At first glance I thought someone made an adapter for s7-fpga.
After that I read Tualatin and I was like 🙁
It's gonna come, someday.

What about FPUs?
Mabe a better performer with new tech?

Nexxen wrote on 2021-12-04, 10:27:

What about FPUs?
Mabe a better performer with new tech?

faster quake on a 486 or maybe even a 386 ? 😁

weedeewee wrote on 2021-12-04, 11:15:

Nexxen wrote on 2021-12-04, 10:27:

What about FPUs?
Mabe a better performer with new tech?

faster quake on a 486 or maybe even a 386 ? 😁

We start a huge campaign for a super performing FPU just to have Cpugalaxy go nuts on OCing a 386 😀 🤣

Nexxen wrote on 2021-12-04, 11:25:

weedeewee wrote on 2021-12-04, 11:15:

Nexxen wrote on 2021-12-04, 10:27:

What about FPUs?
Mabe a better performer with new tech?

faster quake on a 486 or maybe even a 386 ? 😁

We start a huge campaign for a super performing FPU just to have Cpugalaxy go nuts on OCing a 386 😀 🤣

Linky?

Nexxen wrote on 2021-12-04, 11:25:

weedeewee wrote on 2021-12-04, 11:15:

Nexxen wrote on 2021-12-04, 10:27:

What about FPUs?
Mabe a better performer with new tech?

faster quake on a 486 or maybe even a 386 ? 😁

We start a huge campaign for a super performing FPU just to have Cpugalaxy go nuts on OCing a 386 😀 🤣

Instead of the (Intel) Overdrive, you can call it the Warpdrive 😜

Tetrium wrote on 2021-12-06, 08:17:

Nexxen wrote on 2021-12-04, 11:25:

weedeewee wrote on 2021-12-04, 11:15:

faster quake on a 486 or maybe even a 386 ? 😁

We start a huge campaign for a super performing FPU just to have Cpugalaxy go nuts on OCing a 386 😀 🤣

Instead of the (Intel) Overdrive, you can call it the Warpdrive 😜

CpuG: Hello!
I'm here today to overclock the hell out of this new Warpdrive.
I'm expecting an increase around 350% on Quake, making this the fastest Quake run on 386.

Atheatos: hold my greek wine...

This is the kind of war you'd like to see 😀

Anyone who started his computer journey on a 8086 or 186 can relate to this dreaming of faster cpus.

New chipsets t allow higher FSB? 486 @ 87.5mhz 🤣

I can't think of a single platform I know of where the number of CPUs out there do not vastly outnumber the number of motherboards. If you have a motherboard, you can definitely find a CPU but not vice versa, so I don't think this is a real necessity.

That said, some computers/platforms never had wide spread CPU upgrade options, and for things like that I can see this happening. I guess PiStorm is a perfect example of such.

appiah4 wrote on 2021-12-06, 11:01:

I can't think of a single platform I know of where the number of CPUs out there do not vastly outnumber the number of motherboards. If you have a motherboard, you can definitely find a CPU but not vice versa, so I don't think this is a real necessity.

That said, some computers/platforms never had wide spread CPU upgrade options, and for things like that I can see this happening. I guess PiStorm is a perfect example of such.

By “new” the implication was faster.

There is also a thread about “new motherboards for old sockets”.

Considering the pinout of the 486 and/or Pentium and considering the adapters available or Overdrive adapted cpus, would be that difficult to use the FPU only of a 486/Pentium and somehow driven by the 80386 cpu leaving the 486 cpu in a sleep mode? I suppose it's difficult if not impossible but it'd sound great to see a powerful FPU into a 387 socket.. 😉
I can't imagine a 386 config running Quake at something like @5-10 frames/s with some VLB video cards too. 😁

There's already dozens of 486DX33s in the world that are fast enough to be too slow to play Quake on.

That's true. 😉

So let's use the Pentium Overdrive 5V FPU running @40Mhz.. 😁.. or maybe is better to wait someone realizing a project that might use some compatible FPU chip in some micro package and an adapted PCB. Maybe a FPGA version might be something realistic even if I suppose very difficult.

You can already drop the PODP5V down to the speed of the FSB on the motherboard.

For super slow fun, I have a board that I can drop a Cyrix 5x86 down to 2Mhz.

What exactly would be the point of running a 386 with a more powerful FPU? Anything that needs a more powerful FPU is going to work fine on a system with a CPU with a more powerful FPU.

What would be useful is having a system with a real-time programmable clockgen. I've actually though about modifying a Socket 3 board with something like this.

Interesting. How fast it would runs Quake @ 40Mhz compared to the 487/387 anyone tested?

A programmable clock would be very interesting too!

386sx... don't you realize that a 487 is just a plain 486dx ?

I was calling it 487 just to focus on the FPU part of the 80486DX core. 😉 It's still faster than the 80387 for sure at the same clock, isn't it.

Intel atom simulated on a fpga plugged into socket 5/7 motherboard

Many former chip fabs were scrapped, one larger Motorola fab and complex was gutted and scrapped.

One has to wonder since labor/ r&d is the main expense to making new masks if a late 90’s / 2k fab could be sourced and minimized for job lot production levels aka converted to a smaller process so you don’t pop hundreds or thousands of cores at once

MOS ran like a job lot for close to a decade, “testing “ a proto mask was $10,000 in just materials
gotta wonder if that could be translated to slightly obsolete fabs?
Clean Raw materials is a problem , size/power requirements of equipment the other .

Having the real deal would allow you to right size and burn custom silicon without resorting to expensive programmable logic.

Legal ramifications of just copying a core to use for hobby use should be controllable if you limit to 25 year plus designs .

rmay635703 wrote on 2022-01-01, 04:36:

Many former chip fabs were scrapped, one larger Motorola fab and complex was gutted and scrapped. […]
Show full quote

Many former chip fabs were scrapped, one larger Motorola fab and complex was gutted and scrapped.

One has to wonder since labor/ r&d is the main expense to making new masks if a late 90’s / 2k fab could be sourced and minimized for job lot production levels aka converted to a smaller process so you don’t pop hundreds or thousands of cores at once

MOS ran like a job lot for close to a decade, “testing “ a proto mask was $10,000 in just materials
gotta wonder if that could be translated to slightly obsolete fabs?
Clean Raw materials is a problem , size/power requirements of equipment the other .

Having the real deal would allow you to right size and burn custom silicon without resorting to expensive programmable logic.

Legal ramifications of just copying a core to use for hobby use should be controllable if you limit to 25 year plus designs .

Seems like fpga is the future of this idea

Unfortunate that we are seeing such low mhz though. That atom project is super cool but only 50mhz kinda sucks.

Sphere478 wrote on 2022-01-01, 04:42:

rmay635703 wrote on 2022-01-01, 04:36:

Many former chip fabs were scrapped, one larger Motorola fab and complex was gutted and scrapped. […]
Show full quote

Many former chip fabs were scrapped, one larger Motorola fab and complex was gutted and scrapped.

One has to wonder since labor/ r&d is the main expense to making new masks if a late 90’s / 2k fab could be sourced and minimized for job lot production levels aka converted to a smaller process so you don’t pop hundreds or thousands of cores at once

MOS ran like a job lot for close to a decade, “testing “ a proto mask was $10,000 in just materials
gotta wonder if that could be translated to slightly obsolete fabs?
Clean Raw materials is a problem , size/power requirements of equipment the other .

Having the real deal would allow you to right size and burn custom silicon without resorting to expensive programmable logic.

Legal ramifications of just copying a core to use for hobby use should be controllable if you limit to 25 year plus designs .

Seems like fpga is the future of this idea

Unfortunate that we are seeing such low mhz though. That atom project is super cool but only 50mhz kinda sucks.

First High end FPGA has to be so common it’s pennies for modern speeds or the labor to setup is hard to justify .

If the circuitry is moderately simple even fairly ancient silicon can support 100mhz +
All the added complexity within an FPGA limits its ultimate top speed, you don’t have this when you burn real silicon.

Been waiting 20 years for FPGA to become obsolete tech found in kids toys , it’s getting closer but still not cheap by any stretch.