December '97 3D Video Accelerator Comparison

How I Tested

• Each benchmark was run 3 times, with the final result being an average of the three.  If any of the three test runs resulted in a failure the entire score was made void and the set of three tests was run again until a reliable set could be achieved or until the benchmark tests failed 3 times in a row.

• The FIC PA-2012 VP3 Based Socket-7 AGP Motherboard was used for all Socket-7 CPU tests and the AOpen AX6L LX Based Slot-1 AGP Motherboard was used for all the Pentium II tests.  Both test systems were configured identically with no differences other than the motherboard and processor.

• A full 400MB Quake 2 installation was performed, as well as a full installation of Winstone 97 and Winbench 98.  The CDROM drive present was an AOpen 24X.

• The Quake 2 Timedemo Frame rates were achieved by typing the following in the Quake 2 Console (hit the '~' key to activate):
                                  TIMEDEMO 1
                             DEMOMAP DEMOx.DM2

       Where x is the corresponding demo number (either 1 or 2) .  All Quake 2 tests were run at 640 x 480

• All other tests were run at 1024 x 768 x 16 bit color

• No foreign drivers were present in the test system other than those required for the system to function to the best of its ability

• All foreign installation files were moved to a separate partition during the test as to prevent them from effecting the test results

• The ATI Xpert@Work used the beta OpenGL ICD from ATI

• All 3Dfx based Voodoo accelerators used Glide 2.43

• All 3Dfx based Voodoo 2 accelerators used Glide 2.5

• All NEC PowerVR PCX2 based accelerators used the standard PowerVR OpenGL Port

• All 3DLabs Permedia 2 based accelerators used the default OpenGL drivers

• All Rendition Verite V2100 based accelerators used the beta OpenGL drivers for the Verite chipset

• All Riva 128 based accelerators used the Beta nVIDIA OpenGL drivers for the Quake 2 Tests

Now on to the comparisons...