mpe wrote on 2020-02-04, 12:19:

75 MHz @ 50 MHz bus with 25 MHz PCI

Are you sure about PCI running at 25 MHz?
It was common for PCI to run asynchronously, so standard 33 MHz.
It's surely running at 25 MHz when the FSB is also 25 MHz, as in 486SX-25, but with FSB >33 MHz I would rather expect PCI at 33 MHz.

Ahem, 50 Mhz bus has 2/3 divider for PCI. Intel weren't that dumb. As for the bus difference itself, early Pentiums had quite anemic performance, which couldn't use its at full potential anyway. That's why Pentium Overdrive for Socket 3 doesn't feel as complete outsider, compared to P66 or P75, even with doubled L1 cache considered.

This is certainly going to be an interesting benchmark but I wouldn't say it's level ground for the processors. One advantage of the P75 is its compatibility with a more mature platform. Don't take this wrong, I get what you're saying and trying to find out. 🙂👍
With standard settings I guess the 75MHz chip will still be the overall winner but I wouldn't bet on it.
It'd be cool if you could also do benchmarks and include results of 133MHz (66fsb*2) vs. 150MHz (50fsb*3) on the Intel Plato to see wether/how the outcome changes with higher multipliers.

There's no 3x multiplier. So such comparison can be only done with 100Mhz Pentium.

I never had a proper P66 to test but my P75 was so slow compared to my P100 that I presumed they would be roughly equal at performance even if I had one.

The Serpent Rider wrote on 2020-02-04, 12:45:

Ahem, 50 Mhz bus has 2/3 divider for PCI. Intel weren't that dumb. As for the bus difference itself, early Pentiums had quite anemic performance, which couldn't use its at full potential anyway. That's why Pentium Overdrive for Socket 3 doesn't feel as complete outsider, compared to P66 or P75, even with doubled L1 cache considered.

Grzyb wrote on 2020-02-04, 12:42:

Are you sure about PCI running at 25 MHz?
It was common for PCI to run asynchronously, so standard 33 MHz.

I am almost sure that 25 MHz PCI was a thing in Intel's Pentium chipsets and at least boards based on their reference designs.

They mention 25 MHz PCI thorough datasheets (incl. as late as TX) and mandate that PCLK is divided by 2 from HCLK and must be driven by host clock.

But it could be that some mb vendors or 3rd party chipsets tailored that through some clock generator triks. But it wasn't that important as the 75 MHz was the only CPU to use the 50 MHz bus so IMHO it wasn't worth it. The async PCI was more important later in SS7 era. But there was no Intel chipsets then...

The Serpent Rider wrote on 2020-02-04, 13:17:

There's no 3x multiplier. So such comparison can be only done with 100Mhz Pentium.

I have a set of overdrives with locked multipliers and know how to solder too 😀

Doornkaat wrote on 2020-02-04, 13:11:

This is certainly going to be an interesting benchmark but I wouldn't say it's level ground for the processors. One advantage of the P75 is its compatibility with a more mature platform.

But it wasn't until at least spring 1995 when those mature platforms become available. So first 75/90/100 customers had to run it on Neptune and were comparing them against P60/P66 systems.

The Serpent Rider wrote on 2020-02-04, 13:17:

There's no 3x multiplier. So such comparison can be only done with 100Mhz Pentium.

I'm not sure I get your point.
Using a P54CS chip and setting it up to run at 133/150MHz should be possible on this board. Afaik it's the same core but smaller, isn't it? So the comparison would be valid?
Just connect BF1(X34) to Vss (X36 or others) on the board or CPU to get the needed 3x multiplier.
What part am I missing?

I chose 133/150MHz because they have the same relative difference in core speed (150MHz is ~13% faster than 133MHz, 133MHz is ~11,34% slower than 150MHz) as 66/75MHz (again ~+13% or ~-11,34%) but keep the same FSB speed. 100/133MHz makes for a +33% or ~-25% difference. That makes comparing them to 66/75MHz more complicated.
Is it understandable where I am aiming at? I'm not good at explaining this in English I'm afraid.😅

mpe wrote on 2020-02-04, 13:46:

Doornkaat wrote on 2020-02-04, 13:11:

This is certainly going to be an interesting benchmark but I wouldn't say it's level ground for the processors. One advantage of the P75 is its compatibility with a more mature platform.

But it wasn't until at least spring 1995 when those mature platforms become available. So first 75/90/100 customers had to run it on Neptune and were comparing them against P60/P66 systems.

Good point. 👍

Can’t you force a Pentium Overdrive into 1x multiplier by disabling the fan? If that’s the case then you could run it at 66mhz.

clueless1 wrote on 2020-02-04, 20:14:

Can’t you force a Pentium Overdrive into 1x multiplier by disabling the fan? If that’s the case then you could run it at 66mhz.

It is funny, because on Pentium overdrives it doesn't disable multiplier (like the POP5v-83 does). It just applies some throttling so you get performance roughly equivalent to 486DX2-50. (when using 180 MHz OD). The clock cannot be measured with TSC.

The Plato motherboard is a strange one. It is a very early Socket 5. It only has a settings for 75 and 90 MHz CPUs. Not even 100 MHz. Need to find details about the clock chip to see if it really has 66 MHz bus by some undocumented setting.

It does take overdrives though. So I can ran a 180 MHz MMX overdrive (on 90 MHz setting) without problem.

OK After some sloppy soldering, my Plato board is working and the results are in.

It basically confirmed my expectation. The Pentium 66 is quite a bit faster that the P75 when compared on a like chipset. Clearly the bus speed does matter when other things are comparable.

It only stress out how important the launch of the Triton chipset was for the Pentium platform. In fact P 75 on 430FX produces almost similar performance as P90 on Neptune. The pipelined-burst cache and EDO RAMs changed the landscape a bit. The Triton II (430HX) then provided even a bit tighter memory timings and deeper write buffer than the 430FX, but it wasn't as big step as going from LX/NX to FX.

I love this forum, every post looks like a university computer science lesson

mpe wrote on 2020-02-04, 13:28:

But it wasn't that important as the 75 MHz was the only CPU to use the 50 MHz bus so IMHO it wasn't worth it.

This may not be *entirely* true. As far as I recall there was a second (far less common) CPU from intel that used the 50MHz bus...a gimped version of the P100 that ran at 2x50 (probably a substandard part). I'll have to check CPU World again for the S-spec, but I think this CPU might have also run at a higher non-standard voltage too.

So this means that on the 430NX the PCI does run at 25MHz? Really weird for Intel to have done this. I'm curious to see pure CPU benchmarks as well, something that does not rely on the graphics card/PCI bus too much.

Also on a completely unrelated subject, I wish those stacked bar charts get banned 🙂 You need extra effort to extract the information when adding apples to oranges. At least on these you did it in the best possible way writing the individual scores on each rectangle

mpe wrote on 2020-02-04, 22:57:

The Plato motherboard is a strange one. It is a very early Socket 5. It only has a settings for 75 and 90 MHz CPUs. Not even 100 MHz. Need to find details about the clock chip to see if it really has 66 MHz bus by some undocumented setting.

Really strange. The manual states there are "design differences" between 75/90MHz and 100MHz boards. On 100MHz boards J1H4 has to be set to 1-2 and J1H3 has to be removed. Have you tried wether this works on your board (despite "design differences"), mpe?

There is an ebay listing for two very similar DELL branded boards jumpered for 100MHz here. Right away I can see there are additional parts below the cache area and it looks like they have to do with voltage regulation. I also notice an extra capacitor and a diode next to the big coil at the top of the board.
Since there are 100MHz P54C chips requiring more than 3.3V but all 75/90MHz chips are specced for 3.3V, could it be that the 75/90MHz board doesn't have the required VRM to drive the lesser 100MHz Pentiums that need 3.4-3.6V so Intel decided to officially not support any 100MHz CPUs albeit 3.3V parts will run? Could you test this?

Edit: There are also additional parts near the right edge between the CPU socket and speaker connector. Another transistor, some diodes, a small IC and some other parts.

konc wrote on 2020-02-05, 07:59:

So this means that on the 430NX the PCI does run at 25MHz? Really weird for Intel to have done this. I'm curious to see pure CPU benchmarks as well, something that does not rely on the graphics card/PCI bus too much.

Not just the 430NX. But all Intel's chipsets up to 430TX drive PCI @25 with Pentium 75 due to the fixed 1/2 divider. Should anyone know about a motherboard that can override this please share it. All my Socket 5/7 boards all have dividers for ISA bus only.

I think the PCI doesn't hurt performance that much. The cache/mem latency and 50 MHz bus is more important factor IMHO for frame rates in Quake.

Pure CPU benchmarks (like Dhrystone) are obviously proportional to CPU clock - chipset has no effect here.

Doornkaat wrote on 2020-02-05, 08:30:

Really strange. The manual states there are "design differences" between 75/90MHz and 100MHz boards. On 100MHz boards J1H4 has t […]
Show full quote

Really strange. The manual states there are "design differences" between 75/90MHz and 100MHz boards. On 100MHz boards J1H4 has to be set to 1-2 and J1H3 has to be removed. Have you tried wether this works on your board (despite "design differences"), mpe?

There is an ebay listing for two very similar DELL branded boards jumpered for 100MHz here. Right away I can see there are additional parts below the cache area and it looks like they have to do with voltage regulation. I also notice an extra capacitor and a diode next to the big coil at the top of the board.
Since there are 100MHz P54C chips requiring more than 3.3V but all 75/90MHz chips are specced for 3.3V, could it be that the 75/90MHz board doesn't have the required VRM to drive the lesser 100MHz Pentiums that need 3.4-3.6V so Intel decided to officially not support any 100MHz CPUs albeit 3.3V parts will run? Could you test this?

Edit: There are also additional parts near the right edge between the CPU socket and speaker connector. Another transistor, some diodes, a small IC and some other parts.

Yes. I tried the undocumented 100 MHz setting and it doesn't work. Perhaps there are indeed missing parts.

My board started its life as OEM board in a Gateway 2000 machine, although it has been upgraded back to generic using standard Intel firmware. AFAIK there are at least 4 different PCB versions of this boards and only two of them have the support for 100 MHz. It is certainly not the power delivery limit as the board can handle 200 MMX overdrive (although that one has its own regulator, but still draws all the power from the board). It might be the 3.3V vs 3.45V thing. Perhaps they only wanted to support 100 MHz if having the voltage setting.

Similarly my Batman's Revenge board doesn't have the official support for Pentium 66. Only boards with 5V -> 5.27V regulator got the certification. My Batman doesn't have one (see the empty area next to the power connector). Yet, I was unable to find a Pentium 66 that wouldn't work reliably in it.

It looks like that the voltage was a big deal in early Pentium days and I am sure some early chips were really picky.