are we going to see another huge leap in WU times? Wouldn't it be better for total work speed/throughput to include more crunchers and keep them happier? Keeping the WU sizes reasonable and perhaps relaxing the 24hr time a bit would go a long way towards doing that IMO.

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 30216 - Posted: 22 May 2013 | 18:01:21 UTC - in response to Message 30206.

Linux is the fastest OS, then XP. 11 of the top 20 machines use Linux or XP. So many of the top/elite crunchers appear to go out of their way to accommodate the project (and their credits). Perhaps this will continue.

The Linux/xp popularity drops when you expand to the top 250 systems, it's 65/250 - so 26%. However it remains at that even when you go past the 1000th top system - from 1000 to 1100 its 27% Linux and XP.

I have no doubt XP usage will continue to fall with time, but Linux seems to be picking up a lot of the slack (vaguely remember looking at this in the past).

The non project elite crunchers tend to use the less productive Operating systems and this continues through to the occasional cruncher and crunchers with bad setups or entry level cards.

While these lesser systems still belong to crunchers, how much work they do/can do, how important is it to accommodate them, and how much effort it takes is really a question that the project team has to continuously ask itself.

Ideally we could all crunch what we wanted, without problems, but this is more of a hands on project that is always changing. While I don't have the answers (just opinions)t I do have some data that's worth noting:

GPUGrid's total daily RAC is 191,926,972.
The top 20 systems contribute 20,507,084 (just over 10%)

The top 20 crunchers contribute 43,518,798 (23.9%)
The top 100 crunchers contribute 92,956,103 (48%)

The 1000th cruncher has a RAC of ~22,000 and by the time you get to 2000, the RAC is only ~1200. You could get a RAC of >22K by just running one long WU every 6th day.

There is two ways to look at this data, depending on human resources; we don't have a big enough team to accommodate everybody so it's best to concentrate on those that help the most. Our team is big enough to develop and expand to better accommodate more crunchers and new crunchers.

I doubt that the researchers will immediately go by the non-mainstream GK110 Titan when setting task steps against complexity (which equates to task duration on the top GPU), though it is a GeForce GPU. At present the number of crunchers attached to this project with a Titan is likely to be a single figure. It would make more sense to go by the GK104 GTX680 and then the GK104 GTX770 (which might be 10 or 20% percent faster), after there are plenty of crunchers who have the GPU.

____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30266 - Posted: 23 May 2013 | 15:58:37 UTC - in response to Message 30197.
Last modified: 23 May 2013 | 15:59:21 UTC

On XP the GTX680 takes the following times for different Long WU's:

SDOERR_2HDQd ~25,700sec (7.1h) 135,000
NOELIA_klebe ~27,500sec (7.6h) 127,000 (yes, there is a slight credit disparity, but you could argue that the SDOERR WU's are the ones that overpay)
NATHAN_KIDc22 ~36,500sec (10.1h) 167,550 (smack in the middle of the 8-10h estimate)

NATHAN_dhfr36 ~13,600sec (3.8h) 70,800 {these are the old NATHAN WU's and would normally have been in the short queue if there wasn't a shortage of WU's}

Then again, maybe the SDOERR WUs are actually longer than the NOELIAs but run more efficiently. Looking at my GPUs the NOELIA WUs are running at 88-89% usage and the SDOERR WUs are running at 92-94% usage.

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 30273 - Posted: 23 May 2013 | 19:04:47 UTC - in response to Message 30266.

Then again, maybe the SDOERR WUs are actually longer than the NOELIAs but run more efficiently. Looking at my GPUs the NOELIA WUs are running at 88-89% usage and the SDOERR WUs are running at 92-94% usage.

That's the case; the more complex the model the less efficient it becomes. Of course the credit argument could be expanded to entail WU complexity, or lack of. Anyway, their model of assigning credit is reasonably accurate, ~14% difference throughout the range of long WU's, which includes 4 different research lines/models.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30283 - Posted: 23 May 2013 | 23:49:07 UTC

Or they could award credits alphabetically.

matlock
Send message
Joined: 12 Dec 11
Posts: 34
Credit: 86,423,547
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwat
Message 30284 - Posted: 24 May 2013 | 1:34:34 UTC

My last NATHAN_KIDc22 WU completed in 48,862 seconds:
http://www.gpugrid.net/workunit.php?wuid=4476442

This is with an Asus GTX 660 and Linux using the 304.88 drivers.

From looking at other results with Win7, it seems you would need a 660 Ti to get this running 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 30285 - Posted: 24 May 2013 | 9:38:55 UTC - in response to Message 30284.

XP is 11% faster than Vista, W7 and probably W8 (though I haven't measured it).
Linux is faster than XP (up to 5% if it's setup correctly, but it might vary a lot on the system).
I've seen a GTX660Ti on Linux that is 25% faster than my FOC GTX660Ti on W7.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Stefan
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 5 Mar 13
Posts: 348
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 30289 - Posted: 24 May 2013 | 12:20:49 UTC - in response to Message 30285.

skgiven, do you maybe have an idea if the performance difference is project dependent? So do other projects (and other gpu projects) also run that much faster on XP, in which case it would be an OS issue (I know that Win 7 is in general a bit slower than XP), or does this only happen with GPUgrid?

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30297 - Posted: 24 May 2013 | 15:17:02 UTC - in response to Message 30289.
Last modified: 24 May 2013 | 15:18:52 UTC

skgiven, do you maybe have an idea if the performance difference is project dependent? So do other projects (and other gpu projects) also run that much faster on XP, in which case it would be an OS issue (I know that Win 7 is in general a bit slower than XP), or does this only happen with GPUgrid?

If I may jump in. Sometimes there is a small difference between XP and W7 on other projects, but generally not and the difference is not always in the favor of XP. Between Linux and W7: it's a mixed bag on other projects with W7 coming out on top as often as not. As far as the 25% listed above, no way IMO. You can generally tell the GPU efficiency by the percent of utilization (87-88% for me in W7-64 on NOELIA & NATHAN WUs). A 25% difference would most likely be due to something like an extreme OC on an exceptional card and perhaps good liquid cooling. Many other projects running CAL or CUDA run at 99% usage. OpenCL is often less and tends to need more CPU support than CUDA or CAL.

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 30313 - Posted: 24 May 2013 | 19:29:28 UTC - in response to Message 30297.
Last modified: 24 May 2013 | 19:32:23 UTC

Beyond has covered the comparison to other projects.

The issue here is with the Windows Display Driver Model (WDDM) in Vista, W7, W8 and 2008+ servers. XP and Linux are not affected. It might be the case that there is even more of a gap now between XP and Linux when using CUDA 4.2.

From using MSI Afterburner it's obvious that Windows basically grabs a chunk of the display GPU's GDDR and keeps that for itself. It may control other GPU resources too. Don't know if this is a set/fixed amount, based on GPU model, CC or GDDR amount. It may be the case that if there are 2 GPU's in use, the one supporting the monitor under performs - but you would really need two identical GPU's and careful WU monitoring to determine that. Perhaps using the on-die GPU's prevents this 11% loss, but I can't test that (you need a motherboard that supports that to test it).

It was the Vista WDDM that introduced the idea of being able to restart the display driver, without restarting the OS.

http://msdn.microsoft.com/en-us/library/aa480220.aspx

As for relative performances, I have a reasonably FOC on my GTX660Ti. Sometimes it runs at 1202MHz, but at present it's operating at 1189MHz. If Linux is 5% faster than XP and XP is 11% faster than W7, then that's 16.55% faster, so to get to 25% faster you just need to up that by 7% (1267 or 1280MHz, without a GDDR increase). Readily achievable using liquid cooling, and possible with a GDDR tweak on air, with a good GPU. Another possibility is that on Linux the GPU utilization and power target is higher; 95% GPU usage over 88% is an 8% increase.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Profile Mumak
Avatar
Send message
Joined: 7 Dec 12
Posts: 92
Credit: 225,897,225
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 30315 - Posted: 24 May 2013 | 20:48:29 UTC

Higher latency can be caused by WDDM, since drivers are split between kernel and user mode (unlike XP, where all drivers were in kernel only). But exact numbers depend on particular implementation (of CUDA especially).
Another performance loss might be caused by using the Aero interface.

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 30321 - Posted: 24 May 2013 | 21:54:47 UTC

I suspect it can be called a difference of "more closer to metal" with XP and Linux, whereas more abstraction in WDDM allows advanced features at the expense of performance.

SK wrote:
the more complex the model the less efficient it becomes

I think it'S the other way around. That's why GPU utilization on short queue tasks drops. The more atoms / pixels the task contains, the easier it is to keep more shaders busy concurrently.

MrS
____________
Scanning for our furry friends since Jan 2002

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 30414 - Posted: 26 May 2013 | 13:32:46 UTC

Hi guys,

I have completed two NATHAN_KIDc22 WUs and wanted to ask you guys how you find my GTX 650Ti's times:

Received: 25 May 2013 11:18:10 UTC WU: I65R8-NATHAN_KIDc22_2-1-8-RND5776_0 Run Time: 81,090.74 Credit: 167,550.00
Received: 26 May 2013 09:50:00 UTC WU: I29R3-NATHAN_KIDc22_2-2-8-RND5999_2 Run Time: 81,103.17 Credit: 139,625.00

Note: Boinc 7.0.65 on Ubuntu 12.04 x86_64 with Nvidia driver 319.17

Looking at some more powerful cards' times, I believe the times are just about where they should be, am I correct?

Also, isn't the credit difference a little bit strange? The WUs look to be the same type and the run times are virtually the same, shouldn't the credit be the same as well? Except if they just adjusted the credit gain.

flashawk
Send message
Joined: 18 Jun 12
Posts: 297
Credit: 3,572,627,986
RAC: 0
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 30415 - Posted: 26 May 2013 | 14:58:28 UTC - in response to Message 30414.
Last modified: 26 May 2013 | 15:25:20 UTC

Received: 26 May 2013 09:50:00 UTC WU: I29R3-NATHAN_KIDc22_2-2-8-RND5999_2 Run Time: 81,103.17 Credit: 139,625.00


I've never seen that point award before for any NATHAN_KIDc22's, 135,000 is the amount I get for completing a SDOERR_2HDQd. That's strange, is it from the short queue perhaps? I haven't done any of those in quit sometime.

Edit: I took a look at the wu and it was sent on the 25th, I thought you had still managed to complete it within the 24 hour bonus period, you made the 48 hour bonus though.

Profile Mumak
Avatar
Send message
Joined: 7 Dec 12
Posts: 92
Credit: 225,897,225
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 30416 - Posted: 26 May 2013 | 15:18:12 UTC

My runtime on a 650Ti (slightly OCed) is 76,764s, credit 167,550.
On a 660Ti they complete in ~48,100s, credit is always the same.

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 30420 - Posted: 26 May 2013 | 17:22:22 UTC - in response to Message 30415.

I've never seen that point award before for any NATHAN_KIDc22's, 135,000 is the amount I get for completing a SDOERR_2HDQd. That's strange, is it from the short queue perhaps? I haven't done any of those in quit sometime.

Edit: I took a look at the wu and it was sent on the 25th, I thought you had still managed to complete it within the 24 hour bonus period, you made the 48 hour bonus though.

Yes, I figured that out myself after posting. I missed the 24h bonus by 2 hours. The WU had finished, but the client didn't report it in time.. Maybe the 0.2+0.2 cache rule is a bit on the edge for my 650Ti. I adjusted it to 0.12+0.28, let's see how that goes.

The WU was from the long queue, I only take from the long queue.

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 30421 - Posted: 26 May 2013 | 17:28:42 UTC - in response to Message 30416.

My runtime on a 650Ti (slightly OCed) is 76,764s, credit 167,550.
On a 660Ti they complete in ~48,100s, credit is always the same.

My 650Ti is stock-clocked. Can your OC be responsible for the ~4500sec difference?

Not that I'm dieing to squeeze the latest ounce of performance out of the card, I just want it to perform as it should.

flashawk
Send message
Joined: 18 Jun 12
Posts: 297
Credit: 3,572,627,986
RAC: 0
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 30423 - Posted: 26 May 2013 | 17:36:05 UTC - in response to Message 30420.

That should do the trick, others say they set theirs to 0.1 and it works good too. The NATHAN_KIDc22 are the longest running at this time (for me anyway), so you should be able to do any of the 3 in under 24 hours with you're 650Ti.

Profile Mumak
Avatar
Send message
Joined: 7 Dec 12
Posts: 92
Credit: 225,897,225
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 30425 - Posted: 26 May 2013 | 18:59:14 UTC - in response to Message 30421.

My 650Ti is stock-clocked. Can your OC be responsible for the ~4500sec difference?

Not that I'm dieing to squeeze the latest ounce of performance out of the card, I just want it to perform as it should.


Indeed, it can. Moreover, the 650Ti is running on XP, which performs slightly better.

Post to thread

Message boards : Number crunching : NATHAN_KID WUs

//