What are your points per [day per] watt (PP[DP]W)?

A forum just for SPCR's folding team... by request.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

scdr
Posts: 336
Joined: Thu Oct 21, 2004 4:49 pm
Location: Upper left hand corner, USA

Post by scdr » Sun Oct 14, 2007 12:45 pm

VanWaGuy wrote:Thanks scdr for the upbeat message.
You're welcome
VanWaGuy wrote:I see that even you turned on an evil power wasting device to be able to browse this thread and post a reply.
Hmm - I don't see computers as evil power-wasting devices, but to each their own.
I do think people should consider the consequences of their actions, and it seems worthwhile to try to raise awareness of those consequences.
Besides, running back of the envelope calculations like this is fun.
VanWaGuy wrote: Tuning my computers, and seeing the results of the point output, and a little friendly competition amongs the other folders is enjoyable.
Enjoy.

scdr
Posts: 336
Joined: Thu Oct 21, 2004 4:49 pm
Location: Upper left hand corner, USA

Post by scdr » Sun Oct 14, 2007 1:02 pm

In the PPDPW reporting, it might also be worth also noting the CPU
voltage you are using. (whether stock voltage, or give actual voltage if under/over volted). Since power usage depends on CPU voltage.

viewtopic.php?t=41883&postdays=0&postorder=asc&start=0

A question that came to mind while reading that thread is, how
does PPDPW (e.g.) vary with under volt/clocking. Do you get the most
performance per watt when undervolted at stock speed, or can one do
better with minimal stable voltage and under (or over)clocking?

One of the strange ideas that seems to be mentioned often in folding
discussions is this concept of "wasted cycles." As if there were all these
cycles that are just going into the bit bucket if they don't get used.
Since most recent CPUs and operating systems execute low-powered
instructions when they are doing nothing, and many use lower power
states when not busy (reduced clock frequency, reduced voltage, etc...)
the idea of wasted cycles is a bit of an anachronism. Certainly an idle
computer still uses power, but it uses even more when it is busy.
In terms of minimizing electric usage, you want to keep it off as much as
possible, and when on want to keep in low power state.

But if going to do heavy crunching with it - where is the sweet spot - where it gets the most done for least power?

VanWaGuy
Posts: 299
Joined: Mon Sep 11, 2006 1:01 am
Location: Vancouver Wa USA

Post by VanWaGuy » Mon Oct 15, 2007 5:29 am

scdr,

yes, technically you are correct in that most newer CPUs and OSes together let the computer idle at lower power, so for some systems, the "wasted cycles" are not completely free. As you suggest, from one perspective, that phrase is getting a bit dated.

I do not have one, but I have heard for example though that a PS3 uses about 200 watts whether it is idle or folding, so the phrase may not be totally obsolete yet.

From another perspective, while my computer is on, it could be doing productive calculations or idling. Folding certainly keeps the computer doing something arguably more valuable than idling, although as you say, on systems with lower power idle, it is not free.

Still, that potential to be doing work is being wasted, so wasted cycles may still apply.

I have been a computer hobbyist for over 30 years. During the first many years of that time, we believed that leaving the computer on was cheaper than turning it off. Between the mechanical wear to switches (replaced the power switch on my Apple II+ three times) and hard drives, also just the in-rush currents and powering up seemed to be where a lot of the failures occured, and so the cost of the extra power was believed to be less than the price of replacing hardware.

Now, I might be a couple decades out of date with that belief, but I am still more comfortable leaving my main system on.

I should browse a bit and see the recent thinking on this. Drives will spin themselves down now anyway, and monitors will dim themselves, so powering down a system is probably not as bad as it used to be, and power costs more and hardware costs less.

I may keep folding, but you made some good points, and got me thinking more about this.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Sat Oct 20, 2007 4:33 pm

This is odd. After updating the BIOS on my AMD X2 I did not manage to get it to OC any faster (the BIOS came from a advocate OC website). Other managed large gains after the BIOS updates. What I did manage was 2 things with the same speed settings for the CPU and memory:

1. My checkpoints have slowed, take longer.

2. I'm running on less power, from 111 down to 95 watts, fully loaded.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Tue Dec 04, 2007 6:58 am

aristide1 wrote:This is odd. After updating the BIOS on my AMD X2 I did not manage to get it to OC any faster (the BIOS came from a advocate OC website). Other managed large gains after the BIOS updates. What I did manage was 2 things with the same speed settings for the CPU and memory:

1. My checkpoints have slowed, take longer.

2. I'm running on less power, from 111 down to 95 watts, fully loaded.
11.35 for this folder. But I plan on installing window and running SMP as I have more OC options in Windows. The BIOS will promptly go into it's own overly protective "safe mode" if I OC any higher than the 230MHz FSB I have now. In Windows, people boot at a lower FSB to get passed safe mode and then OC using ClockGen or tools that came with the motherboard.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Tue Dec 04, 2007 7:09 am

scdr wrote:I am glad you started this thread. With the advent of the computers and environment forum, I was wondering if the SPCR folding folks were looking at environmental impact of the project. (Just a little looking on the folding boards suggests they haven't considered it much.)

Some back of the envelope calculations based on the figures here -
If we assumed that 10 PPDPW was typical (probably overestimating efficiency - especially for older Folds).
Then the 64.5 million Points produced so far by the SPCR team (at the time of this writing) translate to 6.45MW-days (64.5M points/10PPDPW)

Converting to MW-Hours gives 155 MegaWatt-Hours (6.45MWdays * 24hrs/day)

Assuming electricity costs $0.15/KW-Hour (per Folding at Home site)

Team SPCRs contribution comes to something like $23,000
and about 100 Metric tons of CO2

A rough eyeball of the overall Folding statistics
suggests that overall total is currently about 6 billion points

(I didn't find where in the statistics they summarize total points - but I probably just didn't look in the right place).

Figuring that as about 100 times the
SPCR teams contribution - the total comes to
about 15GW-Hours, $2million, and 10,000 Metric tons CO2
(to say nothing of energy cost for communications, their servers, etc.)

Seems like the folding stats sites should also point out the financial and environmental impact of this affair.
(I am not saying it isn't worth while, but it should be more clearly addressed. They even try to disguise their electricity cost estimate of $130/year by breaking it down to a daily figure.

I can think of plenty of things with surer payofs for humanity than Folding to donate $130/year to.)
I realize it's late to respond to this thread but some points I would like to make:

1. If the PC is on for other reasons (like a server) then the costs of FAH is simply power at full usage minus power at idle. SPRC staff, can you hear me now?

2. All comments about the impact of this electrical usage applies ot all pc's. Gamers, one could argue, are not contributing to anything positive for society, and they have 1000+KW power supplies, dual GPU cards that suck up 200 watts each.

3. PS3s have horrible numbers.

4. While WalMart has gone on a campaign (as has the industry in general) about energy efficient light bulbs millions not only continue to pc's with non-APFC power supplies, but they aren't even aware of APFC power supplies, or they are not offered as an option. Dell, can you hear me now?

stevew
Posts: 6
Joined: Tue Jul 24, 2007 8:35 am

Post by stevew » Tue Dec 04, 2007 8:25 am

While I've yet to see much that comes close to a Mac mini 1.86 GHz running SMP cores in PPD/watt, how about some (back of envelope) calculations regarding FLOPS/watt. My quick look at the http://fah-web.stanford.edu/cgi-bin/mai ... pe=osstats page tells me that the PS3s are doing about 24x more FLOPS than the PCs. Yes each cpu has advantages and PS3s can only do a narrow slice of the computing needed; but they're really fast.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Wed Dec 05, 2007 7:47 am

While we dealing with effiencies we could also factor in actual costs by including the price of a KW of electricity. Of course our Canadian friends would have a real advantage on that front (not blue). 8)

For those of us who may possibly have dual rates, high peak rates and low off peak and weekend rates, one could conceivably fold only during off peak rates, especially if a quad core could complete a WU overnight, and then be shut off all day during peak hours.

scdr
Posts: 336
Joined: Thu Oct 21, 2004 4:49 pm
Location: Upper left hand corner, USA

Post by scdr » Wed Dec 05, 2007 3:34 pm

aristide1 wrote: 1. If the PC is on for other reasons (like a server) then the costs of FAH is simply power at full usage minus power at idle. SPRC staff, can you hear me now?
Yes it would be sort of nice to have stats on the idle power to go along with the PPDPW power as a comparison. Might help to clarify some of the anachronism of "unused cycles."
Though seems like people could do even better by shutting of servers when they are not needed. See various posts about folks running home servers 24/7, when most of the applications don't need that.
(Wake on LAN and scheduled wakeups could be used more effectively to allow servers to be off more.)
aristide1 wrote: 2. All comments about the impact of this electrical usage applies ot all pc's. Gamers, one could argue, are not contributing to anything positive for society, and they have 1000+KW power supplies, dual GPU cards that suck up 200 watts each.
Yes, environmental impact of gaming would also be a nice/handy analysis. Know any good sources of stats to start from? Might be easiest to take on some of the game consoles - more standardized hardware (though the display would be a big variable.)
It would be fun if the games/consoles had an energy counter - initialized to the manufacturing/recycling cost of the console. Ticks up as you use the thing, etc.
Though even cooler might be a calorie counter - displays the time you have been playing (session and total), how many calories you burned, how many calories Laura Croft (or April Ryan, or whoever) burned, and how many you would have burned if you spent the time walking/cycling/weight training. (And, if it is calculable - about how much that equals in weight loss or muscle built ;-). Could also include how much you could have read of the complete works of W. Shakespeare, M. Twain, T. Pratchett or other great literature you could have read in that time (to give equal time to the health of the mind).
But that gets a bit OT for this forum. (Well if it got folks more active, it might have a more immediate impact on raising health than folding. So maybe it isn't that far OT. ;-)

Post Reply