Rosie RAC Harvest - Status/DiscussionThread

[baditude_df]

STATUS UPDATE!
Team RAC----> 82,712 and #10 in the WORLD! Up from 48,546 on the 1st of the month! Only 17,288 to go our goal of 100K!
Distance to Team China---> 694727pts 15.47 days to goal at our current rate.

That's a 4173 point jump in RAC, 95689 total points and two places in the World RAC Standings in one day!
HOLY COW!

 

[drshivas]

great job folks! keep crunching!

 

[Bobnova]

The x2 4400+ is starting to turn results in, i'm happy

 

[QuietIce]

I don't know what scheme Rosetta uses for RAC but if it's anything like SETI then we've still got a big increase coming. We're showing an RAC of ~84K right now but BOINCStats has us at a DAC* of 97K and growing. With the SETI RAC equation it takes about 2 weeks for a machine to start winding up and another week or so to hit full stride. I originally started posting the DAC so that any changes will show up sooner. As time went by I found that DAC was close to a week ahead of RAC, not perfect but a fair reflection. Rosetta may calculate RAC differently, though - the SETI equation is a monster that must have been a nightmare of a CalTech mathematician.


*DAC is a term I coined for BOINCStats "Credit/week" divided by 7, which yields a Daily Average Credit over the past week.

 

[baditude_df]

Daily Production(Daily Average Credit) from what you describe, and RAC in the Rosetta World sound like they would be pretty darn similar, if not exactly the same as Seti then.

 

[drshivas]

As a data point to this discussion, when I crunched in the Gauntlet this year, my RAC peaked around 19K RAC. In Rosetta, I probably won't break 11K with these same rigs. So there's definitely a difference somewhere.

As a side note, my two Win7 Pro x64 quads are cranking out 500 RAC more than they did with Win2K3 32bit. In fact, you can see it in my stats now: http://boinc.bakerlab.org/rosetta/hosts_user.php?userid=284352&show_all=0&sort=expavg_credit (Two are clocked at 3.4, one at 3.6, with the 3.6 being one Win7)

Anyone else finding Win7 x64 is a better crunching OS?

 

[hokiealumnus]

It's better than XP 32-bit, I can tell you that much. Still not as good as Linux (at least with F@H, never crunched Rosetta in Linux and it's not on any systems here presently).

 

[muddocktor]

I just cranked up a Q6600 system for the contest and 30 minutes after firing it up the damn D4 pump on the system crapped out and started making some awful noises. Luckily I had a spare D4 and just brought the system to the kitchen and pulled the pump out of the case and changed the pump motors out without having to take the housing loose from the tubing. After swapping the motors out and replacing the little bit of water lost doing that, I have the system back up and crunching some Rosie units again. And I swapped my Mac Mini over to Rosie too.

 

[Bobnova]

Got an Athlon II 245 headed my way destined for my mom. Guess what it's going to do till Dec. 1st?

 

[nzaneb]

Our DAC has us at 7th in the world, and moving in on Canada, to take 6th.

 

[baditude_df]

As a data point to this discussion, when I crunched in the Gauntlet this year, my RAC peaked around 19K RAC. In Rosetta, I probably won't break 11K with these same rigs. So there's definitely a difference somewhere.

As a side note, my two Win7 Pro x64 quads are cranking out 500 RAC more than they did with Win2K3 32bit. In fact, you can see it in my stats now: http://boinc.bakerlab.org/rosetta/hosts_user.php?userid=284352&show_all=0&sort=expavg_credit (Two are clocked at 3.4, one at 3.6, with the 3.6 being one Win7)

Anyone else finding Win7 x64 is a better crunching OS?

What about the WU's? That would be my first guess right off the bat. But maybe I'm way off base there as I know nothing about Seti other than it is a BIONC based application.

I am crunching Windows 7 x64 RC1 (7100) and this 950 at 4.2GHzHT is pulling 4000RAC by itself and rising fast. I have no reference though, since I have only ever used Windows7 x64 with this 950 (or any i7 for that matter).

PS: THX FOR THE EXTRA HORSEPOWER MUD and BOB!

 

[QuietIce]

As a data point to this discussion, when I crunched in the Gauntlet this year, my RAC peaked around 19K RAC. In Rosetta, I probably won't break 11K with these same rigs. So there's definitely a difference somewhere.

As a side note, my two Win7 Pro x64 quads are cranking out 500 RAC more than they did with Win2K3 32bit. In fact, you can see it in my stats now: http://boinc.bakerlab.org/rosetta/hosts_user.php?userid=284352&show_all=0&sort=expavg_credit (Two are clocked at 3.4, one at 3.6, with the 3.6 being one Win7)

Anyone else finding Win7 x64 is a better crunching OS?

The difference is the optimizers available for SETI crunching. If you had been running only the base SETI app you wouldn't have chalked up nearly the RAC you did. AFAIK no project has code to match the optimized SETI apps with CPU specific routines and all that goes with them. From what little I've read Einstein comes closest to SETI only because the project leaders there update their app frequently and attempt to take advantage of SSE et al as much as possible ...

 

[baditude_df]

STATUS UPDATE!
Team RAC----> 87,488 and #8 in the WORLD! Up from 48,546 on the 1st of the month! Only 12,512 to go our goal of 100K!
Distance to Team China---> 593097pts 12.27 days to goal at our current rate.

That's a 4776 point jump in RAC, 101630 total points and two places in the World RAC Standings since my last update 24.5hrs ago!
WELL DONE, Keep it coming!

PS: Congratulations Pixie on becoming our first 3MIL+ Cruncher! You ROCK!

2) pixie  	3,072,302  	8,461.94  	United States

 

[Transparent Elf]

nice one pixie! where do you get all those quads?!

 

[Sir-Epix]

My RAC is simply a flash in a pan. It'll be gone (or drastically cut) before you know it. But It sure is fun to put some distance between us and VTC. Man, those guys are rocking! Hopefully, their RAC will take a nosedive when the holiday season starts.

When does Thanksgiving start again?

Anyway, all rigs switched to R@H except for 2. I have some leftover WUs from the Gauntlet that needs finishing. I switched to F@H and forgot all about them. Man, they have very long due dates. I was hoping they would've been recycled by now.




Hey Sir-Epix. Sorry to hear about the pesky trojans and HW issues.(must've been watching more than Hulu lately ... j/k Pls. don't shoot me. )

Yeah I was watching Youtube.com watching a tutorial about making an external USB HD bootable to reinstall vista. Alsa they somehow got an automatic download to go and well you know the rest!

 

[nzaneb]

The difference is the optimizers available for SETI crunching.

What about the WU's? That would be my first guess right off the bat. But maybe I'm way off base there as I know nothing about Seti other than it is a BIONC based application.

It's the optimizers, like QuietIce said. Seti's work units grant more credit, on a credit per time crunched ratio, because of this. Some are optimized so efficiently for SSE3, that they grant MUCH more credit. A Seti AstroPulse (or AP) WU, grants almost double the credit compared to an already optimized MultiBeam WU (normal), using the credit granted/minutes crunched ratio. Granted, a single WU will take 10-13 hours on a 4ghz overclocked i7 thread, and up over 48 hours on some rigs; but the credit payout is quite worth it (usually 1200+ credits per AP WU). As of late they are pretty hard to come by though.
From what I've read, Rosetta does not utilize the SSE instruction sets, so optimizing is next to impossible.

 

[drshivas]

Ah, makes sense, QuietIce and nzaneb. I forgot about that I optimized when I ran the Gauntlet. I wonder if Rosetta will ever consider using such instruction sets. I guess the project would have to grow in such a way that they would have no choice but to write new code to take advantage of such instruction sets.

 

[QuietIce]

Ah, makes sense, QuietIce and nzaneb. I forgot about that I optimized when I ran the Gauntlet. I wonder if Rosetta will ever consider using such instruction sets. I guess the project would have to grow in such a way that they would have no choice but to write new code to take advantage of such instruction sets.

SETI@Home doesn't use the advanced instruction sets at all - it's all done by 3rd party. The guys over at Lunatics and Crunch3r (Alex Klein? from the AK_v8 SETI optimizer) are responsible for most of them and have done a great job over the years! All it takes for Rosetta to have optimizers is for someone to sit down and write them. If Rosetta attracts enough lunatics I'm sure we'll eventually see Rosetta optimizers ...

 

[baditude_df]

STATUS UPDATE!
Team RAC----> 91,060 and #8 in the WORLD! Up from 48,546 on the 1st of the month! Only 8940 to go our goal of 100K!
Distance to Team China---> 509018pts 9.68 days to goal at our current rate. WE'RE NOW IN STRIKING RANGE, AND OVERTAKE IS IMMINENT.

That's a 3572 point jump in RAC, 84079 total points and only 5.3K from #7 in the World RAC Standings since my last update 22.85hrs ago!
KEEP CLIMBING!

 

[Bobnova]

Go us!
I hope to have a nice shiny new Athlon II 245 going before the end.