Performace Q

If you're new to FAH and need help getting started or you have very basic questions, start here.

Moderators: Site Moderators, PandeGroup

Performace Q

Postby Bill H » Tue May 30, 2017 7:14 pm

I use a 4790 k intel precessor GT 610 nvida z97 m b with 1 stick of 8 g ram 1600 with a passkey, when iam on lite with no issues about 3300 ppd, when on medium its gets buggy like internet issues but i can manage ,on full it even takes away points with severe bugs ,does memory affect ppd this mother board combo came with cpu ,memory it was a new egg shell shocker deal plus a better video card is needed for V R.
Bill H
Posts: 1
Joined: Tue May 30, 2017 6:16 pm

Re: Performace Q

Postby JimboPalmer » Tue May 30, 2017 9:32 pm

I am going to ramble.

Folding at home hates large* Prime numbers of CPUs and multiples of large Primes. Having 7 CPUs working on a WU almost always fails so recent server side code tries not to assign WUs to large Primes and multiples of large primes. The latest client code (7.4.16) automatically adjusts the number of CPUs downward so the server does not refuse to assign primes.

The latest GPU client lowers the CPU count by one so the GPU client has all the CPU resources it needs. (If you have more than one GPU, it decrements more than one CPU)

Your 4790k has 4 physical CPUs but does hyperthreading so it has 8 logical CPUs.

Because you have a GT610 GPU, Folding at home thinks you have 7 CPUs (This is bad)
On Light, it only tries to use half, so only 4 are used.
On Medium it tries to use all but one, so 6 in your case, this 'should' work.
On Full it tries to use all 7, that always fails as 7 is prime.

You can try two scenarios
1) in Advanced Control, Configure, Slots, CPU: the default number is -1, which means automatically use the rules above, you can set it to 6 CPUs then try Full.

2) Because the GT610 is so wimpy and the 4790k so strong, You may make more points with 8 CPUs on Full and delete the GPU slot altogether.

* for our purposes 7 is too large, and 3 is no problem, and 5 is iffy.
Tsar of all the Rushers
I tried to remain childlike, all I achieved was childish.
A friend to those who want no friends
Posts: 619
Joined: Mon Feb 16, 2009 4:12 am
Location: Greenwood MS USA

Re: Performace Q

Postby jrweiss » Wed May 31, 2017 3:32 pm

Another workaround is to set the max cores for your CPU slot to 6. Then you'll have the equivalent of 1 virtual core working with the GPU slot, and 1 virtual core open for other tasks if the F@H Client is slow to release resources.

Turn OFF "hardware acceleration" in your browser, so it doesn't have to wait for the GPU.

Also, are you monitoring your temperatures (CPUid HWMonitor or similar)? You may be overheating your CPU or GPU when running both at full load. Open your system and clean it out.

Depending on background tasks and your foreground apps, 8 GB RAM may be marginal, causing swapping to pagefile, slowing the system during HDD writes. An investment in another 8 GB may be wise.
i4770K/MSI GTX 1050ti
i7-3770S/XFX RX460 fanless
User avatar
Posts: 1104
Joined: Tue Dec 04, 2007 6:56 am
Location: @Home

Re: Performace Q

Postby bruce » Thu Jun 01, 2017 2:05 am

FAH doesn't really use very much RAM. It will use all the CPU resources it can get once the issue of "prime numbers" is alleviated.

The GT 610 is, in fact, pretty wimpy (and it's really a rebranded GT 520) which is capable of 156 Single Precision GFLOPS. Almost everybody around here upgrades to the best GPU they can afford .. but I do run a pretty wide variety of GPUs, including some that semi-wimpy GPUs that others turn their noses up. NVidia introduced (limited) double-precision hardware with their 700 series and although FAH's performance depends almose entirely on Single Precision, most projects need at least a tiny bit of Double performance.
Posts: 21233
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Return to New Donors start here

Who is online

Users browsing this forum: No registered users and 1 guest