Tertz wrote:

sunaiac wrote:

In Phil's benchmark suite

Where is it from?

France.

(It's mine)

edit : It's on a P55T2P4, with 256KB cache, 32MB EDO, from a standard 5GB IDE drive. The one in my sig.

sunaiac wrote:

It's mine

If there is your article/theme with testing a link would be interesting.
Speed Test has P200, but not P200 MMX for comparision. Unlike Speedsys it seems as FPU independent, hence its results may be useful as addition to yours. P233 MMX would be good to check too.

idspispopd wrote:

I thought that 3D hardware acceleration means that the program is doing lots of floating point calculations.

Not necessarily floating point, it could also be fixedpoint (a game like Doom or Descent is entirely in fixedpoint for example, and doesn't even require an FPU to be present at all), and even performed by MMX.
These calculations have to be done with software rendering anyway, so nothing changes there.

But I was talking about feeding the data to the 3D card.
Namely, your graphics API, be it D3D or OpenGL, is not designed with a particular low-level data format in mind.
So you will have to reorder the data coming in from the API to the particular format that the graphics processor understands. MMX could be a huge help here, with its packing and unpacking instructions.

Tertz wrote:

sunaiac wrote:

It's mine

If there is your article/theme with testing a link would be interesting.
Speed Test has P200, but not P200 MMX for comparision. Unlike Speedsys it seems as FPU independent, hence its results may be useful as addition to yours. P233 MMX would be good to check too.

Right now it's just a bunch of random tests.
I have the same kind for a large bunch of 486 setups, and I'm currently doing the same on slot A systems (with more modern benches in addition of course).
I'l try to organize it a bit more and make a good topic here.
In the mean time, I'll try to get that machine running this week end so I can add speedtest to the results. Depends on how much time baby gives me 😁

sunaiac wrote:

Right now it's just a bunch of random tests.

If while torments of your PCs by benchmarks your hands will go to P200 MMX, it's interesting to see its results also in DOSBox. When we discussed, some dudes said doubts P200 can handle even level of XT there. To check this can be used inside of DOSBox the tests: MIPS, Checkit, Speed Test
Settings for DOSBox: scaler=none, core=dynamic, cputype=486_slow, cycles=max; other are default. You may make screenshot of results or write by text. In MIPS params general, overal; in Checkit param cpu.
Additionally, in case you'll get > 10 XT speed, would be interesting to see realtics in Doom: doom -timedemo demo3 > result.txt

Err, I'll try that.
DosBox should run on which OS to be at its best on a Pentium ? Win 95 ?

sunaiac wrote:

DosBox should run on which OS to be at its best on a Pentium ? Win 95 ?

Pentium MMX had good support from OSR2. DOSBox works in Win95 with Desktop Update (part of IE 4.01 SP2, maybe found at oldversion site) and in Win98. Results should be similar in different Win9x, but it's not exactly.

I am too interested in this topic.

Anyone encountered games that misbehave on a 233 MMX but working good on say a Pentium 133?

I want to bump this thread once again for some numbers.

486 DX2 66: August 1992.
Pentium 66: March 1993.
Pentium 90: March 1994.
Pentium 133: June 1995.
Pentium 166: January 1996.
Pentium 200: June 1996.
Pentium MMX 166-200: January 1997.
Pentium MMX 233: June 1997.

Doom: December 1993.
Hexen: October 1995.
Warcraft 2: December 1995.
Duke Nukem 3D: January 1996.
Quake: June 1996.
Tomb Raider: October 1996.
Blood: May 1997.
Carmageddon: June 1997.
GTA: October 1997.

This arises the very obvious question, what CPU did the developers use when they programmed the game and did they consider the game to run fast on it?
When a game was released it was considered to run on a CPU already on the general marked at least for a year, no game is developed instantly.

I picked up a Pentium 200 SY045 from ebay to play with the multipliers 100, 133, 166, 200 on my Socket 7.
The MMX is quite faster even on 133Mhz because of the larger L1 cache.

James-F wrote:

I want to bump this thread once again for some numbers.
This arises the very obvious question, what CPU did the developers use when they programmed the game and did they consider the game to run fast on it?
When a game was released it was considered to run on a CPU already on the general marked at least for a year, no game is developed instantly

It depends. Games that "we're it's time ahead" require a fast computer for it's time. For example Doom, Quake and don't forget Ultima Underworld 1 and 2 needed (for it'day) a fast system.

There is also a difference between 'programmed for' and 'programmed on'. Doom was programmed on NeXT systems but ran on regular x86 systems. I guess it needs a view on development and the available resources: how fancy can you create it (limitless fancy? 😀) but still make it playable on an average or high-end system. To be future proof; aim at high-end and your game won't be outdated in the next year to come.

I've also got this problem 😀

My first PC was Am486 DX4-100, that was intermediate 486-Pentium solution. Despite being nostalgic I wanted to play with Intel in my retro builds (my main home computer always was the AMD since 1997 till today).

So I planned to play enough with VLB that I've never assembled and even seen before and spend more time on Win 3.11. I have everything now and will spend my weekend putting it all together.
For the latest DOS games and early 3D that I missed (college, work and slow 486 didn't allow that) I will assemble either PII or early Slot 1 PIII with Voodoo 2 in SLI. That will be WIn98 SE machine.

There is something that I need to put in-between for games like Fallout, Warcraft 2 and Starcraft, Heroes, Doom, Duke3D, NFS, Diablo and everything I am (was) used to. That was working pretty well on my DX4/100 but definitely will be lagging on DX2/66.
My initial choice was average Pentium. I already have Socket 5 MB and Pentium 100 with 64 (or even 80) MBs of SIMM memory, ATI Rage 3D, ZIP drive and probably will buy 18GB SCSI. But friend of mine is going to send me his old Pentium 166MMX on Socket 7 MB. So now I'm confused. P100 is just an improved version of the stuff I once had, good candidate for classic Win95 with Plus! and everything to make me cry of nostalgia. P166 is definitely something better but it's from some other world, closer to Win98 era, Voodoo or even Voodoo 2 that I want to put in PII. Even three PCs is a lot of space and efforts but four (and so close ones) is way too much. I would better go down to 386, terra incognita to me.

So would would you prefer in the sequence?

386 DX-33, 486 DX2-66, P-100(or even 133), PII-400
or
386 DX-33, 486 DX2-66, P-166MMX, PII-400 (or even PIII-500, 550, 600)