2 Phase to 3 Phase VRM

[Mustang]

There has been much discussion about VRM's on this forum recently. It lead me to thinking about the viability of upgrading an older motherboard to handle an Athlon 2500 XP. The board up for surgery is a MSI KT3 Ultra ARU. It has 2 phase power but space for the 3rd leg. The board has been recapped and is running the 2500 XP at 333 FSB which is an unsupported option in the latest BIOS. I just question its stability with only 2 sets of mosfets and think a 3rd would add a bit more power handling as well as stabilize the voltage peaks. Is this a feasable option?

Close-up of the board power circuitry. The arrow on the top shows where the HIP 6302 controller chip has been spreadeagled to fit the footprint of the much larger 6301 programmable multiphase controller. The arrowheads indicate missing components or more correctly components that were part of the reference design but have been eliminated by the (invisible on a red PCB) red pen (1: cap, 2: driver chip, 3: MOSFETs, 4: choke).
photo and description courtesy of lostcircuits.com

[gg1978]

Re: 2 Phase to 3 Phase VRM

That's interesting that they just stuffed 2 of the 3 phases on that board. Have you had a chance to put a scope on the VRM output while the CPU is under heavy stress?? I never thought the Barton core chippies were too demanding on core current, so i would think that 2 phase would be fine, provided there's enough output capacitance to keep Vcore from drooping too much due to the high di/dt's exhibited by all modern CPU's.

As for feasibility, as long as you're sure that the 6301 is the proper VR controller to go on the PCB, and make sure that the added FET's are a close match to the existing ones. I personally wouldn't bother, provided the system can hold up to at least 72 hours of continuous Prime95 stress testing.

[Oklahoma Wolf]

Re: 2 Phase to 3 Phase VRM

I never thought the Barton core chippies were too demanding on core current, so i would think that 2 phase would be fine, provided there's enough output capacitance to keep Vcore from drooping too much due to the high di/dt's exhibited by all modern CPU's.

I'm a little unclear on that myself - my AK79G-Max has a 3 phase VRM, yet when I push the Vcore to 1.85v and then push the CPU to 2.4 GHz my software readings go nuts and set off low 5v warnings. Then again, the board does have Lelon and Samsung for caps... can't wait to see how it does with Panasonic FM's all around.

[gg1978]

Re: 2 Phase to 3 Phase VRM

I've got a Tyan S2098 (i845GL, 400FSB only, flexATX) board that has a two phase VRM, and i've run a 2.6Ghz/400FSB Northwood on it, with no issues with Vcore. Of course the Tyan board has all panasonic. rubycon, sanyo, and a couple of small Teapo's on it. Not to mention the toroid inductors are a good bit bigger than of the board in Mustang's post.

Northwood P4 at 2.6Ghz is 1.5V at probably 60A or so, if i had to hazard a guess, without looking at the Northwood datasheet. If i want to really put a Socket A board i'm testing through the wringer, i've got a 1.33Ghz AYHJA Tbird that runs quite happily at 1.5Ghz, 1.8Vcore. I figure if the board can keep the power going to the Tbird, it'll hold up to pretty much any other socket A chip, except in extreme overclocking cases..

As for your software alarms, the onboard voltage monitoring chip's aren't exactly the most accurate things.. I would recommend checking the +5V with a DMM before assuming you've got that much of a voltage drop. Could be oxidation on your ATX connector too..

[Mustang]

Re: 2 Phase to 3 Phase VRM

There's quite a few MSI designs that skimp on the VRM circuit and just use larger caps to smooth out the power. I also have a newer MSI K7N2G with the same setup(2 phase). It can't even run a mobile at 200 FSB (locks up on boot).
I don't have a scope so no way to monitor Prime95 in DOS mode that I'm aware of. It probably wouldn't pass anyway.
The 6301 controller is 3 phase but the board has a 6302 which I'm not sure if it's able to handle 3 phases.
You're probably right it may just be a waste of time. Maybe just soldering larger caps might help it along enough.

[gg1978]

Re: 2 Phase to 3 Phase VRM

I've got an Abit NF7-S (one of the Abit boards that actually has all GOOD caps on it!!!), and it's 3 phase VRM, powered off the P4 4 pin connector. The mobile Athlon XP's work really well with it, even when overclocked. Of course the NF2/400 chipset was never known for extreme FSB overclocking, so the tweakable multipler on the Mobile XP is quite nice. I wish i would have bought about 2 or 3 more of the NF7-S's when newegg had them on sale.

[Fatal0E]

Re: 2 Phase to 3 Phase VRM

8 Phase power

http://usa.asus.com/products4.aspx?l...44&modelmenu=1

[Mustang]

Re: 2 Phase to 3 Phase VRM

I've got an Abit NF7-S (one of the Abit boards that actually has all GOOD caps on it!!!), and it's 3 phase VRM, powered off the P4 4 pin connector. The mobile Athlon XP's work really well with it, even when overclocked. Of course the NF2/400 chipset was never known for extreme FSB overclocking, so the tweakable multipler on the Mobile XP is quite nice. I wish i would have bought about 2 or 3 more of the NF7-S's when newegg had them on sale.

Same setup here, Athlon 2500 mobile on a NF7-S with all Rubycons .It's a great board
I can't get more than 190 FSB w/o stability issues on my kid's MSI board though. He has the same 2500 mobile as well. You get what you pay for although the NF7-S for $79 at Newegg was a bargain Wish I bought a few more too.

[Oklahoma Wolf]

Re: 2 Phase to 3 Phase VRM

I would recommend checking the +5V with a DMM before assuming you've got that much of a voltage drop. Could be oxidation on your ATX connector too..

I usually do it that way - the board can't even strain my Fortron 530W... still annoys me though watching the software and BIOS readings take that much of a dive when the overclocking starts. Goes from 4.82 all the way down to 4.44 with the PSU measuring a solid 5.09. I think the lowest I ever saw was 4.39.

Maybe the board's getting the readings after those Lelon caps... hope the Panasonic FM's help. If I can get them to fit that is (replacing 10mm with 12.5mm). Never really noticed that much fluctuation in Vcore even at 2.4 GHz.

[Mustang]

Re: 2 Phase to 3 Phase VRM

8 Phase power

http://usa.asus.com/products4.aspx?l...44&modelmenu=1

Very nice board. If ASUS would add 8 Phase to this board http://usa.asus.com/products4.aspx?l...87&modelmenu=1
with an Opteron, it'd be the top enthusiast board on the market today IMHO.

[Rainbow]

Re: 2 Phase to 3 Phase VRM

Check the temperature of the mosfets when the CPU is fully loaded. If they don't get very hot and there are no stability problems, I'd leave it as is.
See this PCPartner K333ASA-270 board - there are 5 mosfets in the CPU VRM. They're probably arranged as 2 phases somehow. And those 5 caps in a row are 4700uF so a total of 23500uF

[kc8adu]

Re: 2 Phase to 3 Phase VRM

look for a high resistance connection at the atx connector.
thats a big drop.measure after the atx.hint.whatever is causing the drop will get hot and will eventually burn.

[Per Hansson]

Re: 2 Phase to 3 Phase VRM

Mustang; The only real benefit except cleaner vcore you get from using more MOSFETS is that the load on them naturally decreases so they will not become so damn hot, saving the life of the capacitors around them...

I've seen MOSFETS stay at 100°C on some less genious designs by Epox a few years back, loaded with CPU's that of course did not exisist when the boards where originally built...

[Mustang]

Re: 2 Phase to 3 Phase VRM

Mustang; The only real benefit except cleaner vcore you get from using more MOSFETS is that the load on them naturally decreases so they will not become so damn hot, saving the life of the capacitors around them...

I've seen MOSFETS stay at 100°C on some less genious designs by Epox a few years back, loaded with CPU's that of course did not exisist when the boards where originally built...

That's basically why I'd like to add the 3rd phase. But as Rainbow suggests, I'll check the mosfet temps under load and if ok, will leave as is or pop in some higher rated caps.

[gg1978]

Re: 2 Phase to 3 Phase VRM

That PC partner board has quite a different layout that most VRM's do. Usually there's a pair of FET's for each phase, one's the top FET, that's connected to the VRM input, and the bottom FET, which goes to ground, with the area in between being the part that goes to the toroid output filter. The top fet pulls the voltage up to the input VRM voltage, while the bottom FET pulls it down. Hence, by varying the duty cycle of the FET's, the output current can change while holding Vcore stable.


The reason i say it's weird is that the three FET's right by the ATX connector appear to be "top" FET's but they appear to be connected in parallel, so they'll just work as one big FET, not three individual ones, while the other two appear to be a pair of "bottom" FET's, one per phase, which are right by the two toroid inductors. Hence, it's more like a single phase VRM than a two phase one, just going by looks alone. You'd have to check between the 3 FETs by the ATX connector and verify if they're in parallel to see if that's true or not.

I have seen VRM's that have more bottom FET's than top FET's, since the bottom FET's get more of a workout, since Vcore is so low compared to the VRM input voltage, especially with +12V input VRM's being common now.

[Rainbow]

Re: 2 Phase to 3 Phase VRM

The reason i say it's weird is that the three FET's right by the ATX connector appear to be "top" FET's but they appear to be connected in parallel, so they'll just work as one big FET, not three individual ones, while the other two appear to be a pair of "bottom" FET's, one per phase, which are right by the two toroid inductors. Hence, it's more like a single phase VRM than a two phase one, just going by looks alone. You'd have to check between the 3 FETs by the ATX connector and verify if they're in parallel to see if that's true or not.

I know that it's weird design - that's why I posted it here.
I examined it when I had the board but I forgot it since I only have a more detail scan of the board, hopefully enough to read the marking on the VRM controller chip.

[Mustang]

Re: 2 Phase to 3 Phase VRM

Found out a few things. The Intersil HIP 6302 is a dual phase controller..so much for adding caps and a choke.
3 caps by the VRM are 1500uf while the other 5 by the cpu are 2200uf. My fried ASUS A7N8X board is also dual phase but has 8 3300uf caps. Maybe go with that value.
Running Prime95 so far no problems. Mosfets are just warm to touch but I did put heatsinks on them.

[gg1978]

Re: 2 Phase to 3 Phase VRM

Found out a few things. The Intersil HIP 6302 is a dual phase controller..so much for adding caps and a choke.
3 caps by the VRM are 1500uf while the other 5 by the cpu are 2200uf. My fried ASUS A7N8X board is also dual phase but has 8 3300uf caps. Maybe go with that value.
Running Prime95 so far no problems. Mosfets are just warm to touch but I did put heatsinks on them.

THat's why the spot the 6302 is soldered to has some extra pads close by. The three channel PWM controller most likely has more pins, for the control of one or two more phases. I have seen some VRM controllers that support up to 4 phases, but they can be configured to operate less. I've got an Epox board i've been contemplating recapping for a while..

[Mustang]

Re: 2 Phase to 3 Phase VRM

The board ran 20 hours of Prime95 without an error. The wife needed to do a bit of shopping ...so that cut testing short.

Considering it's intended usage, It's probably stable enough. Thanks for all the tips and ideas

[Per Hansson]

Re: 2 Phase to 3 Phase VRM

I've seen MOSFETS stay at 100°C on some less genious designs by Epox a few years back, loaded with CPU's that of course did not exisist when the boards where originally built...

Just for kicks I wanted to see how much power I could get my comp to draw.

As a side effect of this the temp of my MOSFETS increased slightly, when I ran the system at 1.60v 2.3ghz they never went above 70°C. So I increased the vcore to 1.80v and CPU speed to 2.5ghz, the MOSFET temp recorded by the thermsistor I placed in between two of the 6 MOSFETS was 93°C! So it's a good thing Epox choose to use only high quality GSC caps for the 12v input and TEAPO for the output caps and then place the caps so far away from the MOSFETS... Mind you the VRM design on this board is 3 phase, Many of AMD's processors require 1.75v vcore so my little test is really not that unnatural... So if I where you I would keep a close eye on the temp of that 2 phase VRM design Mustang!