kixs wrote on 2020-11-12, 08:05:
Can't agree with that 🤣
When talking performance/benchmarks only hardware matters.
Er, if that were the case we'd all be flying Concorde and Betamax would have won the home video wars, not to mention the fact that we certainly wouldn't all be using x86 or ARM hardware in 2020.
Yes, hardware is vital, but only if it supports your killer app. There were far more capable 3D accelerators out there than 3DFx's Voodoo in the mid 1990s, but they were too expensive and/or made for non-PC platforms (SGI...) so it was the less capable 3DFx that actually helped you play Quake at a price you could afford that won out.
Same with VLB cards: the cards that perform best in DOS games are actually some of the most primitive designs out there, lacking all but the most basic acceleration features of their more advanced counterparts. There's far more advanced hardware out there, capable of much higher performance - but that added functionality simply isn't supported in our games so it is wasted or even reduces performance (S3 9xx-range vs 8xx-range). If it were about the hardware, we'd test with the software that best shows off that hardware, i.e. in AutoCAD or Windows 3.x (or indeed 95) benchmarks, and then things like the Weitek 9000 and Appian AGC 98032 would leave our beloved ET4000 and 90C33-based cards behind in the dust.
But it's not just about hardware, it's about applications, so we choose the cards that are fastest in our preferred application instead.