Ghetto Mods

[delshay]

Have you got an oscilloscope?

I'd look at the inductor waveforms, and then the gates on all the mosfets.

See if there's a gate resistor that's cracked, or open.

The 4 phases on that chip should be 90 degrees from each other, and it uses 2 mosfets per phase, but the 4 mosfets on the lower side won't run as hot, IMHO.

See how the mosfets are connected to the chip on P23 of the PDF file above.

Half the mosfets being cool may not be an issue; on the P5wd2 premium I'm working on, the lower mosfets aren't heatsinked, but I'm adding small ones.

If you hit the current limit for the switching, it will turn off the phase early; you can go too much capacitance.

Take a look at Table 10 in the pdf; see what kind of current limit it's running.


If the cables get hot from the PSU, you can buy pins and wire and make your own cables; they have pins that handle 16AWG wire, but each pin is only rated 9A.
That's why there are multiple wires per voltage. Most PSU's use 18awg cable.
Solder another connector on the bottom of the board, on a short pigtail, if you need more current than the pins can handle.

Here's a link to all the different power supply cables, pinouts, ans Pin numbers:

All about the various PC power supply cables and connectors

All about the various PC power supply cables and connectors

www.playtool.com

These pins are for 16awg wire:
Molex 39-00-0168 Fem WM3115-ND at digikey

Yes, I have a scopemeter. I will attach it sometime time during the week to see if all is ok. It can also be a faulty controller which I was not aware of as I got the motherboard off Ebay 2 years ago. I have spare controllers here if I need to change it.

The motherboard was operating more or less the same before i changed the MOSFETS, this why i think it's a controller fault.

 

[Grog6]

Compare the current sensing resistors to each other; there's apparently 2 per driver set, and if one of those has opened or changed value, it will alter that phase.

Since all the phases are feeding the same output line, they should all look pretty much the same; If I read it correctly, that chip uses VFO for regulation, changing the frequency to add power.

There's 68k of those chips at a place in hong kong, but I have bad luck ordering stuff from such places.
The last time I ordered anything from china, I got the parts, but the card was cancelled in 10 minutes for fraud other than the real transaction, lol.

 

[delshay]

Compare the current sensing resistors to each other; there's apparently 2 per driver set, and if one of those has opened or changed value, it will alter that phase.

Since all the phases are feeding the same output line, they should all look pretty much the same; If I read it correctly, that chip uses VFO for regulation, changing the frequency to add power.

There's 68k of those chips at a place in hong kong, but I have bad luck ordering stuff from such places.
The last time I ordered anything from china, I got the parts, but the card was cancelled in 10 minutes for fraud other than the real transaction, lol.

Resistors were binned before I installed them. It maybe possible I may of overheat one or two, but all checks show all resistors are ok. I will wait until I get the scopemeter in there to check. Performance has gone up in undervolt/overclocking so it maybe ok.

I get some of my parts from dead motherboards (de-soldering), then solder them to working motherboards to check all is ok. I already tested two controllers which I know is working.

MOSFETS get roasting hot with extreme undervolting 1.225v (2.6GHz), near shutdown temperature, but any fan across it will keep it cool, but still very hot.

Since you like PDF Docs here are the MOSFETS, (check-out the switching speed) & I know it's not for this application but it works. https://www.mouser.co.uk/ProductDetail/Infineon-Technologies/IPD70N03S4L-04?qs=sGAEpiMZZMshyDBzk1/Wi5ISJmYFIFJTw4HQ4M3X3E4=

 

[Grog6]

Those are pretty heavy duty; they shouldn't fail due to overcurrent.

The thing to watch when you replace mosfets is the gate capacitance, and the gate charge; you want them to be close (less than 2x) of the original mosfet's rating.

Otherwise, you need to change the gate resistors, and up the size of the bootstrap capacitor.

The bootstrap supplies the higher voltage for the high side mosfets, above supply voltage, and if they run out of charge the mosfets only turn partially on, melting them, lol.

(I built a fan controller last year, and the biggest power draw at startup was the bootstrap circuit. The gate capacitance on the 300A mosfets was 12nF, tho.

https://www.mouser.com/datasheet/2/196/Infineon-IAUT300N10S5N015-DS-v01_00-EN-1276063.pdf

I milled a stock copper base intel Heatsink to miss the source and gate pins, and soldered one mosfet to a heatsink, connected in a bridge setup.
Doesn't even need a fan, it's total overkill.)

I like pdf's; once an Engineer, always an engineer. I've built a bunch of this kind of stuff; the magnetics are the hard part.

If you saturate the magnetics, it's all over; it'll pass 12V to the load. That's pretty spectacular.
That's the bummer with a buck converter.

That's cool if you have other converter chips; those aren't too hard to remove/replace.
Kapton tape shielding all the stuff around it, and a focused heat gun is what I use, mostly.

If it's too close to other stuff, cutting all the pins off flush with the case and pulling them off the board with the tip of an iron works, but the chip is wasted.

 

[delshay]

Those are pretty heavy duty; they shouldn't fail due to overcurrent.

The thing to watch when you replace mosfets is the gate capacitance, and the gate charge; you want them to be close (less than 2x) of the original mosfet's rating.

Otherwise, you need to change the gate resistors, and up the size of the bootstrap capacitor.

The bootstrap supplies the higher voltage for the high side mosfets, above supply voltage, and if they run out of charge the mosfets only turn partially on, melting them, lol.

(I built a fan controller last year, and the biggest power draw at startup was the bootstrap circuit. The gate capacitance on the 300A mosfets was 12nF, tho.

https://www.mouser.com/datasheet/2/196/Infineon-IAUT300N10S5N015-DS-v01_00-EN-1276063.pdf

I milled a stock copper base intel Heatsink to miss the source and gate pins, and soldered one mosfet to a heatsink, connected in a bridge setup.
Doesn't even need a fan, it's total overkill.)

I like pdf's; once an Engineer, always an engineer. I've built a bunch of this kind of stuff; the magnetics are the hard part.

If you saturate the magnetics, it's all over; it'll pass 12V to the load. That's pretty spectacular.
That's the bummer with a buck converter.

That's cool if you have other converter chips; those aren't too hard to remove/replace.
Kapton tape shielding all the stuff around it, and a focused heat gun is what I use, mostly.

If it's too close to other stuff, cutting all the pins off flush with the case and pulling them off the board with the tip of an iron works, but the chip is wasted.

I'm very good with soldering this is why even 0402 resistors is easy to change.
I used Kapton tape once, & I hate it. I use steak pie foil which you can get from the chip shop. One side of the foil is silver & the other side is gold in colour. Face the gold colour of the foil facing the infrared lamp to reflex the heat. I'v been doing this for years even on my R9 Nano here on TPU.

It must be cheaper than Kapton tape & you get to eat the pie too. The foil works best for me, & it's quicker too. Even near the PCI-e slot, zero heat damage, not even a hint of damage or dis-colour. A single layer of foil is all that is needed, but if I am unsure, I add a second foil to the area I want to protect. The foil I have here must be over 6 years old & still going strong. You need around four of them, so you need to buy four pies.

 

[Grog6]

"steak pie foil" must be a UK thing; I can't find anything but aluminum foil coming up in a search.

Got a link? I'm all for new stuff.

Damn; now I'm wanting fish and chips, with a side of Guiness, lol.

 

[delshay]

"steak pie foil" must be a UK thing; I can't find anything but aluminum foil coming up in a search.

Got a link? I'm all for new stuff.

Damn; now I'm wanting fish and chips, with a side of Guiness, lol.

I found this out by looking at the lamp on my Aoyue BGA 9000A after a few months of using it. It uses a gold colour reflector behind the lamp. I think you will find all infrared workstation has this. I got mine from my local chip shop. Look at the pie if it's silver in colour it will still work but not as good as the gold one. The foil absorb the heat spreading it, but the gold ones reflex some heat back.

 

[silkstone]

"steak pie foil" must be a UK thing; I can't find anything but aluminum foil coming up in a search.

Got a link? I'm all for new stuff.

Damn; now I'm wanting fish and chips, with a side of Guiness, lol.

 

[delshay]

View attachment 136179

That's it. Always face the gold side facing the lamp or pre-heater & bend the foil around the area you want to protect. You can also use the foil at the preheater end to block heat. This is how I stop BGA R9 Nano chip from de-soldering when removing the capacitors, which are very near the chip.

in-short you can also use it to block off heat from the pre-heater also. So you don't have to heat the complete bottom of the PCB. Tear the foil in half if you don't have enough or if you are working on something very small.

When blocking heat from the pre-heater I place the foil on the pre-heater itself, not on the PCB. So part the pre-heater is blocked off with foil.

 

[Grog6]

Ok, I understand.

 

[erixx]

The provided MSI support arm was useless due to flexing of case structure (just a little, enough to make it useless.)
So I digged in my nest and found something (its an old time invoice-pin plus an audio adapter):

 

[BrainMuncher]

Got a shitty mouse with a cable so stiff you could dig a hole with it? Easy solution, fold it hard putting sharp kinks in it in a zig-zag pattern. This eliminates almost all stiffness and drag, cable gives way instead of pushing the mouse.

 

[Ruru]

Got a shitty mouse with a cable so stiff you could dig a hole with it? Easy solution, fold it hard putting sharp kinks in it in a zig-zag pattern. This eliminates almost all stiffness and drag, cable gives way instead of pushing the mouse.

View attachment 136254

Damn, that's pretty genious to be honest.

OK; I ended up using the big iron, a heat gun as a preheater, and my biggest soldapult solder sucker, and I cleared about half the holes.

"Thermal Armor" indeed, lol.

I've used the method you mention, but with multilayer boards you run the risk of pulling the barrels out of the via holes.

If the connection to the interstitial layers goes bad, the board is dead.

At least these don't use the "Shadow Process" to make vias, those would be gone already.

Isn't the Thermal Armor just a plastic shroud? At least with Z87 Sabertooth it was.

 

[Grog6]

You may be right; there was a label that mentioned thermal armor on the back side of the PCB, but I think I'm talking about this:

"Stack Cool 2
Stack Cool 2 is a fan-less and zero-noise cooling solution. It effectively transfers heat generated by the critical components to the other side of the specially designed PCB (printed circuit board) for effective heat dissipation. "

Either "Words" work for me, lol.

The reality of it was a thick plane layer with no thermal reliefs at the pads, making them almost impossible to get out.

I cut the plastic "detents" that keep you from putting a socket 771 in this bad boy, I still need to apply the "Adapter" to the Xeon chip.

Then I have to find my HS compound; I found the 20yo jar of the white stuff; It's still good, lol.
I won't be using that.

I removed and lapped all the other heat sinks on the PCB; they're like tiny mirrors.

I'm putting 2 kilos of Thermatake heat pipe heatsink on the xeon. Tower 112

It's already lapped; I hate to lap all the info off the Xeon, but...

 

[jaw shwaa]

So this is an older mod I did . But ghetto non the less.

I got a cool-it aio second hand that only came with an intel mount and I wanted to put it in an amd rig. (it stayed through a couple upgrades all the way from a 790 board , to a 970 fx build)
So whats the only thing laying around thats the same dimensions as an amd mount? An amd back plate of course . the pics show what happened next . This is probably the quickest mod ive ever done .please ignore the dirty copper XD .

 

[Deleted member 185158]

6950 with heat issues, overclocking near impossible.
Removed the stock cooler and found bad coverage. Some burnt.
Used old Swiftech CPU waterblock. Drilled a couple of holes and mounted to the card.
OC increase ability 200%.

 

[lexluthermiester]

6950 with heat issues, overclocking near impossible.
Removed the stock cooler and found bad coverage. Some burnt.
Used old Swiftech CPU waterblock. Drilled a couple of holes and mounted to the card.
OC increase ability 200%.


View attachment 136513View attachment 136514

Oh yeah, that'll be a problem.

 

[Deleted member 185158]

Oh yeah, that'll be a problem.

Not any more!!

Water Delta is 9c (tad under) that GPU is 32c load at 1300mv!
Base clock 800, now clocks 1050+.

 

[jaw shwaa]

6950 with heat issues, overclocking near impossible.
Removed the stock cooler and found bad coverage. Some burnt.
Used old Swiftech CPU waterblock. Drilled a couple of holes and mounted to the card.
OC increase ability 200%.

ha! two 6950s in a row , I painted the red on mine blue . have you tried unlocking the card to a 6970 yet? I still have an actual 6970 water block from ek on my display shelf with the cards

 

[Deleted member 185158]

ha! two 6950s in a row , I painted the red on mine blue . have you tried unlocking the card to a 6970 yet? I still have an actual 6970 water block from ek on my display shelf with the cards

I did try to unlock it, GPU-Z reports 1408 shaders so it doesn't take hold.
The Max O.C. increase did hold, where I can top out memory at 1600mhz which is useful.

 

[lexluthermiester]

Not any more!!

Water Delta is 9c (tad under) that GPU is 32c load at 1300mv!
Base clock 800, now clocks 1050+.

Nice!

 

[freeagent]

Burnt TIM.. impressive!

 

[jaw shwaa]

I did try to unlock it, GPU-Z reports 1408 shaders so it doesn't take hold.
The Max O.C. increase did hold, where I can top out memory at 1600mhz which is useful.

thats a bummer , I"ve had 3 6950s and they all unlocked no problem.

 

[Deleted member 185158]

thats a bummer , I"ve had 3 6950s and they all unlocked no problem.

It could be that GPU-Z just doesn't read it correctly??
Not sure if one could even tell the difference with 140? shaders (roughly) extra.

I'll try to get a better picture of the waterblock. That one is crappy.

 

[micropage7]

Just bought mousepad online and the stupid seller packed it in small box and squeeze the mousepad into it and create unfixable dent, than throwing it into bin i use it as doormat