SIS 496 don't have integrated IDE. It's provided via Winbond/UMC I/O chip. Which is, presumably, connected to ISA bus.

I think it does include IDE controller. From the datasheet:

The SiS85C496 PCI & CPU Memory Controller (PCM) integrates the Host Bridge (Host Interface), the cache and main memory DRAM Controller, the PCI Bridge, the built-in IDE Controller, and the FS-Link Bus (Fast Slow Link Bus).

Surprisingly it is connected via local bus (even on PCI-only boards). Presumably as PCI IDE Controller spec came quite late for these early PCI chipsets.

ph4nt0m wrote on 2020-01-19, 15:07:

3Dfx Voodoo Banshee works fine on the 66MHz PCI bus just like Voodoo 3, and there is no practical performance difference between these on 486. Another challenge is to get a 66MHz capable IDE or SCSI controller because the integrated one in SiS 496 is slow and somewhat buggy.

I think Promise Ultra100 adapters are PCI 2.1 compliant, so should be 66 MHz capable. Will the Voodoo2 run at 66 MHz? What about Matrox G200? If so, I could try disabling my 1/2 FSB-to-PCI multiplier.

Will the Voodoo2 run at 66 MHz?

Potentially yes. There was Obsidian2 AGP card, which seemingly works without any bridge chip of sorts.

feipoa wrote on 2020-01-19, 21:26:

I think Promise Ultra100 adapters are PCI 2.1 compliant, so should be 66 MHz capable. Will the Voodoo2 run at 66 MHz? What about Matrox G200? If so, I could try disabling my 1/2 FSB-to-PCI multiplier.

Quick OT: Does PCI revision > 2.0 dictate 66MHz capability or does it merely specify how 66MHz are to be integrated?
Is every PCI device > rev.2.0 66MHz capable or are most of them 33MHz devices but the same specification allows for 66MHz on devices that support it?
Thanks!

Wasn't the 66MHz operation introduced with PCI 2.2?

No it was in 2.1

download/file.php?mode=view&id=74915

Doornkaat wrote on 2020-01-20, 08:05:

Quick OT: Does PCI revision > 2.0 dictate 66MHz capability or does it merely specify how 66MHz are to be integrated?
Is every PCI device > rev.2.0 66MHz capable or are most of them 33MHz devices but the same specification allows for 66MHz on devices that support it?
Thanks!

Anyone?

66MHz is an optional feature that can be enabled if all the devices on the bus can handle it, including chipset itself. There's a pin on the slot that tells if something cannot do 66MHz, if that pin is active then 66MHz won't be enabled on the bus.

Thanks, Tiido! :-)

What socket 3 , VLB motherboard and chipset do you recommend ?

I am shopping for a better motherboard and I want a good fast reliable one for an Intel 486dx4-100 CPU.

Probably PCCHIPS M912 Rev 1.7, the one without fake cache soldered on.

The Serpent Rider wrote on 2020-01-22, 23:49:

Probably PCCHIPS M912 Rev 1.7, the one without fake cache soldered on.

VLB, Not PCI.

How is this motherboard ?
Is this “Symphony” chipset any good ?
I want to use it with an intel 486dx4-100 overdrive CPU.

mpe wrote on 2020-01-19, 15:52:

I think it does include IDE controller. From the datasheet:

The SiS85C496 PCI & CPU Memory Controller (PCM) integrates the Host Bridge (Host Interface), the cache and main memory DRAM Controller, the PCI Bridge, the built-in IDE Controller, and the FS-Link Bus (Fast Slow Link Bus).

Surprisingly it is connected via local bus (even on PCI-only boards). Presumably as PCI IDE Controller spec came quite late for these early PCI chipsets.

This makes sense, local bus on a 486 is simply a direct connection to the CPU, rather than through some bus bridge. Making the IDE controller on this old chipset PCI based would reduce performance, increase complexity and make compatibility with ISA based IDE more difficult.

VLB, Not PCI.

M912

Is this “Symphony” chipset any good ?

Nothing spectacular. Also SIMM30, which is a pain.

PCChips M912 aka Cheetah has a very thin PCB. Especially the last v1.7. The capacitors are no good either. The good thing is that it can handle 1Mb of cache and 40MHz FSB.

ECS UM4980 is about the same with a much thicker PCB, but it doesn't come with a voltage regulator. Either you DIY or just use a 5V capable CPU including overdrives. Also no RTC battery, so you have to attach an external one.

ASUS VL/I-486SV2GX4 should be nice too.

So I've been doing some comparisons with my Iwill 430FX board with Voodoo 3 under Windows 98 SE. And apparently Pentium 75 with async cache is more or less equal to my Am5x86 180Mhz setup. MDK2 scored only 5.16 fps, they evened out in Quake 2 with ~13.5 fps and Incoming was slightly faster (~0.4 fps) on a Pentium setup.
Disabling asynchronous L2 cache in Windows only worsened my results, but only slightly. Surprisingly, enabled DMA didn't helped much with OS responsiveness.

PCPlayer and Build engine wise, overclocked Am5x86 is slightly faster than Pentium 100 (Async L2). One can only wonder how fast 486DX could have been with pipeline burst cache.

I'm finding this difficult to follow. Are you saying that the Am5x86-180 was about the same speed as a Pentium 75 in MDK2, Quake2, and Incoming. I figured the 180 would have been faster by a respectable amount. But with PCPlayer Bench, the Am5x86-180 is slightly faster than a Pentium 100?

Are you saying that the Am5x86-180 was about the same speed as a Pentium 75 in MDK2, Quake2, and Incoming.

And it would be slower if Pentium 75 had pipeline burst cache.

But with PCPlayer Bench, the Am5x86-180 is slightly faster than a Pentium 100?

Integer only or mostly integer tests - faster than P75 Async L2. FPU heavy tests - tied of slightly slower to P75 Async L2.

EDIT:
According to this testing (160Mhz Am5x86 vs Pentum 75), I'm not very far from correct numbers: https://www.phantom.sannata.org/viewtopic.php?t=30648

So 60Mhz FSB and +20Mhz to Am5x86 are helping close the gap between this two platforms (barely).