Folding farm power consumption

[pdf27]

Out of boredom more than anything else last night, I decided to sketch up a plan for a folding farm. Since I'm currently an impecunious student I'm not planning on actually building it, but it was interesting to do none the less.

The major point that struck me was that (assuming electricity at 8p/kWh) a cheap blade could well use it's initial purchase price in electricity over the course of a year to 18 months.

This then led me on to the obvious next question - what chip is most efficient in terms of points per kWh? It isn't something I can really answer myself but it's really been bugging me so I thought I'd post it here and see what everyone thinks!

[NullObject]

Out of boredom more than anything else last night, I decided to sketch up a plan for a folding farm. Since I'm currently an impecunious student I'm not planning on actually building it, but it was interesting to do none the less.

The major point that struck me was that (assuming electricity at 8p/kWh) a cheap blade could well use it's initial purchase price in electricity over the course of a year to 18 months.

This then led me on to the obvious next question - what chip is most efficient in terms of points per kWh? It isn't something I can really answer myself but it's really been bugging me so I thought I'd post it here and see what everyone thinks!

In terms of overall powerconsumption and point production, I believe dual Xeons are the most efficent. In general Intel chips are more energy/ppw efficent than AMDs. sbabb mentioned that a few months ago. In this [H]ard OCP thread the poster compares 2 year costs for producting 50,000 ppy for various Intel and AMD processors. Using his cost assumptions, Intel gains the advantage sometime between year 1 and year 2.

The number 1 daily point producer, OC-AMD, switched his farm to Intel from AMD to cut down on energy costs. Thread here

[dukla2000]

Dont know what the Intel folks will offer, but in AMD land you want the most MHz at the lowest VCore.

Just to put some boundaries in: as this is a farm I would assume a no-frills mobo that doesn't provide overclocking or VCore changes.

Then my guess would be a Duron 1800 or a mobile Barton. Most likely the Barton cause it is 1.45VCore, but based on the price premium you could get a pair of Durons and change. Even more interesting would be the most points/£ over say a 3 year life and includes the purchase price!

Not sure what your numbers were: my electricity is 6.17p/kW and at that rate I reckon it costs £130 to keep a (120W AC) box folding for a year.

[haysdb]

I need to put the Kill-a-Watt meter back on my Athlon systems to verify this, but I seem to remember my 3200+ Barton drawing around 135W, vs. 118W for a P4 3.1GHz (with the same model power supply). Since the P4 produces more PPW and uses less electricity, at some point the P4 would pay for itself. But the Intel blade was considerably more expensive, ~$375 vs $200, so I'm not sure I can justify buying any more of them, but I haven't really analyzed the costs over a period of time either.

Another wildcard is that right now, HT enabled P4's have a 15% to 20% advantage over an otherwise equivalent Athlon, but Vijay Pande has dropped a strong hint that soon, running more than one client, even on an HT enabled system, could result in LESS points than a single client. Only Vijay knows what this means. Call it the "X factor" when doing a cost/performance analysis.

Someone will have to convince me that a dual Xeon system is cost effective. I have an open mind on the subject, but I am skeptical.

David

[NullObject]

I'm running one dual Xeon machine, and while its producing quite a lot of points, I do not think its that cost effective, I think AMD wins the points-per-$ race actually. Note: I don't run any AMD-machines; currently, one P4 2.4C, one P4 2.8C and one dual Xeon 2.6, all with HT enabled.

What Vijay said about HT machines getting less points running clients on a virtual cpu is worrying (<-- is that a word?), and will definately affect me should it come to be true.

I agree with haysdb about a dual Xeon being cost-effective, I'm sceptical. But its definately effective on a ppw basis, disregarding the cost of the cpu's and mobo (and power consumption).

Yes AMD wins the pp$ race when it comes to initial costs, but you have to run more AMD machines to produce the same amount of points. This results in more $ in electricity. I have done neither any calculations nor any testing to see what the results would be between various Intel and AMD setups. In this 2cpu.com seti electrical costs thread, OC-Xeon, who folds as OC-AMD, had the electrical company come out and take measurements without his folding farm running and with his folding farm running. I think this was before he switched over to Intel because he lists mostly AMD processors. At the time his electrical costs went up $.17 an hour.

In referring to a dual Xeon setup, I wasn't including initial costs but only referring to electrical costs. Dual processor computers use less electricity than 2 computers of equal power. The same would probably be true of a dual AMD setup vs. 2 AMD computers.