How much juice does folding use?

[haste.]

Hello everyone,

I'm a little curious on how much power folding uses? Does it tax the CPU and GPU at 100%? I have a Phenom II 960t and GTX 580 rig that sits idle 98% of the time. I already pay $200/month in utilities and was curious how much it would go up...

I've decided I can't handle Tom's anymore as it's become 3 days late and 90% Apple fanboyism. I've decided to take on [H] as my new hardware go to. Figured I could contribute with a bit of folding for team #33 if I can afford it...

 

[rflcptr]

Depends on:

a) what your PC draws from the wall under load
b) the energy rates ($ per watt/hr)
c) how long you plan to run it each month

 

[LigTasm]

Yes it will tax your hardware to 100%. You will want to choose either GPU folding or SMP (CPU) folding, not both. More than likely folding the 580 would net you the best PPD, but it will also run your power bill up much faster than folding the CPU alone. For questions along this line, you can also go to the Distributed Computing forum at the bottom of the page.

 

[haste.]

Ok so it runs at 100% on one of the two.

I just realized I reference an AMD processor in the Intel forum. I only have one AMD rig and I didn't even think about it till after posting. I apologize!

 

[RamonGTP]

Yes it will tax your hardware to 100%. You will want to choose either GPU folding or SMP (CPU) folding, not both. More than likely folding the 580 would net you the best PPD, but it will also run your power bill up much faster than folding the CPU alone. For questions along this line, you can also go to the Distributed Computing forum at the bottom of the page.

You would absolutly want to do both unless you're running a sinlge core CPU.

 

[LigTasm]

You would absolutly want to do both unless you're running a sinlge core CPU.

Not on that setup for sure, and usually not on any unless you dedicate a core to GPU use. Trying to run both at the same time will cause both to lose performance. I have dedicated GPU folding rigs so this isn't an issue, but when you combine the two the total isn't high enough compared to SMP only or GPU only to warrant the extra wattage used.

Example: My 3770k machine with a GTX 680. The 3770k can average around 30k PPD, and the 680 can do about 20k by itself. If I fold both at the same time the 3770k drops to around 16k, but it increases my power draw from 160w to 420w while only increasing total PPD by 6k, not to mention the longer turn in times which can cause you to miss the qrb on A4 units.

 

[jojo69]

ALL TEH JUICE!!!

 

[RamonGTP]

Not on that setup for sure, and usually not on any unless you dedicate a core to GPU use. Trying to run both at the same time will cause both to lose performance. I have dedicated GPU folding rigs so this isn't an issue, but when you combine the two the total isn't high enough compared to SMP only or GPU only to warrant the extra wattage used.

Example: My 3770k machine with a GTX 680. The 3770k can average around 30k PPD, and the 680 can do about 20k by itself. If I fold both at the same time the 3770k drops to around 16k, but it increases my power draw from 160w to 420w while only increasing total PPD by 6k, not to mention the longer turn in times which can cause you to miss the qrb on A4 units.

Interesting, I have an identical setup and getting 40k ppd with both on and significanly less with just one on at any given time.

 

[LigTasm]

Interesting, I have an identical setup and getting 40k ppd with both on and significanly less with just one on at any given time.

Depends what WU you're getting, and what client you're using. 40k combined is pretty much the highest I could see with both going, however SMP only I can get up to around 50k with good WU, possibly 55-60k running native linux. I'm not saying you're not making more total, I just don't like to justify the extra expense in power for the small-ish gain. I would rather put the GPU's on a separate rig and fold a bunch of them at once.

For OP's setup, the AMD quad-core could put out some decent SMP numbers, but I think folding the 580 on top will just kill it. A 960T if unlocked could bring in around 22-24k with good projects, but the 580 can just about match that and qrb seems to be coming soon for GPU folding as well.

 

[jojo69]

You would absolutly want to do both unless you're running a sinlge core CPU.

not my experience

I get better ppd on just the CPU, with AMD cards anyway firing up a GPU kicks the CPU in the balls

 

[RamonGTP]

Well it's more of a gaming machine than a folding one. I'll fold every now and again when I need to heat up the room.

 

[RamonGTP]

not my experience

I get better ppd on just the CPU, with AMD cards anyway firing up a GPU kicks the CPU in the balls

Now that you mention that, yes, when I used to fold on AMD cards it would use a full core to keep it going. My experience with nVidia cards is they use a fraction of a core.

 

[Tsumi]

If you're folding on an nVidia card (in my experience), the best way to completely utilize your system is to fold on one less core than what you have (not sure if this would include logical cores, I think it would), and then set the affinity of the nVidia clients to that unused core.

From what I can tell, SMP needs to have all threads working at the same pace. If one is slower than the others, it slows them all down, so leaving one core dedicated to the GPUs circumvents this problem.

In my roommate's experience who was folding on a core i5 750.. he got far more PPD using only his CPU than using a 4850x2 (this was 3 years ago).

 

[lilfiend]

don't forget the bonus points are exponential, even just browsing the web on your computer can make a huge difference in ppd, but I'd say fold on both with one core less on the cpu (so their is one free for the GPU) that way you can do as many work units as possible.

 

[tangoseal]

Folding is cool... but I wonder if your folding is adding to the evil segment of society that uses the results for making super megadeath bio weapons.

 

[DeathPrincess]

No it doesn't have to use 100%.

SMP units can be assigned to a number of threads, so if you have 4 cores, you could set it to use 2 (it's best to keep the numbers even for some reason) and still have 2 spare to broswe and do other stuff. GPUs can also be set usage levels or temperature targets (keep it under 70c/only use 80% etc.) though that doesn't work on all series of cards. Most the time when i'm working on stuff i'll leave it on SMP 4 (using four threads) and then switch to SMP 6 on downtime when it's just browsing/research/low CPU stuff, and 2 during heavy compute stuff.

 

[Nathan_P]

Folding is cool... but I wonder if your folding is adding to the evil segment of society that uses the results for making super megadeath bio weapons.

Given that one of folding's primary projects at the moment is a therapuetic drug for Alzheimers i would say probably not... However if you are unsure either visit the DC sub forum or google folding@home and see what they have already done and are trying to achieve

 

[ccityinstaller]

No it doesn't have to use 100%.

SMP units can be assigned to a number of threads, so if you have 4 cores, you could set it to use 2 (it's best to keep the numbers even for some reason) and still have 2 spare to broswe and do other stuff. GPUs can also be set usage levels or temperature targets (keep it under 70c/only use 80% etc.) though that doesn't work on all series of cards. Most the time when i'm working on stuff i'll leave it on SMP 4 (using four threads) and then switch to SMP 6 on downtime when it's just browsing/research/low CPU stuff, and 2 during heavy compute stuff.

I have folded SMP with the -bigadv flag set to "7" threads with 2 different heavily overclocked i7 quadcores with HT, and my i970 @ 4.2Ghz was set to "11" with no issues what so ever..this has been over a period of years..

 

[jojo69]

Folding is cool... but I wonder if your folding is adding to the evil segment of society that uses the results for making super megadeath bio weapons.

This is not an uncommon fear, I think Stanford is actually doing what they say myself or I wouldn't be folding, and I am a pretty paranoid dude.

I will be quite upset if I learn someday I have been processing echelon WU.

 

[magoo]

If you're folding on an nVidia card (in my experience), the best way to completely utilize your system is to fold on one less core than what you have (not sure if this would include logical cores, I think it would), and then set the affinity of the nVidia clients to that unused core.

From what I can tell, SMP needs to have all threads working at the same pace. If one is slower than the others, it slows them all down, so leaving one core dedicated to the GPUs circumvents this problem.

In my roommate's experience who was folding on a core i5 750.. he got far more PPD using only his CPU than using a 4850x2 (this was 3 years ago).

I folded up over 2 million points in about 3 months using two GTX 285s and a Q6600.
Your roommates problem was/is AMD was not good at folding at that time, whereas nvidia GPUs were absolute monsters.

As far as costs.....I think my electric on that set-up was about 50-60 bucks a month. The utilization was 100% on both GPUs and 75% on the CPU, 24/7.

 

[jojo69]

I folded up over 2 million points in about 3 months using two GTX 285s and a Q6600.
Your roommates problem was/is AMD was not good at folding at that time, whereas nvidia GPUs were absolute monsters.

As far as costs.....I think my electric on that set-up was about 50-60 bucks a month. The utilization was 100% on both GPUs and 75% on the CPU, 24/7.

"at that time"?

I was unaware that PG had crawled out of NV's pocket and acceptably supported AMD hardware to this day.

 

[extide]

Actually yeah ATI cards still suck, and nVidia cards are going to be getting a rather large boost it appears. Fold with the 580 for sure, and then 3 of your 4 threads on the CPU. That will be the best ppd.