I think the 1200 Celeron should be fine as-is; just drop the Ti 4200 in there and it should be set, unless (like swaaye pointed out) you need 1600x1200+ or something, in which case the FX 5900 or some similar card would be more appropriate. The MX 440 and FX 5200 will perform about the same in DX7 if memory serves, and the FX 5200 is kind of a dog when it comes to most games that require pixel shaders.

Thank you all. You helped me make my mind. I'm going to keep the celeron and add the ti4200. No more messing around with random hardware (for now) 😉
Amazing job on the wiki! Thank you.

I'd recommend a geforce 1 or 2, so you can use the older detonator driver sets.
Later nvidia drivers have text issues with some games.
Its not a big issue but if you want 100% no problem DX7 only system, if that is your goal.

Yes, that is indeed my goal. A system only for DX7 games, well... and everything prior I guess. I also have the GTS model of the Geforce2, so what detonator drivers do you recommend my friend? 😀

The 6.50's, is the version I use the most on a Geforce1/2. Most stable in my experience, and has the SSAA modes that are missing from the newer drivers. If you don't use Riva Tuner that is 🤣.

Finished the download of Detonator 3, version 6.50! Thanks for your help. Will try the geforce2 setup.
Nice tip about this Riva Tuner. Now for some gaming!

Looking good 😀

Intel chipsets are always the way to go.

Voodoo 2 cards scale very well with CPU power, but it all depends on the graphics settings.

You will see better scaling at 512 x 384 in older Glide games compared to 800 x 600 in Unreal.

Make sure you have a heatsink on that voodoo II if your putting it in a system that can use it to its fullest... To fast a system can cause the main center processing unit to come unsoldered. Half the pins pulled back on it like a squashed spider.....
It was a 3dfx brand V2 1000 that did this, from being used in a AMD athlon 64 3700+ nforce3 heavily, it could run 98se 😲.
The faster the CPU the harder the v2 has to work.
I epoxied on heatsinks on the rest of my V2's after this little experience.

The lead/tin solder on a Voodoo2 melts at over 180 degrees C. That seems unlikely in normal operation 🤣. Maybe something went wrong with the board that caused it to cook.

Could of simply been wear and tear from expansion\contracting from hot\cold cycles, which could of caused the solder points\joints coming unsoldered from the pcb.
Still there's no doubt it ran hotter on the A64 and was likely a factor in its demise.
I haven't had a V2 in a system above 700mhz since either 🤣.

swaaye wrote:

In practice a K6-2 500 MHz is going to be around a P2-266 for 3D game performance.

😲 That's roughly double the performance. For another system I'm building would a 400MHz k6-2 have the performance of a P2 at 200MHz? What would a P2 at 200MHz be equivalent on Pentium MMX terms?

It is difficult to compare these things directly. Games that use certain features will run differently from games that do not use those features. It has already been suggested that DX7 games in particular will not run well on a K6-2; if you are running games that can take advantage of MMX features, then those will obviously run faster when those features are available.
http://www.mobygames.com/attribute/sheet/attributeId,1478/

Since everyone is talking about choosing a better CPU, I would get an Athlon Slot/Socket A
Something in the 600-1000MHz range sounds good. Also, go with a ATi/early NVIDIA card. Either an 7000VE
or a TNT2.
I recommend the 7000VE, worked well for what I do with it on my 533MHz K62

Jorpho wrote:

It is difficult to compare these things directly. Games that use certain features will run differently from games that do not use those features. It has already been suggested that DX7 games in particular will not run well on a K6-2; if you are running games that can take advantage of MMX features, then those will obviously run faster when those features are available.
http://www.mobygames.com/attribute/sheet/attributeId,1478/

Well said!

Really looking at this from a game by game perspective seems to work well.

Glide games for example run very well on a fast K6-2+ or K6-III+ and can easily keep up with the PIII. For other games it doesn't look as good.

I did a few tests last autumn, you guys can have a look:

PII/PIII VS K6-2/III+

F2bnp wrote:

I did a few tests last autumn, you guys can have a look:

PII/PIII VS K6-2/III+

Excellent work!

F2bnp wrote:

I did a few tests last autumn, you guys can have a look:

PII/PIII VS K6-2/III+

Damn, couln't stop reading the pdf till the end, very interesting results 😀
Must have Take you a while to complete all that, congrats 😎

F2bnp wrote:

I did a few tests last autumn, you guys can have a look:

PII/PIII VS K6-2/III+

The synthetic benchmarks make the K6-III+ look more competitive than it really is. That's why I don't pay much attention to them. The actual game based benchmarks are more accurate. For a gaming rig, it's how it performs in-game that matters because you're not going to sit around running benchmarks all day.

Also remember not to consider K6-2+/III+ 550 as similar to what a K6-2 550 would do. K6-2 550 is severely memory bottlenecked.

http://www.thg.ru/cpu/20000713/print.html

Indeed! K6-2+ is usually just as fast as K6-III+ though, except for the odd game that loves more cache. I think Quake 3 and MDK2 are like that.
From my tests, I have determined that a K6-III+ 550 is right between the Pentium II 350 and 400. I believe a K6-III+ 600 should be just as fast as a Pentium II 400. I'd love to hit 600MHz with that chip someday, it has proved to be a fantastic performer for what I want it to do 😀.

Of course, adding a GeForce card in there should probably pronounce an even bigger gap between the two architectures. Perhaps I should give it a try next Autumn 🤣 .