Thanks! I'm glad to see that even the slowest cards were producing playable framerates at up to 1024x768. Are there more graphs coming? I noticed the Ti4200 and Ti200 were not in the first 6 graphs.
The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks
wrote:Thanks! I'm glad to see that even the slowest cards were producing playable framerates at up to 1024x768. Are there more graphs coming? I noticed the Ti4200 and Ti200 were not in the first 6 graphs.
Yes these are the charts from the 3dfx video, which has a reduced range of NV cards to keep it simpler. The GF4 was overall slower, but stronger with high AA at high resolutions.
This project just gets bigger, the next time I'm doing such a project I will set it up a bit differently, just to make the charts and all of that easier to produce.
How to improve graphs : Think about limiting information on graphs.
Something like this (example) :
You know what is where (name wise), and seeing what exact performance X card get isn't too difficult.
There should be option to make result values show on both top and down of the grapgh... don't quite know how to do it yet tho 😉
Adding table with results to graph can help as well, but I don't think it's nessessary.
Revealing is fun 😀
Searching for cards name in the bottom isn't.
I think that cards name is more important than actual score (in this case), and that's why I used cards name on the result bar.
wrote:Well it seems you already tweaked SS7 to the max :) […]
Well it seems you already tweaked SS7 to the max 😀
Yea, none of my tests use any 3DNow! patches, MiniGL or anything performance boosting.
Crikey that MiniGL is fast! With the AmigaMerlin drivers in GLQuake Demo 1 640x480x16 the score improved from around 117 to 176!
And Quake II from around 66 to 82! The 3Dnow! 3DfxGL however gave me a grey screen.
Now with the sound, the Quake games are not very demanding, Incoming however is, especially if you run it default, it will cost quite some CPU power. That is going to be the reason of the discrepancy.
EDIT: Half Life comes with 1.45 MiniGL, I wonder how much of a boost 1.47 gives, because I think that game is really heavy on the CPU.
Okay, that explains a lot then! I think the MVP3 may be a slightly faster platform than Aladdin V overall. Imperious provided some tests and just check out Forsaken for example, it really distances itself. I'd love to get my hands on a top MVP3 motherboard and fool around with it at some point (perhaps something to consider for the last quarter or 2017 😀 ).
Half-Life is indeed a very CPU demanding game. Blowout still couldn't do steady 60FPS on a Pentium III 750 (1GHz running at 100MHz FSB 😁) and Voodoo5 last I tried. I don't think the newer MiniGL will help out a whole lot, unfortunately. They started implementing a lot of 3DNow! optimizations, in the MiniGLs, which is why it makes such a big difference on the K6 CPUs, I guess.
3DNow! was pretty cool, it's kind of a shame developers didn't support it more. For example, Monolith was updating their LithTech engine to take advantage of the instructions and was about to include a 3DNow! (both Software and D3D) renderer for Shogo and Blood 2 on some of the latest patches. They postponed it (the readme claims you should be able to get the new renderers from their website in the coming weeks 🤣), although they were like 90% ready, but I think they crashed a lot and performance gains probably weren't worth it ( I seem to remember claims of 10-15% best case uplift ).
wrote:It usually was all about Write Combining, I had to fool around with that a lot, until I had a setting that worked optimally. That tweaking Super Socket 7 video sounds awesome!
Got to check out this Write Combining stuff. What tool did you use?
And I got the 3DNow! Quake 2 executable and MiniGL working together. The machine performed a blistering 100 FPS at 640x480x16! That's amazing. There is so much untapped potential it seems with the K6-III and Voodoo 3. Just imagine if all games had such optimisations...
wrote:The thing with the graphs is that they are part of a video with me talking about them. I reveal line, by line... Well, you'll see what I mean 🤣
Next time try to keep the colors consistent between the graphs if you're posting a bunch of them in a row, provided that it makes sense?
"I see a little silhouette-o of a man, Scaramouche, Scaramouche, will you
do the Fandango!" - Queen
Stiletto
wrote:wrote:The thing with the graphs is that they are part of a video with me talking about them. I reveal line, by line... Well, you'll see what I mean 🤣
Next time try to keep the colors consistent between the graphs if you're posting a bunch of them in a row, provided that it makes sense?
Yup, noted, Elinda also mentioned this 😀 This is an easy fix, rather than deleting the cards that can't do 32 bit, just zero them.
wrote:Revealing is fun 😀
Searching for cards name in the bottom isn't.
I also struggled a bit with this. Especially in the more crowded graphs where there were a lot of very similar colors.
You won't lose the suspense of the reveal by adding the name of the card directly to each bar.
Finally, what I'd like to see is results with more common CPUs like K6-2 333-450, since that better represents the kind of machine someone would have had back in the day.
wrote:wrote:It usually was all about Write Combining, I had to fool around with that a lot, until I had a setting that worked optimally. That tweaking Super Socket 7 video sounds awesome!
Got to check out this Write Combining stuff. What tool did you use?
And I got the 3DNow! Quake 2 executable and MiniGL working together. The machine performed a blistering 100 FPS at 640x480x16! That's amazing. There is so much untapped potential it seems with the K6-III and Voodoo 3. Just imagine if all games had such optimisations...
100FPS on Quake 2, that's astounding!
I used the ALi AGP utility, but I'm pretty sure it's only for Aladdin V chipsets. I think swaaye had a download link for it somewhere around here...
Are the driver versions you test with helping or hindering the results for the Nvidia cards?
It seemed back in the day they were making constant improvements each detonator release and then sudden broke everything in the later releases.
wrote:Are the driver versions you test with helping or hindering the results for the Nvidia cards?
It seemed back in the day they were making constant improvements each detonator release and then sudden broke everything in the later releases.
The Nvidia drivers do have 3DNow! optimizations. I had to use newer drivers with the faster cards, they did show a bit of a slow-down at times, but nothing massive. You can use the 3DNow! game patches of course, but you you cannot use the MiniGL drivers. They are basically wrappers, translating a subset of OpenGL commands required for the games to Glide.
With those post-99 video cards i'd stay clear of GLQuake and Quake2. Both use really old extensions and ignore the advances like compiled vertex arrays. Relying on GLQuake alone would skew results, and be unreliable since there has been many official builds of GLQuake with implementation differences, with varying distribution (like an outdated august 97 build with a 2001 retail Quake for example)
Either the TNT1 or 2 has an obscure compatibility issue with Ultima VII - I'm not sure if even Exult handles it correctly. When the cyclops are visited, the screen is supposed to shake. This works with an S3 card, but not the TNT.
