Good question ! Explains why a PDOP was never really much faster than a DX4, if you include the 486 chipset limitations too. Maybe the chipset limitation is why it was disabled. Thanks, interesting.

I love the blog and respect its author, but I don't think the above statement is true. Or better said, I am not quite sure if the pipeline can be disabled or not but there is plenty of solid evidence (in Intel datasheets, on all schemes, in press coverage, as well as actual in my own benchmarks) that the V-pipeline is there and active.

When profiling software on PODP and monitoring Pentium performance counters you can easily see that the expected proportion of instruction is executed on the V-Pipe. Also, arguably no chance the chip would be able to reach performance it is delivering with only one unit. Futhermore, would be a massive fraud from Intel if the chip was lacking the principal feature that was a centrepiece of all Pentium marketing. Surely this would have to be mentioned in datasheets which are quite detailed and list even subtler changes between Pentium and Pentium Overdrive.

See my own test - https://dependency-injection.com/the-perfect-pentium/

Nevertheless, it had the Pentium FPU with all it's glory ..

My personal opinion is that the Pentium Overdrive is in fact a super scalar CPU. A regular Pentium is going to have a 66MHz FSB and 64-bit path to memory. Those are both cut in half in the case of a Pentium Overdrive. That is a lot to overcome when running benchmarks. Much like an 8088 vs 8086 or a 386SX vs 386DX. Just my two cents.

Some really solid evidence from Pentium Overdrive datasheets.

It's still doesn't make you run Quake so theres no point in this CPU. Usually see like 19-20 FPS in quake, its probably not fast enough to run duke nukem 3d at reasonable FPS either. Maybe if you have a PCI board and slot a voodoo 1 u can run quake 🤣.

That is a lot to overcome when running benchmarks.

And surprisingly Pentium Overdrive is faster than classic Pentium 66, even with gimped front side bus.

Warlord wrote on 2020-08-24, 22:18:

It's still doesn't make you run Quake so theres no point in this CPU. Usually see like 19-20 FPS in quake, its probably not fast enough to run duke nukem 3d at reasonable FPS either. Maybe if you have a PCI board and slot a voodoo 1 u can run quake 🤣.

Bit late to the party, but I managed to increase the performance in duke3d from ~12fps to ~40fps in the opening scene on a VLB system.

shock__ wrote on 2020-09-12, 22:14:

Warlord wrote on 2020-08-24, 22:18:

It's still doesn't make you run Quake so theres no point in this CPU. Usually see like 19-20 FPS in quake, its probably not fast enough to run duke nukem 3d at reasonable FPS either. Maybe if you have a PCI board and slot a voodoo 1 u can run quake 🤣.

Bit late to the party, but I managed to increase the performance in duke3d from ~12fps to ~40fps in the opening scene on a VLB system.

My PCI DX4 gets about 30-33fps at the start of E1M1 in Duke 3D. 10FPS in quake. Pretty sure a POD should do a bit more, much more in Quake compared to any 486.

I think my DX4 is a bit gimped though being only a write through model.

We really need benchmarks to clarify the differences in frames between write through/write back DX2/DX4 chips and the like.

There is too much hype surrounding pod83. I've seen very rare case where the chipset of the host computer is just amazing, and the board supports EDO interleave and has either Non buggy PCI, and you have like a voodoo, or you have some amazing VLB card.

Jokes aside for 99% other systems out there pod83 doesn't jump you from having a 386 or a 486 to having a socket 4-5 Pentium class computer that can run those games at acceptable frame rates, or any other DOS games that are as demanding or more demanding than them. You cant magically get Pentium 233mx performance out of a 486.

What you end up with is a 486 that is now too fast to run some games, and too slow to run others as de turboing it doesn't slow it down enough. 100mhz overdrives suffer the same problem where now de turboing doesn't really help you.

Just my opinion it is better to just accept a 486 or a 386 for what it is.

I do think something like a 66mhz overdrive is a good cpu though for such a system. As it gives u a lot of flexibility through the use of turbo and cache disabling to get the thing to slow 386 performance, and enough horse power to run win 3.11 and some later dos games, and it should run the bus at 33 mhz, since its clock doubled.

I do think something like a 66mhz overdrive is a good cpu though for such a system.

120 Mhz / 60 Mhz FSB (with some voltmod) would be fun.

The Serpent Rider wrote on 2020-08-25, 02:09:

That is a lot to overcome when running benchmarks.

And surprisingly Pentium Overdrive is faster than classic Pentium 66, even with gimped front side bus.

It's actually not too surprising...because when the P60 and P66 came out there wasn't proper memory or cache to handle the faster bus speed. So what was done instead was manufacturers just added an assload of wait states and called it a day. Most of the Socket4 Pentiums I've seen have 486 class memory scores, which really hold them back.

Yes. On paper 64bit/66 MHz of Pentium sounds like a great deal compared to 32bit/33 MHz. But it is more complicated.

A fast modern 486 motherboard talks to cache at 2-1-1-1. A typical early Pentium socket 4/5 is lucky to do 4-2-2-2. Both with async SRAMs. (sync SRAM or pipelined burst can do 3-1-1-1, but they weren't common until 2 years after Pentium launch).

This extra latency effectively negates any benefits of 66 MHz bus over 33 MHz.

Yes. Pentium has 64 bit, but it also needs to refill cache lines that's double in size compared to 486/Pentium Overdrive, which is not necessarily a benefit for 1993-94 software. That's why fast 486 board actually deliver higher memory throughput that early Pentium boards.

Warlord wrote on 2020-09-13, 01:30:

Jokes aside for 99% other systems out there pod83 doesn't jump you from having a 386 or a 486 to having a socket 4-5 Pentium class computer that can run those games at acceptable frame rates, or any other DOS games that are as demanding or more demanding than them. You cant magically get Pentium 233mx performance out of a 486.

What you end up with is a 486 that is now too fast to run some games, and too slow to run others as de turboing it doesn't slow it down enough. 100mhz overdrives suffer the same problem where now de turboing doesn't really help you.

I think most of the people who would seek out a POD83 probably already have a multitude of different retro systems to choose from. Honestly who really cares about trying to slow down a 486 when you can just use your AT or XT?

The POD is great if you look at it in the right way. Will it make games that are unplayable on 486 playable? No, not really. But it will make games that do run tolerably on a 486 run much better.

The POD is great if you look at it in the right way

Eh, not really. If you already have multiple systems, you don't need that stopgap solution for Socket 3 platform. And in many cases using overclocked AMD 5x86 would be more beneficial for stuff that actually CAN run decently. So let's be real, people buy POD this days to satisfy their benchmarking itch. That's about it.

mpe wrote on 2020-09-13, 11:17:

Yes. On paper 64bit/66 MHz of Pentium sounds like a great deal compared to 32bit/33 MHz. But it is more complicated. […]
Show full quote

Yes. On paper 64bit/66 MHz of Pentium sounds like a great deal compared to 32bit/33 MHz. But it is more complicated.

A fast modern 486 motherboard talks to cache at 2-1-1-1. A typical early Pentium socket 4/5 is lucky to do 4-2-2-2. Both with async SRAMs. (sync SRAM or pipelined burst can do 3-1-1-1, but they weren't common until 2 years after Pentium launch).

This extra latency effectively negates any benefits of 66 MHz bus over 33 MHz.

Yes. Pentium has 64 bit, but it also needs to refill cache lines that's double in size compared to 486/Pentium Overdrive, which is not necessarily a benefit for 1993-94 software. That's why fast 486 board actually deliver higher memory throughput that early Pentium boards.

Yep, but still great speed for only 66mhz if you run it on a good board.

Apologies for ressurecting the thread, but thought this question might be better answered here:

I recently acquired a podp5v83, and put it into my old childhood 486DX2 66Mhz-build, just to play what if… the board is from ‘93 and I have not been able to ID it, seems like it is a taiwanese clone board..anyway the system boots but the Overdrive is reported as a 486DX2 at boot, is this expected behaviour?

abasak wrote on 2022-03-28, 15:57:

Apologies for ressurecting the thread, but thought this question might be better answered here:

I recently acquired a podp5v83, and put it into my old childhood 486DX2 66Mhz-build, just to play what if… the board is from ‘93 and I have not been able to ID it, seems like it is a taiwanese clone board..anyway the system boots but the Overdrive is reported as a 486DX2 at boot, is this expected behaviour?

Did you change any of the jumper settings on the board?
Unless those jumper settings are silkscreened on the board or, you are able to find those jumper settings, then you're out of luck (since there are normally quite a bit of jumpers to change on those older boards).

Try looking here: https://www.ultimateretro.net/en/

There’s no multiplier on the board, only for clockspeed, which I left at 33mhz. Maybe that is the issue.