AMD Athlon 64 running 36 h and 61-62 h to completion?

Terry
Terry
Joined: 18 Feb 07
Posts: 7
Credit: 9229
RAC: 0

???So slow under windows???

Message 63830 in response to message 63829

???So slow under windows??? I have an AMD system very similiar to Kirstins and I have had to end my association with E@Home as well due to the fact that upon startup E@H gobbled close to 100% of my CPU time, severely slowing my system. When I first began using E@H I did not have this issue but for the last month or so it has been an annnoying problem Deleting BOINC and E@H resolved this issue. It was E@H that was robbing my CPU time to extremes. Perhaps I will use boinc with another project that is less cpu intensive. By the way, I have had absolutely no issues with my AMD system under windows before or after E@H was on my system. Its too bad, it was such a worthy project,
--Terry N.

Quote:

Hi!

I wonder if Kirsten is still around here, now that we figured out why the AMDs are so slow under Windows.

CU

BRM


--Terry photostuff.org

Lloyd M.
Lloyd M.
Joined: 24 Apr 07
Posts: 24
Credit: 2493855
RAC: 0

RE: Well, I have detached

Message 63831 in response to message 63806

Quote:

Well, I have detached the project. My temperament is not for 100 hours long work units, which should not be running for more than around 30 hours on my type of machine: an AMD 64 3200+.

This machine is quite fast and in SETI@home a 9.704 second WU (=2 hour 41 min) is 'rewarded' with 64,20 credits. Even if I am not in BOINC for the credits, I would not like to waste my idle CPU cycles on a questionable work.

That is a lot. My Athlon 64 3700+ takes about 44 CPU hours/WU maximum. It errored out after processing for over 40 hours one time, and that's another story.

What I would like to know is how I just got allocated 8 WU's that are supposed to take 31 hours each, with a deadline of June 6, when the resource allocation for E@H is a little under 16.5%?

Lloyd M.
Lloyd M.
Joined: 24 Apr 07
Posts: 24
Credit: 2493855
RAC: 0

RE: I wondered, of

Message 63832 in response to message 63814

Quote:

I wondered, of course, whether BOINC made a correct estimation of time to completion or not, as the estimation at first was 67 hours. After 36 hours of crunching the estimation was only reduced to 62 hours, though. Therefore my question as I smelled a bug.

Off topic: It would have been nice to know that the estimated runtimes could be inaccurate *before* I detached the project and thereby wasted 36 hours of crunching. IMO I might have received a faster (and probably the quoted correct) answer to my original question if I have had the right to ask questions in the correct forum. To me a bug and problems forum is meant for newbies and/or beta testers.


I run several BOINC projects, and they're all pessimistic about processing times, especially before you've processed a few WU's What I often see is that, as each WU is processed, the estimated completion time for remaining WU's trends downward. The estimated time never seems to go as low as the actual processing time, and this makes sense to me to allow for some "wiggle room" for unexpected outages. If the "real" times were used, it would be far more likely that your system might download more WU's than it can process in the allotted time.

Lloyd M.
Lloyd M.
Joined: 24 Apr 07
Posts: 24
Credit: 2493855
RAC: 0

RE: Does the project have

Message 63833 in response to message 63808

Quote:


Does the project have the servers, manpower, etc. to do it as a traditional? Think about the project resources, etc. They may not be able to do things the way other people may think they should.

I personally do not have any problems with Einstein on any of my 5 machines. No errors have occurred, the data is getting done, and I am getting credit.

I have had considerable trouble (at one point, two WU's errored out, out of four total, one of which errored out after processing for about 40 hours) with my one machine, and it is beginning to appear that it is because it is an AMD machine.

Lloyd M.
Lloyd M.
Joined: 24 Apr 07
Posts: 24
Credit: 2493855
RAC: 0

RE: [snip] It was E@H that

Message 63834 in response to message 63830

Quote:

[snip] It was E@H that was robbing my CPU time to extremes. Perhaps I will use boinc with another project that is less cpu intensive. By the way, I have had absolutely no issues with my AMD system under windows before or after E@H was on my system. Its too bad, it was such a worthy project,
--Terry N.

Terry,
On my AMD box, SETI takes about 2.5 CPU hours/WU. If you want something really speedy, PrimeGrid only takes about 10 CPU minutes per WU!

I had a few 40+ hour E@H WU's, one of which errored out shortly before it was finished processing (to my considerable annoyance). Somehow or another Rosetta gives you credit, even if the WU errors out.

Lloyd M.
Lloyd M.
Joined: 24 Apr 07
Posts: 24
Credit: 2493855
RAC: 0

RE: Hi! I wonder if

Message 63835 in response to message 63829

Quote:

Hi!

I wonder if Kirsten is still around here, now that we figured out why the AMDs are so slow under Windows.

CU

BRM

I'll bite. Just why are "AMDs so slow under Windows"? My <$100 (brand new, before the prices went down) single core AMD processor manages 2555.91 MIPS floating point, and 4777.84 MIPS integer on Windows XP.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 757465201
RAC: 1157036

RE: RE: Hi! I wonder if

Message 63836 in response to message 63835

Quote:
Quote:

Hi!

I wonder if Kirsten is still around here, now that we figured out why the AMDs are so slow under Windows.

CU

BRM

I'll bite. Just why are "AMDs so slow under Windows"? My Pentium III"). Athlon 64 and Operons under Windows suffer a ~ 30% performance penalty for E@H under Windows because of this bug.

Look at the list of top computers. The first AMD box under Windows is ranked at position 180 or something like this at the time of writing. All hosts up to that rank are either Intel or AMD under Linux!

The bug in the math library is relatively easy to correct so I guess we will see an improved Windows app at least as a beta test app in the near future.

CU

BRM

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.