My PPDs seem a bit low. Is my configuration sub-optimal?

[sqook]

Looking around at some of the other threads in the new donors forum, I get a bit jealous looking at the high PPD's reported by other members with hardware similar to mine. My PPD's seem a bit low in comparison, have I done something wrong? I'm folding primarily on 2 machines, both are working basically 24/7 though I do pause F@H when I need them for other work.

Machine 1: Ubuntu 14.04
Intel i7 4790 @ 3.6Ghz, 4 cores (8 virtual), 32Gb RAM, SSD
GPU folding disabled as the card isn't quite fast enough to complete WUs
2 slots, one configured for 2 cores, the other for 6
PPD: ~2500

Machine 2: Windows 8.1
Intel i5 3300 @ 3.0Ghz, 2 cores (4 virtual), 8Gb RAM, SSD
NVidia GTX 760 GPU
1 CPU slot set to "medium" (effectively this uses 2 cores, leaving 1 for GPU. CPU usage is around 75% on this machine when folding)
PPD: ~11000

2k seems a bit low for the Linux machine with an 8-core i7 CPU, and regarding the Windows machine I've seen similar people folding with NVidia 7xx cards that are getting 100k-200k PPD. Have I misconfigured my slots somehow?

[P5-133XL]

Please supply the logs including the system and config portions (Click refresh before copy & paste).

[sqook]

Here's the logs for my Linux workstation, they're pretty big because the machine has been up for awhile and I was messing around with the GPU configuration:

https://gist.githubusercontent.com/nikr ... 3e/raw/log

The end part is probably most useful.

[P5-133XL]

Add a Passkey! No passkey means no quick return bonus points (QRB)and the vast majority of the points most people get come from the QRB rather than base points. Note that QRB points will not happen immediately upon getting a passkey for it needs to be qualified by completing 10 QRB WU's and maintaining an 80% successful return rate.

Why split the CPU cores into two slots? The faster a WU completes the higher the PPD especially with QRB points. It increases exponentially. So allocate the entire CPU to a single WU.

If you are worrying about keeping some CPU for your own use -- Don't. Folding runs at a very low priority so whenever some application needs the CPU folding merely steps out of its way and returns folding when there are CPU resources free again. The exception is when CPU folding is competing with Nvidia GPU folding and then each GPU needs a full CPU core all for itself. Here, folding is competing with folding (same priority) so what happens is nothing gets out of the way and the GPU is starved of its needed CPU time.

[sqook]

Thanks, I just added a passkey, that seems to be the problem! Now my PPDs on Linux are at 15k and rising as the client recalculates... this was definitely the issue. Doh! Such a n00b.

The reason I split the Linux machine into 2 slots is because I'm a developer and require a rather high amount of CPU (but unevenly distributed in big spikes) from my machine during working hours and usually 1-2 cores freed for regular work. My IDE runs horribly if I don't set aside at least 2 cores for Java, but I only need all 8 for large compiling jobs, so I set up the 6 core job to run during the day during normal work and pause it for rebuilds. Without the 2 slot configuration, these extra 6 cores were mostly idle for most of the working day.

The F@H client on Linux is also not very good at reacting to idle, and I have come to this configuration after much experimenting. Overnight I let both configurations run at 100% CPU. On my Windows machine at home these issues are not really present, so I just let F@H decide this automatically and just pause it when I want to play games.

[bruce]

The F@H client on Linux is also not very good at reacting to idle, and I have come to this configuration after much experimenting. Overnight I let both configurations run at 100% CPU. On my Windows machine at home these issues are not really present, so I just let F@H decide this automatically and just pause it when I want to play games.

The Linux scheduler is configured to minimize (background) task switching; the WIndows scheduler is normally configured for maximum interactive response. I know how to change the preferences in Windows and I'm certain there's a similar setting in Linux, I just don't know how to do it (yet). As a developer, you're probably already doing as well as you can.

I only need all 8 for large compiling jobs,

Since you've obviously done a lot of experimentation, I have a question. Next time you're in this situation, run the 2-core slot concurrently and see how much CPU it uses. Also compare idle time both with it running and with both CPU slots shut down. (Just out of curiosity)