feipoa wrote:

I no longer recall exactly how far I tried to take the original C6, but I think it didn't go much past its rated frequency. The 233 MHz rated Winchip2 seemed to run OK at 262 MHz.

You had better luck than I did with one. About 12 years ago, I used one of these to upgrade a P5-75 MHz machine that belonged to a friend. It was one of the 233 MHz rated models, and it would run fine at 240 (60*4), or at 225 (75*3), but not at 266 (4*66) or at 263 (75*3.5). Your benchmarks suggest that the 240 MHz configuration is fastest, but I remember in my own real-world testing after I installed it, that the 225 MHz configuration was noticeably faster than stock 233 or running at 240.

Later on, I used a few more of them to upgrade P5-based 3.3v systems, and always used them at that setting if the board could handle it. They weren't very fast, but they were a pain-free way to speed up a P5.

There was the IDT Winchip C6, the Winchip2, Winchip2A, and Winchip2B. I think you need the Winchip2A to overclock beyond 240 MHz. Only in the last few months did I acquire the Winchip2B, which one day I'll get around to benching.

First of all: Great (!!!) work on all those different levels. I really admire your passion to assemble all this data.

That said, two things I'm wondering about:

1.) Did you establish a performance upper bound for the Matrox G200 in the OpenGL test? I.e. from what point on does it limit the CPUs' performance.

2.) I wonder what the possible explanation for the absolute (not 133-MHz-normalized) Quake 1 results could be:
From your PDF
CPU Quake 1
Intel Pentium II OD, 333 MHz, 66 / 5.0x 27.3
Intel Pentium II OD, 300 MHz, 60 / 5.0x 24.6
Intel PIII - Cop, 600 MHz, 100 / 6.0x 22.2
Intel PIII - Kat, 600 MHz, 100 / 6.0x 22.1
AMD K6-3+, 600 MHz, 100 / 6.0x 22

It cannot, like in the Doom results be attributed to high bus speeds and also architecture-wise, the Overdrive CPUs should not have much of an advantage. I wonder if a timer problem could be the reason for this.

Thank you for your comments. I do not have time to look deeper into your second question, but question 1 is pretty easy to answer.

The plot of QuakeII OpenGL Fps vs. CPU Frequency for the Pentium MMX series of CPUs does not show the slope reaching a plateau, even up to 300 MHz.

For the PII, on the other hand, there are three distinct regions of slope.

For 133-300 Mhz, we see about a 2.5 FPS gain for every 33 MHz increase in CPU speed. For comparison, the Pentium MMX's had about 2.1 FPS gain for every 33 Mhz increase in CPU speed.

For 300-400 Mhz, we see about a 0.9 FPS gain for every 33 MHz increase in CPU speed.

For 400-600 MHz, we see about a 0.4 FPS gain for every 33 MHz increase in CPU speed.

From these results, I'd say that a PII at 300 MHz is the GPU-limit for this Matrox G200 PCI card. There is still acceptable gain up to 400 MHz, but for a 350-400 Mhz PII, its best to upgrade your graphics card.

If the PII-300 is about the limit for the G200 PCI, then this equates to about an AMD K6-3-450.

Very, very interesting and well-done article and comparison. I don't think even the tech publications of the day did such high-quality comparisons. 😎

I have several Pentium Pro 200s inside a couple of IBM Netvista systems. I never realized how nice of a chip they really are until now. Especially since all the IBM systems can run dual Pentium Pros. I need to find some time to play with that.

From what I remember when I was looking to upgrade the processors on these, there was a 400Mhz overdrive available for the Pentium Pros. I remember finding some in CDWs outlet a few years back and almost pulled the trigger. Those processors would be an interesting comparison in the mix. 😀

I thought the fastest overdrive for the socket 8 was a PII-333?

feipoa wrote:

I thought the fastest overdrive for the socket 8 was a PII-333?

its not a "true" overdrive. It is a socket 370 > socket 8 converter. You could then use most 66mhz fsb celeron processors.

link: http://hexus.net/tech/reviews/cpu/125-powerleap-pl-proii/

I might have seen ONE on ebay in recent memory. 😜

Interesting! I wonder what is the fastest stable CPU which will work with such an adapter...

well the fastest 66mhz celeron is the 766 if memory serves... But I think it was mostly made for the 4-533 mendicino cores TBH. I'm not interested in haveing one, but I also wouldn't turn one down.

That link suggests that you could only go up to Celeron 533 MHz for SMP mode because after this speed SMP in Celerons was disabled. However, I am left wondering if a PIII-1100/100 chip might work. If this adapter works with a Celeron-700 at 1.75 V, could it not work with a PIII-1100/100? The PIII-1100/100 runs at 1.75V and provides double the L2 cache. This would give the old socket 8 a max of 733 MHz for SMP operation. Or am I missing something fundamental here?

feipoa wrote:

Or am I missing something fundamental here?

bios support. From my research, it seems that often the L2 cache doesn't get enabled, and has to be worked around in the OS. It might be fun to fool around with a pair of 370>8 converters but honestly i'd rather spend my money on a BP6 or something similar.

Those socket 8 to 370 converters aren't that easy to find and usually not cheap. It would probably be cheaper just to buy a socket 370 motherboard to put your socket 370 CPU into rather than trying to upgrade a socket 8 system to use it.

sliderider wrote:

Those socket 8 to 370 converters aren't that easy to find and usually not cheap. It would probably be cheaper just to buy a socket 370 motherboard to put your socket 370 CPU into rather than trying to upgrade a socket 8 system to use it.

I think everyone here already knows that, its not like its 1996 around here... 🤣 I think the interest stems more from a "what-if" mentality than anything else.

luckybob wrote:

its not a "true" overdrive. It is a socket 370 > socket 8 converter. You could then use most 66mhz fsb celeron processors. […]
Show full quote

feipoa wrote:

I thought the fastest overdrive for the socket 8 was a PII-333?

its not a "true" overdrive. It is a socket 370 > socket 8 converter. You could then use most 66mhz fsb celeron processors.

link: http://hexus.net/tech/reviews/cpu/125-powerleap-pl-proii/

I might have seen ONE on ebay in recent memory. 😜

Yep, that was it! I remember it by picture. It would be interesting to get a hold of these adapters and some processors. All my IBMs could be upgraded. 😎

sliderider wrote:

Those socket 8 to 370 converters aren't that easy to find and usually not cheap. It would probably be cheaper just to buy a socket 370 motherboard to put your socket 370 CPU into rather than trying to upgrade a socket 8 system to use it.

Yep, it's never been easy to upgrade anything that wasn't mainstream in a cost-effective manner. 🙁 Still, a dual 766 Celeron would be interesting to run. Of course, I think the RAM in my IBMs tops out at 128MB, 🤣.

Excuse me, where can I find the benchmark software of the first post? Is there a download section?
Thank you!

pep_one wrote:

Excuse me, where can I find the benchmark software of the first post? Is there a download section?
Thank you!

There is no downloads section on this website. At least officially. That said all the benchmarks can be found for free if you google them. The exception to this is MDK, but copies of this game are rather cheap on ebay last I looked.

pep_one wrote:

Excuse me, where can I find the benchmark software of the first post? Is there a download section?
Thank you!

It will take me some time to read through all the licence agreements before I would feel comfortable posting any benchmark suites. For abandonware, one may also need to see if another company bought out the rights. At any rate, all this software can be found using google with some time and diligence. BYTEmark was the most difficult for me to find.

I seem to recall another post here on Vogons whereby user Mau1wurf1977 compiled a great deal of retro benchmark programs for download. Anyone recall that link?

Thank you so much for reply. it would be great if you found this post with the collection of software!

Took me 2 seconds to find it through a Google search of "Mau1wurf1977 benchmark"

VOGONS Benchmark archive?