| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: x64 version of BOINC 5.9.xx |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 20
|
|
| Author |
|
|
cio_redulla
Advanced Cruncher Philippines Joined: Apr 24, 2006 Post Count: 130 Status: Offline Project Badges: 
|
6.04 (2 X 3.02) is an effective speed. both cores are being used at 100% WAB  , I still don't agree. |
||
|
|
wabr101
Cruncher Joined: Dec 16, 2005 Post Count: 16 Status: Offline Project Badges:
|
It's OK...we can agree to disagree and still be contrinuting to a good cause.
AMD's speed and rating system are usually relative anyways. the 5200+ assumes that each core (really a CPU) is 2.76Ghz. I have it OC'd to 3.02 per core, effectively giving me an AMD 6040. Either way...two cores would have been equivalent to two machines--and faster. I look forward to the AMD quad core. Nonetheless..cio_redulla I wish you the best. |
|||
|
||||
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Probably, in translation, wabr101 means to say that it's equivalent to a 1 core CPU on 6.04ghz.... some tweak the software to make it appear they have a 10ghz box, except with BOINC, the actual CPU details are transmitted, full detail, so the real numbers gets divulged.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Rickjb
Veteran Cruncher Australia Joined: Sep 17, 2006 Post Count: 666 Status: Offline Project Badges: 
|
The BOINC client does not use much of your CPU time - it's the project-science number-cruncher(s) that are downloaded & run by BOINC that use almost all of the CPU time. At the moment, the latter are all 32-bit (as far as I know), so there's not much to be gained from running a 64-bit BOINC client.
As for BOINC 5.9.x, I'm running 5.9.4 (32-bit) under 32-bit Win2k, and it works OK with WCG, though I haven't pressed all the buttons - just the ones I need. |
||
|
|
cio_redulla
Advanced Cruncher Philippines Joined: Apr 24, 2006 Post Count: 130 Status: Offline Project Badges:
|
Probably, in translation, wabr101 means to say that it's equivalent to a 1 core CPU on 6.04ghz.... some tweak the software to make it appear they have a 10ghz box, except with BOINC, the actual CPU details are transmitted, full detail, so the real numbers gets divulged. Yes, if that is the case, then I'd agree. But there's no significance in giving the "effective speed" since it is common knowledge that clock speed does not necessarily equate into performance. Doubling the clock speed does not indicate an equivalent doubling of performance. The performance of a 3.02GHz dual core processor is not comparable to the performance of a 6.04GHz single core processor. If I have a dual core processor (which I do have), I don't see the importance of computing the effective speed. Saying that you have a dual core AMD at 3.02GHz is enough. cio_redulla |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
5.9.5 is out and should now support alternate platforms(Projects still need to update the code on their server side).
----------------------------------------http://boinc.berkeley.edu/download_all.php Howdy Folks, This release includes a bunch of fixes around the various setup issues. This release should support the alternate platform mechanism that has been built into the latest server releases. Basically is the project doesn’t support the x64 platform it should download the 32-bit x86 client instead. I have disabled the x64 uppercase on alpha so we can test it out. Tomorrow or the day after we’ll re-enable the 64-bit uppercase. In order to uninstall the 5.9.4 build you’ll probably need to re-download the MSI file since it wasn’t cached for future use. You can find it here: http://boinc.berkeley.edu/dl/boinc_5.9.4_windows_intelx86.msi http://boinc.berkeley.edu/dl/boinc_5.9.4_windows_x86_64.msi The last release was the first version of BOINC we released that was built with VS 2005, as such the benchmark scores were quite a bit different than what you would previously been used too. See http://boinc.berkeley.edu/trac/ticket/122 for more details. I’m interested in whether this build closes the gap in benchmark scores between 2003 and 2005. Please log the test results to http://issac.ssl.berkeley.edu/alpha and log any bugs to http://boinc.berkeley.edu/trac. ----- Rom Here is a list of changes: http://boinc.berkeley.edu/trac/milestone/5.10 Earlier in this thread knreed noted that WCG is already updated so I guess it should kinda work right now(Still only alpha, so be cautious!). Now, Suppose you have a 64-bit OS and a 64-bit client, But you're running a project that does not have a 64-bit app and it sends out a 32-bit app instead, would it be in the same quorum group of the 32-bit client? If so, then it could lead to a nasty issue with the credits(If we naturally suppose the 64bit client will produce better benchmarks than the 32bit client). If it will be paired with 32bit clients, it could lead to the result being discarded from quorum average, Or if the the gap would be _really_ big, it could lead to the result being penalised. So perpahs a seperate quorum for the 64bit clients would be more suitable, Or a different solution etc.. etc.. just so it won't get ugly.  [Edit 1 times, last edit by Former Member at May 3, 2007 4:30:27 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
If 64bit clients benchmark higher, they should also complete work faster. Benchmark * time should always be a constant.
In practice, there is some variation - but the new credit adjustment system should level out these benchmark inconsistencies once and for all. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
This is what knreed said:
----------------------------------------We have the code already on our servers that will process the 32bit tag and send the 32 bit version (since we do not have a 64 bit version available). The new BOINC credit algorithm does not penalise any more. It normalises! Added: If results from 64bit clients would be chucked out as 'too different' if in quorums with 32 bit clients, obviously it would have to be considered to create a separate Homogeneous Redundancy class. Those running 5.9.x alpha's have to carry the reporting responsibility if it so comes.
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at May 3, 2007 4:55:54 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
If 64bit clients benchmark higher, they should also complete work faster. Benchmark * time should always be a constant. Ahhh.. Interesting. I mean I thought it would complete work faster only if there's a native 64-bit app. It benchmarks higer becasue the client is optimized for 64bit, but if the project sends 32-bit app instead, would it actually perform calculations faster? if it does then ofcourse it's all good and the credits shouldn't be that much of an issue I hope. @Sekerob : Whoops.. Thanks for the fix and news, didn't know that. Good news  .[Edit 1 times, last edit by Former Member at May 3, 2007 6:05:00 PM] |
||
|
|
wabr101
Cruncher Joined: Dec 16, 2005 Post Count: 16 Status: Offline Project Badges:
|
yes if it were compared to a single core, the dual 3.02 would be a single at 6.04GHz
it's what i meant |
||
|
|
|