How's the cache?

The TX97-E has 512kb, that I didn't upgrade so it was 512 on both the 75/166.

Nice! Yes going from a 75 to a 166 meant double the performance 😀

Another thing you do with Quake is get a 3D accelerator card like a 3dfx Voodoo.

*Voodoo2. 'muh bottleneck its pointless' one would say, however the support of multitexturing is a bus congestion ease when it's available 😀

I probably wouldn't stick a 3D accellerator in the machine as it is, maybe a V1 if I dropped in an MMX CPU. I have a K6-2 300/66 CPU waiting for this board, along with 128mb EDO (2x64), but I'm trying to find a deal on a Voodoo 2 or Voodoo 3 PCI. In the mean time, I'm just enjoying the 166 as is.

If you want to remember how much difference (an easy) overclock meant; try running the P75 at 90MHz or 100MHz and retry Quake. A lot of games run on the P75, but they run so much better on a P90 or P100.

Besides what Leileilol said, Voodoo1 lacks full triangle setup hardware and so has slightly more CPU usage than Voodoo2+.

A P75 isn't going to rock Quake super well but it should be acceptable.

you should really get a voodoo card

Thandor wrote:

If you want to remember how much difference (an easy) overclock meant; try running the P75 at 90MHz or 100MHz and retry Quake. A lot of games run on the P75, but they run so much better on a P90 or P100.

I have both a P100 and P120 but felt a good noticeable speed bump, to me anyways was 166.
I couldn't see a P100 making much of a difference but maybe I'm wrong...

I did actually buy a Maxi Gamer Voodoo 2 8mb a few months back, off eBay, but I have been unable to get
Windows to recognize the card. I get nothing, but the pass through cable works fine.
I made a thread on here about he issue, and I've tried the card in a P3 box with the same result. Pass though works fine on this box, and the P3 box but no matter what I try windows can't see it. And it doesn't show up on the list of found hardware when the PC first turns on.
As someone suggested, I check the tiny pins around the V2 chips but none seem stuck together, and I see no broken parts, caps, regulators, etc, and not a single scratch.

dave343 wrote:

I hate having a K6-2 system with 512mb ram, and a 1mb video card.

🤣

My cousin's old Pentium 3 1GHz Coppermine-T had 256MB DDR and a 1MB Trident 9440... Not shitting you..

i looked up my benchmark results and it recorded that i got 30.5 fps with pmmx-133, and 56fps with pmmx-300(overclocked). i don't have non-mmx pentium though.

noshutdown wrote:

i looked up my benchmark results and it recorded that i got 30.5 fps with pmmx-133, and 56fps with pmmx-300(overclocked). i don't have non-mmx pentium though.

All the Quake results one will ever need 🤣

https://docs.google.com/spreadsheet/ccc?key=0 … TWc&usp=sharing

philscomputerlab wrote:

All the Quake results one will ever need 🤣

https://docs.google.com/spreadsheet/ccc?key=0 … TWc&usp=sharing

yeah but google is blocked in china because they would not cooperate with chinese government. 🤣

noshutdown wrote:

i looked up my benchmark results and it recorded that i got 30.5 fps with pmmx-133, and 56fps with pmmx-300(overclocked). i don't have non-mmx pentium though.

Quake does not make use of MMX technology, so that wouldn't matter.

It doesn't take advantage of the MMX instructions, but it sure loves the extra cache available in every MMX Pentium.

noshutdown wrote:

yeah but google is blocked in china because they would not cooperate with chinese government. 🤣

Excel version attached!

F2bnp wrote:

It doesn't take advantage of the MMX instructions, but it sure loves the extra cache available in every MMX Pentium.

Well, if I look at phil's table: https://docs.google.com/spreadsheet/ccc?key=0 … TWc&usp=sharing
I see Pentium 166s and their MMX-cousins scoring within 1 fps of eachother. Which could just as well be explained by a slightly different motherboard/videocard.
So I don't think the extra cache did a lot in Quake (it was probably explicitly optimized hard to fit in the L1 cache of the original Pentium).

It does make a difference, especially at low resolutions such as 320x200, which is what you'll be realistically playing at with these CPUs.

philscomputerlab wrote:

Excel version attached!

good job but i test quake a bit differently. i used a more intensive bigass demo which is about 25% slower than quake's builtin demos, and i used winquake which may provide better performance than the dos version, especially on newer computers.
you can also have a look at my tests here:
my socket7 cpu benchmark results