A short WU needs different time on Mac and PC

Omikronman
Omikronman
Joined: 23 Nov 06
Posts: 33
Credit: 83254
RAC: 0
Topic 192231

I don´t know if anyone had a try with BOINC running on a Intel Mac to compare between MacOs and Win. I had a look at this. The MacOs version of BOINC needs 35 minutes for a short WU, the Win BOINC on the same machine needs 37 minutes for a short WU. BOINC was the only working thing in this test. Why is the Win BOINC slower?

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9352143
RAC: 0

A short WU needs different time on Mac and PC

Looking over the your results from the test, you can't really say the Mac version is any faster or not, for several reasons.

First, you didn't really run enough results on Windows to make for a good statistical sample.

Second, 120 seconds is well within the runtime deviation I've logged on the same datapack, and certainly within the deviation you can expect between different datapacks.

Third, you can't make a direct short term comparision since your test was comparing results from different datapacks. For reference, there is no way you can do that (easily that is), since the project won't send the same datapack to more than one host on any given account.

Probably the easiest way to make the analysis your trying is to log a good number of results for each OS (~100 each at least) and then do a manual computation of the credit rate and compare that.

One last thing. Running tests is fine, but what are you going to do about the other 152 results you DL'ed for Windows? If you don't go back and run them out, or at least abort the unrun ones, most everyone who's paired up with you will be waiting until the second week in January at least to clear these WU's off their machines.

Keep in mind that EAH is not like SAH anymore, with an extra result generated initially to "cover" for the possibility that a host may "poof out" on any given WU. It's just common courtesy to the project and other participants to make sure experimentation doesn't cause unnecessary delays or extra work. Especially with the project seeming to be having continuing fileserver issues, having to resend results only exacerbates the problems they have to deal with at the moment.

Alinator

Omikronman
Omikronman
Joined: 23 Nov 06
Posts: 33
Credit: 83254
RAC: 0

Please don´t worry about the

Please don´t worry about the other work unites I have DL´ed. I will think about doing that work quickly or stop it if I will not be able to do it during the next fiew days. I think that work should easy be done because that are all short WUs. If I give BOINC the full capacity of the Mac it finishes two short works in 35 minutes.

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4267
Credit: 244932206
RAC: 16344

There are two main reasons

There are two main reasons why I would indeed believe that the MacOS Intel version is slightly faster than the Windows one:

1. You usually have something more running in the background on Windows than on MacOS (virus scanner, application-level firewall etc) which "steals" some time from BOINC.

2. As the Intel Macs all have the same CPU (or at least the same CPU features) we don't need to detect CPU features and switch codes optimized for different CPUs there, which is what we do in the Windows App. However I doubt that this alone is responsible for a speed difference as large as you observed.

BM

BM

Comment viewing options

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