Incorrect CPU time?

Message boards : Number crunching : Incorrect CPU time?

Author	Message
Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29432 - Posted: 13 Apr 2013 \| 1:57:22 UTC
	Curious as to why my 660Ti runs Nathan tasks with 2344secs of CPU time, yet my 560Ti is showing over 18 000. I understand it is a slower card, but that is a huge difference. I am wondering if BOINC is not reading correctly because I don't think that the card is actually using that much CPU
	ID: 29432 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29433 - Posted: 13 Apr 2013 \| 2:01:57 UTC
	hmmm. according to Task Manager, it is using that much CPU. Very strange, and that GPU is not the one connected to the monitor either...
	ID: 29433 \| Rating: 0 \| rate: / Reply Quote

skgiven Volunteer moderator Volunteer tester Send message Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level Scientific publications	Message 29436 - Posted: 13 Apr 2013 \| 12:20:56 UTC - in response to Message 29433.
	"Curious as to why my 660Ti runs Nathan tasks with 2344secs of CPU time, yet my 560Ti is showing over 18 000." You got that back to front; it's your GTX660Ti that's using 18000s of CPU time. One of my Long Nathan_dhfr36 WU's on a GTX660Ti: I54R3-NATHAN_dhfr36_5-3-32-RND8170_0 4359051 13 Apr 2013 \| 5:52:36 UTC 13 Apr 2013 \| 11:09:27 UTC Completed and validated 18,283.69 18,255.97 70,800.00 Long runs (8-12 hours on fastest card) v6.18 (cuda42) Your GTX660Ti is a Kepler and the other is a Fermi. Very different architectures. The researchers have deemed it necessary to allocate a full CPU core to these specific tasks. They don't do it for every type of task. ____________ FAQ's HOW TO: - Opt out of Beta Tests - Ask for Help
	ID: 29436 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29439 - Posted: 13 Apr 2013 \| 14:40:05 UTC - in response to Message 29436.
	"Curious as to why my 660Ti runs Nathan tasks with 2344secs of CPU time, yet my 560Ti is showing over 18 000." You got that back to front; it's your GTX660Ti that's using 18000s of CPU time. One of my Long Nathan_dhfr36 WU's on a GTX660Ti: I54R3-NATHAN_dhfr36_5-3-32-RND8170_0 4359051 13 Apr 2013 \| 5:52:36 UTC 13 Apr 2013 \| 11:09:27 UTC Completed and validated 18,283.69 18,255.97 70,800.00 Long runs (8-12 hours on fastest card) v6.18 (cuda42) Your GTX660Ti is a Kepler and the other is a Fermi. Very different architectures. The researchers have deemed it necessary to allocate a full CPU core to these specific tasks. They don't do it for every type of task. Thanks, that would also explain why I had to free up another core to keep the core usage steady. It's amazing the variation in technologies; the Kepler is 50% faster to complete the unit, but requires 7 or 8 times as much CPU to drive that performance. I've been away from this project for a while, it seems the workunits have gotten a little shorter than what I remember. I quite like this though, they seem to be quite stable and returning a nice chunk of points too :)
	ID: 29439 \| Rating: 0 \| rate: / Reply Quote

dskagcommunity Send message Joined: 28 Apr 11 Posts: 460 Credit: 841,848,862 RAC: 1,608,462 Level Scientific publications	Message 29440 - Posted: 13 Apr 2013 \| 15:16:40 UTC Last modified: 13 Apr 2013 \| 15:18:46 UTC
	That seems strange because my 560TI (384 cores, Gaming System) use only ~1400 CPU Time (E8400, no dedicated cores) and the other 560TI (448 cores, BOINC only System) uses only ~3000secs (in an ultraslow old Pentium4 System!! that kicks out 270k RAC ;)) ____________ DSKAG Austria Research Team: http://www.research.dskag.at
	ID: 29440 \| Rating: 0 \| rate: / Reply Quote

ExtraTerrestrial Apes Volunteer moderator Volunteer tester Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications	Message 29443 - Posted: 13 Apr 2013 \| 20:45:27 UTC - in response to Message 29440.
	Why would it seems strange? Simba's got an i7 and probably has HT enabled, so each core is relatively weak. MrS ____________ Scanning for our furry friends since Jan 2002
	ID: 29443 \| Rating: 0 \| rate: / Reply Quote

skgiven Volunteer moderator Volunteer tester Send message Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level Scientific publications	Message 29445 - Posted: 13 Apr 2013 \| 20:49:33 UTC - in response to Message 29440. Last modified: 13 Apr 2013 \| 20:54:24 UTC
	How saturated the CPU's were could have influenced things. Also, Simba123's CPU is an i7-2600K and each core has two threads (and each thread is perceived by Boinc as being a core). Your E8400 has a full 3GHz CPU core and 6MB cache to play with. Your 448 is CC2.0, while the 336 is CC2.1. Doesn't make for easy comparison, but it's interesting. A big part of the research here is getting the code to work on different types of GPU. At present there are the somewhat legacy CC1.3 cards, CC2.0 and CC2.1 Fermi's (GF400's & GF500's), the GF600 Kepler's (CC3.0), and they are trying to facilitate the Titans (CC3.5). Merely enabling new cards is part of the challenge. Just missed MrS's input. Would add that a 4 core with HT is ~1.5 times as fast as without (ballpark). Though comparing between CPU generations is tricky. ____________ FAQ's HOW TO: - Opt out of Beta Tests - Ask for Help
	ID: 29445 \| Rating: 0 \| rate: / Reply Quote

dskagcommunity Send message Joined: 28 Apr 11 Posts: 460 Credit: 841,848,862 RAC: 1,608,462 Level Scientific publications	Message 29446 - Posted: 13 Apr 2013 \| 21:00:33 UTC Last modified: 13 Apr 2013 \| 21:06:35 UTC
	I only wanted to say with that, i never had such big times (GPU Time = CPU Time) with any GPU Card and any Processor (tested both 560ti´s on the P4) nevertheless it was in use by CPU Units or not. on the older GPUGrid app (what was more CPU challenging when i remember right?) i had running CPU Units on the P4 HT Processor too with not much more CPU Time then now. Perhaps it is an XP/Win7 Difference ;) The only time i had such an issue on BOINC was a driverfailure with an ATI Card on Seti. Thats why i found it strange and not normal ^^ Its only my personal experience. ____________ DSKAG Austria Research Team: http://www.research.dskag.at
	ID: 29446 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29449 - Posted: 14 Apr 2013 \| 2:29:09 UTC
	For the curious, my system specs 17-2600K @ 4.5 - Hyperthreading is enabled Windows 7 Server Pack 1 64 Bit Nvidia driver version 314.07 The 660Ti is running at 1097/3019 (90-92% usage) The 560Ti is a 384 core version running at 918/2020 (93% usage) The CPU is running 6 Threads of Docking@home 2 cores free for GPUGrid. The CPU is not maxed in this configuration.
	ID: 29449 \| Rating: 0 \| rate: / Reply Quote

ExtraTerrestrial Apes Volunteer moderator Volunteer tester Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications	Message 29450 - Posted: 14 Apr 2013 \| 7:59:42 UTC - in response to Message 29446.
	@DSK that's because you don't have a Kepler GPU yet: SK, 3rd post wrote: Your GTX660Ti is a Kepler and the other is a Fermi. Very different architectures. The researchers have deemed it necessary to allocate a full CPU core to these specific tasks. They don't do it for every type of task [even if the GPU is a Kepler]. @Simba: I hope your question is answered by now? MrS ____________ Scanning for our furry friends since Jan 2002
	ID: 29450 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29451 - Posted: 14 Apr 2013 \| 10:08:12 UTC - in response to Message 29450.
	@DSK that's because you don't have a Kepler GPU yet: SK, 3rd post wrote: Your GTX660Ti is a Kepler and the other is a Fermi. Very different architectures. The researchers have deemed it necessary to allocate a full CPU core to these specific tasks. They don't do it for every type of task [even if the GPU is a Kepler]. @Simba: I hope your question is answered by now? MrS Still a little confused. So is the Kepler (660Ti) supposed to use more CPU time due to its' architecture, or is it the workunit that effects this, or both. It's annoying that, as far as I can see, there is no way to actually tell from the returned results list, which card they were completed on. I have a unit on the 560Ti about to finish, I'll make note of it before it uploads so I can double check the results. i57r8-nathan_dhfr36_5-5-32-rnd2437_0 is currently running on my 560 i27r74-nathan_dhfr36_3-29-32RND1789_0 in on the 660 (another 1.5hrs on that one)
	ID: 29451 \| Rating: 0 \| rate: / Reply Quote

skgiven Volunteer moderator Volunteer tester Send message Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level Scientific publications	Message 29452 - Posted: 14 Apr 2013 \| 10:35:40 UTC - in response to Message 29451.
	So is the Kepler (660Ti) supposed to use more CPU time due to its' architecture, or is it the workunit that effects this, or both. Both; The Kepler's will use a full CPU core for some task types; the NATHAN_dhfr36 and NATHAN_RPS1_respawn3 WU's, but not the NOELIA_Klebe_Equ WU's. I would expect most future WU types to use a full CPU core, but not necessary all. It's annoying that, as far as I can see, there is no way to actually tell from the returned results list, which card they were completed on. Run Time is a dead give away - your GTX660Ti is a good bit faster than your GTX560Ti, but yes it would be nice to have a more verbose Stderr output. ____________ FAQ's HOW TO: - Opt out of Beta Tests - Ask for Help
	ID: 29452 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29453 - Posted: 14 Apr 2013 \| 12:02:19 UTC
	The Times are in! The 560 http://www.gpugrid.net/result.php?resultid=6754491 660 to come soon
	ID: 29453 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29454 - Posted: 14 Apr 2013 \| 13:09:40 UTC - in response to Message 29453.
	The Times are in! The 560 http://www.gpugrid.net/result.php?resultid=6754491 660 to come soon http://www.gpugrid.net/result.php?resultid=6744685 seems to confirm things for this workunit type at least. Thanks for your replies. on an unrelated note, just passed 16 Million credits :)
	ID: 29454 \| Rating: 0 \| rate: / Reply Quote

Post to thread

Message boards : Number crunching : Incorrect CPU time?

	About	Science	Volunteers	Performance	Forum	Join us	Donate

Author	Message
Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29432 - Posted: 13 Apr 2013 \| 1:57:22 UTC
	Curious as to why my 660Ti runs Nathan tasks with 2344secs of CPU time, yet my 560Ti is showing over 18 000. I understand it is a slower card, but that is a huge difference. I am wondering if BOINC is not reading correctly because I don't think that the card is actually using that much CPU
	ID: 29432 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29433 - Posted: 13 Apr 2013 \| 2:01:57 UTC
	hmmm. according to Task Manager, it is using that much CPU. Very strange, and that GPU is not the one connected to the monitor either...
	ID: 29433 \| Rating: 0 \| rate: / Reply Quote

skgiven Volunteer moderator Volunteer tester Send message Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level Scientific publications	Message 29436 - Posted: 13 Apr 2013 \| 12:20:56 UTC - in response to Message 29433.
	"Curious as to why my 660Ti runs Nathan tasks with 2344secs of CPU time, yet my 560Ti is showing over 18 000." You got that back to front; it's your GTX660Ti that's using 18000s of CPU time. One of my Long Nathan_dhfr36 WU's on a GTX660Ti: I54R3-NATHAN_dhfr36_5-3-32-RND8170_0 4359051 13 Apr 2013 \| 5:52:36 UTC 13 Apr 2013 \| 11:09:27 UTC Completed and validated 18,283.69 18,255.97 70,800.00 Long runs (8-12 hours on fastest card) v6.18 (cuda42) Your GTX660Ti is a Kepler and the other is a Fermi. Very different architectures. The researchers have deemed it necessary to allocate a full CPU core to these specific tasks. They don't do it for every type of task. ____________ FAQ's HOW TO: - Opt out of Beta Tests - Ask for Help
	ID: 29436 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29439 - Posted: 13 Apr 2013 \| 14:40:05 UTC - in response to Message 29436.
	"Curious as to why my 660Ti runs Nathan tasks with 2344secs of CPU time, yet my 560Ti is showing over 18 000." You got that back to front; it's your GTX660Ti that's using 18000s of CPU time. One of my Long Nathan_dhfr36 WU's on a GTX660Ti: I54R3-NATHAN_dhfr36_5-3-32-RND8170_0 4359051 13 Apr 2013 \| 5:52:36 UTC 13 Apr 2013 \| 11:09:27 UTC Completed and validated 18,283.69 18,255.97 70,800.00 Long runs (8-12 hours on fastest card) v6.18 (cuda42) Your GTX660Ti is a Kepler and the other is a Fermi. Very different architectures. The researchers have deemed it necessary to allocate a full CPU core to these specific tasks. They don't do it for every type of task. Thanks, that would also explain why I had to free up another core to keep the core usage steady. It's amazing the variation in technologies; the Kepler is 50% faster to complete the unit, but requires 7 or 8 times as much CPU to drive that performance. I've been away from this project for a while, it seems the workunits have gotten a little shorter than what I remember. I quite like this though, they seem to be quite stable and returning a nice chunk of points too :)
	ID: 29439 \| Rating: 0 \| rate: / Reply Quote

dskagcommunity Send message Joined: 28 Apr 11 Posts: 460 Credit: 841,848,862 RAC: 1,608,462 Level Scientific publications	Message 29440 - Posted: 13 Apr 2013 \| 15:16:40 UTC Last modified: 13 Apr 2013 \| 15:18:46 UTC
	That seems strange because my 560TI (384 cores, Gaming System) use only ~1400 CPU Time (E8400, no dedicated cores) and the other 560TI (448 cores, BOINC only System) uses only ~3000secs (in an ultraslow old Pentium4 System!! that kicks out 270k RAC ;)) ____________ DSKAG Austria Research Team: http://www.research.dskag.at
	ID: 29440 \| Rating: 0 \| rate: / Reply Quote

ExtraTerrestrial Apes Volunteer moderator Volunteer tester Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications	Message 29443 - Posted: 13 Apr 2013 \| 20:45:27 UTC - in response to Message 29440.
	Why would it seems strange? Simba's got an i7 and probably has HT enabled, so each core is relatively weak. MrS ____________ Scanning for our furry friends since Jan 2002
	ID: 29443 \| Rating: 0 \| rate: / Reply Quote

skgiven Volunteer moderator Volunteer tester Send message Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level Scientific publications	Message 29445 - Posted: 13 Apr 2013 \| 20:49:33 UTC - in response to Message 29440. Last modified: 13 Apr 2013 \| 20:54:24 UTC
	How saturated the CPU's were could have influenced things. Also, Simba123's CPU is an i7-2600K and each core has two threads (and each thread is perceived by Boinc as being a core). Your E8400 has a full 3GHz CPU core and 6MB cache to play with. Your 448 is CC2.0, while the 336 is CC2.1. Doesn't make for easy comparison, but it's interesting. A big part of the research here is getting the code to work on different types of GPU. At present there are the somewhat legacy CC1.3 cards, CC2.0 and CC2.1 Fermi's (GF400's & GF500's), the GF600 Kepler's (CC3.0), and they are trying to facilitate the Titans (CC3.5). Merely enabling new cards is part of the challenge. Just missed MrS's input. Would add that a 4 core with HT is ~1.5 times as fast as without (ballpark). Though comparing between CPU generations is tricky. ____________ FAQ's HOW TO: - Opt out of Beta Tests - Ask for Help
	ID: 29445 \| Rating: 0 \| rate: / Reply Quote

dskagcommunity Send message Joined: 28 Apr 11 Posts: 460 Credit: 841,848,862 RAC: 1,608,462 Level Scientific publications	Message 29446 - Posted: 13 Apr 2013 \| 21:00:33 UTC Last modified: 13 Apr 2013 \| 21:06:35 UTC
	I only wanted to say with that, i never had such big times (GPU Time = CPU Time) with any GPU Card and any Processor (tested both 560ti´s on the P4) nevertheless it was in use by CPU Units or not. on the older GPUGrid app (what was more CPU challenging when i remember right?) i had running CPU Units on the P4 HT Processor too with not much more CPU Time then now. Perhaps it is an XP/Win7 Difference ;) The only time i had such an issue on BOINC was a driverfailure with an ATI Card on Seti. Thats why i found it strange and not normal ^^ Its only my personal experience. ____________ DSKAG Austria Research Team: http://www.research.dskag.at
	ID: 29446 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29449 - Posted: 14 Apr 2013 \| 2:29:09 UTC
	For the curious, my system specs 17-2600K @ 4.5 - Hyperthreading is enabled Windows 7 Server Pack 1 64 Bit Nvidia driver version 314.07 The 660Ti is running at 1097/3019 (90-92% usage) The 560Ti is a 384 core version running at 918/2020 (93% usage) The CPU is running 6 Threads of Docking@home 2 cores free for GPUGrid. The CPU is not maxed in this configuration.
	ID: 29449 \| Rating: 0 \| rate: / Reply Quote

ExtraTerrestrial Apes Volunteer moderator Volunteer tester Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications	Message 29450 - Posted: 14 Apr 2013 \| 7:59:42 UTC - in response to Message 29446.
	@DSK that's because you don't have a Kepler GPU yet: SK, 3rd post wrote: Your GTX660Ti is a Kepler and the other is a Fermi. Very different architectures. The researchers have deemed it necessary to allocate a full CPU core to these specific tasks. They don't do it for every type of task [even if the GPU is a Kepler]. @Simba: I hope your question is answered by now? MrS ____________ Scanning for our furry friends since Jan 2002
	ID: 29450 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29451 - Posted: 14 Apr 2013 \| 10:08:12 UTC - in response to Message 29450.
	@DSK that's because you don't have a Kepler GPU yet: SK, 3rd post wrote: Your GTX660Ti is a Kepler and the other is a Fermi. Very different architectures. The researchers have deemed it necessary to allocate a full CPU core to these specific tasks. They don't do it for every type of task [even if the GPU is a Kepler]. @Simba: I hope your question is answered by now? MrS Still a little confused. So is the Kepler (660Ti) supposed to use more CPU time due to its' architecture, or is it the workunit that effects this, or both. It's annoying that, as far as I can see, there is no way to actually tell from the returned results list, which card they were completed on. I have a unit on the 560Ti about to finish, I'll make note of it before it uploads so I can double check the results. i57r8-nathan_dhfr36_5-5-32-rnd2437_0 is currently running on my 560 i27r74-nathan_dhfr36_3-29-32RND1789_0 in on the 660 (another 1.5hrs on that one)
	ID: 29451 \| Rating: 0 \| rate: / Reply Quote

skgiven Volunteer moderator Volunteer tester Send message Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level Scientific publications	Message 29452 - Posted: 14 Apr 2013 \| 10:35:40 UTC - in response to Message 29451.
	So is the Kepler (660Ti) supposed to use more CPU time due to its' architecture, or is it the workunit that effects this, or both. Both; The Kepler's will use a full CPU core for some task types; the NATHAN_dhfr36 and NATHAN_RPS1_respawn3 WU's, but not the NOELIA_Klebe_Equ WU's. I would expect most future WU types to use a full CPU core, but not necessary all. It's annoying that, as far as I can see, there is no way to actually tell from the returned results list, which card they were completed on. Run Time is a dead give away - your GTX660Ti is a good bit faster than your GTX560Ti, but yes it would be nice to have a more verbose Stderr output. ____________ FAQ's HOW TO: - Opt out of Beta Tests - Ask for Help
	ID: 29452 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29453 - Posted: 14 Apr 2013 \| 12:02:19 UTC
	The Times are in! The 560 http://www.gpugrid.net/result.php?resultid=6754491 660 to come soon
	ID: 29453 \| Rating: 0 \| rate: / Reply Quote

Simba123 Send message Joined: 5 Dec 11 Posts: 147 Credit: 69,970,684 RAC: 0 Level Scientific publications	Message 29454 - Posted: 14 Apr 2013 \| 13:09:40 UTC - in response to Message 29453.
	The Times are in! The 560 http://www.gpugrid.net/result.php?resultid=6754491 660 to come soon http://www.gpugrid.net/result.php?resultid=6744685 seems to confirm things for this workunit type at least. Thanks for your replies. on an unrelated note, just passed 16 Million credits :)
	ID: 29454 \| Rating: 0 \| rate: / Reply Quote