VOGONS


Reply 22 of 29, by appiah4

User metadata
Rank l33t++
Rank
l33t++

I remember PCI becoming the BIG THING in late 1995. I moved on to PCI in late 1996 when I upgraded from a DX4-100 to a Pentium 133 thanks to being exposed to Quake by a friend who had a Pentium 90 that danced circles around my PC. My first PCI peripheral was an S3 Trio64 graphics card.

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 23 of 29, by 386SX

User metadata
Rank Oldbie
Rank
Oldbie
imi wrote on 2020-04-05, 00:26:

my first PCI board was an Asus Socket 7 with a Pentium 90 in 1996 if I recall correctly, a much needed upgrade from our previous 386DX40, skipped 486 altogether back then.

My first PCI board was a Socket 7 ALi chipset based with a K62-350 coming from the previous 386DX-20. I skipped every config in the middle of them but back then I couldn't ask for a new computer; I really pushed the 386SX up to its limits and the jump to the K6-2 350 was impressive (but soon not that much in 3D games considering it came with a S3 Trio3D..). So I could say I did jump from ISA to AGP. 😁
Back then a new P1 PCI based computer was not much cheaper of buying a used car.

Reply 24 of 29, by vetz

User metadata
Rank l33t
Rank
l33t

Didn't really know what PCI was back in those days, it wasn't untill the Voodoo started taking off that I cared about hardware. So for me, first time experiencing PCI was when I installed my Orchid Righteous card during christmas of 97 in a Compaq Presario 166MMX. Before that the only Pentium machines I had experienced in the flesh was my dads laptop and desktops in stores for demo purposes.

3D Accelerated Games List (Proprietary APIs - No 3DFX/Direct3D)
3D Acceleration Comparison Episodes

Reply 25 of 29, by imi

User metadata
Rank Oldbie
Rank
Oldbie
386SX wrote on 2020-04-07, 11:21:
imi wrote on 2020-04-05, 00:26:

my first PCI board was an Asus Socket 7 with a Pentium 90 in 1996 if I recall correctly, a much needed upgrade from our previous 386DX40, skipped 486 altogether back then.

My first PCI board was a Socket 7 ALi chipset based with a K62-350 coming from the previous 386DX-20. I skipped every config in the middle of them but back then I couldn't ask for a new computer; I really pushed the 386SX up to its limits and the jump to the K6-2 350 was impressive (but soon not that much in 3D games considering it came with a S3 Trio3D..). So I could say I did jump from ISA to AGP. 😁
Back then a new P1 PCI based computer was not much cheaper of buying a used car.

heh

I remember when I bought Duke Nukem 3D and it wouldn't run on my 386, had to go over to a friends house with a 486 to play it ^^
also command&conquer was almost unbearably slow... I did play descent on the 386 though and enjoyed it very much, it took up almost all of our HDD back then ^^

getting that pentium sure was a relief.

also I thought this might be relevant to the thread 😀

showtime.jpg
Filename
showtime.jpg
File size
422.74 KiB
Views
167 views
File license
CC-BY-4.0

Reply 26 of 29, by NJRoadfan

User metadata
Rank Oldbie
Rank
Oldbie

PCI was adopted widely because it was cheaper to implement than VLB and still had a backwards compatible option of using ISA and EISA slots by using a bridge chip. Being a clean sheet platform independent "mezzanine" bus not hampered by a decade of kludges helped too as eventually Apple, DEC, and Sun adopted it for their machines as well. It managed to quickly kill off VLB, MCA, and NuBus (also a mezzanine bus, but expensive). EISA hung around until the 440FX, mostly in servers.

ISA never really went away on PCs, but that's due to architectural reasons. The only kludge that held PCI back on PCs was IRQs, simply because it took so damned long to support IRQ steering (sharing) and later APIC.

Reply 27 of 29, by mpe

User metadata
Rank Oldbie
Rank
Oldbie
kjliew wrote on 2020-04-06, 01:36:

SiS was a more submissive chipsets vendor compared to UMC/OPTi, but SiS501 was indeed a "legacy approach" with 3-chip design and had both the Cache/Memory controller and PCI LocalBus Data controller hooking on the FSB compared to Intel's single PCI host bridge designs. The i430FX Triton PCIsets marked an unprecedented milestone, a high-integration 2-chip PCIsets and price that reached mainstream, together with cheaper and higher volume 2nd generation Pentiums, thanks to manufacturing process improvements.

I've been thinking about this. But is the integration really such a big improvement of the 430FX. Why do you call it 2-chip?

Technically 430FX is a 4-chip solution as you need 2437FX System Controller (TSC)􏰂, two 82438FX Data Paths (TDP)􏰂 and the 82371FB PCI ISA/IDE Xcelerator (PIIX)􏰃 . It took another iteration step to scale it down (Triton II).

The SiS 501 was indeed a 3-chip solution (without I/O and IDE). I don't find it more legacy than the Triton as the architecture was almost identical. They needed a single 502 chip in place of Intel's two 82438's.

UMC8891 was also a 3-chip solution. But they had built-in IDE and much less extra discreet chips around than anyone else.

Previous 430LX/NX were also 4 chip solutions (+ IDE and I/O + EISA).

Technically ALi had the most integrated first-gen Pentium chipset of the era. The original Alladin was 2-chip already in 1994. Also without I/O and at @ 430LX performance level.

Blog||486DX-50|NexGen 586|S4

Reply 28 of 29, by kjliew

User metadata
Rank Oldbie
Rank
Oldbie
mpe wrote on 2020-04-07, 22:51:

I've been thinking about this. But is the integration really such a big improvement of the 430FX. Why do you call it 2-chip?

Technically 430FX is a 4-chip solution as you need 2437FX System Controller (TSC)􏰂, two 82438FX Data Paths (TDP)􏰂 and the 82371FB PCI ISA/IDE Xcelerator (PIIX)􏰃 . It took another iteration step to scale it down (Triton II).

The SiS 501 was indeed a 3-chip solution (without I/O and IDE). I don't find it more legacy than the Triton as the architecture was almost identical. They needed a single 502 chip in place of Intel's two 82438's.

Thanks for the correction, never noticed about the 2 TDPs. But that does not compare to 502, the TDPs are interfaced between DRAM and cache. SiS501 also requires 3 DRAM address buffers. As far as core logics wise, Intel 430FX is 2-chip vs SiS501 3-chip design.

SIS501.png
Filename
SIS501.png
File size
22.41 KiB
Views
98 views
File comment
SIS501
File license
Fair use/fair dealing exception

And, SiS501 overloaded the CPU FSB with SRAM cache and a separate cache/memory controller. A very 486-like approach.

Reply 29 of 29, by mpe

User metadata
Rank Oldbie
Rank
Oldbie

Now compare that with the architecture of the Triton. Which is IMHO almost identical to the SiS 501 (and not that dissimilar to late 486 PCI chipsets for that matter):

FX.png
Filename
FX.png
File size
267.43 KiB
Views
80 views
File license
Public domain

TDP need to be considered as integral part of core logic and every Triton motherboard need two of these. These are roughly equivalent of a single SiS 502. There are extra 244 series transceivers on either 501 or FX motherboards to meet electrical needs of the particular board. Unlike the TDP these are non-core parts.

On both chipsets the Cache and DRAM control is part of the controller chip and the cache sits on the host bus (just like on a 486). Backside cache bus only came with the Pentium Pro architecture if this is what you mean by overloading of the FSB

Blog||486DX-50|NexGen 586|S4