[zombie67]

So... I have a bizarre variety of machines crunching BOINC, with a wide range of OS. And every time I try to figure out how to optimize their deployment, I need to try to remember which projects support which mix of all of that. (And then I need to understand which combinations work best for each project. But that's another thread.) So, I put this table together. Blue background lines are Mac-Friendly. Red font means no work available.

I am looking for feedback. What's wrong? What's missing? What is not clear?

http://homepage.mac.com/zombie67/BOI...c_Friendly.png

[beadman]

One comment on Einstein@Home:
While it will run on both PPC and Intel Macs, I wouldn't call it "Mac-friendly".
A sampling of both SETI and Einstein made back in May from my MBP shows that SETI takes an average of 110 CPU-Sec/Credit, while Einstein takes 186 CPU-Sec/Credit, resulting in Einstein delivering 60% of the credit that SETI does for the same amount of work (on a MBP). This is apparently because Einstein has "deflated" credit across the board so as to be equivalent to SETI in credit granted. There were some discussions about this, and I pointed out to Einstein that while the standard Intel PC was now receiving equivalent credits for equivalent work from both SETI and Einstein, the Macs were being artifically penalized. I was ignored by those running Einstein.

My MBP on SETI will earn roughly 1350 credits/day, while on Einstein it only earns around 800 credits per day.

beadman

[Billy]

In my time zone, Eastern North America, SIMAP seems to be giving out work on the last day of the month, not the first. It may be because the project is located in Europe plus they seem to give out the work a bit earlier, rather than later.

[zombie67]

Originally Posted by beadman 

One comment on Einstein@Home:
[...]
My MBP on SETI will earn roughly 1350 credits/day, while on Einstein it only earns around 800 credits per day.

Interesting. Is this with or without alexkan's app on SETI?

[beadman]

I'm running seti_enhanced-i386-v7.2-coreduo-nographics on the MBP. When I just checked to make sure, I noticed that setiathome_5.18_i686-apple-darwin had just downloaded today. Guess I need to check my app.info.xml file.

<edit> Found my app.info file had been renamed a while back (my me) and hadn't been changed back yet. <edit>
beadman

[zombie67]

So what is the credit comparison, using stock applications?

[beadman]

Originally Posted by zombie67 

So what is the credit comparison, using stock applications?

Sorry - I may not have been clear - I made the comparison a month ago, using Alex's enhanced client for SETI vs. the stock E@H client. I haven't tried comparing stock to stock.

beadman

[zombie67]

Ah. So then is E@H really so mac-unfriendly? Perhaps S@H w/ alexkan's application is just....overly mac-friendly?

[beadman]

As I remember, they've been adjusting the granted credit downward across the board until a "standard" Intel PC will achieve roughly the same credit/cpuhour for E@H as for SETI. This is all well and good, as it helps prevent a few people from searching for the biggest bang for the buck and abandoning one project for another. However, when the multiplier is applied to all computers/OSs, it's no longer a good thing, IMO. Before they made the adjustments, my MBP was taking just about the same about of time for a credit on both E@H and SETI, while (apparently) most of the Intel PCs were taking less time per credit on E@H than on SETI. So they adjusted everything, instead of just the one type of processor. As I remember, the AMDs were affected similarly to the Macs. That's why I abandoned E@H: I can routinely earn around 1300 credits per day on SETI, but E@H gives me around 800 credits per day.

Take a look at the E@H top computers. The top Mac is a quad, MacPro, and is only getting 2700 credits per day, vs. the quads on SETI, which are getting well into the 4000's (Bad To The Bone's Quad is at 5800). So - my conclusion is that E@H will run on a Mac, but I don't think the credit awarded is very user-friendly. Too bad, too; I like the idea of searching for gravity wave phenomena.

beadman

[Billy]

I think that there is a Climate Prediction Beta for Mac.

[zombie67]

Yeah, I left off CPDN Beta because they run it only occasionally, they don't want the URL publicized, and it's not tracked on sites line BOINCStats. However, their Mac Intel application has been released to the regular production project, and is shown accordingly on the matrix. Thanks for the feedback!

[Billy]

Originally Posted by Billy 

In my time zone, Eastern North America, SIMAP seems to be giving out work on the last day of the month, not the first. It may be because the project is located in Europe plus they seem to give out the work a bit earlier, rather than later.

What they say:
June 29, 2007
New workunits: The calculation of similarities and features of the approx. 127.000 new sequences, that were imported from the databases PDB, Uniprot and RefSeq in june, will start in the evening of june 30th.

May 31, 2007
New workunits: The calculation of similarities and features of the (only) approx. 33.000 new sequences, that were imported from the databases PDB and Uniprot during may will start on may 31st.

[zombie67]

Updated list. Lots of projects added. Feedback/corrections appreciated!

http://homepage.mac.com/zombie67/BOI...Friendly.xhtml

[Gulliver64]

Originally Posted by beadman 

One comment on Einstein@Home:
While it will run on both PPC and Intel Macs, I wouldn't call it "Mac-friendly".
A sampling of both SETI and Einstein made back in May from my MBP shows that SETI takes an average of 110 CPU-Sec/Credit, while Einstein takes 186 CPU-Sec/Credit, resulting in Einstein delivering 60% of the credit that SETI does for the same amount of work (on a MBP). This is apparently because Einstein has "deflated" credit across the board so as to be equivalent to SETI in credit granted. There were some discussions about this, and I pointed out to Einstein that while the standard Intel PC was now receiving equivalent credits for equivalent work from both SETI and Einstein, the Macs were being artifically penalized. I was ignored by those running Einstein.

My MBP on SETI will earn roughly 1350 credits/day, while on Einstein it only earns around 800 credits per day.

beadman

This is correct in parts only. It depends on which science app you use. AT the time when you tested it there was a Mac-cient that was not optimized for Mac at all and therefore was very slow compared to the Windows version (-30%). Then they released version 4.34 which did much better and in fact was appr. 20% faster than the Windows/Intel version. The latest versions are 4.40 (Windows/Intel), 4.41 (Mac PPC) and 4.42 (Mac/Intel). These versions are way slower than the previous ones but are comparable between each other. (All percentages calculated on the basis credits/hour of comparable machines)

[beadman]

Originally Posted by Gulliver64 

This is correct in parts only. It depends on which science app you use. AT the time when you tested it there was a Mac-cient that was not optimized for Mac at all and therefore was very slow compared to the Windows version (-30%). Then they released version 4.34 which did much better and in fact was appr. 20% faster than the Windows/Intel version. The latest versions are 4.40 (Windows/Intel), 4.41 (Mac PPC) and 4.42 (Mac/Intel). These versions are way slower than the previous ones but are comparable between each other. (All percentages calculated on the basis credits/hour of comparable machines)

If you read further down the thread, I mention in later posts that I was running Alex's enhanced client. Using the current latest version of Alex's enhanced client on my G4, G5, and MBP, the SETI and Einstein clients give roughly the same amount of credit per CPU hour. Using the latest stock SETI client, my machines do much better with Einstein, so I continue to run Alex's enhanced client. On my old iMac G3 running 10.3.9, Einstein gives roughly twice the credit per CPU hour as SETI. I attribute this to the need for lots of L2 Cache and RAM for SETI, and my poor old G3 only has 256MB of RAM and a very small L2, so I get more credit running Einstein on it.

beadman

[zombie67]

SETI RAM requirements are very low. Somewhere less than 150MB, if I recall correctly.

[Gulliver64]

Originally Posted by beadman 

If you read further down the thread, I mention in later posts that I was running Alex's enhanced client. Using the current latest version of Alex's enhanced client on my G4, G5, and MBP, the SETI and Einstein clients give roughly the same amount of credit per CPU hour. Using the latest stock SETI client, my machines do much better with Einstein, so I continue to run Alex's enhanced client. On my old iMac G3 running 10.3.9, Einstein gives roughly twice the credit per CPU hour as SETI. I attribute this to the need for lots of L2 Cache and RAM for SETI, and my poor old G3 only has 256MB of RAM and a very small L2, so I get more credit running Einstein on it.

beadman

beadman,

you are mixing 2 different projects. Einstein@Home is different from SETI@Home and they have nothing to do with each other besides the fact that both are BOINC projects (and they also do not influence each other unless you run both projects in parallel with minimum RAM and choose to keep the results in memory).

Credit/hour is influenced by 3 different facts:
1. The speed of your computer
2. The amount of optimization of the science application for a specific hardware platform
3. The way the project assigns credit (e.g. based on benchmarks or calculations/hour)

[zombie67]

Updated. Feedback is good.

http://homepage.mac.com/zombie67/BOI...Friendly.xhtml

[zombie67]

latest update here: BOINC Project List

[Shaktai]

Looks real good. That is a big help.

[Billy]

Originally Posted by zombie67 

Updated. Feedback is good.

http://homepage.mac.com/zombie67/BOI...Friendly.xhtml

SIMAP is now continuous, at least to the end of January. It gives good points on the Mac for the HMMER application.

PostPosted: Tue Dec 11, 2007 8:21 am Post subject: Reply with quote
Yes, as Jonathan said. Just look at Redirecting to SIMAP - the webportal of SIMAP for the scientific community who is using this resource. There you will find the release and usage statistics directly on the portal start page.

The current number of sequences due to the import of the environmental sequences (pre-processed by SIMAPBOINC during the last months) is some more than 17 million. And only 6.3 million of the have been already processed. Therefore these 11 million sequences have to be calculated - which will be continuous work until january; due to the quadratic increase of calculation effort I would expect even until the end of january.

Thanks a lot for crunching!
Regards
Thomas

[zombie67]

Updated!

http://homepage.mac.com/zombie67/BOINC_Projects.html

The spreadsheet is also now available on my Downloads page.

[Billy]

Originally Posted by zombie67 

Updated!

SIMAP has been continuous since about Nov. 30, 2007.

[Billy]

Simap will be sending out work units again on March 31. It is back to being a monthly project.

[Billy]

Docking is reporting a forecasted restart.

Apr 29, 2008 8:20:00
It was a while ago when we updated our volunteers. In the meanwhile Arun Rajendran has joined our group as a post-doc and will dedicate his time to make sure that D@H will be back for this summer. Stay tuned!

[beadman]

Crunch3r has just compiled an optimized version of the MilkyWay@Home app for Intel Macs. You can get it here: calbe.dw70.de - View topic - Optimized applications...

Credits per day comparable to SETI running Alex's latest. This is your opportunity - there are only two Macs listed in the top 200. Project does occasionally have hiccups, so don't run this one exclusively - keep another app going at a minimal percentage for the down times.

beadman