jakethompson1 wrote on 2023-09-10, 02:12:

Even without the stick of the Designed for Windows 95 program, a "carrot" is that DOS/WfW 3.11 would not have taken advantage of any PCI features like grabbing the vendor/device ID and finding or asking for drivers, but Win95 would. So extra motivation to get PCI in place before Win95 whereas WfW 3.11 users wouldn't notice.

Has the PC industry ever actually done any forward planning of the "we are going to include thingy X that's useless today because Microsoft is going to roll out Windows Z in 6 months that runs happier with X" kind?

Generally speaking, I would say it's the other way around - the PC industry delays including thingy X as long as they can and forces Microsoft to design Windows Z to run just fine without X. Which then, if anything, causes the roll-out of X to be even more delayed because, well, Windows Z doesn't actually perform visibly better with X.

I'm sure we can think of a hundred examples - the one that comes to my mind as the most egregious is how the PC industry shipped tons of systems with integrated graphics that couldn't do Vista Aero Glass in 2006. Everybody knew the next version of Windows was going to require those GPU features, people who were on top of things (e.g. I had a Dothan laptop from summer 2005 with the optional-not-that-expensive ATI discrete graphics chip and that could do Aero Glass just fine) had video cards with those features, but if anything, Intel got Microsoft to manipulate the logo programs so that they could ship as many lousy integrated GPUs as possible. Then, of course, it was Microsoft who was left holding the bag when 2-month-old systems couldn't use one of Vista's marquee features.

(When you look at it with the benefit of hindsight and how, 15 years later, the lousy Vista reception was basically the end of innovation in the Windows desktop PC world, I hope Intel is really proud of themselves. They got to sell a year's worth of i915Gs and now, well, I hope they enjoy selling N4500s and seeing people keep their C2D/C2Qs/sandy bridges 10+ years. And when they realize that Chrome is the new OS for anything designed in the last 10-12 years and that Chrome doesn't exactly require an x64 Intel chip to run... and in fact can be very happy on TSMC-made ARM chips... they should remember that this is all the legacy of them torpedoing Vista and creating a culture where no new software can require Windows version X until Microsoft stops supporting Windows version X.)

I'm not the best judge of that stuff as, when I first have to use a Windows system, the first thing I do is revert as many of the UI changes since Win95 as possible. So the more terribly specced the typical Windows system is, the better for me, because the faster it operates when the new junk is turned off on a fast system 😁

I found the "PC 95" standards requied for Windows 95 logo program from the MSDN library. I'm surprised how lenient they are:

Plug and Play 1.0a-compliant BIOS 386 or higher processor 4MB RAM Flat-framebuffer display with 640x480 @ 8 bpp or better Dedica […]
Show full quote

Plug and Play 1.0a-compliant BIOS
386 or higher processor
4MB RAM
Flat-framebuffer display with 640x480 @ 8 bpp or better
Dedicated mouse port or integrated pointing device
One serial port
One parallel port

Has anyone ever seen a 386 system with a PnP BIOS? Maybe a 486SLC?
In practice, I think the PnP BIOS requirement rules out 386es and ISA/VLB 486 systems unless they received a BIOS upgrade and a way of flashing themselves for ESCD.

JustJulião wrote on 2023-09-10, 02:01:

I'm a bit concerned about the I/O card, I suppose I'll have to add its drivers on one of the DOS install floppies. The other thing about I/O is that I plan to overclock to 40MHz, and I know that PCI IDE controllers don't like overclocked buses, I expect even more issues on a VLB I/O controller. I'll have to choose it carefully.

I think it is opposite. Most main boards were designed to run the FSB from 25 to 50 MHz and also the VLB bus. You might have had to add wait states, and not everything was always stable at those higher clocks, but I am running a factory 486 that was sold at 33 MHz bus at 40 MHz with everything as fast as the BIOS allows me to set. So that is an n=1 example that you might get away with faster bus speeds. But you must be willing to gamble a bit with those VLB systems 😉

jakethompson1 wrote on 2023-09-10, 02:58:

I found the "PC 95" standards requied for Windows 95 logo program from the MSDN library. I'm surprised how lenient they are: […]
Show full quote

I found the "PC 95" standards requied for Windows 95 logo program from the MSDN library. I'm surprised how lenient they are:

Plug and Play 1.0a-compliant BIOS 386 or higher processor 4MB RAM Flat-framebuffer display with 640x480 @ 8 bpp or better Dedica […]
Show full quote

Plug and Play 1.0a-compliant BIOS
386 or higher processor
4MB RAM
Flat-framebuffer display with 640x480 @ 8 bpp or better
Dedicated mouse port or integrated pointing device
One serial port
One parallel port

Has anyone ever seen a 386 system with a PnP BIOS? Maybe a 486SLC?
In practice, I think the PnP BIOS requirement rules out 386es and ISA/VLB 486 systems unless they received a BIOS upgrade and a way of flashing themselves for ESCD.

It's basically the minimum requirements for Windows 95 if you ignore the 'must be fully PNP bit'. When these reqs rolled out, and 95 was finalized and out, it wouldn't have been a hard sell really to make all the PC makers that wanted that sticker on their box basically burn into the BIOS some program code that was dubious really if it ever made that much difference. I don't think there would be that much difference, especially to Win95 that really didn't care and had to just work, from a system that did not have a PNP BIOS, but had working PNP cards, vs a system with a PNP BIOS and can do whatever its magic was automatically with whatever cards supposedly could allow that. So yeah, not a large hurdle in 1995 to meet that. It was likely more like a marketing gimmick to get people to buy that Windows 95 royalty paid machine. Designed for Windows 95 AMD chip, anyone? Surely that means AMD is better than Intel, right? That AMD 5x86 socket 3, 486 class chip -- I wonder if that would be installed in a fully compliant PNP BIOS motherboard...

the3dfxdude wrote on 2023-09-10, 14:05:

Designed for Windows 95 AMD chip, anyone? Surely that means AMD is better than Intel, right? That AMD 5x86 socket 3, 486 class chip -- I wonder if that would be installed in a fully compliant PNP BIOS motherboard...

I always figured AMD got the Designed for Windows 95 logo so as to help counteract any FUD about compatibility.
At least for the MB-84333UUD system I used, it was a fully compliant board 😁

JustJulião wrote on 2023-09-10, 02:01:

For DOS itself, I'll use tools and packs of tools that will make my life easier, like Phil's starter pack or "REMOVED" on REMOVED that looks promising. I'll test FreeDOS too at some point.
I'll go deeper afterwards but I want to get to know DOS by the easier end.

If you want to indeed dive deeper with DOS, you will find out how much you can actually customize how it works and looks like, mainly using some command.com replacements and ANSI.SYS. You can really improve the visuals from the usability stand point of view with different colors or go totally bonkers with everything. Me and couple of my friends almost competed with different things we could do back in the day to DOS shell. And when I got back to retrocomputing and DOS, I have always customized DOS running on every machine, albeit I nowadays use pretty much same configuration with everything instead of fooling around with it. Vanilla DOS just feels unnecessarily clumsy to me, but if you tweak it a bit you can get some same QoL features for it that many modern-ish *NIX shells have.

I learned my ropes with PCs using DOS and boy the transition to Windows95/98 took time and sucked. I hated how those newer OSs hid and spread the relevant information from user to registry and different configuration menus. And you still had the plain old DOS under everything so you needed to handle both. With DOS, configuration could be tweaked easily with text editors and troubleshooting was, and is, generally much easier. And DOS is almost impossible to get to a point where reinstalling is needed unless you really want to break it. It is the epitomy of what you see is what you get without the complexity of *NIX OSs.

I guess I still like DOS for same reason. The raw simplicity and limitations of DOS also mean highly robust system, at least to me. Every time I use DOS, it feels like returning to home in the context of computing.

There were only few VLB Pentium systems out there, but most of them suffered from poor bus performance.
Like the TMC board Iown.
They might be as rare as VLB networking boards. Yes, even these really exist.
Bus speed was higly dependant from the number of VLB slots and the used boards.
While most VGA boards work flawlessly with 40 or even 50 MHz many controllers boards caused data corruption on higher frequencies than 33 MHz.
Early PCI cards were as big or even bigger than VLB. I own a 1 MB S3 graphic board similar to the Miro S8 PCI which is almost full length.
Chipsets were probably the most costly part of early PCI boards, but at this time there were many systems sold with EISA VLB as hight performance solutions.
Early PCI systems suffered from the buggy IDE chips mentioned allready and therefore wererarely used for high performance systems.

VivienM wrote on 2023-09-10, 02:45:

Has the PC industry ever actually done any forward planning of the "we are going to include thingy X that's useless today because Microsoft is going to roll out Windows Z in 6 months that runs happier with X" kind?
....
Generally speaking, I would say it's the other way around - the PC industry delays including thingy X as long as they can and forces Microsoft to design Windows Z to run just fine without X. Which then, if anything, causes the roll-out of X to be even more delayed because, well, Windows Z doesn't actually perform visibly better with X.

The installed user base is both a blessing and curse for Wintel.

These look like cases where Microsoft and Intel successfully encouraged the PC world to navigate to a new standard, even if it took a little while in some cases:
1) PCI and along with it USB
2) PCI-E (Replaced AGP pretty quickly)
3) UEFI and Trusted Platform Module (TPM) version 2.0.

Things got easier once the PC world got device drivers for everything instead of BIOS roms.

douglar wrote on 2023-09-12, 11:39:

These look like cases where Microsoft and Intel successfully encouraged the PC world to navigate to a new standard, even if it t […]
Show full quote

These look like cases where Microsoft and Intel successfully encouraged the PC world to navigate to a new standard, even if it took a little while in some cases:
1) PCI and along with it USB
2) PCI-E (Replaced AGP pretty quickly)
3) UEFI and Trusted Platform Module (TPM) version 2.0.

1) PCI, sure, but it took, oh, 6-8 years or so for the last ISA slot to fall off motherboards? (Compare that with an Apple-style transition - none of Apple's PCI machines had any NuBus slots. Zero. Hope you didn't have a NuBus card you were attached to.)
USB, well, sure, USB was added, but it's worth noting, PS/2 ports remain on some motherboards to this day. I'm not sure whether parallel/serial ports are completely gone... I feel like I haven't seen any on any newer things in a long time, but that doesn't mean that some random things don't still have them. Worst Buy still sells desktops with DVI ports.

And lots of brand new machines still don't have USB-C, the latest USB speeds, etc.

2) That one I will have to give you. While everybody made both AGP and PCI-E products for a few years, somehow, the PCI-E side of things managed to get established despite having zero backwards compatibility with AGP.

3) UEFI was largely invisible to users, though, with enough compatibility mechanisms. As for TPM 2.0, well, how many of those TPM 2.0s out there are the built-in processor ones? Easy enough to drive adoption of something if you throw it into something they're already buying.

There's another interesting transition which is PATA to SATA, especially for optical drives. I remember when Intel took the PATA out of the ICH... 8 or 9 or whichever one it was that came with the first C2D 965 chipsets. Enthusiast computer stores at the time did not offer a single SATA optical drive. The Taiwanese motherboard manufacturers all stuck at least one third-party PATA controller on their motherboards for at least 2-3 generations as a result. But somehow, that gave the ODD industry the hint to start producing and selling to non-large-OEMs SATA products.