Example frequency 754.85 fastest WUs: 58 175 292 409
Am I correct?
Yes, you are correct! Also because a trough on the cycle graph has a gentle slope on either side of the very bottom values you quoted, you could expect sequence numbers in the range of say 48-68, 165-185, 282-302, 399-419 to be almost equally as fast as each other.
So if you were wanting to check out the affect of some tweaking of your machine and you just happened to have sequence numbers that were adjacent and within these ranges, you could safely assume that any significant change in crunch time was probably due to the tweak rather than the change in sequence number.
i lost 10% of speed with this new app. on work unit below 800mhz
Probably not, you were just unlucky. The runtime of the workunits vary up to +/- 10% and sometimes more, and the workunits you got with the new version happened to be near the top end of the runtime. You can read this thread to learn all the details about this phenomenon.
I guess that you PC is actually crunching 5...10% faster now, which means that the old app would have spent about 38,000 .. 40,000 seconds for these particular workunits, even tho it did others in just 33,000 seconds. Hang on in a while and you should see the new app crunching other workunits in about 30,000 seconds.
Testing this on BOINC 6.1.7 (official alpha).
I especially went for it to see what it did with the "graphics in a separate program ("BOINC APIv6")" part. The graphics are showing up all right, even when I didn't set the service to interact with the desktop on the local system account.
But that's all it does. The separate graphics thread takes up 100% of the CPU cycles. No progress is made in the actual task. Where it would checkpoint every minute or so without graphics on, it has now been doing nothing for the past 15 minutes that I have the graphics window open.
Close the graphics window and the 'normal' application takes over in full force.
The graphics application lives on Normal priority, whereas the cruncher application lives on Low. So that has to change.
The above test was done on my IGP (S3/VIA Unichrome with 64MB), because at the time my video card was giving me problems.
Yesterday I fixed my video card, an ATI 9600 Pro with 128MB DDR, AGP8x card. I got a couple of new capacitors, took the old ones off my card, soldered the new ones on and went to test the card. It plays Oblivion 1.204 and Painkiller with settings set to high without a hitch, so I guess that's fixed.
This moment I am testing the graphics, still on BOINC 6.1.7, without the graphics ability to use local system account with interaction with the desktop.
The computation thread is getting all the CPU cycles, whilst the graphics thread gets none. Graphics are showing, though.
This leads me to believe that BOINC 6 needs an external graphics card to be show the screen saver/graphics. Anything on-board will always take up the CPU, be it by half (see Kathryn's post) or by full (mine). Perhaps an early warning is in place?
Originally not planned for a release I now published the 4.26 as first "official" "separate graphics" App. I hope that we'll see the benefits of this approach in this App soon, i.e. graphics on Vista working and a decrease of computation errors as graphics problems shouldn't affect the computation any more.
BM
PS: BOINC currently doesn't compile with VS2008, I'm afraid we'll have to stick to VS2003 for a while.
Originally not planned for a release I now published the 4.26 as first "official" "separate graphics" App. I hope that we'll see the benefits of this approach in this App soon, i.e. graphics on Vista working and a decrease of computation errors as graphics problems shouldn't affect the computation any more.
As an example of "graphics problems shouldn't affect the computation any more", I would guess the following that happened to me while I was sleeping would be an example:
2/5/2008 2:45:13 AM|Einstein@Home|[error] einstein_S5R3 not responding to screensaver, requesting exit
2/5/2008 2:45:14 AM|Einstein@Home|Task h1_0779.45_S5R2__77_S5R3a_0 exited with zero status but no 'finished' file
2/5/2008 2:45:14 AM|Einstein@Home|If this happens repeatedly you may need to reset the project.
2/5/2008 2:45:14 AM|Einstein@Home|Restarting task h1_0779.45_S5R2__77_S5R3a_0 using einstein_S5R3 version 426
2/5/2008 4:49:06 AM|Einstein@Home|Computation for task h1_0779.45_S5R2__77_S5R3a_0 finished
As an example of "graphics problems shouldn't affect the computation any more", I would guess the following that happened to me while I was sleeping would be an example:
2/5/2008 2:45:13 AM|Einstein@Home|[error] einstein_S5R3 not responding to screensaver, requesting exit
2/5/2008 2:45:14 AM|Einstein@Home|Task h1_0779.45_S5R2__77_S5R3a_0 exited with zero status but no 'finished' file
2/5/2008 2:45:14 AM|Einstein@Home|If this happens repeatedly you may need to reset the project.
2/5/2008 2:45:14 AM|Einstein@Home|Restarting task h1_0779.45_S5R2__77_S5R3a_0 using einstein_S5R3 version 426
2/5/2008 4:49:06 AM|Einstein@Home|Computation for task h1_0779.45_S5R2__77_S5R3a_0 finished
This error is new to me. What the OS, Client version and installation type is this?
It's bad that in this case the Client terminates the App, but it does this gently and should simply restart it, instead of the graphics crashing the whole task.
As an example of "graphics problems shouldn't affect the computation any more", I would guess the following that happened to me while I was sleeping would be an example:
2/5/2008 2:45:13 AM|Einstein@Home|[error] einstein_S5R3 not responding to screensaver, requesting exit
2/5/2008 2:45:14 AM|Einstein@Home|Task h1_0779.45_S5R2__77_S5R3a_0 exited with zero status but no 'finished' file
2/5/2008 2:45:14 AM|Einstein@Home|If this happens repeatedly you may need to reset the project.
2/5/2008 2:45:14 AM|Einstein@Home|Restarting task h1_0779.45_S5R2__77_S5R3a_0 using einstein_S5R3 version 426
2/5/2008 4:49:06 AM|Einstein@Home|Computation for task h1_0779.45_S5R2__77_S5R3a_0 finished
This error is new to me. What the OS, Client version and installation type is this?
It's bad that in this case the Client terminates the App, but it does this gently and should simply restart it, instead of the graphics crashing the whole task.
BM
The "no finished file" message has been around for a while and in multiple projects. Mostly when I see it, I ignore it. The result validated without any problems, so dunno what the deal was... Thinking back, I may not have been sleeping... I may have been doing something on here... Let me think about that or see if I can reconstruct what was going on at the time...
As for the particulars, this was in XP Pro (32-bit) with BOINC 5.8.16 as a single-user...
Edit: No, I was not in bed yet. At the time I have a bunch of port 80 requests in my router log, so I was viewing web pages...
Also, here's what happened in the output for the result. As you said, it played nice and restarted just fine from the checkpoint. I don't know if this is a "problem".
[pre]
955, 956, 957, c
958, 2008-02-05 02:45:15.0000 [normal]: Built at: Jan 21 2008 15:44:15
2008-02-05 02:45:15.0000 [normal]: Start of BOINC application 'projects/einstein.phys.uwm.edu/einstein_S5R3_4.26_windows_intelx86.exe'.
2008-02-05 02:45:15.0156 [debug]: Set up communication with graphics process.
2008-02-05 02:45:15.3750 [debug]: Reading SFTs and setting up stacks ... done
2008-02-05 02:45:34.0937 [debug]: Successfully read checkpoint
2008-02-05 02:45:34.0937 [debug]: Total skypoints = 1203. Progress: 958,
$Revision: 1.82 $ OPT:0 SCV:9, SCTRIM:8
959, c
960, 961, 962, c
[/pre]
RE: Ah, thanks a lot. Will
)
Yes, you are correct! Also because a trough on the cycle graph has a gentle slope on either side of the very bottom values you quoted, you could expect sequence numbers in the range of say 48-68, 165-185, 282-302, 399-419 to be almost equally as fast as each other.
So if you were wanting to check out the affect of some tweaking of your machine and you just happened to have sequence numbers that were adjacent and within these ranges, you could safely assume that any significant change in crunch time was probably due to the tweak rather than the change in sequence number.
Cheers,
Gary.
i lost 10% of speed with this
)
i lost 10% of speed with this new app. on work unit below 800mhz
RE: i lost 10% of speed
)
Probably not, you were just unlucky. The runtime of the workunits vary up to +/- 10% and sometimes more, and the workunits you got with the new version happened to be near the top end of the runtime. You can read this thread to learn all the details about this phenomenon.
I guess that you PC is actually crunching 5...10% faster now, which means that the old app would have spent about 38,000 .. 40,000 seconds for these particular workunits, even tho it did others in just 33,000 seconds. Hang on in a while and you should see the new app crunching other workunits in about 30,000 seconds.
CU
Bikeman
ok. thanks for your reply!
)
ok. thanks for your reply!
Ageless wrote:Testing this on
)
The above test was done on my IGP (S3/VIA Unichrome with 64MB), because at the time my video card was giving me problems.
Yesterday I fixed my video card, an ATI 9600 Pro with 128MB DDR, AGP8x card. I got a couple of new capacitors, took the old ones off my card, soldered the new ones on and went to test the card. It plays Oblivion 1.204 and Painkiller with settings set to high without a hitch, so I guess that's fixed.
This moment I am testing the graphics, still on BOINC 6.1.7, without the graphics ability to use local system account with interaction with the desktop.
The computation thread is getting all the CPU cycles, whilst the graphics thread gets none. Graphics are showing, though.
This leads me to believe that BOINC 6 needs an external graphics card to be show the screen saver/graphics. Anything on-board will always take up the CPU, be it by half (see Kathryn's post) or by full (mine). Perhaps an early warning is in place?
Originally not planned for a
)
Originally not planned for a release I now published the 4.26 as first "official" "separate graphics" App. I hope that we'll see the benefits of this approach in this App soon, i.e. graphics on Vista working and a decrease of computation errors as graphics problems shouldn't affect the computation any more.
BM
PS: BOINC currently doesn't compile with VS2008, I'm afraid we'll have to stick to VS2003 for a while.
BM
RE: Originally not planned
)
As an example of "graphics problems shouldn't affect the computation any more", I would guess the following that happened to me while I was sleeping would be an example:
2/5/2008 2:45:13 AM|Einstein@Home|[error] einstein_S5R3 not responding to screensaver, requesting exit
2/5/2008 2:45:14 AM|Einstein@Home|Task h1_0779.45_S5R2__77_S5R3a_0 exited with zero status but no 'finished' file
2/5/2008 2:45:14 AM|Einstein@Home|If this happens repeatedly you may need to reset the project.
2/5/2008 2:45:14 AM|Einstein@Home|Restarting task h1_0779.45_S5R2__77_S5R3a_0 using einstein_S5R3 version 426
2/5/2008 4:49:06 AM|Einstein@Home|Computation for task h1_0779.45_S5R2__77_S5R3a_0 finished
RE: As an example of
)
This error is new to me. What the OS, Client version and installation type is this?
It's bad that in this case the Client terminates the App, but it does this gently and should simply restart it, instead of the graphics crashing the whole task.
BM
BM
RE: RE: As an example of
)
The "no finished file" message has been around for a while and in multiple projects. Mostly when I see it, I ignore it. The result validated without any problems, so dunno what the deal was... Thinking back, I may not have been sleeping... I may have been doing something on here... Let me think about that or see if I can reconstruct what was going on at the time...
As for the particulars, this was in XP Pro (32-bit) with BOINC 5.8.16 as a single-user...
Edit: No, I was not in bed yet. At the time I have a bunch of port 80 requests in my router log, so I was viewing web pages...
Also, here's what happened in the output for the result. As you said, it played nice and restarted just fine from the checkpoint. I don't know if this is a "problem".
[pre]
955, 956, 957, c
958, 2008-02-05 02:45:15.0000 [normal]: Built at: Jan 21 2008 15:44:15
2008-02-05 02:45:15.0000 [normal]: Start of BOINC application 'projects/einstein.phys.uwm.edu/einstein_S5R3_4.26_windows_intelx86.exe'.
2008-02-05 02:45:15.0156 [debug]: Set up communication with graphics process.
2008-02-05 02:45:15.3750 [debug]: Reading SFTs and setting up stacks ... done
2008-02-05 02:45:34.0937 [debug]: Successfully read checkpoint
2008-02-05 02:45:34.0937 [debug]: Total skypoints = 1203. Progress: 958,
$Revision: 1.82 $ OPT:0 SCV:9, SCTRIM:8
959, c
960, 961, 962, c
[/pre]
Well, this seems to now be
)
Well, this seems to now be the standard App? No announcement. I have all WU now running this App, and I didn't use the test App.
Although I do not notice much, if any speed increase, yet.