Re: FPS performance increase when deliberately spinning the CPU with an unrelated task

[Peter Clifton <pcjc2@cam.ac.uk>]

On Mon, 2010-10-25 at 13:14 -0700, Eric Anholt wrote:

I'm taking a bet on memory bandwidth being the constraint for fill
limited applications. glxgears manages 59fps at full screen (minus title
bar), which is 1680x1024, giving a pixel rate of approx 100Mpix/second

(Ignoring over-fill)

Assuming 2 bytes per pixel (ignoring possibility of depth / stencil
writes), that is 200M bytes / second.

Now I don't know what the likely memory bandwidth of the GMCH is, but I
bet with scan out and other activity going on, we'll be pushing it quite
hard to sustain that framerate.

I'm targeting a desired ~15-20fps at nearly full-screen for my circuit
board design app, so I will need to be careful with my code!


This said, some top-range current cards from other manufacturers are
claiming fill-rates in 10s of BILLIONS per second, and memory bandwidth
of 10s up to ~100 GB/sec! (Although it comes with more than half the ram
I have in this laptop, and probably costs about 1/2 - 3/4 what the
laptop did).

-- 
Peter Clifton

Electrical Engineering Division,
Engineering Department,
University of Cambridge,
9, JJ Thomson Avenue,
Cambridge
CB3 0FA

Tel: +44 (0)7729 980173 - (No signal in the lab!)
Tel: +44 (0)1223 748328 - (Shared lab phone, ask for me)