cyclone3d wrote:

Couldn't the Intel 386-33 just have been easily overclocked to 40?

Possibly, but the 33 MHz version dates from 1989, so the manufacturing process was somewhat outdated compared to the one AMD used, or the one used by Intel for the 486 at 50 MHz.
Intel could certainly have made a 40 MHz model, if not by overclocking the older one, then by a die-shrink to bring it to the newer manufacturing process. It may even have been possible to go beyond 40 MHz.

The one thing the 486SX has going for it is that it allowed localbus. But the chips were more expensive to make than 386DX ones, so who knows what the exact logic behind Intel's decision was.
It looks like Intel just re-purposed all 386DX production facilities to make the slower 486 models, and used the newer manufacturing for the high-end 486es. And Intel apparently didn't reverse that decision and put the 386DX back in mass production.

Plenty of speculation there. Local Bus, in particular VLB was certainly not 486 specific. VLB with 386DX-40 ?

Even EISA 386 systems existed. The ALR Powerpro: - Dual 386 EISA System

Caluser2000 wrote:

Plenty of speculation there. Local Bus, in particular VLB was certainly not 486 specific. VLB with 386DX-40 ?

DXSLC_Brochure_1992.pdf

Yea, every rule has its exceptions.
But these were extremely rare and generally quite buggy (seeing as VLB is a direct connection to the 486 bus and technically *is* 486-specific. There were also Pentium boards with VLB but they never worked quite right either, for the exact same reason).

Caluser2000 wrote:

Even EISA 386 systems existed. The ALR Powerpro: - Dual 386 EISA System

Sure, but EISA is not localbus.
Also, 386DX-40 was a budget option, and EISA most certainly was not budget. You only found it on expensive workstation and server machines. So not in the market for the 386DX-40/486SX-25 budget machines for consumers.

Scali wrote:

Caluser2000 wrote:

Plenty of speculation there. Local Bus, in particular VLB was certainly not 486 specific. VLB with 386DX-40 ?

DXSLC_Brochure_1992.pdf

Yea, every rule has its exceptions.
.

No. It plainly points out you were incorrect on your assumption.

Caluser2000 wrote:

No. It plainly points out you were incorrect on your assumption.

No I'm not.
Whether or not some esoteric 386 boards with flaky VLB exist or not does not change anything about the fact that VLB was a standard option for most, if not all 486SX machines (perhaps on very late ones it was replaced by PCI). And on these 486SX machines, VLB works like a charm and gives a very significant performance advantage.
386 boards with VLB on the other hand are extremely rare, and even if you manage to find one and get it working, the performance gains are marginal at best.
I'd say that perhaps 95% of all 386DX-40 users would have an a ISA-only system. No VLB, no EISA. So they are irrelevant.

You're just being argumentative, and I have no idea why. Just stop it.
Clearly you should have gotten the gist of what I was saying, so taking things out of context and trying to pull them apart in this way is rather pathetic and annoying (just like your continued insistance that the Am486DX did not start out as an illegal copy of an Intel chip, infringing on patents and whatnot).
I never claimed that 386 with localbus didn't exist anyway (and obviously I knew they did, because I actually commented in that topic you linked).

Scali wrote:

But these were extremely rare and generally quite buggy (seeing as VLB is a direct connection to the 486 bus and technically *is* 486-specific. There were also Pentium boards with VLB but they never worked quite right either, for the exact same reason).

Happy Nx586 VL-Bus user here...

I think it is better to say that the VL-Bus has been modelled after 486 CPU bus, but supports both 386 and 486 as well as other CPUs if implemented properly. There are some differences when using non-486 (such as handling of reset or 16-bit word steering), but VESA compliant devices should know about these and work with any CPU type. If not it is just 486 Local bus device and not VESA.

mpe wrote:

I think it is better to say that the VL-Bus has been modelled after 486 CPU bus, but supports both 386 and 486 as well as other CPUs if implemented properly.

Not really.
The VLB *is* the 486 bus. It connects directly to the CPU.
A 386, Pentium or whatever other CPU has a different bus, and as such will require a separate bus interface chip to translate the signals.
The problem here is that it's not impossible to design and make such a bus interface chip. It was however not economically interesting to the larger chip manufacturers.
So this opportunity was mainly jumped on by smaller Asian companies, who managed to reverse-engineer and hack together a solution. Which is why it's generally so buggy in practice. They are just poorly engineered (and even on real 486 machines, VLB isn't exactly perfect. It is inherently a very timing-sensitive and electrically sensitive design).

Over the years I've seen many such contraptions, and they never worked quite right.
I also recall a 486 board with both VLB and PCI that a friend of mine had. Sure, there was VLB on there, and it actually sorta worked. The problem was that they had apparently messed up the timing somewhere, so if you inserted a VLB card, all ISA cards stopped working.
The board worked fine as long as you only used the VLB slot, and none of the ISA slots. But once you wanted to use both (say, install a sound card), everything went wrong very badly.

Scali wrote:

Not really.

That was actually a quote from VL Bus specs which goes into details about what the differences between VESA Local bus and 486 Local Bus are 😀

mpe wrote:

That was actually a quote from VL Bus specs which goes into details about what the differences between VESA Local bus and 486 Local Bus are 😀

Oh really?
Well, even if they say that, it's misleading (what it basically translates to is exactly what I said: you need additional logic in the chipset to mimic the 486 bus behaviour... which in theory can be fine, but in practice is mainly done by hackjob chipsets from vague Asian companies, doesn't meet the 'implemented properly' criterion).
Fact is that the VLB connects cards directly to the 486 memory bus, and has very little additional logic in the chipset.
Fact is also that it didn't translate to Pentiums well for this exact reason. The same applies to 386 of course.
Since the Nx586 is its own thing, it could well be that it used a 486-like bus (or else solved in its custom chipset/board design). But we weren't talking about Nx586 anyway, we were talking about 386DX.

Ahh goodness. This gets funnier and funnier 🤣. Ahhh the wonderful AMD386DX-40.Crusher of the Intel 386SX line he he.

Scali wrote:

.
I'd say that perhaps 95% of all 386DX-40 users would have an a ISA-only system. No VLB, no EISA. So they are irrelevant..

And what metric do you base that on? C'con figures big boy.

Caluser2000 wrote:

And what metric do you base that on? C'con figures big boy.

You're the one who claims 386DX with VLB is relevant. You provide the figures. Burden of proof is on you. As evidenced by the very thread you linked to, common perception is that 386 VLB systems are very rare.

Scali wrote:

Caluser2000 wrote:

And what metric do you base that on? C'con figures big boy.

You're the one who claims 386DX with VLB is relevant. You provide the figures. Burden of proof is on you. As evidenced by the very thread you linked to, common perception is that 386 VLB systems are very rare.

I'm that YOUR assumtion that 486 production was directly related to local base in particular VLB. Which is NOT correct. You also assume just because you never saw a And386DX-40 until you got you you 486DX2/66 which was when? I bet I've seen a damn site mofre than you have because the first time on the job we got very good offers from a local seller to0 purchase them. You probaly were still using your Olivetti then 🤣. No sorry 386sx16 was it?

And I can cdertainly say 100% of 486 systems I've purchased have been upgrade from a 386/486SX class cpu.

hank goodness I took a sleeping pill other wise I'd be up All night chuckling

Caluser2000 wrote:

that YOUR assumtion that 486 production was directly related to local base in particular VLB. Which is NOT correct.

Well no, you have made no compelling argument to argue that it is not correct (aside from the fact that I never actually said this, just re-read my statement. You are creating a strawman argument).
I could repeat what I said, but why bother, just scroll up to see you being refuted.

Scali wrote:

Caluser2000 wrote:

that YOUR assumtion that 486 production was directly related to local base in particular VLB. Which is NOT correct.

Well no, you have made no compelling argument to argue that it is not correct (aside from the fact that I never actually said this, just re-read my statement. You are creating a strawman argument).
I could repeat what I said, but why bother, just scroll up to see you being refuted.

Lol now out comes the ol' strawman. Excuse me while I go to the bath room....

mpe wrote:

I think it is better to say that the VL-Bus has been modelled after 486 CPU bus, but supports both 386 and 486 as well as other CPUs if implemented properly.

Just a quick 2 cents from me. This sentence is somewhat good one to respond to.

The whole thing about VLB is that the cpu itself is the implementation.
And only 486 really implements it.

386 and pentiums need emulating hardware in between or in parallel to help them along as they don't natively fully implement VLB. Remember, CPU hardware IS the implementation, not chipset.

(I bet a good electronics engineer can implement some hardware to support VLB over PCIe on modern machine and do all the translation in between. But 486 cpu is special because it doesn't need anything to help it.)

Caluser2000 wrote:

Lol now out comes the ol' strawman. Excuse me while I go to the bath room....

Yes, your strawman.
I repeat:

The one thing the 486SX has going for it is that it allowed localbus. But the chips were more expensive to make than 386DX ones, so who knows what the exact logic behind Intel's decision was.

Clearly, I merely mention localbus (practically VLB at that time) as a possible reason why 486SX could be more compelling than a 386DX (seeing as the existing 386 ecosystem had no VLB boards and chipsets, after all, VLB was only introduced a few years after the 486 launch, and only then just trickled down into 386DX machines sporadically, which would have been well after Intel's decision was made).
I also mention a downside here: 486 is more expensive.
So I say I don't know what the logic behind the choice was exactly.

Then you go argue this:

YOUR assumtion that 486 production was directly related to local base in particular VLB. Which is NOT correct.

I never said 486 production was directly related to VLB. I merely said that it was possible that Intel factored this in when making their decision (but who knows how many other things they factored in, and how they weighed them. I clearly said I don't know).
So you are indeed attacking a strawman.

Aside from that, you never provided any compelling evidence that this would not be correct. So you are not even attacking it convincingly.
And then you resort to personal digs. We've all seen the pattern before...

I'd say there is a difference between 386 and Pentium/Nx586.

Pentium/Nx586 surely cannot be interfacing directly. If for nothing else its 64-bit 60/66 MHz electricals would be a problem.

The VL-Bus spec states that:

Hosts other than 386- or 486-class CPU may require control translation circuity between the host CPU local bus and the VL-Bus.

In fact even some 486 systems can have indirectly connected VLB devices where the VLB is interfacing to CPU bus indirectly through a controller.

But 386 bus i close enough to the 486 and some subtle differences are accounted for in the specs and are burden of VLB devices to implement (if VESA compliant, there are surely non-VLB local bus or non-compliant devices). The VL-Bus even defines 16-bit operation modes for 386SX CPU and I've seen a plenty of evidence (like Cirrus Logic VGA chips datasheets) that this is implemented.

Do you have a link to that spec? Because I wonder what revision you're looking at.
Sounds to me like that is a very late revision of the spec, with features that were not originally there (for example, in the specs of VLB I've seen, the two least-significant bits of the address lines are missing altogether (no address 0 and address 1 lines), so you can only do 32-bit aligned transfers. Making 16-bit transfers physically impossible on half of the 16-bit aligned addresses).
I've only ever seen specs that detail 32-bit and 64-bit operation of VLB, nothing else.

Gives me a taste of VESA BIOS Extensions. Yes, on paper there were these various revisions of the standard, with various features. In practice, most hardware didn't implement most of them (or implemented them incorrectly), and you had to rely on UNIVBE to fix it.

This is from 2.0 from 1993. I know, what you want to say. But 1.0 spec revision is nowhere to be found and there is a plenty of evidence the 386DX/SX bus support has been always there and is implemented by targets as documented by various VGA controllers datasheets.

The 2.0 is basically added bursting, 64-bit multiplexing, write-back cache support, enhanced timing, 3.3V for onboard devices + other minor corrections.