Best to bookmark the vgamuseum.info domain?

If you get a Mach64 ISA some of the guys here might be overwhelmed with joy. 😀

swaaye wrote:

Best to bookmark the vgamuseum.info domain?

If you get a Mach64 ISA some of the guys here might be overwhelmed with joy. 😀

Bookmarked.

And, ya. I'd like to see the Mach64 ISA 2mb and 4mb benchmarks though I don't really expect them to be faster than the highest rated ISA cards already on there. They are really more for the extra resolutions and color depths than speed. I'd also like to see how the DRAM versions of those cards stack up against the VRAM ones. I also noticed a distinct lack of VL Bus cards in those tests, too.

There are no VLB cards because he used the same system for all of the tests. You can't get AGP+PCI+VLB+ISA boards. I think it is desirable to have all of the cards tested on the same platform because otherwise you'll have new variables in the mix. A VLB system is going to be limited by other factors such as the slow CPU and slow memory performance.

swaaye wrote:

There are no VLB cards because he used the same system for all of the tests. You can't get AGP+PCI+VLB+ISA boards. I think it is desirable to have all of the cards tested on the same platform because otherwise you'll have new variables in the mix. A VLB system is going to be limited by other factors such as the slow CPU and slow memory performance.

There are PCI/ISA/VLB boards, though. Any system with an AGP slot is going to require a PII or higher CPU, too. So that is going to skew the results, also, since machines from the DOS era were much slower and many old games can't run on a CPU that fast so the test results would be useless. Many games won't run on a Pentium let alone a PII and some very old ones won't even run on a 486.

No it won't skew the results to run a CPU / mobo that's relatively super fast. What it does is remove bottlenecks beyond the video card, so you can see the most the video card can possibly put out.

Running a slow CPU will not let the faster cards show their best and will make it appear that you can't benefit from the better cards. Although this can tell you that on a 486 system there may be little point in running the fastest possible card.

swaaye wrote:

No it won't skew the results to run a CPU that's relatively super fast. What it does is remove the CPU bottleneck so you can see the most the card can possibly put out.

Running a slow CPU will not let the faster cards show their best and will make it appear that you can't benefit from the better cards.

But if the games won't run at that speed, what good is the benchmark? It's like building a 1000hp engine for a race car and then putting a restrictor plate on it. You'll never access the full 1000hp when you go to drive the car.

A better question in that case is what is the point of worrying about which video card you have?

Beyond the ultra slow cards like the Trident 8900C, the CPU becomes the performance barrier on a 486 system.

Would you benchmark a Voodoo3 on a 486 and judge the card in that case? A Voodoo1 would show the same ultra slow speed. You would have no idea how fast a Voodoo3 really is because the system is holding it back.

To properly test hardware you have to isolate it as much as you can from other components. Otherwise you aren't testing one piece of hardware but are testing the whole individual platform.

Wow. How the heck did the Rendition PCI and AGP 1x cards manage to score down with the slower ISA cards? Shouldn't the bus itself have pulled those cards much higher? Or do they use some proprietary form of 3d acceleration that the benchmark doesn't test for?

sliderider wrote:

Wow. How the heck did the Rendition PCI and AGP 1x cards manage to score down with the slower ISA cards? Shouldn't the bus itself have pulled those cards much higher? Or do they use some proprietary form of 3d acceleration that the benchmark doesn't test for?

Renditions have some bugs, so they're slow in dos in resolutions under 640x480 = unusable for dos - but aimed for Win 3D aplications, so no problem at that time
fps for quake 320x240, 360x480, 640x480 are...
12fps, 5,5fps and 18,9fps, so its fastest in 640x480....and in GLQuake 640x480 its 31,5fps.

Tests were ment as comparsion of as many cards i can get - dont have time doin year by year perfected test pc's with different cpu for each year. Dont want to know which card was best in every year in last 15 years, but wanna know how much faster is any card than other cards (too bad that voodoo1 isnt working on mine test pc - cant multiply now speeds of other cards by basic speed of voodoo1).
BTW - dont have any mach64 on isa - only mach32 - its there already.
Did some basic testing with vlb too, posted graph already in some other thread - almost none diferences between pci/isa/vl-bus cards - everything was slow because of 486 cpu.

swaaye wrote:

Would you benchmark a Voodoo3 on a 486 and judge the card in that case? A Voodoo1 would show the same ultra slow speed.

Actually i did, because voodoo3 was one of fastest pci cards i have - and results are exactly as you might think...

Verite is quite fast in DOS as long as it is using a VESA mode. It is horribly slow in other modes.

Rendition released a utility called "renutil" that will remap regular VGA to its VESA equivalent. Unfortunately this isn't possible for the very popular Mode X, which games like Doom use.