oh and blame your Admin people they came over and posted on FaD what they were and about and invited us.

We are not complaining :)

More like checking the progress and hoping for more CPU power for Rosetta.

Anders n

---

You got me curious too - I'm overclocked, but with cheap RAM, so not quite as far as you, roughly 2.6GHz (still playing with different settings, so it still may be different every day). ASUS MB, only 512MB of that RAM.

Similar here (2640MHz at the moment) with cheap PC3200 RAM (which is at 240MHz, CAS 3). Have had it running at 2700MHz with the RAM at 193MHz, CAS 2.5 but the best benchmarks I could get with the official BOINC client is 2581/4789 (I think).

---

*** Join BOINC@Australia today ***

the best benchmarks I could get with the official BOINC client is 2581/4789 (I think).

I just downloaded and clean-installed V5.2.8 - and the benchmarks were 2402/4469. Looking around, I find no less than seven "boinc.exe" files on this machine... I've now trashed everything except the new "stock" versions and YAOSCW for SETI. Sigh. If I have to test something on this box again, I'll be sure to backup the WHOLE folder instead of just boinc.exe, and put it back correctly when done.

To anyone who _should_ be in the "top 500" but was delayed getting there by my benchmarks being out-of-whack, my apologies. 2500/6500 vs 2400/4500 isn't nearly as bad as the Trux 5.3.1 (2500/8500) that I _had_ been running for SETI, and removed when I shifted my priorities to Rosetta, but obviously was still higher than it should have been.

Other than making sure I'm "in the ballpark" for claimed credit on other projects, I don't worry about benchmarks. On Rosetta, obviously it makes a difference. I am still hoping that Rosetta goes to the "flops-counting" approach, as the benchmarks are just flat not good enough to use without redundancy, and an optimized client _IS_ necessary for some projects where you're running optimized apps, or if you're running Linux, or you have a HT CPU... The problem is that what is necessary or irrelevant elsewhere sometimes causes inflated scores here. Looking at my average of 18.4 credits per result, I am still not even sure that that was really "inflated". I just don't have enough data to tell. What are the averages for others?

Yoda, thanks for catching this!

EDIT:: Hm. I just realized that I am going to have a problem whenever I put Rosetta back on my Mac Mini. It's running SETI and Einstein Altivec-optimized apps right now, and therefore an optimized BOINC...

---

To anyone who _should_ be in the "top 500" but was delayed getting there by my benchmarks being out-of-whack, my apologies.

Without accusing anyone in particular, there's plenty of others who got there with inflated benchmarks (whether deliberate or by accident) and I'm sure that will continue. I have run optimised BOINC versions myself but since I no longer crunch for SETI, I am removing them.

Looking at my average of 18.4 credits per result, I am still not even sure that that was really "inflated". I just don't have enough data to tell.

In the absence of a "reference unit" and with wildly fluctuating completion times for Rosetta work units, averages probably won't tell much of a story.

Either way, I think we've hijacked this thread for long enough :-) If you want to continue this discussion Bill, you can contact me via the (webmaster) link on my team's website.

---

*** Join BOINC@Australia today ***

Looking at my average of 18.4 credits per result, I am still not even sure that that was really "inflated". I just don't have enough data to tell.

In the absence of a "reference unit" and with wildly fluctuating completion times for Rosetta work units, averages probably won't tell much of a story.

Either way, I think we've hijacked this thread for long enough :-) If you want to continue this discussion Bill, you can contact me via the (webmaster) link on my team's website.

Agreed. The 18.4 looks about right on the basis of what I'm seeing from BoincView for four of my systems. In no particular order, looking at the 4 rosetta WU's currently in flight:

An AMD Athlon 750 MHz with an estimated time of 6H:40M has an est credit of 14.29
An AMD Athlon 1400 MHz with an estimated time of 3H:56M has an est credit of 12.72
A 2.4 GHz Celeron with an estimated time of 10H:18M has an est credit of 23.68
An AMD Athlon 2400+ with an estimated time of 6H:23M has an est credit of 11.97

Your actual mileage may vary. To add to the confusion, the Celeron should really be doing FaD work, but I want to clear a couple of Rosetta WU's that snuck onto it while I was setting up. So it's running 50/50 between Boinc and FaD. AFAIK, that doesn't make any diff under Win XP, since credit is based on actual CPU time.

---

Agreed. The 18.4 looks about right on the basis of what I'm seeing from BoincView for four of my systems. In no particular order, looking at the 4 rosetta WU's currently in flight:

Are you looking at the estimated credits (so far) for an unfinished work unit? That's not a good indicator as the credit estimate is based on CPU time already spent, not on the total time it is estimated to take.

---

*** Join BOINC@Australia today ***

Are you looking at the estimated credits (so far) for an unfinished work unit? That's not a good indicator as the credit estimate is based on CPU time already spent, not on the total time it is estimated to take.

Not sure what BOINCView shows, my 18.4 figure was the average credit for all results still on the server, earlier today, for my one AMD Windows host, per Andy K's stats utility...

http://www.andyk.de/boinc/pc.php?pid=rah&uid=6600&query=user&ev=on&av_seconds=on&avcc=on&avgc=on&trsh=0&tch=on&hsort=not

and replace uid (mine is 6600) with your own... right now it's showing (with 268 results) average 18.71. I'm sure this includes results done under at least 3 different boinc.exe programs. I have noticed that the time per result has been increasing overall, as the newer results take longer (and earn more credit) than the ones I was originally seeing - the average time is up to 1:04 now. BOINC currently estimates 2:39 per result!

EDIT:: I've looked at some other people's results (who have posted in this thread) using this utility... I'm seeing a range per host from 6.6 to 29.6. (Even excluding those with too few results to be significant.) As I kindof expected, it's all over the map. Overall, I'd say that 17 seems to be the most common number.

---

Agreed. The 18.4 looks about right on the basis of what I'm seeing from BoincView for four of my systems. In no particular order, looking at the 4 rosetta WU's currently in flight:

Are you looking at the estimated credits (so far) for an unfinished work unit? That's not a good indicator as the credit estimate is based on CPU time already spent, not on the total time it is estimated to take.

Yes. In my infinite wisdom, I figured that the est credit was based on what the credit would be at completion. It strikes me that the est credit being based on current CPU usage is really rather less than useful. :)

---