On a whim, I tried the oldest SPEC OpenGL benchmark, SPECviewperf 6.
SPECViewperf 6.1.2 lives here: http://www.spec.org/gwpg/pastissues/Feb2_02/opc.static/opcview.htm It's also downloadable from ftp://spec.it.miami.edu/dist/gpc/opc/viewperf/specviewperf612is01.exe
So far, the two tests I've tried seem to run ok, though I don't know if the rendering is correct.
On one machine, $ cd ".wine/drive_c/Program Files/SPECopc/SPECviewperf 6.1.2" $ wine cmd /c RunLight04.bat produced a proud 3 frames per second, woot!
This is pretty slow compared to the published results from 2002, http://www.spec.org/gwpg/pastissues/Feb2_02/opc.data/summary.html even without thinking about how much faster today's CPUs are.
Still, it's nice that it runs. Perhaps we can take some baseline data and start looking for bottlenecks one of these days.
- Dan
First of all these tests should be performed on solid display drivers. The performance of most open source drivers is crap (in case of Intel an upgrade to the new i915tex branch can do miracles it easily doubles performance). For best results use ATI or Nvidia.
Second I expect the performance on a Geforce / Radeon card to reasonable to good. For the best performance you likely need a Quadro or FireGL the reason being that the drivers of those cards have been optimized for CAD-functionality which is basically what the Spec-tests test.
There are also some more consumer orientated OpenGL tests (something a bit like a 3dmark for opengl). Glmark is one though a bit old, much more recent is FurMark though I have never tried it.
Roderick