Voltage Regulator Modules

[jdb2]

Hello. I've been experimenting with some extreme overclocking on an extremely obsolete platform. ;)
Specifically, an Athlon XP-M 2600+ and a DFI Lanparty NFII Ultra B. Now, you might ask why waste the effort. My answer is this is not my primary system ( I have a Pentium-M Desktop Replacement Laptop ), because I can, and I have an obsession/specialize in modding/hacking obsolete hardware. (kind of like the people that overclocked a 1(?)Mhz Z80 to something like 100MHz using LN2 :D ) Anyway, for various reasons, I need a better voltage regulator module than the one integrated into the MB. I've looked around for discrete modules, but I haven't found any that provide the power and voltage requirements I need. At this point if you're asking how one can replace an integrated VRM with a discrete one : male-to-female socket A prototyping socket, male-to-male socket A connector, wire. :) Anyway, I decided that another approach would be to use a used Abit NF7 just for its integrated VRM using the connection technique I previously mentioned. I could just keep the whole MB intact, but I'd rather de-solder and cut/saw off any extraneous circuitry. So, my question is twofold. Is it feasible to slice-and-dice a MB just for its VRM (leaving only the ATX power connector and the CPU socket, plus of course the VRM circuitry) and are there any discrete VRMs that can do up to, say, 2.5V at over 200W.

Thanks for any help,

jdb2

[WrigleyVillain]

Welcome to XS; you definitely came to the right place. Good luck :up:

[jdb2]

There are discrete VRM's for the K7, such as this http://www.fairchildsemi.com/pf/FA/FAN5240.html , but I'd prefer a 3-phase unit that can supply more than 2v . I'm not sure how much power this thing can handle but it probably isn't suitable for OC. Someone has to make a desktop K7 VRM.

jdb2

Edit: In the datasheet it says the max output voltage is 1.6v . :(

[jdb2]

A little googling goes a long way : http://www.google.com/search?hl=en&q=programmable+%222+phase%22+OR+%22two+phase%22+OR +%223+phase%22+OR+%22three+phase%22+controller&btnG=Google+Search

jdb2

Edit: It seems almost every controller has a max output voltage of 1.85v . Crap. :(

Edit #2: The HIP6301 controller used in the Abit NF7 says that it only provides up to 1.85v max output voltage. But, the BIOS lets one set the voltage to 2.0v . (or am I mistaken?) This must mean that the internal DAC continues to operate as normal when given a VID that's outside the range of the standard.

[jdb2]

Well, I think I found what I wanted : http://www.intersil.com/cda/deviceinfo/0,0,HIP6303,0.html .

I'd have to roll my own in terms of the rest of the circuit and it would require some glue logic, but it's pretty much spelled out for you in the datasheet. Heh, sawing the VRM circuitry off of a motherboard -- only I could think of something that retarded - born out of pure laziness. :p: :D

jdb2

[jdb2]

Googling again : http://www.google.com/search?hl=en&q=50mv+%223+phase%22+OR+%22three+phase%22+OR+%22mu lti-phase%22+controller+k7+OR+k6+OR+athlon+OR+k8&btnG=Search
reveals more options that can go up to and over 2v .

Since I've been answering my own questions, I have an actual question for the electronics gurus. Given that the HIP6301 chip in the Abit NF7 and in the DFI Lanparty NFII Ultra B has a max Vout of 1.85v, how the the heck do/did the motherboard manufacturers achieve a Vcore of 2v , at least as indicated in the BIOS?

jdb2

Edit: OK. I'm answering my own questions again. My brain must be out to lunch because the obvious answer is that the MB manufacturer fudges the voltage to the FB (feedback) input of the controller using a (variable) voltage divider eg. resistor + transistor . In fact, you *have* to do this with Intersil's general purpose part. Anyway, I think I have all I need to know. This opens up new possibilities in the volt-modding arena, unless this has already been done, but all I've seen are the Vdroop and OVP mods. Messing around with the FB input could give you a stable, rock-solid and accurate way of getting more juice.

Edit #2: The socket-A riser method I mentioned previously is not feasible nor is it worth the effort. People have added resistors and capacitors to the MB -- why not just take it to the next level and add another driver (or 2) and the required extra circuitry (caps, power transistors, resistors etc.) to get 3 or 4 phase power. It's not as outrageous as it sounds as the HIP6301 was designed to be extended this way and supports up to 4 phases. Most of the circuitry would be on a daughterboard with wires soldered to the MB. Also, the previously mentioned vmod to the FB pin would give you an extra 0.2v .