Honestly there is no need for rewriting the OS. An extended version of Windows 95 would work fine. What would be needed is more RAM and LESS BLOAT

Web browsers that don't take a 1GB of RAM and multi-gigahert processor. Like Opera 9.6 -- very functional with not much resources.

Web sites would need to be coded efficiently (like they used to be)

Video decoding is a tough nut to crack and that is one thing that would benefit from highly specialized coding to optimize for speed.

Then I suppose a 200MHz 486 would be out by the end of 1996 (there are 5x86 samples that can do that, and it would be just a matter of time before the technology would reach that point). The multipliers would go higher than 4x, we'd probably see FSB increase as well. More motherboards would support larger cache for higher cacheable memory limit. Socket 4 would appear as a new 486 platform featuring lower voltage. Intel then would step in and promise a new Overdrive, which would provide an "easy upgrade", but deliver too little too late.

Since the Pentium-class FPU essential to 3D gaming would never be developped, 3D gaming would have trouble evolving. Voodoo Graphics would dominate the 3D market even more due to lower CPU load, yet the market itself would be much smaller. The demand in stronger FPU would probably lead to the return of 487 or similar socket for a third-party FPU, which would turn the stock 486 FPU off. The new 487 socket would become the dividing factor between gaming/performance PC and office PC. We would probably even see top model motherboards with two 487 sockets for parallel tasking (as you remember, the original Socket 4 Pentium featured 2 FPUs). The 487 FPUs would be made by a large variety of companies, starting with Intel and AMD, then IIT, ULSI, SIS, VIA, OPTi.

This is how I see the "hot" system of 1997 if Pentium never arrived:

AMD 486X 240MHz (6x40MHz)*
2xIIT 4C87DX8 200MHz (5x40MHz)
ASUS X-487D Socket 4 + 2xSocket487 motherboard with 512Kb of asynchronous cache upgradable to 1MB
64MB EDO RAM
Matrox Millinium 4MB PCI videocard
2xVoodoo 2 12Mb SLI
Creative Sound Blaster AWE64

*AMD 5x86 would never be named 5x86 if Pentium did not arrive. Instead, it would be named "486DX6". Intel would be the first to make it and would name it in the tradition of bloating. That way x4 would be DX6, x5 would be DX8, x6 would be called DX10 in press releases, but in order to shorten things up, it would be renamed "486X" instead, bringing even more confusion to the masses.

RacoonRider wrote:

Since the Pentium-class FPU essential to 3D gaming would never be developped, 3D gaming would have trouble evolving. Voodoo Graphics would dominate the 3D market even more due to lower CPU load, yet the market itself would be much smaller. The demand in stronger FPU would probably lead to the return of 487 or similar socket for a third-party FPU, which would turn the stock 486 FPU off. The new 487 socket would become the dividing factor between gaming/performance PC and office PC. We would probably even see top model motherboards with two 487 sockets for parallel tasking (as you remember, the original Socket 4 Pentium featured 2 FPUs). The 487 FPUs would be made by a large variety of companies, starting with Intel and AMD, then IIT, ULSI, SIS, VIA, OPTi.

I disagree. It would have struggled but there would still be 3d games. Doom runs on a 486 for example.
And I am sure there would be work on hardware graphics rendering (i.e. GPUs)
Now if GPUs were halted in the same way CPUs were, it might be another story

smeezekitty wrote:

I disagree. It would have struggled but there would still be 3d games. Doom runs on a 486 for example.
And I am sure there would be work on hardware graphics rendering (i.e. GPUs)
Now if GPUs were halted in the same way CPUs were, it might be another story

The 3D would struggle, because, in other words, it would be held back by FPU. That's what I meant 😀

RacoonRider wrote:

The 3D would struggle, because, in other words, it would be held back by FPU. That's what I meant 😀

Yes, but I think smeezekitty means that GPUs have taken over the role of the FPU by the time the GeForce arrived, with hardware T&L.
If GPUs would evolve, they would just take over most of the FPU work, and the 486 may be quite a feasible option for 3d gaming after all.
I wonder how well a PCI-based GeForce card would work in a 486 actually.

moved to marvin.

RacoonRider wrote:

Since the Pentium-class FPU essential to 3D gaming would never be developped, 3D gaming would have trouble evolving.

Yes, but not too much trouble. Even on the eight bit home computers of the 1980s, 3D gaming was possible, both wire frame (Elite, Starglider, etc) and filled (Driller, The Sentinel, etc). And when you look at how 8-bit computer software improved over time on any given popular machine, such as the ZX Spectrum, or Commodore 64, it does make you wonder how software for the PC 486 would have improved if programmers had to always use the same consistent hardware over a period of years.

smeezekitty wrote:

Web browsers that don't take a 1GB of RAM and multi-gigahert processor. Like Opera 9.6 -- very functional with not much resources.

Don't intend to hijack this thread but I had to post this. http://i.imgur.com/wA67J8Q.png

Kerr Avon wrote:

RacoonRider wrote:

Since the Pentium-class FPU essential to 3D gaming would never be developped, 3D gaming would have trouble evolving.

Yes, but not too much trouble. Even on the eight bit home computers of the 1980s, 3D gaming was possible, both wire frame (Elite, Starglider, etc) and filled (Driller, The Sentinel, etc). And when you look at how 8-bit computer software improved over time on any given popular machine, such as the ZX Spectrum, or Commodore 64, it does make you wonder how software for the PC 486 would have improved if programmers had to always use the same consistent hardware over a period of years.

Probably the same way things happen with consoles. I mean, who could imagine something like StarFox when the SNES was released?

jwt27 wrote:

smeezekitty wrote:

Web browsers that don't take a 1GB of RAM and multi-gigahert processor. Like Opera 9.6 -- very functional with not much resources.

Don't intend to hijack this thread but I had to post this. http://i.imgur.com/wA67J8Q.png

Clearly a memory leak 😵

Honestly, if CPU and RAM wasn't growing so fast, it would probably a good thing for software design. It worse programmers to be efficient and not make horrible bloaty code

jwt27 wrote:

smeezekitty wrote:

Web browsers that don't take a 1GB of RAM and multi-gigahert processor. Like Opera 9.6 -- very functional with not much resources.

Don't intend to hijack this thread but I had to post this. http://i.imgur.com/wA67J8Q.png

70 processes?!?! Holy shit. 🤣 That Opera memory leak is just absurd, but at least you could do something about the number of background processes you have running.

SLI VLB videocards with 3d daughterboards 😜

mr_bigmouth_502 wrote:

jwt27 wrote:

smeezekitty wrote:

Web browsers that don't take a 1GB of RAM and multi-gigahert processor. Like Opera 9.6 -- very functional with not much resources.

Don't intend to hijack this thread but I had to post this. http://i.imgur.com/wA67J8Q.png

70 processes?!?! Holy shit. 🤣 That Opera memory leak is just absurd, but at least you could do something about the number of background processes you have running.

You're kidding right?
http://s28.postimg.org/4jyyr0ahp/process_list.jpg

Scali wrote:

I wonder how well a PCI-based GeForce card would work in a 486 actually.

A GeForce 2 MX400 worked well in my tests using an Am5x86-160 on a UMC8881-based Biostar MB-8433UUD. Using graphics cards which are more modern than the socket 3 generation of hardware is well documented in this thread, Modern graphics on a 486

An excerpt is provided below for the graphic card you inquired about. Recently, I tested Half-Life and it seems to run well in my opinion.

1PNY GeForce2 MX400 64MB
2Win95c
3NVIDIA Detonator v12.41 for Am5x86, however I was unable to get Direct3D working (e.g. Outlaws, MDK). OpenGL works with the Am5x86 with v12.41, however Direct3D games will require a POD. NVIDIA Detonator v77.72 for POD.
4
5AMD Am5x86-160 - GLQuake (nosound, 640x480x16) = 26.7 fps [with sound at 1280x1024x32 = 24.4 fps]
6AMD Am5x86-160 - QuakeII (nosound, 640x480x16) = 15.5 fps
7Intel POD-100 - GLQuake (nosound, 640x480x16) = 36.1 fps (vs. 29.2 fps with v12.41)
8Intel POD-100 - QuakeII (nosound, 640x480x16) = 17.5 fps (vs. 17.3 fps with v12.41)
9
10WinNT4
11For an Am5x86, the NVIDIA Detonator v6.50 (GL v1.2.1) display driver works best, however GLQuake gives a Dr. Watson error when exiting. QuakeII yields a BSOD with relation to the sound card. NVIDIA Detonator v12.41 (GL v1.2.2) works fine with the Quakes, however changing resolutions is a little slow.
12
13For a POD, NVIDIA Detonator v29.42 (GL v1.3.1) is the best of the older drivers, however Detonator v77.72 (GL v1.5.4) yields better Quake results.
14
15AMD Am5x86-160 - GLQuake (nosound, 640x480x16) = 28.4 fps [with sound at 1280x1024x32 = 25.6 fps]
16AMD Am5x86-160 - QuakeII (nosound, 640x480x16) = 16.6 fps
17Intel POD-100 - GLQuake (nosound, 640x480x16) = 36.0 fps (vs. 31.3 fps with v29.42)
18Intel POD-100 - QuakeII (nosound, 640x480x16) = 18.6 fps (vs. 18.0 fps with v29.42)
19
20
21NVIDIA drivers do not work with IBM/Cyrix 5x86 CPUs in Win95c and NT4.
22
23When switching pages in the BIOS, the redraw rate is slow.

smeezekitty wrote:

You're kidding right?
http://s28.postimg.org/4jyyr0ahp/process_list.jpg

I've only got 44 processes going at once on a fairly fresh install of Win7, and I haven't even gotten down to disabling services yet! 🤣

http://postimg.org/image/mjn0yclcn/

feipoa wrote:

Scali wrote:

I wonder how well a PCI-based GeForce card would work in a 486 actually.

A GeForce 2 MX400 worked well in my tests using an Am5x86-160 on a UMC8881-based Biostar MB-8433UUD. Using graphics cards which are more modern than the socket 3 generation of hardware is well documented in this thread, Modern graphics on a 486

That's interesting, thanks!
It does not quite show what I was expecting though, since there are some non-T&L cards that deliver similar performance in GLQuake.
I suppose that could be explained by the fact that GLQuake's engine is not designed to take full advantage of a T&L-based GPU. It only sends very small batches of polygons to the GPU at a time, so the GPU can't really flex its T&L muscles.
It would be interesting to do some synthetic tests, such as the ones in early 3DMarks, where they just send large batches of polygons to the GPU:
http://youtu.be/M2qoFiXxCck?t=5m9s
The CPU-overhead for that test is near 0, so it may be possible that a GeForce card performs about the same in a 486 as it does in a faster system, with such tests.

RacoonRider wrote:

(as you remember, the original Socket 4 Pentium featured 2 FPUs).

Hum, not sure what you mean here.
Are you confusing FPUs with pipelines? All Pentiums have two integer pipelines (U and V pipe) and one FPU pipeline. There is no optional FPU socket or chip for any Pentium as far as I know.

Scali wrote:

RacoonRider wrote:

(as you remember, the original Socket 4 Pentium featured 2 FPUs).

Hum, not sure what you mean here.
Are you confusing FPUs with pipelines? All Pentiums have two integer pipelines (U and V pipe) and one FPU pipeline. There is no optional FPU socket or chip for any Pentium as far as I know.

I meant internal FPU, I know there is no 587 socket... Anyway, I revised my sources and it turns out, I did not remember it correctly. Of course, I read about integer pipelines, but for some reason I thought it was about floating point pipelines, which correlated with Pentiums's exceptionally strong FPU very well. My apologies.