PS3 Frankenstein PHAT PS3: CECHA with 40nm RSX

[RIP-Felix]

...would it be possible to nickel plate the contact area of the heatsink to make liquid metal okay to use?

Not stupid at all.

I think the answer is no. If you were to nickel plate it, yes it'd slow the penetration of LM into the aluminum, but it wouldn't prevent it. Eventually it would soak in a little bit and the Nickel plating would crack/flake as the aluminum underneath is attacked by the small amount of Gallium that penetrated. At which point the gallium would have full access to penetrate and destroy.

The reason copper is nickel plated is because the Gallium in LM amalgamates with the copper. It soaks in like a sponge and you have to reapply too often. Nickel plating reduces the amount of amalgamation that can occur, but it doesn't completely stop it. You'll notice that the Nickel plating gets "stained" by LM when you have to reapply in a few years. It slowed and limited the amount of LM that could soak in, but didn't completely stop it. In the case of copper it doesn't reduce the integrity of the metal. So the staining is no big deal. Just reapply when temps begin rising again. Once the gallium has penetrated as far as it's going to, it'll last longer before evaporating away.

That's why SONY globs the LM onto the PS5. They know a fair amount is going to soak in and be lost. So they need excess to compensate. That's why the devised such a complicated gasket method to contain the excess LM. It also makes a seal limiting evaporation.

 

[Srassyfrass]

Not stupid at all.

I think the answer is no. If you were to nickel plate it, yes it'd slow the penetration of LM into the aluminum, but it wouldn't prevent it. Eventually it would soak in a little bit and the Nickel plating would crack/flake as the aluminum underneath is attacked by the small amount of Gallium that penetrated. At which point the gallium would have full access to penetrate and destroy.

The reason copper is nickel plated is because the Gallium in LM amalgamates with the copper. It soaks in like a sponge and you have to reapply too often. Nickel plating reduces the amount of amalgamation that can occur, but it doesn't completely stop it. You'll notice that the Nickel plating gets "stained" by LM when you have to reapply in a few years. It slowed and limited the amount of LM that could soak in, but didn't completely stop it. In the case of copper it doesn't reduce the integrity of the metal. So the staining is no big deal. Just reapply when temps begin rising again. Once the gallium has penetrated as far as it's going to, it'll last longer before evaporating away.

That's why SONY globs the LM onto the PS5. They know a fair amount is going to soak in and be lost. So they need excess to compensate. That's why the devised such a complicated gasket method to contain the excess LM. It also makes a seal limiting evaporation.

Thank you much for the detailed reply!

 

[RIP-Felix]

My modchips came in today. Took about 2 weeks to get here. I bought the minimum qty of 5 for a total of $100 shipped. $20 each when all said and done.

I need to run errands today and make some meals for the week. Summer is over for me and I'm back to work full time starting tomorrow. So the earliest opportunity I have to attempt the frankenstein is next weekend. So Saturday is the plan.

 

[Dariush Aghaee]

Not stupid at all.

I think the answer is no. If you were to nickel plate it, yes it'd slow the penetration of LM into the aluminum, but it wouldn't prevent it. Eventually it would soak in a little bit and the Nickel plating would crack/flake as the aluminum underneath is attacked by the small amount of Gallium that penetrated. At which point the gallium would have full access to penetrate and destroy.

The reason copper is nickel plated is because the Gallium in LM amalgamates with the copper. It soaks in like a sponge and you have to reapply too often. Nickel plating reduces the amount of amalgamation that can occur, but it doesn't completely stop it. You'll notice that the Nickel plating gets "stained" by LM when you have to reapply in a few years. It slowed and limited the amount of LM that could soak in, but didn't completely stop it. In the case of copper it doesn't reduce the integrity of the metal. So the staining is no big deal. Just reapply when temps begin rising again. Once the gallium has penetrated as far as it's going to, it'll last longer before evaporating away.

That's why SONY globs the LM onto the PS5. They know a fair amount is going to soak in and be lost. So they need excess to compensate. That's why the devised such a complicated gasket method to contain the excess LM. It also makes a seal limiting evaporation.

I heard somewhere in psx place that the ps3's IHSs already are nickel plated. They are plated something but I'm not sure what. Also is there any element that wouldn't react to LM?

 

[DoublesAdvocate]

I heard somewhere in psx place that the ps3's IHSs already are nickel plated. They are plated something but I'm not sure what. Also is there any element that wouldn't react to LM?

They are plated with nickel... the material underneath is copper.
There are probably plenty of elements that wouldn't react with liquid metal but the majority if not the entirety of them are not suitable materials to use for a heatsink or IHS.

 

[RIP-Felix]

I heard somewhere in psx place that the ps3's IHSs already are nickel plated. They are plated something but I'm not sure what. Also is there any element that wouldn't react to LM?

Yeah, there is no issue using LM on the IHS/DIE interface. The problem is the Aluminum HS. We can't use LM on the HS/IHS interface.

Nickle plating the aluminum would be difficult. It would limit the diffusion of gallium into it, but not stop it. Gold is inert. It's possable that coating the nickel in hard gold after the nickle plating could work. That's essentially what the pins on cartrige connector are. I just don't see a feasable way to do that to the Heatsink. It might be cheaper to design a new Heatsink from scratch, one with vapor chambers and heatpipes in direct die contact. IDK.

 

[truemaster]

LM between chip die and ihs will be good. but we also need to see what precaution steps sony takes for the LM, the ps5 apu substrate is covered with a sponge like thing to suck possible exess, and avoid furter leak and propably a short circuit. the ihs to heatsink is a large area an mx-5 would be good. also since LM is much durable compared to a thermal paste. we can use a soft glue to keep ihs stuck to chip. by soft glue i meen a glue that will keep ihs steady on chip but will need a very litle force to delit.

 

[truemaster]

another think i want to ask tommorow my frankenstein cecha will arrive it has for psu the aps 226. i also have the aps 231 psu from a non bc phat. now that the console have 40nm would be the aps 231 be enough?

 

[RIP-Felix]

So I have given this some more thought and I think it would be possible to nickel plate the RSX HS.

I could dip it into the plating solution and evenly cover the entire surface. If I do it thick enough, it "might" give it sufficient protection. I am debating on whether or not I should do a copper layer after the Nickel. Then another Nickel layer after the copper, just for added protection from LM diffusion. If the LM penetrated the first barrier of Nickel, it'll soak into the copper layer and spread out like a sponge, but be contained there by the first nickel barrier. There is also a gold rub technique I could use to plate a very thing layer of gold onto it afterwards. That might be enough to create an effective barrier to LM diffusion. In theory.

EDIT:
It may also be possible to bend the heat pipes to get the HS to directly contact the die, so that the IHS isn't necessary. And if I can find a way to dissolve the thermal adhesive that's gluing the heatpipe inside the aluminum block, it should be possible to machine a copper block. I was thinking of bending the heat piped down to sit directly on the die. The copper block would be attached above the pipes using thermal adhesive. That's mainly to contact the RAM and apply pressure. The heat pipes would direct contact the Die and LM could then be used.

 

[RIP-Felix]

another think i want to ask tommorow my frankenstein cecha will arrive it has for psu the aps 226. i also have the aps 231 psu from a non bc phat. now that the console have 40nm would be the aps 231 be enough?

I'd stick with the APS-226. It's the most efficient PSU and a lower load will have a few beneficial effects. First, since there is reduced load, there'll be less ripple and noise. That'll increase the filter's effectivness (noise reduction). Second, the PSU will not get as hot. The higher the PSU spec, the better it's efficiency under load (less heat). The closer the load is to the PSU's maximum rating, the less efficient it is (more heat). Less heat inside the console = better. That's why Microsoft was smart to put the XB360 power supply on a power brick outside the console, where it doesn't contribute to the systems total heat budget. It's the same principal with computer PSU. You want calculate the total Watt's your system will draw and choose a PSU that can deliver double that. 80+ Gold or platinum is an efficiency rating. That's based on how much energy is wasted as heat and noise.

Less PSU heat/load means the PSU will last longer too. Fringe benefit to the Frankenstein mod.

 

[sandungas]

So I have given this some more thought and I think it would be possible to nickel plate the RSX HS.
View attachment 34911 View attachment 34912
I could dip it into the plating solution and evenly cover the entire surface. If I do it thick enough, it "might" give it sufficient protection. I am debating on whether or not I should do a copper layer after the Nickel. Then another Nickel layer after the copper, just for added protection from LM diffusion. If the LM penetrated the first barrier of Nickel, it'll soak into the copper layer and spread out like a sponge, but be contained there by the first nickel barrier. There is also a gold rub technique I could use to plate a very thing layer of gold onto it afterwards. That might be enough to create an effective barrier to LM diffusion. In theory.

EDIT:
It may also be possible to bend the heat pipes to get the HS to directly contact the die, so that the IHS isn't necessary. And if I can find a way to dissolve the thermal adhesive that's gluing the heatpipe inside the aluminum block, it should be possable to machine a copper block. Possably also use indium to join them. I thought about brazing them, but the ends of the heatpipe are soldered (I think). That would melt it and the vacuum would be lost. That would be bad.

Thats too much imo, have you thought in replacing the squared metal block ? (on that photo seems relativelly easy, in other heatsink models maybe not so much)

I mean... the problem of the liquid metal attacking chemically the metal block is happening because the metal block is made of alluminium... so replace the metal block to solve the problem at its root

The first challengue is how to separate the original squared metal block from the original pipes... but i see two ways to do it (i would use a combination of both)
1) by deeping it inside some aggresive solvent for some days (paint removal, etc... something derivated from petrol)
2) by using a metal saw and files to "cut" it in small bits

After that you can take a copper block (of higher height, this way you dont need the IHS) and again... a metal saw and files to "cut" it... and drill the holes + one of those tools to create the threads for the bolts in the holes



Edit:
An improvement of this method is (after you have the new metal block ready) you are going to need to "glue" it to the pipes... and you can cover the pipes completly with a higher quality thermal glue
The new thermal glue is going to result in a better thermal transfer between the metal block and the pipes than the original

 

[RIP-Felix]

Yes, that's what I mean. But Instead of putting the pipes inside of the copper block, I would place the block on top of the pipes. The pipes themselves would direct contact the die. Excuse the terrible photo shopping, this is just a representation of what I mean.

 

[DoublesAdvocate]

I hope you realize how pointless this all is, there's no point trying to optimize a cooling solution that has been shown over the years to simply not be good enough.
Fundamentally the issue is that there just isn't enough space inside the console for a heat sink and fan assembly that will cool it to a consistent ~50-60C while also leaving the fan speed at near idle.
We're talking about 160-220W of cooling capacity that needs to fit in an area roughly 22x20x3cm, while also including space for the fan...

 

[RIP-Felix]

I hope you realize how pointless this all is, there's no point trying to optimize a cooling solution that has been shown over the years to simply not be good enough.
Fundamentally the issue is that there just isn't enough space inside the console for a heat sink and fan assembly that will cool it to a consistent ~50-60C while also leaving the fan speed at near idle.
We're talking about 160-220W of cooling capacity that needs to fit in an area roughly 22x20x3cm, while also including space for the fan...

There are efficiency improvements to be made. 2-3C from LM on the die alone is a significant improvement. If we can get LM on the HS/IHS interface, or remove the IHS entirely, that would be significant.

 

[RIP-Felix]

PS3#7 - Update
(Frankenstein Attempt #2)​

Since I've already explained my process, I skip the details.

I got the RSX off just fine. I noted there are some missing pads, but @squeept did say this MB has been through a bunch of reflow cycles before I got it. And I reflowed once too. I mean, it was working before (with the pressure test). So I "assumed" they were already gone. Since I was already committed, I decided to continue anyway. I took a picture and figured I check it later. We'll see if that bites me later.

That's when disaster struck! I knocked the 40nm RSX just after it had flowed! I had just nudged the chip to ensure it had flowed and it did. All I had to do was lift the Hot air out of the way. I mised while grasping it and it knocked the chip, joining the balls and ruining the attempt.

That was my last NOS 40nm with per loaded balls. I only had 1 more, but it was some reballers old stock that he had pre-balled with lead balls. This chip ohm tested fine, but looked bad. A couple of balls were missing and the solder balls didn't look great. I had no choice but to try it anyway.

So I cleaned and prepped the MB BGA again. Sucks to do it once, let alone twice! The chip went on this time and flowed. Definately lead balls this time. They flowed at a lower temp. So at least he didn't lie. It sprang back after nudging, so I "hope" that means none of them joined underneath. Not that I'm confident at all!

The board is cooling now. I figured I'd look up those missing pads and write this post in the mean time.
Looks like the missing pads are...

• V39 (RSX VI10), No VIa.
  
• AV34 (FBC DBG), No VIA.
  
• AW22 (FLD O0), No VIA.
  
• AU20 (NC)
• AW15 (NC)
• AU16 (NC)
• BA6 (MOS I0), Does have a VIA!
  
• AW9 (PM DT06), No VIA
  
• AV9 (PM STRB), No VIA

Most of these are either not connected. The labeled ones not connected to a VIA don't connect to a trace or internal layer. So they are essentially NC. Not sure why they are labeled, but they are. That only leaves BA6 (MOS I0). It does have a via. I'm assuming it is Master Out Slave In. That's an SPI connection to communicate with something. Right next to it is MIS O0. Which I guess is Master In Slave Out. This is equivalent to RX/TX, just using different terminology. The question it what is this SPI port for? And will it screw me? I'm not entirely sure it's missing, it's hard to tell from my picture. It looks missing, but it might just be the angle.

What do you think? What are the chances this works? It feels like a lost cause at this point and I would have said there's about a 0% chance of it working, but I felt the same about the last Frankenstein. I was shocked when it fired up! So I'll give this one a 10% chance. That feels more than optimistic!

Next I need to install the MOD chip and resistor mod. I'd need to find another MOSFET from a slim to do the VDDR voltage mod though. I already grabbed the one from the only donor board I have (KTE-001). So either I need to remove it from PS3#8 (Frankenstein #1) or I need to find another. Are there more of the same type on a KTE-001? IDK, I haven't looked. I can test without it and it should work (if the mod is going to). Eventually I will need to perform it.

 

[M4j0r]

That only leaves BA6 (MOS I0). It does have a via. I'm assuming it is Master Out Slave In. That's an SPI connection to communicate with something. Right next to it is MIS O0. Which I guess is Master In Slave Out. This is equivalent to RX/TX, just using different terminology. The question it what is this SPI port for? And will it screw me? I'm not entirely sure it's missing, it's hard to tell from my picture. It looks missing, but it might just be the angle.

This is what connects to Syscon (and what the modchip intercepts). It's definitely needed.

 

[RIP-Felix]

This is what connects to Syscon (and what the modchip intercepts). It's definitely needed.

I read a short across VDDC anyway. So some of the balls must have joined underneath the BGA. I'd say that, combined with BA6 being missing, is the final nail in the coffin. Frankenstein attempt #2 is officially a failure!

Perhaps in the future I'll try my hand at microsoldering and pad repair, but I gotta say...I just do not enjoy this sort of work. It's nerve racking every time. It takes up way too much space. It's a long process. There are dangerous fumes and chemicals. I mean, this kind of mod work just plain SUCKS!

I still have the E01 that I may make an attempt on, but probably not until December.

 

[DeadEnd]

I read a short across VDDC anyway. So some of the balls must have joined underneath the BGA. I'd say that, combined with BA6 being missing, is the final nail in the coffin. Frankenstein attempt #2 is officially a failure!

Perhaps in the future I'll try my hand at microsoldering and pad repair, but I gotta say...I just do not enjoy this sort of work. It's nerve racking every time. It takes up way too much space. It's a long process. There are dangerous fumes and chemicals. I mean, this kind of mod work just plain SUCKS!

I still have the E01 that I may make an attempt on, but probably not until December.

I can subscribe to every word. I've been trying to do a few more of these mods myself, but there are way too many steps. There are too many nuances that can easily go wrong. Reballing a 40nm takes a lot precision, there is always something that causes at least one of the balls to roll no matter how steady you move your hot air gun. The fumes are awful and you still end up inhaling some of them no matter how good the air extraction is...As for the balls joining underneath the chip, that has happened several times even without knocking the chip until I finally readjusted the board and applied less flux. They usually merged around the corners. You can check the resistances based on a picture Vyktor once shared.

The mod is ridiculous work and for this reason I've changed my mind about doing it for others. I will just order a few japanese CECHAs and mod them when I feel like it. Then I can just sell them off.

 

[devbug_mind]

Another proud owner here. I own a PS3 (CECHA01 model) with an official refurbished (COK-001) board by SONY. I thought the PS3 model/board was just like any other (COK-001) board of the same console model. I found out that it was a very different board after checking the NAND dump logs with (PS3DumpChecker) and seeing something different or some error report.

Did some research about it minutes later and found a comment here which I believe was from the user littlebalup saying that such report or error meant that it was an official refurbished board from SONY. This was more than two years ago. I really wasn't aware of such RSX of that size on a (CECHA01) model even after opening the console to remove both IHS.

I've been lurking and checking this thread since it was created and finally decided to sign-up. I opened the console once again a few days ago to check if it did came with such RSX of that size and yes it did. Now I know why the temperatures of this model were very low. I'm not sure if I should create another thread or to continue here since I need some help with the board.

 

[marciolsf]

i think here would be great! I'm sure there will be requests for pictures and such, so it would help with documenting it. Hey, @sandungas, would it make sense to start a page on the dev wiki about the frankenstein? it would at least make it easier to consolidate data.