I think there's definitely some tinkering with the credits on the ppc end, too. Take a look at this person's Macs. Both are Power Macintosh PowerMac 7,3, which should correspond to a 2xG5 tower, but one is running Darwin 7.5.1 and gets 4328.53 million ops/sec, while the other is running Darwin 8.0.0b and getting [/i]11309.41[/i] million ops/sec???

First off, I'd be laughing my butt off Darwin 8 actually was available before its correspondind OS X incarnation (Tiger/10.4) were released. But more importantly, that's gotta be a helluva compiler he's got to increase his ints by 250%.

There's an odd whiff of trout in the atmosphere...

My 2.66(online since Sept 18) is getting 319.58 RAC and is requesting 513.75 credits.The 3.00 (online since Oct 28) gets 220.92 RAC and is requesting 513.09 credits.And both climb up every day.

---

I Refuse to hold myself responsible for any of my actions.

si hoc legere scis nimium eruditionis habes

> > >
> > > Answer: Any programmer who wants one. Source for Boinc and
> Seti@home is
> > open
> > > source and downloadable.
> > >
> >
> > Yes, have more than once pointed to the source in my posts so knows this.
> ;)
> > But you're indicating one or more of these computers in this thread have
> a
> > compiled client with artificially increased benchmark...
> > Not having any P4 maybe overlooks something...
> >
>
> If you run a homebrew client (as I do) that is noticeably faster than the
> plain one, you end up with claiming ridiculously low amounts of credit.
>
> I guess it's fair to increase the benchmark values in a way that the
> claimed credits stay "middle-of-the-road".
>
> Any opinions on that?
>
> Regards Hans
>
> P.S:
>
> If the benchmarking wasn't broken in the first place, we wouldn't have
> this discussion.
>
>

I think it's all somewhat moot. Sure you could run an optimised boinc client and claim rediculously high credit but because of the averaging system used it doesn't really matter if you claim 50, 100 or 10,000 credits as the highest claimed value is always discarded and the median value awarded. The only time it would make a significant difference is if two or more computers are doing the same thing on the same unit which is statistically improbable given the number of active client computers. It may slightly increase your RAC score, but not to this extend. Besides, the benchmark scores for that machine looks perfectly reasonable as do the requested credits.

My fastest machines running optimised clients still only manage 400 RAC at this time. I think the answer in this case is elsewhere.

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

> I get about 420 Rac a day out of my one pc.I have ht.
>
At 4:20 I engage hypertoke , ht. always works to get my Rac higher.

---

> >
> > Answer: Any programmer who wants one. Source for Boinc and Seti@home is
> open
> > source and downloadable.
> >
>
> Yes, have more than once pointed to the source in my posts so knows this. ;)
> But you're indicating one or more of these computers in this thread have a
> compiled client with artificially increased benchmark...
> Not having any P4 maybe overlooks something...
>

If you run a homebrew client (as I do) that is noticeably faster than the
plain one, you end up with claiming ridiculously low amounts of credit.

I guess it's fair to increase the benchmark values in a way that the
claimed credits stay "middle-of-the-road".

Any opinions on that?

Regards Hans

P.S:

If the benchmarking wasn't broken in the first place, we wouldn't have
this discussion.

---

>
> Answer: Any programmer who wants one. Source for Boinc and Seti@home is open
> source and downloadable.
>

Yes, have more than once pointed to the source in my posts so knows this. ;)
But you're indicating one or more of these computers in this thread have a compiled client with artificially increased benchmark...
Not having any P4 maybe overlooks something...

> > What I think you're omitting is the recompiled boinc client, with higher
> > benchmark scores.
>
> Maybe overlooks something, but what computer have a recompiled client?
>

Answer: Any programmer who wants one. Source for Boinc and Seti@home is open source and downloadable.

---

> What I think you're omitting is the recompiled boinc client, with higher
> benchmark scores.

Maybe overlooks something, but what computer have a recompiled client?

The only Cheat that i need is a real good Money Cheat for Game Morrowind !

> theres no way i have a dual 2.4ghz xeon server w/ ht that only hits 558.03
> with roughly 25,000 credits. Now this is a production server but a single proc
> machine ..theres no way
>
What I think you're omitting is the recompiled boinc client, with higher benchmark scores.

If you were to use such a client on the dual xeon server, with modest benchmark multiple so it wasn't allways the highest result, it would get higher RAC.

Hmm, just look the results: Most are calculation times are around 5 hours. So there is most likely around 4 machines or more. I guess the RAC gets a bit messed when merging similar computers. Or there is some other bug in merge.

---

> My star cruncher is a DUAL AMD Opteron 248 system with 1 gig
> matched pair low latency memory, and it sits at 814.80 RAC.
> This is running WinXP Pro SP2. So the 1137.68 RAC for a
> single P4 seems pretty far fetched to me, even if there
> are 4 of them merged...
=========

I wouldn't say that, my one P4 3.4 was up to 420 RAC at the LHC Site and still climbing so if you merged 4 of them together then 1137.68 isn't un-realistic at all. Off course it's not going to stay there because it won't be able to sustain that high a RAC.


It will be interesting to see what that Computer can do here at the Seti Site now that I'm back Crunching over here again.

---

I get about 420 Rac a day out of my one pc.I have ht.

---

......In Space No One Can Hear You Scream......

2 different hardware.

Was not attached to SETI for a while.
Reput some clients a day ago (m$, 4.53)
Reading this thread, went to my hostfile.

Never say never, BOINC is offering "merge identical hosts"
This right now taken hardcopy shows, BOINC can offers a BAD selection.

Of course I canceled the action.

It looks like, the client identification subroutine should be reviewed to
better discern the clients.

My star cruncher is a DUAL AMD Opteron 248 system with 1 gig
matched pair low latency memory, and it sits at 814.80 RAC.
This is running WinXP Pro SP2. So the 1137.68 RAC for a
single P4 seems pretty far fetched to me, even if there
are 4 of them merged...

---

Good catch Legacy. It is possible to run up a lot of credit in a short period of time (I have personally had 39,000 credits granted in a single day, shortly after a server outage), but in light of no recent problems, it does seem fishy. I have about 30 active PC's, and I earn roughly 17000-18000 credit per day, but no single machine has a RAC higher than 285. One thing to keep in mind here, most of my PC's are work systems as well, meaning they are in use all day long....

---

<a><img src="http://boinc.mundayweb.com/stats.php?userID=525">

<a><img src="http://www.wombatradio.com/stats/rtb/sig.php">

> Created 9 Oct 2004 12:56:53 UTC
> Total Credit 148169.72
> Recent average credit 894.74
> http://setiweb.ssl.berkeley.edu/show_host_detail.php?hostid=239985

That one is interesting. I am checking his returned results. Am now on offset=1300 and still the download files are errors on the 10th of November. It almost looks as if the download limit doesn't apply for this person. Over 2600 units downloaded and returned with a Client Error? Come on!

---

One: Yes it appears the person has merged several PCs whos CPUs are same speed and using same O/S. For my estimate I would guess 4 similar machines.

Two: Person has recompiled boinc client to grant higher benchmark scores. Optimized, changed source possibly. No way to be sure.
The calculations for those benchmark scores (1287 Float, 3819 integer) computes correctly when multiplied by CPU time. Which means with client 4.05 typically this "computer" was claiming 45 to 57 credits per WU.
What I don't understand is how those numbers were not allways the highest (and thus thrown out) for credit granting. But checking a few WUs it seems they were not allways highest.

The seconds per WU seems like a reasonable number for a machine of that CPU and O/S (2.53 MHz Pentium 4, non-HT, Windows XP) for a 4.05 client.

The Recent Average Credit (1123.54) is VERY unusual. Even if the user had not submitted results for 2 weeks and then submitted them all at once (and they were then verified all at once) the RAC would not be that high. So again I think maybe 4 machines.

How they get around the max 50 WUs per host #, I don't have any clue.

---

> >
> > No way in hell that machine can do 1123 RAC!!!
> >
>
> Then you merge 2 machines the RAC is artificially increased also.
>

Ingleside

You are very well informed and most of the time you are correct. I always pay attention to your comments and I have learned from them. In this occasion I believe you are wrong.

When merging machines the Total credit is the sum of both machines, but the RAC does not change and the RAC of the original machine remains the same. for example:

Machine A: Host=12; Total credit=24; RAC=8 merging with
Machine B: Host=22; Total credit=3; RAC=2, gives
Machine C: Host=22; Total credit=27; RAC=8

Is this correct?

---

In that case, SETI should change "Top Computers" to "Who Merged The Most Computers"

---