gdjacobs wrote:

That VRM might offer additional protection if:
1) The VRM includes protective mechanisms like a PTC, fuse, etc.
2) Internal impedance of the supply limits output current, thus taking a few seconds longer for damage to be dealt (I^2 * t threshold is usually considered to be constant for a given component)

rmay635703 wrote:

A voltage regulator is a very simple device not unlike a comparator […]
Show full quote

A voltage regulator is a very simple device not unlike a comparator

You should be able to tap a few pins off the regulator to get any voltage you want

And again I would wire the 5v input on the vrm straight to a psu wire for added protection.

I would brush up on basic electricity and buy the cheapest VRM with the highest amp rating, then if you want to experiment with high power draw CPUs or overclocking you are set and won’t have to revisit this issue

I see.

Anyway, since a K6-2 has 2.2V core voltage but 3.3V I/O voltage, will the HP 0950-3265 VRM mentioned above (2.0 volt output voltage) be able to provide the 3.3 I/O voltage needed by the K6-2 CPU?

VIO should just pass through from the onboard regulator
AKA the vrm doesn’t drive it.

One way around the issue of weak onboard regulators would be to wire the 3.3 right off the atx supply to the pass through pins if amp draw worries you

rmay635703 wrote:

VIO should just pass through from the onboard regulator
AKA the vrm doesn’t drive it.

One way around the issue of weak onboard regulators would be to wire the 3.3 right off the atx supply to the pass through pins if amp draw worries you

Thank you.

By the way, is it okay to power 2.2V K6-2 using the VRM's 2V output voltage? I don't know how stable such retro CPU when undervolted.

Yes it’ll work fine
Many “laptop “ k6-2’s ran on 1.6-2.0 Volts but at reduced clock speeds

These laptop chips had an identical core to a standarddesktop k6-2 but were tested at lower voltages and speeds.

rmay635703 wrote:

http://datasheets.chipdb.org/Intel/x86/Pentium/24318702.PDF […]
Show full quote

http://datasheets.chipdb.org/Intel/x86/Pentium/24318702.PDF

A VRM “should” follow a standard pinout

Older motherboards sometimes had the pinout in the manual.

Otherwise you would need to find ground, 5v and the the pins that go to the vi/o vcore pins of the cpu

0A545FB9-4210-4FA4-BBDE-964C6F588CE4.png

Great document. With the dimensions for the connector, this should enable one to retrobuild VRMs for Pentium class motherboards ..

rmay635703 wrote:

One way around the issue of weak onboard regulators would be to wire the 3.3 right off the atx supply to the pass through pins if amp draw worries you

Where are you guys getting these ideas from 😁? You cannot directly tap unregulated power source to a sensitive device such as CPU.

I see regulators on the 3.3 volt rail in my atx supply

Anyway
As designed a vrm does nothing to the onboard vio sends it right through and takes the 5 v and regulates it down to whatever you set forvcore

In my mind the vio shouldn’t be touched since it’s likely a lower power draw than the core but if it were an issue my pcchips m590 used to send 4.2 Volts to the io and my k6-2 survived many years at that voltage.
I doubt any decent atx supply would send anything other than 3.3 Volts off the rail

Pardon the necro.

I recently bought XVI 073-20742-30 VRM, whose output voltage is 1,3-3,5V. At first, I was happy with the specs, since its output voltage seems to be able to acommodate a wide range of CPU (AMD K6-2, being 2.2v, falls into the range).

But much to my dismay, the VRM's input voltage is 12V instead of 3.3V, and I cannot cancel my purchase.

So, will it work with my Socket 7 mobo?

Kreshna Aryaguna Nurzaman wrote on 2020-08-27, 16:34:

Pardon the necro. […]
Show full quote

Pardon the necro.

I recently bought XVI 073-20742-30 VRM, whose output voltage is 1,3-3,5V. At first, I was happy with the specs, since its output voltage seems to be able to acommodate a wide range of CPU (AMD K6-2, being 2.2v, falls into the range).

But much to my dismay, the VRM's input voltage is 12V instead of 3.3V, and I cannot cancel my purchase.

So, will it work with my Socket 7 mobo?

The pinouts appear to be completely incompatible.
VRM header for Socket 7:

VRM header for P2/P3 (rev 8.2):

Getting the guts of that IBM module to work would likely be a significant job of re-engineering.

gdjacobs wrote on 2020-08-28, 05:26:

The pinouts appear to be completely incompatible. VRM header for Socket 7: Screenshot_2020-08-28_00-23-18.png VRM header for P2/ […]
Show full quote

Kreshna Aryaguna Nurzaman wrote on 2020-08-27, 16:34:

Pardon the necro. […]
Show full quote

Pardon the necro.

I recently bought XVI 073-20742-30 VRM, whose output voltage is 1,3-3,5V. At first, I was happy with the specs, since its output voltage seems to be able to acommodate a wide range of CPU (AMD K6-2, being 2.2v, falls into the range).

But much to my dismay, the VRM's input voltage is 12V instead of 3.3V, and I cannot cancel my purchase.

So, will it work with my Socket 7 mobo?

The pinouts appear to be completely incompatible.
VRM header for Socket 7:
Screenshot_2020-08-28_00-23-18.png
VRM header for P2/P3 (rev 8.2):
Screenshot_2020-08-28_00-20-38.png

Getting the guts of that IBM module to work would likely be a significant job of re-engineering.

I see. Thanks, buddy.

It seems every VRM I can found on ebay is 40-pin instead of the proper 30-pin suitable for Socket 7 motherboard (and I have to kick myself for forgetting to count the bloody pins before buying the item). Worse, 30-pin VRM are almost impossible to find. In fact, I haven't been able to find any.

What would you think about not using VRM at all? According to PTherapist, he used AMD K6-2 500 CPU on the exact same motherboard as mine, without any modifications to the board nor any special adapters. I assume he didn't use VRM either. It should be noted he also underclocked the CPU to 300 Mhz by setting the FSB to 50MHz. Otherwise the system becomes unstable in DOS.

PTherapist wrote on 2019-06-18, 04:45:

I have that exact same motherboard. I've been running an AMD K6-2 500 CPU underclocked to 300MHz for years, that's with FSB set to 50MHz. I have had it clocked slightly higher in the past, but it lead to stability issues in DOS (Windows was fine).

Over the years I've ran several K6-2 CPUs, a K6-III and several Cyrix CPUs. It worked for me with all of them. That's all without any modifications to the board nor any special adapters.

Oh and as already said, the cache on that board is definitely fake. I still haven't gotten around to getting a COAST module for mine yet.

So I assume AMD K6-2 (and K6-III, and Cyrix, for that matter) can tolerate overvolting, as long as you underclock the CPU. Am I correct?

Underclocking generates less heat so an overvolt can be countered by an underclock to some extend.
Too much of an overvolt will damage the CPU's transistors regardless. I wouldn't overvolt the 0,25 µm dies above 2.7V for long and only with good cooling. The 0,18 µm dies I wouldn't run above 2.3V for long. Maybe that's a bit conservative but I'm rather safe than sorry.😄
System stability also depends on your VRM's capability of delivering sufficient power. Lower clocks mean less power consumption, so maybe that's the reason for stability @50MHz fsb. I don't think any board with linear VRM design will be able to handle K6-III CPUs at their rated speed even though some will provide compatible voltages.

I wouldn't overvolt the 0,25 µm dies above 2.7V for long and only with good cooling.

Large lithography of that time is usually quite robust. AMD K6-2 (250nm) can tolerate default PMMX voltage (2.8v) without any problems.

I don't think any board with linear VRM design will be able to handle K6-III CPUs at their rated speed even though some will provide compatible voltages.

It's about VRM specs primarily. Linear VRM will just generate more heat and probably will require additional active cooling.

On my board I have the jumper for voltage regulator set to 3.3 Volt STD/VR. Doesn't let you go any lower.

I could never get the board to POST with the CPU at it's highest clock speed (think that may have been using the 2.0x setting or 1.5x, I forget now). With the 500, I could get it to run at around 333MHz but it wasn't stable with DOS games and higher than that was even less stable overall. 300MHz at 50MHz FSB was the best I was getting out of it.

I've since removed the 500 from that PC to use in a different system. In it's place I put in an AMD K6-2 400MHz CPU, which I played around with at different settings. It did not work correctly with the FSB set to 66MHz on any clock speed setting, the mouse would lag and keyboard would eventually lock up. Managed to squeeze 210MHz out of the 400 CPU at 60MHz FSB, though haven't tested DOS games on it yet.

Also should note on mine that the JP10 AT Bus Clock Selector jumper had no effect. It's supposed to be set depending on speed, ie. 50MHz or 60/66MHz, but didn't make a difference to stability.

Basically the K6-2 500 is probably the sweet spot for the fastest clock speed you'll pull from this board. If you could track down a faster 533 or 550 CPU, I have no idea how they would perform on this board. It's also been too long since I tried running a K6-III on it, so I don't recall how that was other than noting that the K6-III got fried by this board after using it for several months back in the early 2000s, so I probably wouldn't recommend trying a K6-III without some serious cooling.

A word of warning, those heatsinks immediately below the CPU will get blisteringly hot. For longevity I'd probably recommend adding a fan, though I personally never bothered and it ran for years like that and is still working well today.

The Serpent Rider wrote on 2020-08-28, 11:25:

It's about VRM specs primarily. Linear VRM will just generate more heat and probably will require additional active cooling.

Of course it's about VRM specs - I recently aquired an interposer board with a linear transistor rated for 10A typical/12A max - but I'm saying I don't think there are boards with linear VRM design made that would handle those CPUs. Motherboard manufacturers simply switched to switching regulators to deliver those currents. I'm happy to be proven wrong. 😉

PTherapist wrote on 2020-08-28, 11:56:

On my board I have the jumper for voltage regulator set to 3.3 Volt STD/VR. Doesn't let you go any lower. […]
Show full quote

On my board I have the jumper for voltage regulator set to 3.3 Volt STD/VR. Doesn't let you go any lower.

I could never get the board to POST with the CPU at it's highest clock speed (think that may have been using the 2.0x setting or 1.5x, I forget now). With the 500, I could get it to run at around 333MHz but it wasn't stable with DOS games and higher than that was even less stable overall. 300MHz at 50MHz FSB was the best I was getting out of it.

I've since removed the 500 from that PC to use in a different system. In it's place I put in an AMD K6-2 400MHz CPU, which I played around with at different settings. It did not work correctly with the FSB set to 66MHz on any clock speed setting, the mouse would lag and keyboard would eventually lock up. Managed to squeeze 210MHz out of the 400 CPU at 60MHz FSB, though haven't tested DOS games on it yet.

Basically the K6-2 500 is probably the sweet spot for the fastest clock speed you'll pull from this board. If you could track down a faster 533 or 550 CPU, I have no idea how they would perform on this board. It's also been too long since I tried running a K6-III on it, so I don't recall how that was other than noting that the K6-III got fried by this board after using it for several months back in the early 2000s, so I probably wouldn't recommend trying a K6-III without some serious cooling.

A word of warning, those heatsinks immediately below the CPU will get blisteringly hot. For longevity I'd probably recommend adding a fan, though I personally never bothered and it ran for years like that and is still working well today.

See, I am a bit conservative about Vcore 😅

Doornkaat wrote on 2020-08-28, 10:47:

Underclocking generates less heat so an overvolt can be countered by an underclock to some extend.
Too much of an overvolt will damage the CPU's transistors regardless. I wouldn't overvolt the 0,25 µm dies above 2.7V for long and only with good cooling. The 0,18 µm dies I wouldn't run above 2.3V for long. Maybe that's a bit conservative but I'm rather safe than sorry.😄
System stability also depends on your VRM's capability of delivering sufficient power. Lower clocks mean less power consumption, so maybe that's the reason for stability @50MHz fsb. I don't think any board with linear VRM design will be able to handle K6-III CPUs at their rated speed even though some will provide compatible voltages.

I see. What cooling would you suggest?

The Serpent Rider wrote on 2020-08-28, 11:25:

I wouldn't overvolt the 0,25 µm dies above 2.7V for long and only with good cooling.

Large lithography of that time is usually quite robust. AMD K6-2 (250nm) can tolerate default PMMX voltage (2.8v) without any problems.

Or, in PTherapist's case, it's 3.3v.

Kreshna, have you got an existing PMMX regulator module? If it's linear, it might be straightforward to modify the set point and add some cooling to the pass element.

PTherapist wrote on 2020-08-28, 11:56:

On my board I have the jumper for voltage regulator set to 3.3 Volt STD/VR. Doesn't let you go any lower. […]
Show full quote

On my board I have the jumper for voltage regulator set to 3.3 Volt STD/VR. Doesn't let you go any lower.

I could never get the board to POST with the CPU at it's highest clock speed (think that may have been using the 2.0x setting or 1.5x, I forget now). With the 500, I could get it to run at around 333MHz but it wasn't stable with DOS games and higher than that was even less stable overall. 300MHz at 50MHz FSB was the best I was getting out of it.

I've since removed the 500 from that PC to use in a different system. In it's place I put in an AMD K6-2 400MHz CPU, which I played around with at different settings. It did not work correctly with the FSB set to 66MHz on any clock speed setting, the mouse would lag and keyboard would eventually lock up. Managed to squeeze 210MHz out of the 400 CPU at 60MHz FSB, though haven't tested DOS games on it yet.

Also should note on mine that the JP10 AT Bus Clock Selector jumper had no effect. It's supposed to be set depending on speed, ie. 50MHz or 60/66MHz, but didn't make a difference to stability.

Basically the K6-2 500 is probably the sweet spot for the fastest clock speed you'll pull from this board. If you could track down a faster 533 or 550 CPU, I have no idea how they would perform on this board. It's also been too long since I tried running a K6-III on it, so I don't recall how that was other than noting that the K6-III got fried by this board after using it for several months back in the early 2000s, so I probably wouldn't recommend trying a K6-III without some serious cooling.

A word of warning, those heatsinks immediately below the CPU will get blisteringly hot. For longevity I'd probably recommend adding a fan, though I personally never bothered and it ran for years like that and is still working well today.

Thanks, PTherapist. Now, what if I use AMD K6-2 300 instead?

I will use the same setting you used; I will set FSB to 50MHz, so the CPU would run at 300 MHz. But unlike you, I will not underclock the K6-2, because it is already a 300 Mhz CPU anyway. As such, I will overvolt the CPU (because I will not use VRM), while not underclocking it, because it's a 300 Mhz CPU anyway.

Will I have problems?

The reason I'm asking this is because AMD K6-2 300 is cheaper and more plentiful than AMD K6-2 500 anyway. Well at least on Ebay.

gdjacobs wrote on 2020-08-28, 15:16:

Kreshna, have you got an existing PMMX regulator module? If it's linear, it might be straightforward to modify the set point and add some cooling to the pass element.

Unfortunately, I haven't. Where could I buy such module? Ebay search yields items like this, but how to use it with a Socket 7 mobo?

EDIT: I suppose I need to buy a 30-pin header connector like this?

Or, in PTherapist's case, it's 3.3v.

Nah, 3.3v is too much for K6-2 for any prolonged use. Short term benchmarking - sure. Staying on 50% over default voltage (2.2v) is dangerous. And realistically it would be not 3,3v, but 3,45v on single rail boards, so already over 50%.

The Serpent Rider wrote on 2020-08-28, 15:33:

Or, in PTherapist's case, it's 3.3v.

Nah, 3.3v is too much for K6-2 for any prolonged use. Short term benchmarking - sure. Staying on 50% over default voltage (2.2v) is dangerous.

Even underclocked?