My guess is that it's not a voltage divider reference, likely a non thermally controlled zener perhaps coupled to a switch mode regulator instead of a linear.

POD MMX documentation:
http://download.intel.com/design/archives/pro … cs/29060701.pdf

So if one were overclocking and wanted to give her a little extra juice, changing the voltage jumpers to 3.6 wouldn't yield a net increase?

That's my guess, but it's not made explicit by the documentation (3.6V by the way is the recommended upper limit from Intel). You could perhaps modify the CPU with a monitor lead on the regulated 2.8V output.

Baby steps. Currently testing it at 225Mhz (75x3, 30Mhz PCI) and it's running stable. The plan is to try 83.3x3 next. If it doesn't POST, I'll try 3.5v (the highest this mobo goes). If not, then oh well. Not important enough to risk the CPU.

Thanks, man.

My understanding is that the Overdrive chips (and definitely split voltage adapters for Socket 5/7) use "standard" voltage regulator ICs, which means it doesn't matter what the supply voltage is, it will always output the specified voltage (i.e. output is specified at 2.8V, output will always be 2.8V regardless if the input is 3.3V, 3.5V, or 3.6V). These types of ICs are the only really reliable way of getting the required voltage at the required amperage. Proportional voltage dividers (like resistor networks) just would not be sufficiently efficient to support the required amperage.

Unfortunately these types of regulators source as much current as they supply. More sophisticated DC-to-DC converters which preserve power would be much more costly and consume a lot of space on the package. This means that the higher the difference in voltage between what is supplied and what is ouput to the chip, the more heat that must be dissapated as waste. As the processor package is designed to accept 3.3V, the thermal properties of the package would be designed to dissapate the waste 4-ish watts that will be produced by the 0.5V difference between 3.3V and 2.8V (assuming 8~9 amp typical draw). Supplying a higher voltage means more heat to dissapate which the factory heatsink and fan might not handle adequately.

Well, linear voltage regulators are just a bleed transistor with feedback from a voltage reference, sometimes with some extra filtering. It wouldn't be hard for Intel to design their own solution if it's beneficial in terms of cost savings and device packaging requirements. It'd be hard to know for sure without popping the headsink off and having a look.

So wouldn't make sense for me to just jumper the motherboard for 2.8v to minimize unnecessary heat dissipation? The mobo does have a setting for 2.8...

Is it a socket 7? If the mobo can do 2.8v why do you need a pod?

And if it is a socket 5/7, why not use a normal mmx and run that at 3,3v? BTW, a normal 200mmx will do 3x83mhz at standard volts no problem.
I get 300mhz (3x100) at a super 7 board with 3,1v

It's outside the roughly 10% VCC tolerance band for operation recommended by Intel, so probably not worth trying.

clueless1 wrote:

So wouldn't make sense for me to just jumper the motherboard for 2.8v to minimize unnecessary heat dissipation? The mobo does have a setting for 2.8...

The excess heat might get outputted by the motherboard instead.

meljor wrote:

Is it a socket 7? If the mobo can do 2.8v why do you need a pod?

And if it is a socket 5/7, why not use a normal mmx and run that at 3,3v? BTW, a normal 200mmx will do 3x83mhz at standard volts no problem.
I get 300mhz (3x100) at a super 7 board with 3,1v

It's Super Socket 7: SiS 530 chipset. It's simple, really. I have the POD. I don't have a regular MMX. I was using the POD in my Socket 5 up until a few days ago. Now I'm testing it in the SiS 530 that I've had in storage for awhile. Actually, testing the POD200 head-to-head with a K6-2 at various frequencies. Clock-for-clock, the POD is stronger, I'm just evaluating the merits of the higher K6-2 frequencies vs the better flexibility of the POD.

gdjacobs wrote:

It's outside the roughly 10% VCC tolerance band for operation recommended by Intel, so probably not worth trying.

So best to leave the voltage at 3.3?

BTW, it's running fine at 250Mhz so far. Nice speed boost, but still have to test if the speed zones are pushed up too high.

clueless1 wrote:

gdjacobs wrote:

It's outside the roughly 10% VCC tolerance band for operation recommended by Intel, so probably not worth trying.

So best to leave the voltage at 3.3?

BTW, it's running fine at 250Mhz so far. Nice speed boost, but still have to test if the speed zones are pushed up too high.

I think so. I doubt the current draw is too big. At the end of the day, what's a few watts between friends.

Speed zones are effected running it at 250Mhz. The slowest it'll go is somewhere between a 40Mhz 386 and 16Mhz 486, so a little too fast for WC1.

Interestingly, even though Ultima 7 is immune to L1 cache disabling, it's not immune to the test registers DCD and CCD. I thought that these two in conjunction were equivalent to L1D, but it turns out that if I do L1D, Ultima 7 does not slow down at all, but if I do DCD+CCD, it slows down to just about perfect speeds. If I add BPD and VPD, it slows down a little more. So, something like
setmul DCD CCD = 486/33
setmul DCD CCD BPD VPD = 486/25

So, the TR12 capable CPUs are perfect processors for Ultima gaming?

gdjacobs wrote:

So, the TR12 capable CPUs are perfect processors for Ultima gaming?

At least Ultima 7: The Black Gate. I will post a TR12 chart in the Setmul thread when I'm done testing.

IIRC, Ultima V also ignored L1 cache settings. I think Origin did a lot of funky low level stuff in terms of platform, so I wonder if this setup might have further use in the land of Sosaria.