I've always been told PCI implementation on Socket 3 and Socket 4 was buggy and not to be trusted. Specially, I should avoid it at all costs. It is my understanding the proper entry point for PCI is Socket 5 and later.
Is this bro knowledge or factual?
Does anyone have any comment on this?
I've also heard such legends, but nothing specific...
Probably the most important limitation of early PCI is the lack of MSI interrupts, a feature required by eg. many SATA controllers...
but it got introduced with PCI 2.2 (1998), so even Socket 7 and Slot 1 boards are affected.
Nie tylko, jak widzicie, w tym trudność, że nie zdołacie wejść na moją górę, lecz i w tym, że ja do was cały zejść nie mogę, gdyż schodząc, gubię po drodze to, co miałem donieść.
It should work fine assuming you have a compatible PCI card, early PCI IIRC was limited in what voltages it could deliver (3.3v only) and had the slot reversed to later PCI 2.3 which could deliver higher voltages (3.3v and 5v).
you can read more about it here https://en.wikipedia.org/wiki/Peripheral_Comp … nt_Interconnect naturally the very early versions of PCI had their problems but so long as you have a 3.3v 33Mhz compatible card it should work with little issue.
The post above adds another issue, but since you are talking about 386/486 class board here I doubt you will even find a 3.3v compatible sata card that doesn't also require 5v so I doubt this issue is a problem for you.
The original PCI is 5 V only.
The 3.3 V option was added later, and 32-bit 3.3 V slots are rare.
64-bit PCI-X slots are 3.3 V.
And there's plenty of 5 V compatible SATA cards that can be installed even in a 486, but won't work stable due to the lack of MSI.
Nie tylko, jak widzicie, w tym trudność, że nie zdołacie wejść na moją górę, lecz i w tym, że ja do was cały zejść nie mogę, gdyż schodząc, gubię po drodze to, co miałem donieść.
Quick rule of thumb for 486 boards--
pico1180 wrote on 2023-03-21, 00:49:I've always been told PCI implementation on Socket 3 and Socket 4 was buggy and not to be trusted. Specially, I should avoid it at all costs. It is my understanding the proper entry point for PCI is Socket 5 and later.
Is this bro knowledge or factual?
Does anyone have any comment on this?
Late 486 UMC and SiS chipsets have very good pci implementation. Good performance and compatibility.
Its not buggy, but its non-PNP.
Pros are compatibility and possibility to configure all PCI options AS YOU want - its a good for OLD IDE cache (or non-cache) controllers.
Cons - it need to understand what you want and what you do)))
Need help? Begin with photo and model of your hardware 😉
Depends on chipset type and actual PCI BIOS implementation. Stuff like IRQ steering and busmastering tends to be broken or missing in early PCI implementations.
I have had good experiences with later chipsets like aforementioned SiS 496/97 and UMC 8881/6.
OTOH, there were a lot of compatibility issues with PCI graphics cards on a board with ALi 1489 FinALi chipset.
Early Intel chipsets (Saturn, Aries) had problems due to just being ancient (I think bus mastering didn't work properly on them sometimes).
Mercury/Neptune Pentium stuff should be mostly fine (never checked properly).
Not sure about VIA chipsets (Pluto, Apollo Master).
...and then there are those boards with an OPTi VL-to-PCI bridge (82C822). These dreadful things are just plain garbage...
pyrogx wrote on 2023-03-21, 09:15:Depends on chipset type and actual PCI BIOS implementation. Stuff like IRQ steering and busmastering tends to be broken or missi […]
Depends on chipset type and actual PCI BIOS implementation. Stuff like IRQ steering and busmastering tends to be broken or missing in early PCI implementations.
I have had good experiences with later chipsets like aforementioned SiS 496/97 and UMC 8881/6.
OTOH, there were a lot of compatibility issues with PCI graphics cards on a board with ALi 1489 FinALi chipset.Early Intel chipsets (Saturn, Aries) had problems due to just being ancient (I think bus mastering didn't work properly on them sometimes).
Mercury/Neptune Pentium stuff should be mostly fine (never checked properly).Not sure about VIA chipsets (Pluto, Apollo Master).
...and then there are those boards with an OPTi VL-to-PCI bridge (82C822). These dreadful things are just plain garbage...
I have two early 486 motherboards with PCI:
https://theretroweb.com/motherboards/s/fic-486-vip (OPTi VL-to-PCI bridge 82C822 PCI 1.0 ?).
https://theretroweb.com/motherboards/s/anigma-bat4ip3e (420EX Aries PCI 2.0 & 3.3v support)
I checked to see which PCI cards would work on those boards a couple years ago.
These video cards worked fine in both:
With these, the computer booted but produced no VGA signal:
I probably need to test with some 1997 boards.
The FIC board with the PCI bridge didn't perform as well with PCI cards. Seemed like the right buy back in 1994 though. We don't always get every purchase right in the moment.
I was not able to get any PCI IDE controller to work on either board. Tried a bunch of different cards, promise, silicon image, and via. Sometimes the Aries motherboard could make it through the drive detection with the promise controllers, but it would lock before post completed. All of the IDE controllers I tried were 1996, ATA-4 and newer. Hard to find ATA-2 PCI IDE controllers from 1994.
I recently was lucky enough to get a UMC 8881/6 board. I'll have to run the compatibility gauntlet on it.
