[OneMacGuy]

So Reader50, anyone on our teams ever reached the magic 1000 Ghz or 1 Thz threshold before? Looks like Mikkyo is about to any day now. Do your scripting formulas consider that high a number?

What kind and quantity of hardware are you doing all this on Mikkyo? Come on, fess up?

[mikkyo]

Quad G5s seem to do quite well

[fran]

Originally Posted by OneMacGuy

So Reader50, anyone on our teams ever reached the magic 1000 Ghz or 1 Thz threshold before? Looks like Mikkyo is about to any day now. Do your scripting formulas consider that high a number?

Where are you seeing this?

[OneMacGuy]

Originally Posted by zombie67

Where are you seeing this?

Mikkyo's Rosetta Stats:

http://teamstats.macnn.com/rosetta1/...page=p2&UID=80

As of 2:00 PM EST on 6/13/2006, Mikkyo was 974 Mhz power rating and still rising. He is number on 20 on the top producers list in Rosetta and has only been doing Rosetta for around a month!

[fran]

Found it! Rosetta. I thought that we didn't have a stats page for Rosetta, because it's not a link at the top of the other stats pages. But if you go to the top of the site at http://teamstats.macnn.com/ , you can see it listed there.

Very impressive!

...which leads me to another question. Why do some people hide their hosts? I assume there is a reason (beyond frustrating me ).

[reader50]

Code:
# Power Project
$temp1 = $UserArray['PowerProject'];
if ($temp1 > 0)
{
  $temp2 = number_format($temp1, 0)." MHz";
  if ($temp1 > 1000) $temp2 = number_format(($temp1/1000), 1)." GHz";
  if ($temp1 > 1000000) $temp2 = number_format(($temp1/1000000), 1)." THz";
  echo 'Power: <b>'.$temp2.'</b>';
}

We can handle the higher numbers. In other areas (the quickstats bar specifically) we are already coded for the Peta suffix to appear as needed.

zombie67, thanks for pointing out that Rosetta is missing from the nav bar. It's not supposed to be. This will be fixed.

As for the THz, Rosetta is almost certainly overestimating Power. If someone will post the stats a given box generates in Rosetta per week, I'll fix the estimate. We could use some current info on stats-per-box on most projects, especially BOINC projects.

[mikkyo]

Opterons and Xeons are pretty fast at Rosetta too

[OneMacGuy]

Originally Posted by reader50

As for the THz, Rosetta is almost certainly overestimating Power. If someone will post the stats a given box generates in Rosetta per week, I'll fix the estimate. We could use some current info on stats-per-box on most projects, especially BOINC projects.

I can give you stats on any type of box you want except a G5, howver you have to specify whether ou want them done using the stock client or one of the optimized clients. My PC's are using the crunch3r optiized clients.

[reader50]

Whichever client version is more typically used. I'll probably use an average rate that falls somewhere between the stock & optimized clients. It needs to be an average anyway so it falls between different CPU architectures.

As each of our BOINC stats were added, we didn't have separate stats-per-box estimates available for each new project. So with the exception of Predictor (our first BOINC stats), I just reused the Predictor estimate for all later BOINC projects.

This assumed that BOINC awarded the same stats per box per day regardless of project. Which may be true with the newer flops-stats system Berkeley is using now, but it's almost certainly wrong for the other projects.

If you can offer some more precise estimates, all our users will benefit. Thanks for anything you can come up with.

[kenn]

There's NO WAY I'm doing 29.8 GHz on my Sawtooth. I may be wearing cement boots, but who else on the team is crunching away on a solitary computer? There's a certain pride in knowing I'm still in the top 10 with my six year old G4.

[reader50]

kenn, what CPU card are you using, and what percentage of the time does your Sawtooth crunch on your 29.8 GHz project? ie - 20 hours per day crunching on SETI.

This is the sort of info needed to update the power estimates. Downside, truthful answers will cause your extra GHz (or THz) to melt away.

[mikkyo]

Id say a 1.5GHz G4 crunching optimized full time does around 600 a day.
A Xeon 2.40GHz or a Dual Core AMD Opteron 270 crunching optimized full time does around 3k a day.
A quad G5 crunching optimized full time does around 4k a day.

[OneMacGuy]

Dual G4 1.2 - Optimized Client - RAC 800
P4 3.0 Ghz HT - crunch3r optimized client - RAC 750
Athlon 64 3200 - crunch3r optimized client - RAC 650
eMac, 1 Ghz G4 with only 256K Cache - Optimized Client - RAC 320.
Athlon XP 3000 - crunch3r optimized client - RAC 630
These are all running 24/7.

[reader50]

I'll do some updates tonight. Hope we have our user accounts back by then, as I post this, I'm a guest too.

On the plus side, our forum may be the only one on MacNN where we can keep posting, because we allow unregistered users to post.

[arkayn]

I wonder what I would be doing if I only ran SETI on my machines.

[reader50]

The navigation bar has been updated.

The SETI-BOINC power estimate has been updated too, with a new compromise estimate for assorted platforms. mikkyo will not be getting his THz label until he turns back on the other half of his farm, and adds a few more boxes besides. The estimate was cut to about 42% of what it was before (cut to 3/7 to be precise).

If anyone has updated crunch rates for projects other than SETI-BOINC, feel free to post them. I'll be happy to clip power overestimates in other projects too.

Thanks to mikkyo and OneMacGuy for the current info. Better thank mikkyo especially, that post cost him ~550 GHz. It takes willpower to make a post like that.