• Please review our updated Terms and Rules here

fast 486 vs. older Pentium

C

Casey

Veteran Member
Joined
May 31, 2016
Messages
621
Location
Fairfield, Ohio
Still don't have a working 486 system in my collection, and have noticed that 486 motherboards on eBay have become ridiculously high-priced; I'm guessing because of the retro gaming craze. Apparently there's a lot of retro gamers who insist on authentic period hardware.

Since early Pentium system boards haven't gotten that expensive, I was considering the possibility of getting one of those instead. Given that later model 486s ran at 66Mz, 80Mz, and faster I was wondering how did they perform compared to an early 60Mz or 90Mz Pentium? Does mainboard cache make much of a difference?

I already have a tweener that runs a 1Gz AMD Athlon XP cpu. Trying to run MIPS or SI on that gives some ... interesting results, shall we say. Even disabling the cache & kicking the bus speed down I only get it down to 700Mz equivalent, from the vintage utilities I've tried. They weren't made for modern processors. Not much of a gamer here, but a lot of older software (eg MS-DOS applications) don't like speeds that high

So would a low end Pentium system be an equivalent until I find an affordable 486, or is the generational jump too great? I was thinking of something 200Mz or less.

Oh, the irony. Back in the early 90s I couldn't wait to dump my 486dx2-66 for a Pentium system. Took me forever to afford one, and a "slow" Pentium MMX 233Mz wasn't fast enough. Now I want them back, and... Heh.
 
A 486DX4-100 would match up fairly evenly with a Pentium-60. Vogons has a couple of lengthy benchmarking threads which might be helpful to you. https://www.vogons.org/viewforum.php?f=46

This can vary depending on cache. No L2 cache on the Pentium and it might be slower than a equivalently clocked 486 with correctly sized cache.
 
A pentium is fine, speed sensitive software usually needs something a lot slower than a 486 to work correctly

Pentium systems main advantage is floating point,
“Other “ performance metrics place more modern 486’s on par with Pentium 60-90mhz when floating point isn’t important.
But the older pentiums should run most anything period correct for a 486
 
Thanks for the link.

It's been a long time since I studied Pentiums in detail. Would L2 cache be the motherboard cache in the older systems?
 
It sounds like a 75Mz or 90Mz would qualify as a fast 486, then; perhaps like the the 5x86 133Mz that RadRacer203 mentioned...
 
Casey said:
I already have a tweener that runs a 1Gz AMD Athlon XP cpu. Trying to run MIPS or SI on that gives some ... interesting results, shall we say. Even disabling the cache & kicking the bus speed down I only get it down to 700Mz equivalent, from the vintage utilities I've tried. They weren't made for modern processors. Not much of a gamer here, but a lot of older software (eg MS-DOS applications) don't like speeds that high
Click to expand...

Reason that disabling cache on the Athlon XP doesn't reduce performance much is because the architecture wasn't reliant on cache for performance. This is why the Duron with only 64 KB of L2 cache was only about 10% slower than the Athlon at the same speed. That and because it used an exclusive cache design where L1 cache isn't duplicated to L2 cache. If you want more control over speed, get an Athlon XP-M; It's the mobile variant of the Athlon XP and uses the same socket. It has unlocked multipliers and a early version of Cool'n'Quiet (requires a driver under Windows) to dynamically adjust the clock speed, like Intel's Speed Step.

Intel's Netburst architecture on the other hand was heavily reliant on both large and fast cache for performance. This is why the Netburst Celerons were such pigs, because they had 128-256k of cache until some of the final models based on the Cedar Mill core that got 512k and had respectable performance. Intel was still scummy though, they disabled speed step on Celerons so they consumed a lot more power at all times, and produced more heat. This was a problem on laptops that used the Celeron M (based on the Pentium M), batteries would be quickly drained dry. My Dell laptop I bought in 2006 came with a Celeron 1.5 and the battery lasted exactly 30 minutes. I swapped it out for a Pentium M 2.0 and the battery life jumped to 2.5 hours.
 
I'd say it depends on what you want to run. I always thought Windows 95 ran better on a Socket 5 Pentium than it did on an AMD 133 MHz 486, even on a PCI Socket 3 board. It'd certainly be cheaper, if you had to buy everything anyway.

The early 60/66 MHz Pentiums were dogs, floating point was broken so you lost one of the big advantages. I'd not bother with one of those.
 
glitch said:
I'd say it depends on what you want to run. I always thought Windows 95 ran better on a Socket 5 Pentium than it did on an AMD 133 MHz 486, even on a PCI Socket 3 board. It'd certainly be cheaper, if you had to buy everything anyway.

The early 60/66 MHz Pentiums were dogs, floating point was broken so you lost one of the big advantages. I'd not bother with one of those.
Click to expand...

The goal is mostly to simulate a fast 486 environment until I can find a 486 system or motherboard I can afford.

The mainboard I had at the time came from CompUSA. I had a choice between getting a 486dx2-66 with ISA slots or a 486dx-50 with some VLBus slots. The latter was more expensive, so I went with the dx2-66. It worked quite nicely with 4Mb of ram under DesqView/386. When Win95 came out I bought a Creative Labs sound card/CD-ROM controller. I wasn't about to install '95 on 32 diskettes, or whatever the number was. Increased the ram to 8Mb, and that kept me going for a long time, until I bought a Pentium-90 motherboard, if memory serves. It was a long time ago.

Basically a system that reproduces a 486 environment and doesn't cause vintage software to blow up. IIRC Turbo Pascal needed a patch when cpu speeds went north of 300Mz. A lot of vintage apps either run far too quickly (being written for 4.77 8088 or 6Mz 286 systems) or just blow up with overflow/underflow errors because the clock checks report impossible results.

I've seen some socket 3 boards for sale, but more frequently socket 7 boards. I have some ISA vga, network, and sound cards to hand, so that's not too much of a problem. Even have a SB Live! PCI card somewhere. :)

P.S. Are there versions of DOSBox that run under XP/98SE? Running that on the 1Gz Athlon is a kludge, but it might be workable.
 
Casey said:
P.S. Are there versions of DOSBox that run under XP/98SE? Running that on the 1Gz Athlon is a kludge, but it might be workable.
Click to expand...

Definitely. I've run a version of DOSBox on a PII 500 under Windows 98. The PII was a bit underpowered but it worked. The Athlon will be fine.
 
In DOSBox, you set the emulation speed, similar to other emulators such as Bochs or PCem.
Especially PCem is very much period-correct. Try it on a modern computer.

But then, even on a modern high-end CPU, DOSBox won't be able to do "fast 486" speeds.
 
Well, if you're just looking for a Pentium because eBay 486s are too expensive, perhaps I can fix that problem for you :p

In the "fast PCI 486 vs. Socket 5 Pentium" debate though, I'd personally go with a Socket 5 system every time. To me, an ideal 486 system is a 486DX2-66 either all-ISA or VLB. A 486DX-33 is also often acceptable, depending on the intended use. I know "ideal" varies wildly from individual to individual! I always felt like a DX2 system with VLB graphics was the quintessential high end 486 box though. And, personally, I'd be running DOS 6.22 and Windows for Workgroups 3.11 if I wanted a Wintel box for games or DOS development or something.
 
I always found the idea of a PCI 486 system appealing because they were somewhat obscure "tweener" systems, and I never had one back in the day
 
maxtherabbit said:
I always found the idea of a PCI 486 system appealing because they were somewhat obscure "tweener" systems, and I never had one back in the day
Click to expand...

486 boards with PCI are a mixed bag. Many newer PCI cards that have firmware on them (eg. video cards, disk controllers, etc.) won't work because the firmware was compiled against a Pentium and won't run on a 486 because of missing instructions. It's the same reason you can't run PCI cards designed for x86 on a PowerPC or vice versa.
 
GiGaBiTe said:
486 boards with PCI are a mixed bag. Many newer PCI cards that have firmware on them (eg. video cards, disk controllers, etc.) won't work because the firmware was compiled against a Pentium and won't run on a 486 because of missing instructions. It's the same reason you can't run PCI cards designed for x86 on a PowerPC or vice versa.
Click to expand...

Very good analogy and can save a lot of hair pulling.
 
Back in the day the 486 DX/2 66 VLB was the most popular DOS gaming machine around. The later 486 chips (especially the 133) were also popular with people who could not afford the new Pentium Machines.

486 PCI is great for overclocking a 133 to 160 (still have mine) and much cheaper to get video and network cards for. I never found any problems with PCI video cards of that era on my motherboards, but later cards are probably a different story.

I would think the very first VLB, PCI, EISA, AGP, USB chipsets and boards were somewhat flacky.

PCI 1.0 was 5V while PCI 2.0 was 3.3 V and the slots were flipped. There are universal cards that work in either slot.
 
Unknown_K said:
486 PCI is great for overclocking a 133 to 160 (still have mine) and much cheaper to get video and network cards for.
Click to expand...

Not on the boards that ran PCI clock the same as the CPU FSB. PCI tends not to be very forgiving of overclocking in my experience and 40 MHz PCI slots usually cause issues.

Unknown_K said:
PCI 1.0 was 5V while PCI 2.0 was 3.3 V and the slots were flipped. There are universal cards that work in either slot.
Click to expand...

PCI 2.0 added 3.3v operation, while PCI 3.0 removed 5.0v completely. In practice, 3.3v slots were almost never used outside high end workstation and servers because of compatibility issues with existing cards on the market. The only mainstream machine I can think that uses at least one 3.3v slot is the Powermac G3 B&W.
 
That is true for number crunching, but not all instructions are created equal. Running Windows 95 within DOSBox is possible and performance feels reasonable, but then run some more demanding games (Anno 1602 comes to mind) and performance goes down greatly, while it would work fine on a similar-spec'd system. I never looked into the reason, though - maybe the CPU is fine and the VGA emulation slows down the whole thing.

This is even more noticeable in PCem.
 
You must log in or register to reply here.
Back
Top