Are there any performance differences between running the client, or running the program in the browser?
I notice I can squeeze 10-20% more performance out of the browser, if I would give nacl64 or nacl_helper in Linux (the ones that eat the CPU) a higher thread priority.
Client vs browser, performance differences?
Moderators: Site Moderators, FAHC Science Team
Re: Client vs browser, performance differences?
There are MAJOR differences in performance. The browser client, NaCl, does all of its processing on your CPUs. The FAHClient creates independent processes (known as slots) on each GPU ... and if there are unused CPU threads, on them, as well so if you have a supported GPU(s), a lot more hardware will be used.
Even if you don't have a supported GPU, the installed FAHClient will make use of more advanced CPU features such a AVX, if your hardware supports it.
Even if you don't have a supported GPU, the installed FAHClient will make use of more advanced CPU features such a AVX, if your hardware supports it.
Posting FAH's log:
How to provide enough info to get helpful support.
How to provide enough info to get helpful support.
Re: Client vs browser, performance differences?
I suppose that's why the work units of the client are larger?
Re: Client vs browser, performance differences?
Not really. The fundamental design goal of NaCl was to give folks the opportunity to fold on a shared computer (like at at library) where all runs must be short because you're expecte to walk away from that computer rather quickly. With the installed FAHClient, it's expected to run for hours or days at a time. Also, assignments for FAHClient earn bonus points as long as you maintain a series of successfully completed WUs.
Posting FAH's log:
How to provide enough info to get helpful support.
How to provide enough info to get helpful support.