2

For a brief intro - I've been collecting old GPUs for the past 15+ years (I have around 100 of them, starting from year 1996 (1MB SiS) till around 2008 (GTX280), representing each generation of released gaming cards and architectures.

The problem is - I have never properly benchmarked them, and finally want to set up a proper benchmark rig(s) to do so. As we know, back in the days of Win98/XP swapping a different vendor (and sometimes simply a different model) GPU most often lead to bluescreen. It was also pain in the a** to properly uninstall and reinstall drivers for a different GPUs through safe mode, and most often replacing GPU also mean reinstall Windows altogether, just to avoid some strange driver conflicts and performance issues later down the road...

So my question is - can you suggest a best, quickest and most painless way to set up a few benchmarking rigs (PCI, AGP and PCIe ones) to test a wide range of GPUs on Win98/XP?

What I currently assume -

  • I need to dig through my stack of PC hardware, and find a few MB+CPU combos that would represent the best performance for each GPU interface (PCI+AGP and PCIe) and eliminate the bottlenecks.
  • Find and download the latest, most compatible and popular driver revisions for ATI, nVidia and 3Dfx cards I'm about to bench (I assume that there will be multiple driver versions, to cover each generation of cards, right?).
  • Install a two separate dual-boot versions of XP (or 98) on each of these PCs, and install a proper latest driver version for ATI and nVidia (each on their own OS, of course).
  • When switching to a different GPU, will have to select a proper Windows version from a multiboot menu.

Maybe there is any other easier way how to accomplish this? Many thanks for ideas in advance!

Artanis
  • 121
  • 2
  • 3
    It might be hard to setup a singe setup for all of them, as graphic cards developed in turn with motherboards and CPU. So for each you may want to use an appropriate CPU/motherboard combo of that time. Using one 'too old' or 'too new' will give quite different results. An AGP RIva 128 in a K6 system of 1997 will barely run Quake, while the same card in a (still AGP) 2007 Athlon 64 will run smooth like a pound of butter... So it might be worth to define the goals of your benchmark first. Authentic performance or tuning in hindsight. – Raffzahn Mar 23 '21 at 18:20

0 Answers0