TR Forums

Actually, I am using the POEM @ Home program which is similar to folding at home but is optimized for AMD video cards. I'm currently running a Dell Workstation with Core 2 Quad Extreme QX9650 @ stock 3ghz with an HD 7850 card. IF I eBay the workstion and run an ivy or sandy bridge rig with this same card 24/7 how much money per month do you guys think I would save on my electric bill?
Using Boinc and POEM@Home this system is generating 114,000 Points per day ( I do not know how much the Folding @ Home equivillant is, as no one seems to know the correct ratio....BUT this card would only do about 10k PPD folding @ home max).

Here's a link:

http://boinc.berkeley.edu/wiki/ATI_Radeon

I'm also interested in the monthly energy savings/cost and where you guys live.

Atlanta

It depends on your setup, but my guess is little or none.

Its like asking about the fuel savings on your range rover when you change tires. sure, if you go from mud tires to street, there could be a noticeable difference (+/- 10), but more than likely you'll get a similar tdp chip that really doesn't effect overhead much.

Your GPU is what sucks power like a mother. for the record, your chip is a 130W, a 3570k is 77W.

77w x 24hrs x 365 days / 12months /1000 w to kw conversion = 56 kWh per month
130 x 24 x 365 /12 /1000 = 94.9 kWh per month

Atlanta charges 0.145 per kwh http://www.bls.gov/ro4/aepatl.htm

so
qx9650 13.76 per month
3570k 8.12 per month

Thanks, the savings of only $5 a month is not worth the added expense. Much appreciated!

You also have to consider the power supply efficiency, which is probably about 80%-85%. So your saving would be about $6 a month.

My Dad is big into folding & other distributed projects that use both CPUs and GPUs for different workloads. On his CPU-bound workloads he recently upgraded from a Core 2 9550 to an Ivy Bridge 3770 (both chips non-overclocked) and got about a 3x performance improvement on the CPU side, plus the machine uses quite a bit less power (it is a dedicated 24/7 machine). He also has a 3570 setup that is about 2.5 times faster than the Core 2 was.

Ivy also has the benefit of PCIe 3.0 so you can probably stack in more GPUs effectively if you have the motherboard slots for them.

chuckula wrote:

My Dad is big into folding & other distributed projects that use both CPUs and GPUs for different workloads. On his CPU-bound workloads he recently upgraded from a Core 2 9550 to an Ivy Bridge 3770 (both chips non-overclocked) and got about a 3x performance improvement on the CPU side, plus the machine uses quite a bit less power (it is a dedicated 24/7 machine). He also has a 3570 setup that is about 2.5 times faster than the Core 2 was.

Ivy also has the benefit of PCIe 3.0 so you can probably stack in more GPUs effectively if you have the motherboard slots for them.

This means that $ savings (in terms of power usage, let's forget initial acquisition costs for a moment) may not be significant, but performance/watt(or $) is much better. So is the added points good vanity enough for you to upgrade? I suppose this depends on how your geek factor is influencing you!

Why is there such a large difference between the core 2 quad and an ivy quad for folding (CPU Client)??????? The IPC isn't that much greater such that a 300% performance boost would occur. Is the client optimized differently for Ivy somehow?

I recently tried to run the Boinc POEM @ Home project (GPU with HD7850 scoring 120,000 PPD) while simultaneously running the Folding @ Home CPU client and only had 500 ppd. Did I do something wrong:


CPU Intel Core 2 extreme QX9650 @ stock 3ghz
AMD HD7850 with 10% overclock
4 gigs ram

P.S.

I definately would not pay a grand for e-pene rites. However a 3 x performance boost might change things a bit. Is this accurate?

I have no idea on the ppd/energy/etc part of things, but there are a few reasons why a Sandy or Ivy quad would run down a C2Q pretty hard, though 3X/300% sounds a bit much.

Simplest one? The Ivy/SB quad has more cache, and better optimized. It also has higher clocks and clock modulation/turbo boost.

Slightly less simple but more important; a C2Q is dual dual-core CPUs. A SB/Ivy quad is a real quad. C2Qs aren't quads, they're two duals, joined only at the FSB, and they happen to share a substrate/socket. They're fully independent dice, though.

Three words: early return bonus.

Because the OP is not talking about F@h, but a program LIKE F@h (Poem) which is GPU based, all of the discussion about performance per watt is moot (unless there is a cpu bottleneck(there shouldn't be)). In fact, honestly the discussion about switching CPUs to save power is a little moot anyway because the CPU is probably not kept at 100% 24/7. it might be kept at 20-40% if that. if the OP could under-volt his CPU, disable HT, down-clock the cpu, etc, he could maximize his savings without spending a penny.

Diplomacy42 wrote:

Because the OP is not talking about F@h, but a program LIKE F@h (Poem) which is GPU based, all of the discussion about performance per watt is moot (unless there is a cpu bottleneck(there shouldn't be)). In fact, honestly the discussion about switching CPUs to save power is a little moot anyway because the CPU is probably not kept at 100% 24/7. it might be kept at 20-40% if that. if the OP could under-volt his CPU, disable HT, down-clock the cpu, etc, he could maximize his savings without spending a penny.

Your point is valid, but HT is already off, EIST is doing undervolting and underclocking automatically.

Core 2 Quad was a bad place to be if you were seeking power efficiency.

Forge wrote:

...Core 2 Quad was a bad place to be if you were seeking power efficiency.

Oh, I just can't wait for my electric bill to arrive!

BIF wrote:

Forge wrote:

...Core 2 Quad was a bad place to be if you were seeking power efficiency.

Oh, I just can't wait for my electric bill to arrive!

Hey now, I wasn't saying it was as bad as Pentium D, just that it wasn't that awesome, either.

They were great CPUs overall. They just weren't awesome at power. In particular, Intel had to dump a lot of the deepest idle/sleep stuff to get two CPUs in one socket to work.

Forge wrote:

...Intel had to dump a lot of the deepest idle/sleep stuff to get two CPUs in one socket to work.

Yep, and yet they did do a good job, because mine's been in service for far longer than any prior system I've ever built/owned. It has been so troublefree that I never ever think of it as being two chips sharing a slot, but that is precisely what it is!