Folding at Home CPU Efficiency: Multi-Core Intel Q6600

In the last post, I showed how environmentally unfriendly it is to run just the uniprocessor client.  In this post, I’ll finish off the study about # of CPU cores vs. folding efficiency.  As it turns out, you can virtually double your folding at home efficiency when you double the amount of CPU cores you are running with. Using the same Intel Q6600 as before, I told the Folding at Home client to ramp up and use three cores.  Then, once I had some data, I switched it to four-core folding.  With the CPU fully engaged, my computer became a bit slow to use, but that’s not a problem since what we are all about here is dedicated F@H Rigs (the only way to fold efficiently is to fold 100%).   If I want to use my computer, I’ll stop the folding to do so, then start it up later.

Here are the results of the 1 through 4 core F@H PPD experiment!


As you can see, both performance (PPD) and energy efficiency (technically efficacy in PPD/Watt) scale with the # of CPU cores being used.  Yes, the system does use more total electricity when more cores are engaged (169 watts vs. 142), but the amount of work being done per day has far surpassed the slight increase in power consumption.  In graph form:

Intel Q6600 Folding@Home Points Per Day / Watt Graph

Intel Q6600 Folding at Home Efficiency Graph

Intel Q6600 Folding at Home Efficiency Graph

In conclusion, it makes the most sense from a performance and efficiency standpoint to use as much of your CPU as you can.  In the next post, I’ll look at a few more powerful CPU-based folding@home systems.


7 responses to “Folding at Home CPU Efficiency: Multi-Core Intel Q6600

  1. Pingback: CPU Folding Revisited: AMD FX-8320E 8-Core CPU | Green Folding@Home

  2. Hi, I think I found the guy (you) who could answer my question 🙂

    Running one WU on multiple cores is definitely quicker and we get more PPD but is running multiple WU (let say 2 WU) on a quad core CPU faster or slower?

    For instance : (1 WU on 2 cores) x 2 = more PPD ??

    Thanks for your time.
    Regards, Loïc

  3. Hi, and thanks for the question. Typically you can earn more PPD when running one work unit on all 4 cores, vs 2 simultaneous work units on 2 cores each. Although theoretically (for an identical molecule being solved) the amount of work being done per unit time is the same (excluding minor additional threading inefficiency in the 4-core case), the 4-core solve will get that single work unit done faster. Folding@home awards additional bonus points for work that is returned quickly (quick return bonus). The clock starts ticking as soon as your folding slot downloads the WU from the server. By solving work units sequentially (where each WU is returned faster due to having all 4 cores on the solve), you get more bonus points. Thus, PPD is increased. Assuming the CPU uses similar wattage in either case (it should, since both cases represent a 100 percent load), efficiently will also be better for the 4-core solve.

    You can actually use the table above to see what I mean. The 4-core solve earned about 6000 PPD. The 2-core solve did about 2000. Thus two 2-core solves at the same time is about 4000 PPD. The 4-core solve still wins.

    Where this might break down is with modern high core count processors, where threading inefficiencies (extra overhead work needed to put parallel work back together) start to add up. I currently have a 16-core Ryzen 3950x that I am investigating this on.

    You can read more about the quick return bonus here:

  4. Hi Chris,

    yes it makes sense theoretically and this behaviour actually happens with my Folding client! Folding so (1 WU on 2 cores) x 2 = 1/4 less PPD ! roughly, from ~40000 to ~30000 PPD on my i5 6500T.

    It would be great you plan to write a post about your trials on 16 core 🙂
    Have a good day!


    • Cool, it’s good to see that hold true for the i5. And yes I can’t wait to post the article on the new Ryzen. I’m running the test right now! Sweeping up from 1 to 32 threads and logging power and PPD numbers. I also want to do this with SMT (AMD’s hyperthreading) turned off, and also one with Core Performance Boost (turbo core) turned off to see the effect. So far, the most interesting thing I’ve seen is that for large prime numbers (7, 11, etc), performance actually decreases.

  5. Pingback: AMD Ryzen 9 3950X Folding@Home Review: Part 1: PPD vs # of Threads | Green Folding@Home

  6. Pingback: AMD Ryzen 9 3950X Folding@Home Review: Part 3: SMT (Hyperthreading) | Green Folding@Home

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s