Message boards :
Number crunching :
Question about faster boxes and credit
Message board moderation
Author | Message |
---|---|
Pascal, K G Send message Joined: 3 Apr 99 Posts: 2343 Credit: 150,491 RAC: 0 |
Was checking pending saw this one , all three have return valid info but still pending. Now the 3 boxes are AMD 2500 = 53.85 credit, Intel 2.53 = 41.18 credit and Intel 1.8 = 34.03 credit....... I thought the slower box would get a higher credit or and I missing something. It seems that the AMD is the slowest as far as time is concerned, something just does not add up...... Disclaimer: I do not care about credits just checking to see if the system is working correctly. Result ID click for details Host Sent Received Server state explain Outcome explain Client state explain CPU time (sec) claimed credit granted credit 9296595 10878 22 Sep 2004 19:36:12 UTC 6 Oct 2004 6:33:02 UTC Over Success Done 17,928.30 53.85 pending 9296596 154788 22 Sep 2004 19:36:31 UTC 26 Sep 2004 0:39:40 UTC Over Success Done 17,716.56 34.03 pending 9296597 28811 22 Sep 2004 19:36:40 UTC 24 Sep 2004 20:37:26 UTC Over Success Done 14,282.45 41.18 pending Semper Eadem So long Paul, it has been a hell of a ride. Park your ego's, fire up the computers, Science YES, Credits No. |
mikey Send message Joined: 17 Dec 99 Posts: 4215 Credit: 3,474,603 RAC: 0 |
> Was checking pending saw this one , all three have return valid info but still > pending. Now the 3 boxes are AMD 2500 = 53.85 credit, Intel 2.53 = 41.18 > credit and Intel 1.8 = 34.03 credit....... I thought the slower box would get > a higher credit or and I missing something. It seems that the AMD is the > slowest as far as time is concerned, something just does not add up...... > > Disclaimer: I do not care about credits just checking to see if the system is > working correctly. > > > > > > Result ID > click for details Host Sent Received Server state > explain Outcome > explain Client state > explain CPU time (sec) > claimed credit granted credit > 9296595 10878 22 Sep 2004 19:36:12 UTC 6 Oct 2004 6:33:02 UTC Over Success > Done 17,928.30 53.85 pending > 9296596 154788 22 Sep 2004 19:36:31 UTC 26 Sep 2004 0:39:40 UTC Over Success > Done 17,716.56 34.03 pending > 9296597 28811 22 Sep 2004 19:36:40 UTC 24 Sep 2004 20:37:26 UTC Over Success > Done 14,282.45 41.18 pending > > I have 7 machines currently crunching for Boinc/Seti and their predicted requested credits according to Boinc View are: #1----2.4 Intel Celeron----63.49 #2----2.5 Intel Celeron----62.23 #3----1.6 Intel P4---------61.32 #4----AMD1700+-------------55.69 #5----AMD2200+-------------61.48 #6----AMD2800+-------------61.94 #7----AMD64 3200+----------60.89 The actual numbers vary per unit but these are the starting points for the credits. The AMD64 3200+ is the only one running 4.13, the rest are all running 4.12, flawlessly I might add! |
EclipseHA Send message Joined: 28 Jul 99 Posts: 1018 Credit: 530,719 RAC: 0 |
Faster boxes request more credit than slower boxes, and that's the way it's been since the beta.. The end result is that the faster boxes get far less credit for a WU if run comapred against two slower.... Also, slower doesn't mean processor... Let's say you have three identical boxes. Two have Linux and one has Winxx. They all crunch the same WU, yet the Linux boxes claim far less than the winxx box... The funny thing here, is all three boxes crunch the WU is just about the same time... |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
> Faster boxes request more credit than slower boxes, and that's the way it's > been since the beta.. The end result is that the faster boxes get far less > credit for a WU if run comapred against two slower.... > > Also, slower doesn't mean processor... > > Let's say you have three identical boxes. Two have Linux and one has Winxx. > They all crunch the same WU, yet the Linux boxes claim far less than the winxx > box... > > The funny thing here, is all three boxes crunch the WU is just about the same > time... These issues, as has been stated have been with us since the Beta project. I know they irk some people because it seems such a small thing to have fixed. But, since it does not affect the science I think it should be one of the last things on the list of things to do ... |
Pascal, K G Send message Joined: 3 Apr 99 Posts: 2343 Credit: 150,491 RAC: 0 |
> Faster boxes request more credit than slower boxes, and that's the way it's > been since the beta.. The end result is that the faster boxes get far less > credit for a WU if run comapred against two slower.... > > Also, slower doesn't mean processor... > > Let's say you have three identical boxes. Two have Linux and one has Winxx. > They all crunch the same WU, yet the Linux boxes claim far less than the winxx > box... > > The funny thing here, is all three boxes crunch the WU is just about the same > time... > I know it is not only speed of the processor, but it just looks funny that a AMD Mobile XP 2500 gets more points than a Intel 1.8 and they are both close in time.... I assume that AMD is a laptop, maybe, and they are slower than dirt....... As I look at Mikey's boxes I am getting points envy. his 1.6 out points my 1.8 by almost double.....I just checked my results and I never get any credits over 51 for my p4 3.0... and again as long as the scientific aspect is fine I do not care if I get points.........MY right arm is long and can pat myself on the back just fine lololol;o) #1----2.4 Intel Celeron----63.49 #2----2.5 Intel Celeron----62.23 #3----1.6 Intel P4---------61.32 #4----AMD1700+-------------55.69 #5----AMD2200+-------------61.48 #6----AMD2800+-------------61.94 #7----AMD64 3200+----------60.89 And I thank Mikey, Azwoody, and Mr Buck for trying to help an old mathmatics challenged dude understand...... Semper Eadem So long Paul, it has been a hell of a ride. Park your ego's, fire up the computers, Science YES, Credits No. |
John Cropper Send message Joined: 3 May 00 Posts: 444 Credit: 416,933 RAC: 0 |
> Let's say you have three identical boxes. Two have Linux and one has Winxx. > They all crunch the same WU, yet the Linux boxes claim far less than the winxx > box... > > The funny thing here, is all three boxes crunch the WU is just about the same > time... > Okay, can somebody explain why a Linux would claim far less than a Win box? Stewie: So, is there any tread left on the tires? Or at this point would it be like throwing a hot dog down a hallway? Fox Sunday (US) at 9PM ET/PT |
RandyC Send message Joined: 20 Oct 99 Posts: 714 Credit: 1,704,345 RAC: 0 |
> > Okay, can somebody explain why a Linux would claim far less than a Win box? > Because the Benchmarks used by Linux are different than the Windoze Benchmarks. |
Toby Send message Joined: 26 Oct 00 Posts: 1005 Credit: 6,366,949 RAC: 0 |
> > Okay, can somebody explain why a Linux would claim far less than a Win > box? > > > Because the Benchmarks used by Linux are different than the Windoze > Benchmarks. No... not different, just less optimized. If you compile the BOINC client yourself with more optimized settings, it will nearly double your benchmark scores and make them comparable to windows boxes which in turn will cause the client to claim higher credit. It is very simple to do. Instructions can be found here. Changes for your specific CPU might be required from the instructions on this website. A member of The Knights Who Say NI! For rankings, history graphs and more, check out: My BOINC stats site |
Ned Slider Send message Joined: 12 Oct 01 Posts: 668 Credit: 4,375,315 RAC: 0 |
> > No... not different, just less optimized. If you compile the BOINC client > yourself with more optimized settings, it will nearly double your benchmark > scores and make them comparable to windows boxes which in turn will cause the > client to claim higher credit. It is very simple to do. Instructions can be > found here. > Changes for your specific CPU might be required from the instructions on this > website. > Yes, that's correct. Also, the release of the new science (SETI) client for windows has compounded issues. For example, on my windows box, the old (fast) v4.03 client was requesting about 30-35 credits, but the new (slower) V4.05 is requesting ~60 credits because the time to process has nearly doubled. However, linux still uses the older (and faster) 4.02 client and only requests credit in the ~30 range with a benchmark optimised client or ~20 with the standard boinc client. With such a varying differential in requested credit depending on hardware/software combination used, it's no surprise it's difficult to make head or tail of what's going on. @ Paul D Buck - I fully appreciate your comments in threads discussing topics like this. But please remember, no one is neccessarily suggesting the developers put this at the top of their "to do list" or urgently spend time on this stuff. It's just that some of us are INTERESTED in understanding more about the project and are interested in discussing these things. I, personally, find this stuff interesting and would like to be able to discuss it freely and openly without people butting in every 5 minutes telling me it's irrelevent and there are more important issues. Yes, you're right, there indeed may well be more important issues, so please feel free to start a thread on them and discuss them there. I appreciate you're just trying to focus the discussions of this forum on to what you consider important but please try and allow those of us that are interested in these issues the freedom to discuss them in the process. Thank you :) Ned *** My Guide to Compiling Optimised BOINC and SETI Clients *** *** Download Optimised BOINC and SETI Clients for Linux Here *** |
Ned Slider Send message Joined: 12 Oct 01 Posts: 668 Credit: 4,375,315 RAC: 0 |
> grrr > i've been reading a few threads about credits , i still canot work it out > > system here is a dual xeon 2Ghz , its crunching 4 units at once and requesting > ~37 credits for each > all still pending so far > > ive just updated to 4.13 and the time thay take has gone up from 9hrs to 10 > hrs but the credits requested still seem to be in the same range about 37 > > only been runing boinc for 2 to 3 days > > so how do you rate machine performance from the credit values ? it seems > better to use the processor time in seconds as a indicator of that > > dave > Requested credit is a product of CPU time taken x benchmark score. If you look at your benchmark scores, you'll notice they are quite low relative to other peoples. That's why your machine is not requesting as much credit. I haven't looked at it in any detail but I did notice Pentium 4 and Xeon chips seem to give relatively low benchmark scores. OS also has an effect with linux scores being lower than those for Windows on the same hardware so be careful when making comparisons. Ned *** My Guide to Compiling Optimised BOINC and SETI Clients *** *** Download Optimised BOINC and SETI Clients for Linux Here *** |
Pascal, K G Send message Joined: 3 Apr 99 Posts: 2343 Credit: 150,491 RAC: 0 |
> > > > No... not different, just less optimized. If you compile the BOINC > client > > yourself with more optimized settings, it will nearly double your > benchmark > > scores and make them comparable to windows boxes which in turn will cause > the > > client to claim higher credit. It is very simple to do. Instructions > can be > > found here. > > > Changes for your specific CPU might be required from the instructions on > this > > website. > > > > Yes, that's correct. > > Also, the release of the new science (SETI) client for windows has compounded > issues. For example, on my windows box, the old (fast) v4.03 client was > requesting about 30-35 credits, but the new (slower) V4.05 is requesting ~60 > credits because the time to process has nearly doubled. However, linux still > uses the older (and faster) 4.02 client and only requests credit in the ~30 > range with a benchmark optimised client or ~20 with the standard boinc client. > With such a varying differential in requested credit depending on > hardware/software combination used, it's no surprise it's difficult to make > head or tail of what's going on. > > @ Paul D Buck - I fully appreciate your comments in threads discussing topics > like this. But please remember, no one is neccessarily suggesting the > developers put this at the top of their "to do list" or urgently spend time on > this stuff. It's just that some of us are INTERESTED in understanding more > about the project and are interested in discussing these things. I, > personally, find this stuff interesting and would like to be able to discuss > it freely and openly without people butting in every 5 minutes telling me it's > irrelevent and there are more important issues. Yes, you're right, there > indeed may well be more important issues, so please feel free to start a > thread on them and discuss them there. I appreciate you're just trying to > focus the discussions of this forum on to what you consider important but > please try and allow those of us that are interested in these issues the > freedom to discuss them in the process. Thank you :) > > Ned > Hey Ned you think this benchmark to low or OK...P4 3.0 --- - 2004-10-14 21:07:17 - Number of CPUs: 1 --- - 2004-10-14 21:07:17 - 1565 double precision MIPS (Whetstone) per CPU --- - 2004-10-14 21:07:17 - 4556 integer MIPS (Dhrystone) per CPU --- - 2004-10-14 21:07:17 - Finished CPU benchmarks --- - 2004-10-14 21:07:18 - Resuming computation and network activity> Semper Eadem So long Paul, it has been a hell of a ride. Park your ego's, fire up the computers, Science YES, Credits No. |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
Ned, > @ Paul D Buck - I fully appreciate your comments in threads discussing topics > like this. But please remember, no one is neccessarily suggesting the > developers put this at the top of their "to do list" or urgently spend time on > this stuff. It's just that some of us are INTERESTED in understanding more > about the project and are interested in discussing these things. I, > personally, find this stuff interesting and would like to be able to discuss > it freely and openly without people butting in every 5 minutes telling me it's > irrelevent and there are more important issues. Yes, you're right, there > indeed may well be more important issues, so please feel free to start a > thread on them and discuss them there. I appreciate you're just trying to > focus the discussions of this forum on to what you consider important but > please try and allow those of us that are interested in these issues the > freedom to discuss them in the process. Thank you :) And nowhere in any of my posts did I suggest otherwise ... Though I will admit to some weariness with the topic ... :) If you go back to the beta site and look through the threads you can find discussion like this (which gives you a bunch of MySQL errors though the thread should come up) ... And I did put back up my page of benchmark <a>data[/url] that we developed during the Beta test ... Enjoy ... The pages won't be available for a little bit ... but they are uploading ... |
Ned Slider Send message Joined: 12 Oct 01 Posts: 668 Credit: 4,375,315 RAC: 0 |
> > And nowhere in any of my posts did I suggest otherwise ... > > Though I will admit to some weariness with the topic ... :) > > If you go back to the beta site and look through the threads you can find > discussion like <a> href="http://setiboinc.ssl.berkeley.edu/ap/forum_thread.php?id=3447">this[/url] > (which gives you a bunch of MySQL errors though the thread should come up) > ... > > And I did put back up my page of benchmark <a>data[/url] that we developed > during the Beta test ... > > Enjoy ... The pages won't be available for a little bit ... but they are > uploading ... > Paul, Really sorry if I sounded harsh - that was out of line. I was having a bad day, but excusses don't make up for it! I guess this is one of the disadvantages of a forum - the tone of the conversation is lost or misrepresented. I hope you accept my appologies :) Ned *** My Guide to Compiling Optimised BOINC and SETI Clients *** *** Download Optimised BOINC and SETI Clients for Linux Here *** |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
Ned, > Really sorry if I sounded harsh - that was out of line. I was having a bad > day, but excusses don't make up for it! > > I guess this is one of the disadvantages of a forum - the tone of the > conversation is lost or misrepresented. I hope you accept my appologies :) I know, and that is why I try real hard to stay straight ... I did see the emote and I knew you said it with a smile and I did not take it harshly... As *I* said (I know bold does not work well on the single i), elsewhere, I cannot always tell if *I* have made a statement that is out of line ... And, with my "limp" being more pronounced, well, it is easy for me to make a mistake ... Anyway, Ned, it is cool ... Ok, back to business ... I noted that my benchmark tests are now alive and posted. Maybe you folks can get something out of them we did not ... anyway, you can see what we tried in late Beta test with regard to HT and benchmark scores and the like. I tried to keep test conditions constant on the machines and ran the tests 10 times to try to avoid single run errors ... I would love it if someone can make sense out of it ... we never did ... Also, look at the discussion thread in the beta board, link is in my prior post. |
Pascal, K G Send message Joined: 3 Apr 99 Posts: 2343 Credit: 150,491 RAC: 0 |
> i seem to remember from some time a go when usung the old seti program > something about credits and them not beeing consistant with processor speed > > an explanation at the time went (from what i remember) something like :- > > higher credits wher given to slower machines as it gave uses of slower systems > a better feeling of worth to the project where those with faster systems had > better chances and apreciation for using seti@home .... > > but this was a couple of years ago before boinc > > bench marks are bench marks what one system gives another should give simlar > results (runing the same software) > > my system has xeon processors , its processing power is about equal to a p4 @ > 4Ghz but benchmarking multiple processors is difrent to single processors in > that it will crunch 4 work units at once it takes it 10hrs per work unit but > it is doing 4 if the program used all 4 cpus to work on the one it would > work out at 2.5hrs each > > anyway it gives :- > > --- - 2004-10-15 09:57:51 - Suspending computation and network activity - > running CPU benchmarks > --- - 2004-10-15 09:58:52 - Benchmark results: > --- - 2004-10-15 09:58:52 - Number of CPUs: 4 > --- - 2004-10-15 09:58:52 - 862 double precision MIPS (Whetstone) per CPU > --- - 2004-10-15 09:58:52 - 888 integer MIPS (Dhrystone) per CPU > --- - 2004-10-15 09:58:52 - Finished CPU benchmarks > > as for Pascal K, G > > and your results:- > >Hey Ned you think this benchmark to low or OK...P4 3.0 > >--- - 2004-10-14 21:07:17 - Number of CPUs: 1 > >--- - 2004-10-14 21:07:17 - 1565 double precision MIPS (Whetstone) per > CPU > >--- - 2004-10-14 21:07:17 - 4556 integer MIPS (Dhrystone) per CPU > >--- - 2004-10-14 21:07:17 - Finished CPU benchmarks > >--- - 2004-10-14 21:07:18 - Resuming computation and network activity > > this would tell me that you have used a compiler optimisation and recompiled > it > as the number of mips is grater than the processor speed you dont have a > machine running above 4.5Ghz. it's because the compiler will optimize the > program, and as the benchmark is a repeted loop, it omits the need to do > repeate the same calculation where the input and output is the same or the > instruction duplicates(that is the case for the Whetstone & Dhrystone > benchmarks) > and because of it the result is meningless as a valid benchmark > > incidently you could get a better result from a slower machine with better > compiler optimization's > > i had some source for the benchmark code some time ago , its commented that > this happens and for results to be a valid figure that you need to disable > compiler optimizations for this reson > > its not to say that optimising the code wont speed up the other parts of the > code not relating to the benchtests but you should build the benchmark > sections without optimizations and the rest with if you want acurate benchmark > data displayed > > yours > Dave > > > > > > HAve not done a thing, that score is right out of the box, I am running xp sp2 4.13 GUI. Semper Eadem So long Paul, it has been a hell of a ride. Park your ego's, fire up the computers, Science YES, Credits No. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.