Message boards :
Number crunching :
AK V8 ported release ap. issues, install, questions etc.
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 · 8 . . . 19 · Next
Author | Message |
---|---|
dnolan Send message Joined: 30 Aug 01 Posts: 1228 Credit: 47,779,411 RAC: 32 |
This is from the results on my computers page, not CPUz... If it sticks around long enough, here is the over-speed-reported one I cut and pasted from below... |
JDWhale Send message Joined: 6 Apr 99 Posts: 921 Credit: 21,935,817 RAC: 3 |
The function called to determine CPU speed is ippGetCpuFreqMhz and is provided by Intel... Here is that functions description... The function ippGetCpuFreqMhz is declared in the ippcore.h file. This function estimates the processor operating frequency and returns its value in MHz as an integer stored in pMhz. Note that no exact value of frequency is guaranteed so that the estimated value can vary depending on the processor workload. Hope this helps, JDWhale |
dnolan Send message Joined: 30 Aug 01 Posts: 1228 Credit: 47,779,411 RAC: 32 |
I guess that makes some sense, I've just always seen it under-reported up until now (which I assumed could be thermal throttling), it's only since using the AK port that I've seen it over-reported (EXCEPT: when I have the multiplier lower than standard, ie. 8 for a Q6600 - in that case it seems like the function assumes stock multiplier when figuring out the frequency, but that's not the case with this particular instance, it's at the stock multi). -Dave |
benkid Send message Joined: 26 Feb 00 Posts: 16 Credit: 1,171,999 RAC: 0 |
Hi All, Brief background:- I'm very interested in the seti@home project and will keep crunching, but I'm so very busy in real life at the moment that I've been "out of the loop" when it comes to checking out updates to the BOINC app and the optimised apps etc... I'm also not up on the recent discussions on these forums, but I do try to check things out every few months or so to avoid getting too far behind the latest client versions. I certainly don't want to be one of those contributers who screws up the credits for other users by running an ancient client version as I know that really irks some folks! Anyway, now I realise it is now time for me to change over from the KWSN apps to the new AK apps, and also upgrade to the latest BOINC client at the same time. (I'm going from 5.8.x to 5.10.x) I have just performed my first update and migration to AK V8 on a SSSE3 system - that has gone flawlessly and it appears to all be running fine. I will shortly do the same on my other SSE3 systems. However I also have some older systems (SSE/SSE2/MMX), mainly AMD. Can anyone tell me if the older 2.4V app versions on Crunch3r's page (which supports these older CPUs) are typically faster than the KWSN app on AMD CPUs? If so I will migrate these systems over to the 2.4V apps too. In fact I want to do this, as I would like to update to the latest BOINC client on all systems. Do the 2.4V apps work with BOINC 5.10.45? Thanks very much for any possible advice! |
RandyC Send message Joined: 20 Oct 99 Posts: 714 Credit: 1,704,345 RAC: 0 |
Hi All, The primary difference between 2.4V and the KWSN apps is that the 2.4V (V=Vista) apps are Vista tolerant. The difference in speed for corresponding apps is minimal. [edit]If you're using XP, then use the one that matches your cpu best. If you're using Vista, by all means use the proper 2.4V app.[/edit] |
benkid Send message Joined: 26 Feb 00 Posts: 16 Credit: 1,171,999 RAC: 0 |
Thanks for the speedy reply! I was using Vista for a while, but I'm now back on XP again due to needing to use some legacy apps, so I don't require the Vista tolerant version at this particular moment. Therefore, I guess I could keep my 5.8.x BOINC/KWSN setup on those older machines for a while longer. But I may migrate to 5.10.45 anyway just for the sake of keeping things up to date. I now realise this question is probably off-topic in this particular thread, but briefly, do you know if the 2.4V apps would work fine under BOINC client version 5.10.45? Thanks Again. |
benkid Send message Joined: 26 Feb 00 Posts: 16 Credit: 1,171,999 RAC: 0 |
I tried it. The 2.4V MMX app initially appears to run fine under BOINC 5.10.45. I had forgotten this is a KWSN app. As I am currently updating my newer systems to AK V8, I am also taking the opportunity to upgrade all of my systems to the latest BOINC client version. This is regardless of which optimised app each PC happens to use, I just like having them all on the same client version. Regards, benkid |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
I now realise this question is probably off-topic in this particular thread, but briefly, do you know if the 2.4V apps would work fine under BOINC client version 5.10.45? Thanks Again. Yes. |
benkid Send message Joined: 26 Feb 00 Posts: 16 Credit: 1,171,999 RAC: 0 |
I now realise this question is probably off-topic in this particular thread, but briefly, do you know if the 2.4V apps would work fine under BOINC client version 5.10.45? Thanks Again. Thanks for the confirmation. Indeed, I've been running one instance each of the 2.4V MMX, SSE and SSE2 versions on three different machines since last night. I've upgraded all six of my PCs to BOINC 5.10.45 and all are running well. I'm also glad that my three newer machines are now running the faster AK V8 SSE3 or SSSE3 apps without any problems. |
Alaskan Send message Joined: 18 May 99 Posts: 18 Credit: 53,634,823 RAC: 0 |
Mr. Whale, to gain stability on the q6600, I had to install the newest version of Boinc along with AK_v8_win_SSSE3x, since then my averages have started to show increased performance. Thank you for your efforts. Hi, after installing this AK_v8_win_SSSE3x on my quad q6600, my computing averages have decreased. Any thoughts on this? |
benkid Send message Joined: 26 Feb 00 Posts: 16 Credit: 1,171,999 RAC: 0 |
I had one final problem, now fixed. Two of my machines at work are behind a HTTP proxy. BOINC 5.10.45 doesn't seem to work correctly using FreeProxy 3.92 BUILD 1392, although BOINC 5.8.16 was always fine with it. The problem was that the new BOINC client would report the seti@home server as returning "no header, no data", yet it could see the reference site OK, and reported that the server was down. I determined that this was purely a problem with FreeProxy, and not with the latest BOINC version. Simply upgrading to FreeProxy 4.00 beta completely fixed this problem. Perhaps this info would be useful to anyone who is currently updating their BOINC setup and uses FreeProxy. |
Iw@n Send message Joined: 9 Oct 99 Posts: 4 Credit: 33,219,222 RAC: 7 |
Hi All, I'm now happily crunching with the AK V8 app for several days! No problems so far, beside a overheating problem of my CPU & MB which is fixed now. As we've a small 'heat wave' in The Netherlands (temperatures of 27C+, while in southern Europe it's only 20C i've had to improve the cooling of my system. After the installation of the AK V8 app the Intel E600 dual core of my 'primary' Windows Vista system went too hot, so I immeditely bought an Zalman CNPS9700 at a local store (the only fan in stock). The temperature of the E6600 decreased with about +/- 10C so i couldn't resist to upgrade with a quad-core Q6600 :-) Although I've a crappy computer case (bad air circulation, need to buy a rear fan) the AK V8 app is now running smoothly 24/7 on my Q6600 :-) My latest BOINC graph: Iw@n |
NewtonianRefractor Send message Joined: 19 Sep 04 Posts: 495 Credit: 225,412 RAC: 0 |
So far on an AMD Turion 64 X2 @ 1.6 Ghz running the SSE3 app I have the following: This is a graph of CPU time vs angle range This is a graph of credit per CPU hour vs angle range This is credit claimed per processing time along with a linear fit of a line with a fixed y-intercept at 0 There are about 150 datapoints on each graph... |
Paul D Harris Send message Joined: 1 Dec 99 Posts: 1122 Credit: 33,600,005 RAC: 0 |
So far on an AMD Turion 64 X2 @ 1.6 Ghz running the SSE3 app I have the following: Impressive graphs how did you get them using Microsoft sql? |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
Could people please remove the [img] links, when replying to posts with heavy graphical content? This is a popular and busy thread, and repeated graphs make it difficult to load and navigate. Also, please remember that not everyone has a large or widescreen monitor, and very wide graphics make the whole thread difficult to read. Another solution is to use an image hosting site which generates linkable preview thumbnails, like Fred has used here. You need to include the direct link as well, because some users have problems with the thumbnail links, but that way the thread is kept small and quick to load, but the fine details and small print are available for people who are interested. |
NewtonianRefractor Send message Joined: 19 Sep 04 Posts: 495 Credit: 225,412 RAC: 0 |
So far on an AMD Turion 64 X2 @ 1.6 Ghz running the SSE3 app I have the following:http://abrau.durso.googlepages.com/sah_boinc07.xls[/url] I used Boinc Log X (an addon program) to gather the AR and time data. It runs in the background and creates a CVS file. Then I manually went to the 'task' website and copied and pasted the amount of claimed credit. I ended up with this excel worksheet: http://abrau.durso.googlepages.com/sah_boinc07.xls |
Stephen R Send message Joined: 25 Aug 99 Posts: 56 Credit: 3,736,096 RAC: 0 |
Hi all, I am running AK opt app SEE3 (64bit amd/intel), wingman is running stock app. My Wu is canonical result, both WU are valid vs each other and get claimed credit. See this WU 263778712 I look at results of each Wu and I see my wingman reports Spike count: 2 Pulse count: 0 Triplet count: 4 Gaussian count: 0 yet my result reports Spike count: 0 Pulse count: 0 Triplet count: 0 Gaussian count: 0 I did have un expected restart [edit](not because of the app) [/edit]... Small issue with the app I am running writing to stderr? No issue as I got full credit for computer work ... (joking of course) my real ? is if my result is canonical result does the result get recorded in the DB as result 0,0,0,0 for above when as per my wingman, there might be something there. As in for the next stage of processing would the result of this WU just get passed up because of seemingly 0,0,0,0 result? (I have to assume that we both have same real results, so I would hope not and this is just a clitch with output to stderr) Sorry if this has been asked before. Kind Regards |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
Hi all, It's good you brought it up. The code base from which Alex worked was branched from the project sources at 5.13, and at that time the individual counts of found signals were not written to the checkpoint file. That means the counts in stderr will only show the number of signals found since the last restart. I know Jason hopes to do some more tuning of the ported apps, we'll try to get some of these cosmetic issues cleaned up for any future updates. It's a cosmetic issue because what gets put into the master science database is the canonical uploaded result file chosen by the Validator, the stderr data is merely included in the report sent to the Scheduler so users have something to look at. Joe |
Stephen R Send message Joined: 25 Aug 99 Posts: 56 Credit: 3,736,096 RAC: 0 |
It's good you brought it up. The code base from which Alex worked was branched from the project sources at 5.13, and at that time the individual counts of found signals were not written to the checkpoint file. That means the counts in stderr will only show the number of signals found since the last restart. I know Jason hopes to do some more tuning of the ported apps, we'll try to get some of these cosmetic issues cleaned up for any future updates. Thanks for the reply Joe, was as I hoped. Regards |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
We will see soon SSE and MMX AK V8 also? (100. post in this thread!) |
©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.