The 486SX has the same bus as the DX, unless you are thinking about the 386sx or I am unaware of something!

brostenen wrote:

The 486slc I had in 1993. Really sucked at games from 1993 and onwards.
On the other hand, it was a really shitty repack of the 80386 cpu.

It was actully a "shitty" repack of a 486 cpu. Internally 486slc/dlc are 486.

mwdmeyer wrote:

The 486SX has the same bus as the DX, unless you are thinking about the 386sx or I am unaware of something!

I'm pretty sure he's thinking of 386SX.

kixs wrote:

It was actully a "shitty" repack of a 486 cpu. Internally 486slc/dlc are 486.

Ahh... I see.
It was basically one of the worst for Doom and games alike.
I remember feeling ripped off, as not even Epic Pinball ran good on that machine.
I had virtually no idea what I bought at that time.
Just remembering that I saw the numers "486" and bought thinking I got a fast pc.
First pc-buy mistake I guess. 😁

I did manage to play the entire first episode of Doom on a 486DLC-40... Though I used it as an upgrade to a 386DX-40 and already had a very fast motherboard and graphics card, my motherboard also supports the Hidden Refresh and such.

As a result, I rather like the 486DLC.

486SX processors should run pretty much identically to their DX counterpart for games bound to Integer Math. Many of them were simply 486DX's with the FPU disabled; this was achieved by cutting away the connections to the die with a laser.

HighTreason wrote:

Many of them were simply 486DX's with the FPU disabled; this was achieved by cutting away the connections to the die with a laser.

Remembering how people back in the early 90's explained this to me, as Intel were not good at producing fail-free DX chips.
If it was a biased argument towards AMD, I simply don't know. Yet people just said that Intel made a lot of mistakes in the
production of 80486-DX chips. Even the magazines hinted that in their articles.

Maybe. I was always told that the SX sold better than expected and Intel simply cut the FPU off the DX to keep up with demand.

I suppose it could also have been that it was cheaper to make only one model, so perhaps it was the other way around and Intel simply made DX chips and cut the FPU off when it was defective or to meet demands and would only make SX chips when they couldn't keep up with demand that way.

Possibly we'll never know for sure. Though given their luck with the 386 and the double sigma thing that arose, your theory has credibility.

Hehe.... Company-secrets, stored in the dark corners of an old archieve.
Makes me think of a plot for a movie about stealing information. 😁

HighTreason wrote:

Maybe. I was always told that the SX sold better than expected and Intel simply cut the FPU off the DX to keep up with demand.

I suppose it could also have been that it was cheaper to make only one model, so perhaps it was the other way around and Intel simply made DX chips and cut the FPU off when it was defective or to meet demands and would only make SX chips when they couldn't keep up with demand that way.

Possibly we'll never know for sure. Though given their luck with the 386 and the double sigma thing that arose, your theory has credibility.

Yes, that is what happened then, and what is still happening now. Which is why I referenced the Celeron.
The FPU was probably the most complex part of the circuit, and the quickest to fail (in the 286-era, Intel actually chose to run the FPU at 2/3rd the clockspeed of the main CPU, probably for the same reasons).
So they built DX chips... tested them, and the ones where the FPU failed, could still have a fully functional CPU otherwise. So, disable the FPU and sell it as 486SX.
These days the same techniques are still used on CPUs and GPUs, usually to disable some caches, some cores and whatnot.

It gets worse actually: a 487SX is not just an FPU. This is again a complete 486DX die. If you place a 487SX in your motherboard, the main 486SX is actually bypassed altogether.
So basically, the 486DX, 486SX and 487SX are all the same chip 😀

And yes, it is possible that they sold 486DX as 486SX if the demand was high enough. This again is something that is still done to this day with CPUs and GPUs.

I had an SX 33 and it was a good chip at the time, not quite up to DOOM but it was playable. I had access to an SX 25 machine too and recall it being notably less impressive.

The Redhill guide has some interesting things to say about the SX and Intel's marketing tactics: http://www.redhill.net.au/c/c-3.html#486remarking

The 486SX was a strange beast. Essentially, it was a 486DX with the maths co-processor disabled. This involved one extra step in the manufacturing process, so the 'cheap' 486SX actually cost more to make than the 'expensive' 486DX! The SX was developed mostly for marketing reasons — to give Intel a relatively cheap chip to equal or better the AMD 386DX-40 but not cut the heart out of their massively profitable flagship, the 486DX-33.

How could it be 'cheap' when it cost more to make than the DX? Because the actual manufacturing cost was so low anyway that they could still sell it at a good margin. An SX-33 used to retail for around A$200, the DX about twice that, but they were both understood to cost less than US$10 to make. So an incremental increase in manufacturing cost just wasn't significant. (Later on, they got around to making SX chips without co-pro, which saved them a few thousand transistors.)

HighTreason wrote:

I did manage to play the entire first episode of Doom on a 486DLC-40... Though I used it as an upgrade to a 386DX-40 and already had a very fast motherboard and graphics card, my motherboard also supports the Hidden Refresh and such.

As a result, I rather like the 486DLC..

The very first PC I bought (in late 1993) was a Cyrix 486DLC-40 with its math coprocessor (it came in one "package"). I remember that I had to choose between an Intel 486DX-33 or the Cyrix 486DLC.
I still have that CPU and the FPU. Recently started it up and it's still going strong.
However, if I had known back then what I know now, I probably would have gone for the Intel 486DX-33 (purely because of the "true" 486 upgrade path).
That being said, I also do like the 486DLC since it offered a great upgrade path for owners of 386 machines and the 486DLC-40's performance was fairly on par with your 486DX-33 (in terms of integer performance).
Back then, with the cost of PC's being very high, it was a very attractive option.

kixs wrote:

It was actully a "shitty" repack of a 486 cpu. Internally 486slc/dlc are 486. […]
Show full quote

brostenen wrote:

The 486slc I had in 1993. Really sucked at games from 1993 and onwards.
On the other hand, it was a really shitty repack of the 80386 cpu.

It was actully a "shitty" repack of a 486 cpu. Internally 486slc/dlc are 486.

mwdmeyer wrote:

The 486SX has the same bus as the DX, unless you are thinking about the 386sx or I am unaware of something!

I'm pretty sure he's thinking of 386SX.

It depends which 486SLC your are referring to (IBM or Cyrix/Texas Instruments).
The IBM 486SLC CPU's were based on the Intel core and actually performed a bit better than the equivalent Cyrix/Texas Instruments SLC CPU's.

However, both CPU's only had a 16-bit external data path, resulting in less desirable performance, compared to your Intel SX CPU's (which had a full 32-bit external data path).
I also wouldn't have been impressed by a 486SLC CPU. However, the 486DLC-40 CPU was a much better performer and was actually able to more or less match the (integer) performance of your Intel 486DX-33.

Cyrix 486slc-33 is on par with 386DX-33, even on 16bit and without external on-board cache and it's around 50% faster than 386SX-33. I had it in 1993 and also have one in my collection. 486dlc is much more powerfull.

Caluser2000 wrote:

mwdmeyer wrote:

The 486SX has the same bus as the DX, unless you are thinking about the 386sx or I am unaware of something!

My thoughts exactly. Some smartallic quotes a book without even reading it.

The books themselves are hard to rely on, don't blame elianda, he's shown himself as a well-educated person before and I'm sure that's an honest mistake. I have a book called "IBM PC для пользователя" from a respacted author Figurnov and it features some lame mistakes. For expample, on the unlucky page 13 he has this line:

Микропроцессор Intel-80486 ничем не отличается от Intel-80386, но его производительность в 3-4 раза выше.

Intel-80486 microprocessor is not in any way different from Intel-80386, but it has 3-4 times better performance.

I guess the reason behind this is the fact that internet was smaller back in the day and the authors had to write books entirely according to their personal experience and the material they find in manuals. The computer world was evolving fast and there was simply nothing to use as reference because once the book on this material is released, the second book from a different author would not sell half as well, especially given how much time there is between a finished manuscript and the first appearance on the shelves.

That's why I removed that post. His ref. may even have been tongue in cheek. A few computer mags of the time had pretty detailed comparisons about the differences between the various cpus. And of course there was usenet.

RacoonRider wrote:

Микропроцессор Intel-80486 ничем не отличается от Intel-80386, но его производительность в 3-4 раза выше.

Intel-80486 microprocessor is not in any way different from Intel-80386, but it has 3-4 times better performance.

Well, in the right context, that may even be true.
For example... If we assume he means a stock 386, without a 387, and compare it to a 486DX, while running FPU-code... Then yes, the 486 would easily be 3-4 times faster, since the 386 will have to use an FPU emulation library.

Also, he doesn't mention clock speeds... So if he's comparing a 386DX-16 to a 486DX2-66, it should also be 3-4 times as fast...
Or a 386DX33 compared to a 486DX4-100 for example.
Add in some extra caches and localbus, and if you want, you can make a worst-case scenario where the difference is huge.

But I agree, without any context whatsoever, the statement seems a bit strange. You'd probably interpret it as "a 486DX-33 is 3-4 times faster than a 386DX-33 at everything".
Which is obviously wrong. A 486 is somewhat faster with most code (which was mostly integer back then), as we know the 486SX-25 was a decent competitor to the 386DX-40, but that's not a factor 3-4.

Actually... That was only about the '3-4 times performance'... I have bigger issues with 'not in any way different'. Because the 486 is a different architecture, that's what makes it faster. Better pipelining of instructions, and the L1 cache. Aside from the obvious FPU integration.

Scali wrote:

Actually... That was only about the '3-4 times performance'... I have bigger issues with 'not in any way different'. Because the 486 is a different architecture, that's what makes it faster. Better pipelining of instructions, and the L1 cache. Aside from the obvious FPU integration.

Exactly. The book is dated 1991 (ISBN 5-279-00900-8), so DX2 and VLB are out of question. Later it states that 486 "supports floating-point operations itself", which adds even more confusion. But then again, back in 1991 the 486 were so new and powerful they had ridiculously high prices and the author probably just could not get his hands on any relevant data.

Depends on the context, but I assume that he was talking about the progress. 286 was 16-bit, 386 32-bit and in this matter he was right, 486 was still "only" 32-bit.

The additional cycles statement is indeed only true for 386SX with a 16 bit bus. So the book is wrong in this point applying that also to the 486SX and I just quoted it without enough reflection. As the 486SX appeared 26th of september in 1991, they relied most likely on preliminary documents.
However the book is still a very good reference for in detail 386/486 CPU programming and gives also background information on design decisions. The author was part of the 386 design team.