The K6 is an interesting CPU to have and would definitely be nice to install in such a setup from a nostalgic point of view.
The K6-233 runs at 3.3V vcore which is the same as the earlier Pentium (P54C) CPU's and hence the motherboard should have no problem running that CPU.

From a synthetic benchmark perspective, the K6 might run faster in certain tests but, the FPU of the Pentium 233 MMX is must faster and, based on other forum posts I've read in the past, the Pentium 233 MMX will outperform the K6-233 to a certain degree.
I think it depends what you intend on using the system for.

I didn't expect the Pentium MMX FPU to be much faster but I imagined it. I sure was expecting (and tested lately) the Pentium II FPU to be much faster than both as it is obviously. What it's immediately felt is how hot the cpu heatsink become on the K6 at that freqs compared to the cold P-MMX even if the typical watts seems similar, the maximun watt requirement seems quite high for these old mainboards.
It'd be interesting to know what that Jp6 jumper reserved for future AMD cpu actually do and if on a voltage level or simply unused.

386SX wrote on 2022-01-22, 21:57:

I didn't expect the Pentium MMX FPU to be much faster but I imagined it. I sure was expecting (and tested lately) the Pentium II FPU to be much faster than both as it is obviously. What it's immediately felt is how hot the cpu heatsink become on the K6 at that freqs compared to the cold P-MMX even if the typical watts seems similar, the maximun watt requirement seems quite high for these old mainboards.
It'd be interesting to know what that Jp6 jumper reserved for future AMD cpu actually do and if on a voltage level or simply unused.

My eye just caught this - might answer some of your questions (there is also a link to a page with the jumper settings): A nice late socket7 motherboard: PCPartner TXB820DS 35-8333-03
Edit: this is probably your board but, the other board is very similar: https://www.ultimateretro.net/en/motherboards/6618

jesolo wrote on 2022-01-22, 21:45:

The K6 is an interesting CPU to have and would definitely be nice to install in such a setup from a nostalgic point of view. The […]
Show full quote

The K6 is an interesting CPU to have and would definitely be nice to install in such a setup from a nostalgic point of view.
The K6-233 runs at 3.3V vcore which is the same as the earlier Pentium (P54C) CPU's and hence the motherboard should have no problem running that CPU.

From a synthetic benchmark perspective, the K6 might run faster in certain tests but, the FPU of the Pentium 233 MMX is must faster and, based on other forum posts I've read in the past, the Pentium 233 MMX will outperform the K6-233 to a certain degree.
I think it depends what you intend on using the system for.

I agree the 233 k6 vs the pentium 233 the pentium should be the better processor.

However it’s a curious experience of mine that even a pentium 1 tillamook at 300mhz is dreadfully slow next to a k6+ at 300mhz but the k6+ feels equal clock to clock to a mII for windows and desktop applications

At OP:

Usually most motherboards of that era support 2.0v or 2.2v by removing all voltage jumpers. Some are just one jumper.

You’ll probably need a volt meter to find the core voltage somewhere between ground and one of the tabs on the fets.

Some regulators support voltages down to 1.2v or so so remove all and see what you get you may have to add one back.

Anyway, reason I mention this is your mobo probably would work with a k62/3+ At 400 to 500mhz

jesolo wrote on 2022-01-22, 22:20:

386SX wrote on 2022-01-22, 21:57:

I didn't expect the Pentium MMX FPU to be much faster but I imagined it. I sure was expecting (and tested lately) the Pentium II FPU to be much faster than both as it is obviously. What it's immediately felt is how hot the cpu heatsink become on the K6 at that freqs compared to the cold P-MMX even if the typical watts seems similar, the maximun watt requirement seems quite high for these old mainboards.
It'd be interesting to know what that Jp6 jumper reserved for future AMD cpu actually do and if on a voltage level or simply unused.

My eye just caught this - might answer some of your questions (there is also a link to a page with the jumper settings): A nice late socket7 motherboard: PCPartner TXB820DS 35-8333-03
Edit: this is probably your board but, the other board is very similar: https://www.ultimateretro.net/en/motherboards/6618

That's the mainboard layout I'd say the same, I think exists different revisions at the end of that number -01 -02 and change the pins configuration for the CPUs at least found some manual pages having different pin configs of what the PCB printed tables describes on my board. The JP6 is seen anyway on the left of the socket and in the manual is said for future cpu. As suggested by @Sphere478 (thanks) I might try removing the voltage jumper to read the voltage where suggested. I was thinking that mosfet might not change directly on its pins the voltage but decreased after on the PCB.

386SX wrote on 2022-01-22, 22:52:

jesolo wrote on 2022-01-22, 22:20:

386SX wrote on 2022-01-22, 21:57:

I didn't expect the Pentium MMX FPU to be much faster but I imagined it. I sure was expecting (and tested lately) the Pentium II FPU to be much faster than both as it is obviously. What it's immediately felt is how hot the cpu heatsink become on the K6 at that freqs compared to the cold P-MMX even if the typical watts seems similar, the maximun watt requirement seems quite high for these old mainboards.
It'd be interesting to know what that Jp6 jumper reserved for future AMD cpu actually do and if on a voltage level or simply unused.

My eye just caught this - might answer some of your questions (there is also a link to a page with the jumper settings): A nice late socket7 motherboard: PCPartner TXB820DS 35-8333-03
Edit: this is probably your board but, the other board is very similar: https://www.ultimateretro.net/en/motherboards/6618

That's the mainboard layout I'd say the same, I think exists different revisions at the end of that number -01 -02 and change the pins configuration for the CPUs at least found some manual pages having different pin configs of what the PCB printed tables describes on my board. The JP6 is seen anyway on the left of the socket and in the manual is said for future cpu. As suggested by @Sphere478 (thanks) I might try removing the voltage jumper to read the voltage where suggested. I was thinking that mosfet might not change directly on its pins the voltage but decreased after on the PCB.

Usually it’s the large tab on one of the fets. If that fails pull the cpu and put a resistor leg in the middle pin on the notch side of the cpu socket and close the lever. Read the leg to ground

Oh one final note, do not change voltage jumpers while the board is on, sometimes the controller will get confused and set line voltage 3.3 or 5v 🤣

Sphere478 wrote on 2022-01-22, 23:22:

Oh one final note, do not change voltage jumpers while the board is on, sometimes the controller will get confused and set line voltage 3.3 or 5v 🤣

Thanks! Great. Sorry for the obvious question but the resistor leg only without the resistor itself right? 😁

386SX wrote on 2022-01-22, 23:31:

Sphere478 wrote on 2022-01-22, 23:22:

Oh one final note, do not change voltage jumpers while the board is on, sometimes the controller will get confused and set line voltage 3.3 or 5v 🤣

Thanks! Great. Sorry for the obvious question but the resistor leg only without the resistor itself right? 😁

It’s about the size of a cpu pin. Just find anything metal that is like a cpu pin.

But usually just measuring from a fet is good enough.

The k6-233 was AMDs attempt to smack down Intel in the MHZ game

It’s factory overvolted, hot and draws too many amps for many antique motherboards.

K6-200 or k6-266 are better choices if your board supports them

Considering the manufacturing process size the power requirement seems really high, at least from the heatsink temp. I don't remember previous cpu having similar specifications even with higher voltages.

I'd prefer the Intel MMX CPU for sure.

386SX wrote on 2022-01-23, 08:44:

Considering the manufacturing process size the power requirement seems really high, at least from the heatsink temp. I don't remember previous cpu having similar specifications even with higher voltages.

Not sure this is what you're referring to, but the K6 does seem to have a much higher transistor count (compared to the Pentium 1)(according to the wiki) which would result in higher power dissipation at similar voltages and frequencies.

It contained 4.5 million transistors and had an area of 140 mm2. It was fabricated in a 0.28 μm CMOS process with the same metal pitches as the previous 0.35 μm BiCMOS process, so Intel described it as "0.35 μm" because of its similar transistor density.[27] The process has four levels of interconnect.[27]

While the P55C remained compatible with Socket 7, the voltage requirements for powering the chip differ from the standard Socket 7 specifications. Most motherboards manufactured for Socket 7 before the establishment of the P55C standard are not compliant with the dual voltage rail required for proper operation of this CPU (2.8 volt core voltage, 3.3 volt input/output (I/O) voltage).
https://en.wikipedia.org/wiki/Pentium_(original)

K6 (Model 6)
8.8 million transistors in 350 nm
https://en.wikipedia.org/wiki/AMD_K6

The difference in the number of transistors is probably mostly because of this:
16KB L1 data cache + 16KB L1 instruction cache for the Pentium MMX
L1-Cache: 32 + 32 KB (data + instructions) for AMD K6

But I think this would be sufficient explanation as to why the K6 233MHz is so much more power hungry compared to the Pentium 1 MMX 233MHz (any Pentium 1 basically 😜 ).

Tetrium wrote on 2022-01-23, 10:55:

Not sure this is what you're referring to, but the K6 does seem to have a much higher transistor count (compared to the Pentium […]
Show full quote

386SX wrote on 2022-01-23, 08:44:

Considering the manufacturing process size the power requirement seems really high, at least from the heatsink temp. I don't remember previous cpu having similar specifications even with higher voltages.

Not sure this is what you're referring to, but the K6 does seem to have a much higher transistor count (compared to the Pentium 1)(according to the wiki) which would result in higher power dissipation at similar voltages and frequencies.

It contained 4.5 million transistors and had an area of 140 mm2. It was fabricated in a 0.28 μm CMOS process with the same metal pitches as the previous 0.35 μm BiCMOS process, so Intel described it as "0.35 μm" because of its similar transistor density.[27] The process has four levels of interconnect.[27]

While the P55C remained compatible with Socket 7, the voltage requirements for powering the chip differ from the standard Socket 7 specifications. Most motherboards manufactured for Socket 7 before the establishment of the P55C standard are not compliant with the dual voltage rail required for proper operation of this CPU (2.8 volt core voltage, 3.3 volt input/output (I/O) voltage).
https://en.wikipedia.org/wiki/Pentium_(original)

K6 (Model 6)
8.8 million transistors in 350 nm
https://en.wikipedia.org/wiki/AMD_K6

The difference in the number of transistors is probably mostly because of this:
16KB L1 data cache + 16KB L1 instruction cache for the Pentium MMX
L1-Cache: 32 + 32 KB (data + instructions) for AMD K6

But I think this would be sufficient explanation as to why the K6 233MHz is so much more power hungry compared to the Pentium 1 MMX 233MHz (any Pentium 1 basically 😜 ).

And then there is the tillamook which I was running without a heatsink the other day… 😂

Only playing in the bios and trying jumper settings but still, it just sat there, barely warm rofl.

Drop the voltage some and frequency and I can totally see heavy load with no heatsink.

Tetrium wrote on 2022-01-23, 10:55:

Not sure this is what you're referring to, but the K6 does seem to have a much higher transistor count (compared to the Pentium […]
Show full quote

386SX wrote on 2022-01-23, 08:44:

Considering the manufacturing process size the power requirement seems really high, at least from the heatsink temp. I don't remember previous cpu having similar specifications even with higher voltages.

Not sure this is what you're referring to, but the K6 does seem to have a much higher transistor count (compared to the Pentium 1)(according to the wiki) which would result in higher power dissipation at similar voltages and frequencies.

It contained 4.5 million transistors and had an area of 140 mm2. It was fabricated in a 0.28 μm CMOS process with the same metal pitches as the previous 0.35 μm BiCMOS process, so Intel described it as "0.35 μm" because of its similar transistor density.[27] The process has four levels of interconnect.[27]

While the P55C remained compatible with Socket 7, the voltage requirements for powering the chip differ from the standard Socket 7 specifications. Most motherboards manufactured for Socket 7 before the establishment of the P55C standard are not compliant with the dual voltage rail required for proper operation of this CPU (2.8 volt core voltage, 3.3 volt input/output (I/O) voltage).
https://en.wikipedia.org/wiki/Pentium_(original)

K6 (Model 6)
8.8 million transistors in 350 nm
https://en.wikipedia.org/wiki/AMD_K6

The difference in the number of transistors is probably mostly because of this:
16KB L1 data cache + 16KB L1 instruction cache for the Pentium MMX
L1-Cache: 32 + 32 KB (data + instructions) for AMD K6

But I think this would be sufficient explanation as to why the K6 233MHz is so much more power hungry compared to the Pentium 1 MMX 233MHz (any Pentium 1 basically 😜 ).

Thanks, yes that seems to be a good reason for such difference. I don't know how much such old mainboards with those voltage regulators (my board has a "big heatsink" for a single dual mosfet anyway) can work with these "high end" cpu but I suppose that was the limit. In fact at the power wall plug I measured a total system peak of almost total 90W of current required with the K6 when such values I was used to see with much faster configs. The Pentium MMX seems to run cool enough on the heatsink temp and asking 70W at the desktop. These values of course are variable but it's not every day to see that on such old configs with few PCI/ISA cards that should ask few watts only.

386SX wrote on 2022-01-22, 21:06:

Hello, […]
Show full quote

Hello,

I've this 430VX board I'm using with a Pentium MMX 233, which should be the maximun if the minimun cpu vcore of the jumper is 2,7v or something like it, and testing this and the K6-233. I'd say maybe the K6-233 in Windows feels a bit faster but its power requirement seems higher too and stressing a bit probably the regulators and the board I suppose. Which cpu do you prefer? And why?
And any possibility to install a K6-2 300 66Mhz model 2,2vcore? Are there any tweak to decrease the cpu vcore to be compatible with those? There's a JP6 jumper that must be left free for future K6 cpus but it's not clear what it's supposed to do. The board is a PcPartners 35-8332-01 Socket7 430VX ISA/PCI.
Thanks

Since you have a board very similar to mine (see the thread A nice late socket7 motherboard: PCPartner TXB820DS 35-8333-03 ), I wonder if you have any Cyrix or IBM 6x86 CPUs on hand to verify if your board does the same undocumented multiplier thing that I found - see the second half of the second page in that thread. Essentially, put a 6x86 CPU in, set JP1 to 2-3, JP2 to 2-3 and see if the 6x86 POSTs with a multiplier of 1x.

stamasd wrote on 2022-01-23, 13:17:

386SX wrote on 2022-01-22, 21:06:

Hello, […]
Show full quote

Hello,

I've this 430VX board I'm using with a Pentium MMX 233, which should be the maximun if the minimun cpu vcore of the jumper is 2,7v or something like it, and testing this and the K6-233. I'd say maybe the K6-233 in Windows feels a bit faster but its power requirement seems higher too and stressing a bit probably the regulators and the board I suppose. Which cpu do you prefer? And why?
And any possibility to install a K6-2 300 66Mhz model 2,2vcore? Are there any tweak to decrease the cpu vcore to be compatible with those? There's a JP6 jumper that must be left free for future K6 cpus but it's not clear what it's supposed to do. The board is a PcPartners 35-8332-01 Socket7 430VX ISA/PCI.
Thanks

Since you have a board very similar to mine (see the thread A nice late socket7 motherboard: PCPartner TXB820DS 35-8333-03 ), I wonder if you have any Cyrix or IBM 6x86 CPUs on hand to verify if your board does the same undocumented multiplier thing that I found - see the second half of the second page in that thread. Essentially, put a 6x86 CPU in, set JP1 to 2-3, JP2 to 2-3 and see if the 6x86 POSTs with a multiplier of 1x.

I don't think I've those cpu while I have a lot of Pentium, MMX, K6-2, etc.. but also that JP6 for future cpu I wonder if actually change something or in the voltage core or in the multiplier.

Sphere478 wrote on 2022-01-23, 11:15:

And then there is the tillamook which I was running without a heatsink the other day… 😂 […]
Show full quote

Tetrium wrote on 2022-01-23, 10:55:

Not sure this is what you're referring to, but the K6 does seem to have a much higher transistor count (compared to the Pentium […]
Show full quote

386SX wrote on 2022-01-23, 08:44:

Considering the manufacturing process size the power requirement seems really high, at least from the heatsink temp. I don't remember previous cpu having similar specifications even with higher voltages.

Not sure this is what you're referring to, but the K6 does seem to have a much higher transistor count (compared to the Pentium 1)(according to the wiki) which would result in higher power dissipation at similar voltages and frequencies.

It contained 4.5 million transistors and had an area of 140 mm2. It was fabricated in a 0.28 μm CMOS process with the same metal pitches as the previous 0.35 μm BiCMOS process, so Intel described it as "0.35 μm" because of its similar transistor density.[27] The process has four levels of interconnect.[27]

While the P55C remained compatible with Socket 7, the voltage requirements for powering the chip differ from the standard Socket 7 specifications. Most motherboards manufactured for Socket 7 before the establishment of the P55C standard are not compliant with the dual voltage rail required for proper operation of this CPU (2.8 volt core voltage, 3.3 volt input/output (I/O) voltage).
https://en.wikipedia.org/wiki/Pentium_(original)

K6 (Model 6)
8.8 million transistors in 350 nm
https://en.wikipedia.org/wiki/AMD_K6

The difference in the number of transistors is probably mostly because of this:
16KB L1 data cache + 16KB L1 instruction cache for the Pentium MMX
L1-Cache: 32 + 32 KB (data + instructions) for AMD K6

But I think this would be sufficient explanation as to why the K6 233MHz is so much more power hungry compared to the Pentium 1 MMX 233MHz (any Pentium 1 basically 😜 ).

And then there is the tillamook which I was running without a heatsink the other day… 😂

Only playing in the bios and trying jumper settings but still, it just sat there, barely warm rofl.

Drop the voltage some and frequency and I can totally see heavy load with no heatsink.

I was using a DX4-100 with just a large passive heatsink and it seemed to run totally fine. However, I did notice that after prolonged use the heatsink would still be rather hot mostly because of how poor the ventilation was in that case (and my juryrigged cardboard sound dampeners could have helped prevent thermal dissipation through the sides of the case and to the rest of the case, but this is just a suspicion on my part).

Tillamook is definitely one heck of a cool chip ( 😜 ) but personally I'd not risk running it without a heatsink for a couple hours on end 😜
Not saying it couldn't work though, but it's easy to underestimate the effect of prolonged use has on poorly ventilated airspaces.

386SX wrote on 2022-01-23, 12:48:

Tetrium wrote on 2022-01-23, 10:55:

Not sure this is what you're referring to, but the K6 does seem to have a much higher transistor count (compared to the Pentium […]
Show full quote

386SX wrote on 2022-01-23, 08:44:

Considering the manufacturing process size the power requirement seems really high, at least from the heatsink temp. I don't remember previous cpu having similar specifications even with higher voltages.

Not sure this is what you're referring to, but the K6 does seem to have a much higher transistor count (compared to the Pentium 1)(according to the wiki) which would result in higher power dissipation at similar voltages and frequencies.

It contained 4.5 million transistors and had an area of 140 mm2. It was fabricated in a 0.28 μm CMOS process with the same metal pitches as the previous 0.35 μm BiCMOS process, so Intel described it as "0.35 μm" because of its similar transistor density.[27] The process has four levels of interconnect.[27]

While the P55C remained compatible with Socket 7, the voltage requirements for powering the chip differ from the standard Socket 7 specifications. Most motherboards manufactured for Socket 7 before the establishment of the P55C standard are not compliant with the dual voltage rail required for proper operation of this CPU (2.8 volt core voltage, 3.3 volt input/output (I/O) voltage).
https://en.wikipedia.org/wiki/Pentium_(original)

K6 (Model 6)
8.8 million transistors in 350 nm
https://en.wikipedia.org/wiki/AMD_K6

The difference in the number of transistors is probably mostly because of this:
16KB L1 data cache + 16KB L1 instruction cache for the Pentium MMX
L1-Cache: 32 + 32 KB (data + instructions) for AMD K6

But I think this would be sufficient explanation as to why the K6 233MHz is so much more power hungry compared to the Pentium 1 MMX 233MHz (any Pentium 1 basically 😜 ).

Thanks, yes that seems to be a good reason for such difference. I don't know how much such old mainboards with those voltage regulators (my board has a "big heatsink" for a single mosfet anyway) can work with these "high end" cpu but I suppose that was the limit. In fact at the power wall plug I measured a total system peak of almost total 90W of current required with the K6 when such values I was used to see with much faster configs. The Pentium MMX seems to run cool enough on the heatsink temp and asking 70W at the desktop. These values of course are variable but it's not every day to see that on such old configs with few PCI/ISA cards that should ask few watts only.

A 20W difference between those 2 chips is not something I'd want to contest, it seems legit to me tbf.
K6/233 was nearly double the transistor count on a higher voltage (3.2v instead of 2.8v) and it was (as someone already mentioned here) basically a factory overclock.
Pentium Pro and P2 Klamath could run even hotter btw.

The big heatsinks were typically used for the voltage regulators and these could run very hot, potentially even failing if used out of spec. Often these tiny (compared to CPU HSFs) heatsinks would have to be cooled by some airflow from the CPU heatsink fan (in a similar way to how this would become somewhat more typical in later years with, say, AM2 boards).
The Pentium 1 MMX was a pretty good chip for its time. It basically helped the PC world open up to gaming and to a lesser extend to the internet (the multimedia PC remember? 😜 ).

Btw I really like measuring power usage from the wall, at some point I started measuring basically everything except for the washingmachine 😜