The Serpent Rider wrote on 2020-01-13, 23:17:

Some of you probably already know, that late AMD 486DX2-66 with 16Kb cache were just rebranded 5x86 ADZ CPUs. But apparently the […]
Show full quote

Some of you probably already know, that late AMD 486DX2-66 with 16Kb cache were just rebranded 5x86 ADZ CPUs. But apparently the story with AMD rebranding doesn't end there.

Meet this CPU:
AM486DX4-100SV8B.jpg

On the first glance, it's nothing interesting. Just your typical semi-late 8Kb DX4 with write back option. But look at the production number in the corner - 25544. It's the same number for all enhanced 486 350nm 16Kb cache CPUs (ADZ, BGZ, etc).
All tests reported 16Kb L1 cache capacity and yes, it has undocumented 4x multiplier. Performance is also exactly identical to 5x86. So it's practically a 5x86-133 ADW!

What does it all mean? Looks like AMD didn't bothered much with a separate production and at some point were just quietly replacing old offerings with the new ones under the hood. It wouldn't be surprising to discover some late NV8T with Write back support or 3.3v DX2 rebranded as 5v DX2 CPU.

Wow, that's really interesting.
A stealth 5x86 CPU.
But I don't think they did rebrand 3.3v cpus as 5v, because they would burn in 5v boards, they would not have taken that risk 😁

Not necessarily, they are quite robust and, as example, we had 3.3v/5v capable EDO and SDRAM.

I have one of those cpus too where it says it is suppose to have 8Kb cache but has 16Kb cache. I think I made a post in this forum about it long time ago. It also runs at 133Mhz well.

Remember this?
Re: 486 OC and cooling.

Do you have the photo of that CPU?

Looks like AMD didn't bothered much with a separate production and at some point were just quietly replacing old offerings with the new ones under the hood.

That's completely normal and still how CPUs are selected today. This was probably going to be a 5x86-133 from the ouset but failed the test for 133 MHz, so it was sold as a DX4-100.

Hmm my V8B has 25498 in the corner, I guess it's a regular potato?

Now I have a proof, that even some 5V CPUs were relabeled:

Production number - 25220. Which is NV8T DX2 series, designed to work at 3V. That particular CPU doesn't work at 3V, but is stable at 4V. Can't be oveclocked to 80Mhz even with 5V though. In other words, it's a faulty 3V CPU, relabeled as 5V.

So it's safe to assume that you can use early DX2 NV8T CPUs in a motherboard without voltage regulator. It will work just fine. Maybe some DX4 3V CPUs as well.

derSammler wrote:

This was probably going to be a 5x86-133 from the ouset but failed the test for 133 MHz, so it was sold as a DX4-100.

It does work at 133Mhz without any voltage increase, but doesn't boot at 160Mhz.

How do you decode the 25XXX production numbers?

Here are the Am486 chips I have:

DX2-50 24361
DX2-66NV8T 25220
DX4-100NV8T 25253
DX4-100SV8B 25498

Any of them hit the jackpot?

http://www.cpu-world.com/forum/viewtopic.php?t=25326

DX4-100SV8B 25498

Possibly 16Kb cache, but may be disabled.

The Serpent Rider wrote on 2020-01-14, 14:29:

DX4-100SV8B 25498

Possibly 16Kb cache, but may be disabled.

How do I check? Just plug into a motherboard and fire it up, then run cachechk?

Yes. Also Speedsys.

The Serpent Rider wrote on 2020-01-14, 10:35:

Do you have the photo of that CPU?

Here is one:

I found another old thread with similar discussion, which never took off: Am486 DX4-100V16BGC Multiplier - 2/3x or 3/4x?

To summarize so far:
All 25544 - 350nm, 16Kb WB L1 cache, 2x/3x or 3x/4x multiplier.
All 25220 - the original N8VT, 700nm, hybrid 5v/3v core.
All 25253 or higher - 486DX4 with 2x/3x multiplier (includes DX2), 500nm.

I also have a DX2-80 that's a DX4-100 under the hood and a DX4-100 that's a 5x86 under the hood (but it can't run at 160MHz, only 133). I'll take a picture of both of them if you want to.

Interesting ! Not in boards at this time I have a 25253 labeled A80486Dx-100NV8T 3v and a 25544 labeled AMD-x5-133ADW 3.45v. Not sure what AMD DX4-100 and DX4-120 are in the boards but will have to check...

The Serpent Rider wrote on 2020-01-14, 13:43:

Now I have the proof, that even some 5V CPUs were relabeled:
Production number - 25220.

I wouldn't conclude it's a relabeled CPU, merely a hybrid process that can work with both 3V3 and 5V. I suppose AMD might have binned the chips as well, depending on performance and electrical characteristics of each die.
I have 486SX2 with 25220 code, and all SX are 5V CPUs. Never tried running it at 3V3.

Personally what I'm interested in is thermals, is 25220 any better than 24361 when it comes to heat at 5V? Both seem to be 700nm process, has anyone seen any 5V AMD CPUs with 25253 or later code?

So that means I can safely use this in 5V only motherboards? I assume I just need to set it as an AMD DX2?

And this can be run at a 2x multiplier? (How? Most motherboards don't have jumpers to differentiate DX2 and DX4..)

And this may actually have 16K WB cache?

appiah4 wrote on 2020-02-12, 14:34:

So that means I can safely use this in 5V only motherboards? I assume I just need to set it as an AMD DX2? […]
Show full quote

So that means I can safely use this in 5V only motherboards? I assume I just need to set it as an AMD DX2?

And this can be run at a 2x multiplier? (How? Most motherboards don't have jumpers to differentiate DX2 and DX4..)

And this may actually have 16K WB cache?

I thought only ones with 25544 in the corner had the 4x multiplier and 16K cache because they had same number as 5x86 133Mhz ones.