VOGONS


First post, by Jo22

User metadata
Rank l33t++
Rank
l33t++

Hello everyone,
I was just tinkering with PCem and noticed that
it supports the POD for 486 machine types.

This makes me wonder, how well did these OverDrive chips
work with real, existing 486 mainboards/chipsets of the time ?

As far as ISA and PCI are concerned, these bus types
are more or less processor independed.

But what about VESA Local Bus, did the POD implement
all i486 signals that were required for it ?

If you're an (ex-) owner of such a setting, please tell.
I'm curious. 😀

Thanks in advance,
Jo22

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//

Reply 1 of 9, by Deksor

User metadata
Rank l33t
Rank
l33t

I don't think that created a lot of issues : there are even some socket 4/5 mobos with VLB !

Trying to identify old hardware ? Visit The retro web - Project's thread The Retro Web project - a stason.org/TH99 alternative

Reply 2 of 9, by dionb

User metadata
Rank l33t++
Rank
l33t++

Yep. The mechanism is the same as on those So4/5 chipsets - the 64b Pentium bus is split and half of it feeds VLB. In the case of the 'native' Pentium chipsets with VLB (which were usually just 486 chipsets tacked onto a Pentium bus, with 32b wide memory access and the awful performance to go with it) the clock also needed to be halved, but the PODP ran at native 486 bus speeds (33MHz x 2.5) so this extra step was unnecessary. Note that if you remove the fan, the multiplier automatically drops to 1x.

Reply 4 of 9, by Scali

User metadata
Rank l33t
Rank
l33t
Jo22 wrote:

This makes me wonder, how well did these OverDrive chips
work with real, existing 486 mainboards/chipsets of the time ?

I have a VLB board on which the POD doesn't work (I have a POD83). I don't think the VLB is the problem though (it simply doesn't boot, even with only ISA cards).
I think the problem is because my 486 predates the POD, and my manual only speaks of the 'P24T' processor, which it should support.
P24T was the codename of the CPU before it arrived.
So the manufacturer probably wanted to support the P24T, and put the new socket with the extra pins on there, and added the required jumper settings... But they could never actually test with a real POD because they weren't available yet, so there's probably a bug somewhere in the board or the BIOS or whatever.

I put the POD in a later 486, a Compaq Deskpro XL466, and it works fine in there (but this is a PCI system).

So there are indeed compatibility problems with 486 motherboards, and since VLB boards are by definition earlier 486 boards, you will be more likely to encounter problems there.
But the POD should be 100% compatible with the 486 bus, including the VLB part, so if you can find the right board, it should work.

http://scalibq.wordpress.com/just-keeping-it- … ro-programming/

Reply 5 of 9, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

From what I understand, there was a sample version of the POD that was available to OEMs for compatibility testing sometime in late 1993 or early 1994. That version was reportedly 5V and ran at a clock doubled 66MHz. I suspect it likely did not support writeback internal cache, and was 16KB like the Socket4 version. As you probably recall, the 5V Pentiums were initially very hard to produce at 66MHz, and the smaller profile of the Socket3 chips likely worsened the heat problems. It's also quite likely that the integer performance of this earlier POD wasn't much better than that of the DX2-66. I suspect these are the reasons why Intel held the POD back for so long, and only released it when threatened with legal action.
The reason why most 486 boards don't fully work with a POD83 was because they were certified to work with the earlier 5V version. From what I've seen the only 486 boards that fully support the POD83 were the ones made in late 1995 or 1996.

Has anyone ever seen or heard about the 5V "POD66"?

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 6 of 9, by Scali

User metadata
Rank l33t
Rank
l33t
Anonymous Coward wrote:

It's also quite likely that the integer performance of this earlier POD wasn't much better than that of the DX2-66.

Why is that likely? The Pentium is a superscalar CPU, which is twice as fast as a 486, best-case, and as fast as a 486 worst-case.
They also included extra cache in the POD, to compensate for the 32-bit bus on a 486 board, so that would only make its performance even higher compared to a regular 486.

http://scalibq.wordpress.com/just-keeping-it- … ro-programming/

Reply 7 of 9, by swaaye

User metadata
Rank l33t++
Rank
l33t++

Integer performance can be matched by a 16K L1 486 120/133 but nothing can touch the 83MHz POD FPU. You can make a 486 platform run Quake pretty well with one. If you have PCI, a Voodoo on a 486 board actually makes some sense with a POD.

But the problem with POD is I've never seen a motherboard that works perfectly with it. The last board I used it in forced the motherboard L2 cache into write through mode with the POD. Even if they say P24T is supported it doesn't guarantee it will work correctly, let alone at it's full potential. One board actually needed physical modifications according to the manual and it was still not quite stable.

Reply 8 of 9, by Scali

User metadata
Rank l33t
Rank
l33t
swaaye wrote:

Integer performance can be matched by a 16K L1 486 120/133 but nothing can touch the 83MHz POD FPU. You can make a 486 platform run Quake pretty well with one. If you have PCI, a Voodoo on a 486 board actually makes some sense with a POD.

Yup, after all, it is a Pentium, not a 486.
I've benchmarked my POD83, and it's pretty much up there with a P75 in terms of processing speed in practice (in theory it is of course a real Pentium at 83 MHz, but the 32-bit bus affects performance somewhat, even in more compute-heavy tasks).
You have less memory and PCI bandwidth, but for 3d-accelerated games that is not necessarily a problem. 3D geometry calculations do not require that much bandwidth (especially with the kind of lowpoly games aimed at P75-ish PCs), and the 3D accelerator stores its textures in onboard memory, so it doesn't rely that much on the bandwidth of the PCI bus.

http://scalibq.wordpress.com/just-keeping-it- … ro-programming/

Reply 9 of 9, by Jo22

User metadata
Rank l33t++
Rank
l33t++

Thanks for all the replies, folks! 😀

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//