In my ideal world, the Windows user interface would have stopped with 95 and all the improvements would have been under the hood. Also, web standards would have remained fully backwards-compatible, with a forward-thinking way in old browsers to ignore newer code.

But the world is not ideal 😠

keenmaster486 wrote:

In my ideal world, the Windows user interface would have stopped with 95 and all the improvements would have been under the hood. Also, web standards would have remained fully backwards-compatible, with a forward-thinking way in old browsers to ignore newer code.

But the world is not ideal 😠

↑ ↑ ↑ Right!!!

I would love for Microsoft to decide that XP and Win 7 should get modern kernel updates. You get to keep the look n' feel you like, and no need to play hide-and-seek with control panel icons.

When I think about VLB, I'm a little amazed that it works at all. It's a giant snake of parallel traces from the CPU, to the chipset, RAM, and slots. Several different points of electrical load, and a bunch of short stubs. Running at 33MHz. The buffered point-to-point schemes we use now are so much more sensible.

Even worse was Slot 1 / A. How did anyone think THAT was the way forward? Wow. Talk about built-in performance caps.

SirNickity wrote:

I would love for Microsoft to decide that XP and Win 7 should get modern kernel updates. You get to keep the look n' feel you like, and no need to play hide-and-seek with control panel icons.

When I think about VLB, I'm a little amazed that it works at all. It's a giant snake of parallel traces from the CPU, to the chipset, RAM, and slots. Several different points of electrical load, and a bunch of short stubs. Running at 33MHz. The buffered point-to-point schemes we use now are so much more sensible.

Even worse was Slot 1 / A. How did anyone think THAT was the way forward? Wow. Talk about built-in performance caps.

😲 your knowledge of that stuff is too extensive for me to appreciate your post fully, but I agree with the XP and 7 thing you mentioned.... maybe how about applying the same thing to Win x Workgroup 3.11 ? 🤣

Unfortunately I've missed the 486' heyday, and my family switched from 286's straight to Pentiums, but this -

keenmaster486 wrote:

In my ideal world, the Windows user interface would have stopped with 95 and all the improvements would have been under the hood. Also, web standards would have remained fully backwards-compatible, with a forward-thinking way in old browsers to ignore newer code.

But the world is not ideal 😠

I second so much!..

aries-mu wrote:

applying the same thing to Win x Workgroup 3.11

And also this 😀

SirNickity wrote:

Even worse was Slot 1 / A. How did anyone think THAT was the way forward? Wow. Talk about built-in performance caps.

I keep finding mentions that the cartridge slot design was due to the yeld/cost problems of on-die L2 cache at that time. So maybe mating a processor to a COAST module was never meant to be a future-proof decision to begin with, and rather just a temporary workaround 😀

Revolter wrote:

Unfortunately I've missed the 486' heyday, and my family switched from 286's straight to Pentiums, but this - […]
Show full quote

Unfortunately I've missed the 486' heyday, and my family switched from 286's straight to Pentiums, but this -

keenmaster486 wrote:

In my ideal world, the Windows user interface would have stopped with 95 and all the improvements would have been under the hood. Also, web standards would have remained fully backwards-compatible, with a forward-thinking way in old browsers to ignore newer code.

But the world is not ideal 😠

I second so much!..

aries-mu wrote:

applying the same thing to Win x Workgroup 3.11

And also this 😀

SirNickity wrote:

Even worse was Slot 1 / A. How did anyone think THAT was the way forward? Wow. Talk about built-in performance caps.

I keep finding mentions that the cartridge slot design was due to the yeld/cost problems of on-die L2 cache at that time. So maybe mating a processor to a COAST module was never meant to be a future-proof decision to begin with, and rather just a temporary workaround 😀

Exactly that. It was a compromise between cost, performance, and the available technology of the time. Intel already had the Pentium Pro with full speed L2 cache on a socket package 2 years before the first Slot 1 Pentium II launched. They knew that was the end goal, but in 1995-1999 the cost of doing am independent L2 cache die or attempting a monolithic 350 nm CPU die with L2 cache would have been too high for consumer systems. That's why the Pro mostly stayed in high end workstations and servers. Slot 1 got the L2 cache from bus speed up to 1/2 the CPU clock. In 1998 the Celeron got 128 KB of full speed on-die L2 on a 250 nm process, and intel rolled out Socket 370 for it quickly thereafter. The Celeron was competitive with the Pentium II and III at equal clockspeed and bus speed, except for a few applications that benefited more from the slower, but larger 512 KB cache of the Pentiums. It was pretty obvious by 1998/99 that the slot form factor would be replaced by sockets in 2000. I personally love the Slot cpu era for this very reason. It was an interesting time with a lot of visible changes in CPU design. Things got pretty boring after Socket 478/754/939.

very interesting discussion, thanks guys

Well not 486 exclusively but here’s what I miss:

1) multiple manufacturers competing on the same socket
2) overdrive cpus - even if there are compelling reasons to change chipset, I loved that new tech was retrofitted into existing platforms. Plus chipset changes nowadays are just frivolous now.

It just seemed like a more exciting era of computing.

Socket 370 should have never existed, celeron should have been socket 8 and the socket improved over time like super 7.

Also 486 class should have gotten the Pentium fpu and bus speed increases.

BinaryDemon wrote:

Well not 486 exclusively but here’s what I miss: […]
Show full quote

Well not 486 exclusively but here’s what I miss:

1) multiple manufacturers competing on the same socket
2) overdrive cpus - even if there are compelling reasons to change chipset, I loved that new tech was retrofitted into existing platforms. Plus chipset changes nowadays are just frivolous now.

It just seemed like a more exciting era of computing.

Right! Overdrives were over-exciting! An attractive idea of past-future blend as much as possible!

rmay635703 wrote:

Socket 370 should have never existed, celeron should have been socket 8 and the socket improved over time like super 7.

Also 486 class should have gotten the Pentium fpu and bus speed increases.

I like the idea of 486s with higher clock and pentiun's FPU!

AT/XT form factor Motherboards and cases should still be a thing. This includes Baby-AT. I like the idea of connecting individual cards so when my USB ports die or my SATA controller goes out, I could just yank it and replace, instead of replace the whole friggin motherboard.

The 486 class to me would still be a viable workstation as an "old world" type system - ie, eventually become it's own genre with all the speed increases. Lately I've been using my 486 more and more for modern tasks save for financial transactions and serious stuff and it's becoming readlily apparent both that a lot of the things we have today are not "too much" for the old systems nor do they really "need" all that additional bloat that's being used. Even my 286 running Arachne does fairly well considering it's age. And a 486 running Seamonkey or Dillo is only a few Javascript errors shy of being "current". I even got an older version of Firefox working in Windows 2000 on the 486 - like circa 2011 or something, and it worked, it was slow, it ate up almost all remaining of the 64MB of RAM, but it worked. Actually, one of the blogs that bugged the heck out of me concerning retro O/S is one where a guy tried to spend a week working in Windows 3.1 - the whole time I wanted to go "uh, dude, you need Windows For Workgroups 3.11, and you need to use something OTHER than what of the 8MB install there is to do it".

Northgate or Avant would have kept making keyboards and made a 101 key USB variant of the OmniKey 102.

IBM Would have released a 30th Anniversary IBM PC limited edition - not an actual 8088, but a Core i7 gamer box that looks like a true blue IBM PC XT - complete with a special edition Windows 10 with a DOS command-line driven console option, or FreeDOS as a pack-in O/S.

aries-mu wrote:

(remember that, besides the clock frequency, Intel 486s were slightly faster than AMDs):

No they were not. Cyrix yes, but AMD no. UMC was even faster clock for clock.

Socket 3 was fairly long lived but it would have been cool to try squeezing even more out of it. How about a Pentium MMX Overdrive 150/160/166? Or a CPU with MTRR/write gathering? Or how about a 486 board with SDRAM? The Sega Saturn used SDRAM in '94 so it was out there at the time.

Back in the day I heard there were a few socket 5 boards with VLB slots and I wanted to get one and try running a 6x86MX to see how fast it could push a VLB video card with write gathering enabled (it made a big difference on PCI cards, at least in benchmarks). I never did get one but later I read that the VLB slots on those boards were slow to begin with.

VLB slots were kind of a disaster though. 16-bit ISA cards can already be a bit of a pain to insert/remove, so making them even longer is regrettable. EISA was also a bit strange with its extra rows of contacts at a different depth. Maybe they could have used a completely different slot instead, as with PCI, or tried to extend the performance of ISA slots by adding an option for higher clock rate or a burst mode.

Socket A is another one that could have seen some more love. Where's a socket A chip with 1MB L2 and SSE2??? 😉

aries-mu wrote:

In the world how I would have wanted it, here's how 486 CPUs would have evolved: […]
Show full quote

In the world how I would have wanted it, here's how 486 CPUs would have evolved:

Intel would have released also the following (remember that, besides the clock frequency, Intel 486s were slightly faster than AMDs):

486 DX 40
486 DX-2 80
486 DX-3 120 (yes, in my ideal world the DX-3 naming wouldn't have been skipped)

With 16 KB L1 cache, write back.

And, of course, VLB and PCI slots perfectly optimized to run fully at 40 MHz with zero wait states, and so the VLB and PCI cards.

Except from the 16kb wb cache, I am running two VLB cards at 0ws on my DX2-80. At least that is how the cards are configured. And it is stable.

creepingnet wrote:

AT/XT form factor Motherboards and cases should still be a thing. This includes Baby-AT. I like the idea of connecting individual cards so when my USB ports die or my SATA controller goes out, I could just yank it and replace, instead of replace the whole friggin motherboard.

The 486 class to me would still be a viable...

I agree!

appiah4 wrote:

aries-mu wrote:

(remember that, besides the clock frequency, Intel 486s were slightly faster than AMDs):

No they were not. Cyrix yes, but AMD no. UMC was even faster clock for clock.

You sure? Because I must have seen a lot of tests and benchmarks published by our fellows here showing that Intel 486s were faster than AMD ones, for example even overclocked DX4-100 to 120 vs Am486 DX4-120.... or something... And I remember somebody noticing in some of those reviews that Intel's CPUs were faster... something like that

bakemono wrote:

Socket 3 was fairly long lived but it would have been cool to try squeezing even more out of it. How about a Pentium MMX Overdrive 150/160/166?

Nice!!!!!

bakemono wrote:

Or a CPU with MTRR/write gathering?

The heck is that???

bakemono wrote:

Or how about a 486 board with SDRAM? The Sega Saturn used SDRAM in '94 so it was out there at the time.

Very interesting, I didn't know that! How about a Socket 6 AND Socket 5 or 7 motherboard, VX chipset, DDR SDRAM, pipeline burst L2 cache, UATA 133, able to host both Pentiums and 486s (and related overdrives 😉 )?

bakemono wrote:

Back in the day I heard there were a few socket 5 boards with VLB slots and I wanted to get one and try running a 6x86MX to see how fast it could push a VLB video card with write gathering enabled (it made a big difference on PCI cards, at least in benchmarks). I never did get one but later I read that the VLB slots on those boards were slow to begin with.

Sounds interesting!

bakemono wrote:

VLB slots were kind of a disaster though. 16-bit ISA cards can already be a bit of a pain to insert/remove, so making them even longer is regrettable. EISA was also a bit strange with its extra rows of contacts at a different depth. Maybe they could have used a completely different slot instead, as with PCI, or tried to extend the performance of ISA slots by adding an option for higher clock rate or a burst mode.

The beauty of that EISA system is that EISA slots were backwards compatible with ISA cards. A different slot would have lost this.
But I agree EISA could have been carried on way longer, simply increasing the clock to 33 MHz, or even to 66 MHz later when technology allowed, so you would still have had ISA-compatible slots, and PCI-like performance. Can you imagine a 32 bit 50 or 66 MHz bus-mstering EISA slot and card?

bakemono wrote:

Socket A is another one that could have seen some more love. Where's a socket A chip with 1MB L2 and SSE2??? 😉

Too modern to spike my interests 🤣

brostenen wrote:

Except from the 16kb wb cache, I am running two VLB cards at 0ws on my DX2-80. At least that is how the cards are configured. And it is stable.

WOOW MAN!!!! I'd love to know EXACTLY:
• What motherboard are you using?
• Would THAT motherboard accept an Enhanced Am486 DX-4 120 WB?
• WHAT VLB CARDS are running at 40 MHz 0WS successfully??????

As for the 16 KB WB cache, you can always rent a clean room, order some silicon, open up the CPU and try to DIY , piece of cake!