waterbeesje wrote on 2024-07-31, 14:18:

In lots of threads we can read the 486 DX3 was planned to run at the 2.5 multiplier, but it was cancelled before launch.

Now my big question is: do engineering samples exist or are they really phantoms that never were?

The ill fated P24t used a 2.5x multiplier in a retail overdrive.

As for 486dx3…

30mhz FSB 2.5x multiplier 75mhz clock? Nope

5 volt 25mhz 2.5x multiplier 63mhz twin to the dx2?

Nope DX3 was on an Intel cpu roadmap around 1991 and gone by 1992.

Later we got such gems as Pentium SX and socket 6 and Pentium Overdrive for DX4

Intels glacial pace at releasing useful overdrives put it behind the competition which meant they skipped a bunch of mostly redundant stuff and CPUs that would not have been price or speed competitive at the time. The Pentium Overdrive for DX4 is a sad casualty because it would have been the top official cpu for that architecture.

But instead…

An expensive 63mhz Pentium overdrive in 1995 was a very perplexing offering , especially considering it was the only speed available until late that year.

Being at least a year late at too high a price with too slow a speed rating meant the faster clock to upgrade dx4’s would have been even later to the party.

https://www.cpu-world.com/forum/viewtopic.php … highlight=p24ct

The 33Mhz bus clock was the most popular bus speed, and the Dx2-66 most popular CPU. Don't know why.

So an upgrade CPU for this at 2.5x multiplier would make a 83Mhz CPU, and not competitive to 4x and 5x multiplier CPUs, unless it had a Pentium core.

MikeSG wrote on 2024-08-01, 07:03:

The 33Mhz bus clock was the most popular bus speed, and the Dx2-66 most popular CPU. Don't know why.

There's an Wikipedia article that reminds me of this.

"The i486DX2-66 was a very popular processor for video games enthusiasts in the early to mid-90s.
Often coupled with 4 to 8 MB of RAM and a VLB video card, this CPU was capable of playing virtually every game title available for years after its release,
right up to the end of the MS-DOS game era, making it a "sweet spot" in terms of CPU performance and longevity.

The introduction of 3D graphics spelled the end of the 486's reign,
because of their heavy use of floating point calculations and the need for faster cache and more memory bandwidth.
Developers began to target the P5 Pentium processor family almost exclusively with x86 assembly language optimizations
which led to the usage of terms such as Pentium compatible processor for software requirements.

An i486DX2-50 version was also available, but because the bus speed was 25 MHz rather than 33 MHz,
this was a significantly less popular processor."

https://en.wikipedia.org/wiki/Intel_DX2

Personally, I think the popularity also was because the 486DX2-66 was compatible with most motherboards that could run with a 486DX-33.
It was like a drop-in replacement or upgrade processor to most users of the time.

Stability and robustness was another factor, maybe.
In practice, the 486DX2-66 ran on both 5v and 3.3v boards and needed no active cooling.

Except if being under constant load, of course. Like a server, for example.
That's when a fan and heat sink was being recommended, at least.

Ah, so there probably were no engineering samples too. Then the mystery remains bigger why the 3x multiplier was not badged DX3 in the end...
As a result the AMD 5x86 was called 'dx5' with its 4x multiplier ofc, an intended mistake too.

waterbeesje wrote on 2024-08-01, 12:46:

Ah, so there probably were no engineering samples too. Then the mystery remains bigger why the 3x multiplier was not badged DX3 in the end...
As a result the AMD 5x86 was called 'dx5' with its 4x multiplier ofc, an intended mistake too.

Doubtful , dx3 was dead before dx2 launch due to competitive pressures and thus before they were even sampling. Does make you wonder if a 2.5x muliplier is wired off as errata in the dx4 silicon as mentioned here.

Re: Is there a way to change the internal multiplier of 486DX/n processors?

Would need to find someone who worked at Intel in that era, I used to work with a former Intel employee but he oddly didn’t work on CPUs if that can be believed.

My guess is it’s a simple case of Intel being Intel when they made their roadmap with the DX,DX2,DX3,DX4 and the associated definitions in 1991 they probably refused the redefine and update the roadmap as to what the as yet unproduced dx3,dx4 definitions were in 1992 when they discovered they had to release faster CPUs early due to competition edging in so
they just depreciated dx3 out of stubbornness so they could keep the same roadmap and cpu definitions.

I haven’t seen the early Intel CPU roadmap releases that covered the 486 generation in 25 years. It would be sort of nostalgic if someone would post the slides from the announcements online, I was lucky enough to bum around a mid size integrator late 90’s and got to see some antique nda material. Sad that’s all gone.

A set of 5v. dx2 / dx3 66mhz / 63mhz twins releasing side by side in 1992 would have been interesting but likely a marketing disaster.

That is my guess as well. The DX3 likely never made it to any production. The idea was likely scrapped early on, but the naming definitions made at the time stuck.
It was likely Intel realized they needed concentrate on the next generation CPUs.

I'm sure Intel was fully aware of the massive hype surrounding upcoming RISC CPUs at the time.
Interestingly enough, the performance gains from most of these RISC architectures proved to be true. But in the end lost anyway for pricing themselves out of the low end computer market.

rmay635703 wrote on 2024-08-01, 04:51:

Later we got such gems as Pentium SX and socket 6 and Pentium Overdrive for DX4

What now? Were they on a road map?

H3nrik V! wrote on 2024-08-01, 16:20:

rmay635703 wrote on 2024-08-01, 04:51:

Later we got such gems as Pentium SX and socket 6 and Pentium Overdrive for DX4

What now? Were they on a road map?

Allegedly The Socket 6 3.3 volt p24CT went by 2 different names Pentium Overdrive for DX4 and Pentium SX, at least per the place I worked in.

None of the p24t stuff let alone p24ct was on the roadmaps im aware of (<1991) but I was told they were a thing on later ones.

There is an actual mechanical sample of the P24CT so regardless of what it would have gotten called Intel was considering a line of 32 bit Pentiums that fit a 486 socket in higher clocks. The myth goes that someone in Intel wanted a high and low end line of Pentiums 32/64 bit
but clock speed pressures from competition killed it