Personally I think that any 486 faster than a 66MHz VLB machine is just a wanna be Pentium, so you might as well go Pentium! I find myself spending a lot more time messing around with my P166MMX than my slower PC's - SVGA at decent frame rates and the hardware is just that much easier to deal with.

It depends on what you want. If you want to play with different hardware, than good 486 "fits the bill". Otherwise you can have a P-II or P-III and be done with it 😉

I have a similar point of view. I've settled on a DX2-66 and Pentium 200, which covers all of my retro games well. The 486 can run at 286 and slow 386 speeds if needed (by de-turboing and disabling caches), and the Pentium can handle the SVGA DOS games well. Plus all the speeds in between that setmul lets me hit. To me, a High End 486 is just like a slow Pentium, and my Pentium 200 can already do that. Most cpus can get close to 486-25 speeds with cache disabling, but DX2-66 is one of those speeds that's hard to hit unless you're running a DX2-66. 😉

Can you replace your SX-33 with a DX2-66 in the same system? That might be your solution.

The effect turbo has on fast 486's can be pretty interesting; it varies by motherboard and chip combo. I can switch between a "slow Pentium" speed of the 133 MHz Am5x86, about a 33-40 MHz 486 speed with turbo, and slower than that by disabling caches.

Thanks everyone for your responses. Basically, what I am saying if I have a 486SX/33 system on the left and a Pentium 133 on the right and they are already setup and ready to go, then is there any reason to have a high end 486 as well, aside from the fact that I want it and it's cool. I can't frankly think of anything which is why I am asking.

Rhuwyn wrote:

Thanks everyone for your responses. Basically, what I am saying if I have a 486SX/33 system on the left and a Pentium 133 on the right and they are already setup and ready to go, then is there any reason to have a high end 486 as well, aside from the fact that I want it and it's cool. I can't frankly think of anything which is why I am asking.

The only thing I can think of is that you can probably get close to 486SX/33 performance on your P133 by disabling L1 cache, but you can't get close to DX2/66 performance by manipulating caches. So if your 486 supports a DX2/66, you can upgrade it, use the P133 with disabled L1 for slow 486 games, and use the 486 for DX2/66 speed games. If that's important to you, that is. 😀

Rhuwyn wrote:

I can't frankly think of anything which is why I am asking.

Well, that's just it, isn't it? Even if there is something out there somewhere (and there probably is), there's no point in worrying about it if you have no intention of ever running it.

Yeah the 486 SX is soldered directly on the board so there's no upgrading it. There is a spot for a math co-processor but I don't see a point for what I wold use it for.

I can't think of anything right now. HOWEVER, I am always interested in figuring out what new old games I'd like to play that I don't know about. So just throwing it out there.

Yeah, you should have things covered pretty well with a mid grade Pentium and a slow 486. One of the really well known speed problems is in Ultima VII, which should be okay on your SX, maybe a tad slow. It's one rare case where you can't 'cheat' by outright disabling caches on a faster machine.

A Pentium 233MMX with L1 cache disabled is very close to a 486DX2 66Mhz.
Anyway a Pentium with the cache disabled should cover huge ground for DOS games.

James-F wrote:

A Pentium 233MMX with L1 cache disabled is very close to a 486DX2 66Mhz.
Anyway a Pentium with the cache disabled should cover huge ground for DOS games.

At least on the cache benchmark database it performs closer to a DX-33. The VIA Cyrix III Ezra-T with L1 disabled is closest, at DX-50 speeds, followed closely by the K6-2 550 and 500 with L1 disabled.

Here is Phil's video comparing 486 DX33 vs DX266: https://www.youtube.com/watch?v=HNlcZetLzY8
And here is a DX2 66MHZ with benchmarks: https://www.youtube.com/watch?v=9bqqpuJ7M_I
I have run some benchmarks without L1 and it's just short of the 66 with some benchmarks and a smidge higher with others.

There are two kinds of benchmarks:
Constant load without graphics like Speedsys, Landmark, Topbench.
Varied load with varied graphics like Doom, Quake, Wolfenstein, 3Dbench.
The former shows stable clock/load measurement, the latter shows and average of varied load over time.

Judging by Doom gameplay with L1 disabled on the 233MMX and Phils video of the 33/66, it is much closer to the 66 just by eyeballing the smoothness of gameplay, the 33 is definitely choppier than what I see with L1 disabled.
The Doom benchmark averages a lot of frames and varies in processor load during the demo for quite a lengthy time, this actually is not a good measurement of CPU power but the system as a whole.
With Wolfenstein I get exactly the same FPS (70) score as the 66, with Quake FPS (6.3) very close to 66, with Doom like the 33, Topbench 181.
Practically all benchmarks score just a hair short of the 66 except Doom for some reason, yet Doom "choppiness" looks closer to the 66 even if the score is like the 33.

Anyway, a Pentium 233MMX with the cache disabled gets me in the ballpark between 33 and 66 which is sufficient to my 486 gaming needs.

Some big picture info that people might find useful with the cache tricks.

The Pentium MMX, as well as the Cyrix, are the fastest chips with caches disabled. They are your best bet to get DX2 speeds with L1 off, but L2 on. I think in my mind I settled for a DX2 50 as a good comparison.

The regular Pentium is an average performer with caches off. And the AMD is the slowest, which makes it very useful for 386 era games, but you're just getting 486DX speed with L1 off and L2 on.

With each architecture, the FSB and memory timings can also help dial in speed.

PhilsComputerLab wrote:

Some big picture info that people might find useful with the cache tricks. […]
Show full quote

Some big picture info that people might find useful with the cache tricks.

The Pentium MMX, as well as the Cyrix, are the fastest chips with caches disabled. They are your best bet to get DX2 speeds with L1 off, but L2 on. I think in my mind I settled for a DX2 50 as a good comparison.

The regular Pentium is an average performer with caches off. And the AMD is the slowest, which makes it very useful for 386 era games, but you're just getting 486DX speed with L1 off and L2 on.

With each architecture, the FSB and memory timings can also help dial in speed.

So a K6 @ 600mhz with cache off is slower than a pentium MMX 233 with cache ofF?

candle_86 wrote:

So a K6 @ 600mhz with cache off is slower than a pentium MMX 233 with cache ofF?

Yes, surprising maybe, but it is what it is 😀

EDIT: I have to clarify that this is at the same FSB speed!

I have a mmx 233 pentium but with a intel se440bx currently so can't disable the cache. Are there any slot 1 mobo's out there that will allow this?

You can't disable cache with that board? Even via software? ICD/ICE, ICACHE, etc.

Have not tried the software route (setmul???) but going into the bios the cache settings are grayed out. I guess if you got a socket 7 variation it's more straight forward to do this.

Just use software. Might have to try a few tools. You can disable L1 cache through software on modern machines 😀