This gets irritating if you run several different computers/OS's. If I set preferences to get enough files to keep the Linux machines running (twice the buffer it would require if the client would calculate the estimated time correctly), my windows machines get buried in tasks. I know I can "work around this" by declaring different locations and preferences for the Windows and Linux machines, but the question is-
Why should I have to??????? It seems that everyone knows that the Linux benchmark is about 1/2 of the real operating speed, is it that hard to fix?

---

Part of an email from David Anderson for upcoming versions of BOINC:

near term: add support for "benchmark workunits".
The time it takes to machine to finish the benchmark WU
determines how much credit per CPU second it gets for subsequent WUs.
If all projects use this, we can get rid of CPU benchmarking
in the core client (which is the root of many problems).

---

This gets irritating if you run several different computers/OS's. If I set preferences to get enough files to keep the Linux machines running (twice the buffer it would require if the client would calculate the estimated time correctly), my windows machines get buried in tasks. I know I can "work around this" by declaring different locations and preferences for the Windows and Linux machines, but the question is-
Why should I have to??????? It seems that everyone knows that the Linux benchmark is about 1/2 of the real operating speed, is it that hard to fix?

Not being a Linux user, I can only give you a generalized answer.

1. Different Linux distros and other OS on different CPUs do produce different Benchmark result.
2. The benchmark is used by some projects to determine claimed credits. The major projects are now employing a different approach in determining actual granted credits. For example Seti now counts flops or actual work being done for claimed credits.
3. Since Boinc Version 4.70 the benchmark is mainly used to determine the initial size of your Work Buffer. It should only have a minimal effect on how much work you receive after a few WU have been processed. Each project, on each machine now has its own duration_correction_factor(DCF) which is used to calibrate the efficiency of your individual machine. The benchmark on one computer has no relevance to how much work another computer on the same account receives.

In summary, the factors effecting how much work you receive from a project can be found on the bottom of each project's Computer Summary Page.
These are
Maximum Daily Work Unit Quota per CPU
Average CPU Efficiency
Result Duration Correction Factor
Additional factor that effect what work you receive are
the 'Connect to network about every' setting in your General preferences (propagates across all your projects) and is cumulative. You get this much work from each project as long as the work scheduler determines your system can handle it and still meet due dates.
the resource share you set for each project.
the debt accumulated by individual project.

Detail information can be found about
Work Buffer
Work Scheduler
Work-Fetch Policy

edited to fix links

---

Boinc V 7.4.36
Win7 i5 3.33G 4GB NVidia 470

This gets irritating if you run several different computers/OS's. If I set preferences to get enough files to keep the Linux machines running (twice the buffer it would require if the client would calculate the estimated time correctly), my windows machines get buried in tasks. I know I can "work around this" by declaring different locations and preferences for the Windows and Linux machines, but the question is-
Why should I have to??????? It seems that everyone knows that the Linux benchmark is about 1/2 of the real operating speed, is it that hard to fix?

I've been dealing with this since I switched to linux last August. The only solution I've found is to use an optimized version compiled specifically for your processor. It eliminates a good portion of the disparity between linux and windows and though some projects frown on it my computers claim credits very close to what is claimed by others running windows.

---

Nothing travels faster than the speed of light with the possible exception of bad news, which obeys its own special laws.
Douglas Adams (1952 - 2001)

Part of an email from David Anderson for upcoming versions of BOINC:

near term: add support for "benchmark workunits".
The time it takes to machine to finish the benchmark WU
determines how much credit per CPU second it gets for subsequent WUs.
If all projects use this, we can get rid of CPU benchmarking
in the core client (which is the root of many problems).

Jord,
Not wishing to be too rude to Dr. A, whose work on distributed computing is in general to be praised. Haven't we be saying this for at least two years. Or did Paul D. Buck just wind him up too much.

Andy

LOL Andy.

I just figure, if you repeat it long enough it'll register that something is wrong. I don't think it was Paul that was behind it finally registering, since Paul's been AWOL for a long time now. I'll try calling him this New Year's and hear his reaction. ;-)

---

I'm not sure what the discussions are but know that WCG and Dr. A. agreed, that the next formal release i.e. 5.8 would have a fixed Linux benchmark to bring the result on par with the Windows benchmark.

In quorum 3 situations like WCG one needs 3 Linux machines to have reasonable claims to arrive at a median, which is awarded to all in the quorum. "Outliers"....those overclaiming get normalised or penalised. A single machine having been 'optimized' in a quorum is thus chucked for median calculation purposes.

A quorum is never a mix of multiple OS results and they give slightly different output.

Oh, and the benchmark for Windows version is the one suspected of having been inflated in the past....Linux benchmark is the one actually close to reality!

I find it a bit disturbing that most of what users say goes ignored yet if WCG wants something it gets immediate attention. It's a bit disappointing that the benchmark disparity has been discussed many times and was always dismissed because "only 10% of the computers are running Linux". Now WCG wants it fixed and it's being addressed. Benchmark work units were suggested a long time ago and dismissed as a bad idea. Now WCG wants them and it's going to be implemented. I think it's obvious who's really in control of BOINC.

---

Nothing travels faster than the speed of light with the possible exception of bad news, which obeys its own special laws.
Douglas Adams (1952 - 2001)

...I find it a bit disturbing that most of what users say goes ignored yet if WCG wants something it gets immediate attention. It's a bit disappointing that the benchmark disparity has been discussed many times and was always dismissed because "only 10% of the computers are running Linux". Now WCG wants it fixed and it's being addressed. Benchmark work units were suggested a long time ago and dismissed as a bad idea. Now WCG wants them and it's going to be implemented. I think it's obvious who's really in control of BOINC.

Which methods of credit generation are good and bad will vary according to which project is involved. So benchmark WUs may have been bad for (say) SETI, but that doesn't mean they're bad for other projects.

CPDN has been using fixed credit per work unit for several years (or, rather, per trickle), it it works really well. But their work units are consistent within 10% for execution time, whereas for other projects it's different. Would you want to run a 3 month-long WU simply to benchmark how much credit you get in the future? :-)

Personally I dislike the benchmark approach to credit generation, it's far too prone to cheats and gives erratic results. And don't get me started on my opinion of quorums! :-) :-)

---

Will someone Please fix the Linux client benchmark!

You could install the new 5.7.5 (alpha) Linux manager.
Benchmarking has been improved and is almost equal to a Win host.

Besides from that... it also looks much better.

;-)

---

Will someone Please fix the Linux client benchmark!

You could install the new 5.7.5 (alpha) Linux manager.
Benchmarking has been improved and is almost equal to a Win host.

Besides from that... it also looks much better.

;-)

Yeah, I noticed the improvement in BOINC Manager's appearance. The standard BOINC manager now looks just like the ones I've been building locally since Sept. 2005. That's the case on my computer running Sabayon 3.2. I haven't tested it on my other computer that runs SuSE 10.1 yet.

---

Nothing travels faster than the speed of light with the possible exception of bad news, which obeys its own special laws.
Douglas Adams (1952 - 2001)