PS3 Frankenstein PHAT PS3: CECHA with 40nm RSX

[squeept]

@Computer Booter Stars 922 thermal plaster is what I settled on and have been using for years. The caps on the tubes are terrible, so expect to only get one or two uses per little tube.

My MG order just got delivered, so I'm gonna play around with the epoxy that @RIP-Felix has been using. I vaguely remember comparing them before and going with Stars because the epoxy was too.... well, epoxy.

 

[squeept]

Anyone need some virgin practice boards for reballing? Cost of shipping rounded up.

I was just processing a pile of G and H models that I got super cheap to look for good 90nm donors. Lots of low ohm 3034's that nobody has touched before. Good for practice delidding, too. USA only please.

 

[sandungas]

Isn't Epoxy a complete mess-up in terms of a possible re-visit, if the new owner decides to re-apply some new paste at some point in the future?

Wouldn't do some basic thermal paste like MX2 instead of epoxy on the RSX-Ram chips do almost the same?
Maybe it drys out quick, fill the gap and may prevent the RSX-PCB from warping too?

I dont like to use the term epoxy neither resine for this case, because for me this kinf of "bi-component" products involves a very strong bounding (the result of most of this products is a material as strong as the rock)

...but the fact is there are some bi-component adhesives that i guess are based in epoxy/resine and the density of the material is a bit flexible... the problem when seleting this kind of products is we need to know very well the specifications of the exact product and the exact brand we are talkig about
I guess the reason why some of them are a bit flesible is because only contains a very small percentage of epoxy resine... lets say 10% (and the other 90% is silicone or whatever)... so it depends a lot of the chemical formula used by the manufacturer
The problem is we should not tell a newcomer simply to use a generic resine/epoxy because if he asks about it in a general store they are going to give them something super hard and the result is going to be a disaster (you are not able to unsitck it ever)

Time ago i bought a product from arctic cooling (i dont remember the name sorry) that was a bi-component in white color (and was smeeling like jummy strawberry) where the resulting material had the density of a blubblegum (a bit harder but clos, it was like rubber, you could press a bit of it with your fingers and it was returning back to his original volume)
It was like a paste to make your own thermal pads

Thats the main goal we need to achieve, because there is a tiny gap in between the RSX RAM chips and the RSX IHS... we cant use standard thermal paste in them because we know along time (when it gets dry) is going to reduce his volume, eventually will lost all humidity and will become dust... and eventually couold fall apart from the RAM chip
The kind of thermal adhesive i mentioned with the density of a bubblegum is perfect for that purpose, because is like a piece of rubber

RIP-Felix was going one step further because he want to improve the mechanical resistance by using something harder (less flexible than the bubblegum i mentioned)
In that case yeah.. you need to be really picky in selecting the exact brand/product suggested by others because you are trying to adjust that density very accuratelly
Otherway... i really think there are lot of silicones intended for industrial use that are going to do the work well and are cheap, the grey silicone used to seal car/bikes engines is fine, his density is good enought to fill the RAM gaps, is not going to decrease in volume, and is designed to handle high temperatures

Additionally... the RSX RAM chips doesnt seems to overheat by themselfs and can handle high temperatures, they could fail because the microbumps under them but i dont think the internal circuitry is going to be fryed
So all this considerations about the thermal compound used in the RSX RAM chips are not much important imo

 

[RIP-Felix]

Right, the thermal expoy should be considered a 1 time thing. That's why I like the idea of using a graphite themal pad, which will never loose performance. You still get the rigidity of the IHS to help prevent warping, without worrying about the paste underneath the IHS.

The epoxy I used does state the glass trasition temperature. So it's supposed get softer at a higher temp (not under normal operation), like the stuff SONY used. It is designed for this sort of application. Not like your store bought stuff.

I'm thinking now that it may not be necessary for the 40nm RSX. But for the 65nm it's a good idea. And it is definately needed for the 90nm RSX.

Silicone doesn't provide enough rigidity. It may provide some tho. That's better than nothing.

 

[Workz_777]

So i read the whole thread, page by page, it was a 3 day journey (in my spare time). I have a question about the voltage mod. Has anyone tried the Sony way for the voltage mod yet? i.e...

BD3504
R1 = R6216 = 3900 Ω 1% or 0.5%
R2 = R6222 = 1800 Ω 1% or 0.5%
R1'= R6214 = 3900 Ω 1% or 0.5%
R2' = R6219= 1800 Ω 1% or 0.5%

RSX PWRGOOD (mod)
R6205 = 27K Ω1% or 0.5%
R6207 = 10K Ω 1% or 0.5%

CELL PWRGOOD (mod)
R6111 = 27K Ω 1% or 0.5%
R6112 = 10K Ω 1% or 0.5%

It would be interesting to see how it compares to the "dirty" voltage mod version currently being employed. Would the Sony way provide better stability and less chance of YLODs after the 40nm has been installed?

 

[DeadEnd]

Well if you have read all of the pages then you should already see that Sony way is the only proper and 'safe' method. At the very least it would give more confidence in the swap working the way it should. Felix and others have outlined/theorized why Sony performed those steps. I imagine an electrical engineer could explain better if the "dirty" mod is breaking stability in some way. But of course, YLODs could happen for many other reasons even after the new RSX...

 

[Workz_777]

Well if you have read all of the pages then you should already see that Sony way is the only proper and 'safe' method. At the very least it would give more confidence in the swap working the way it should. Felix and others have outlined/theorized why Sony performed those steps. I imagine an electrical engineer could explain better if the "dirty" mod is breaking stability in some way. But of course, YLODs could happen for many other reasons even after the new RSX...

Indeed, that's why i was asking if anyone had tried the Sony way for the voltage mod? So far i think you guys have all gone with the dirty version, right?

Edit:- I was hoping someone might have tried the Sony way and we could compare the results.

 

[Workz_777]

I know @RIP-Felix had ordered some BD3504 recently and i was hoping maybe he has tried the Sony voltage mod by now, seeing as he ordered the BD3504 a month or so ago.

Also, i noticed @Computer Booter had done a few Frankies using the "dirty" voltage mod recently and a few YLOD'ed when everything seemed good to go. As pointed out by you guys, the Sony way allows for more ripple tolerance because it has a variable low and high voltage, compared to the fixed 0.95v coming from the "dirty" version.

 

[Computer Booter]

I know @RIP-Felix had ordered some BD3504 recently and i was hoping maybe he has tried the Sony voltage mod by now, seeing as he ordered the BD3504 a month or so ago.

Also, i noticed @Computer Booter had done a few Frankies using the "dirty" voltage mod recently and a few YLOD'ed when everything seemed good to go. As pointed out by you guys, the Sony way allows for more ripple tolerance because it has a variable low and high voltage, compared to the fixed 0.95v coming from the "dirty" version.

With all the stuff I have been doing, I've heard of the "Sony" way of doing the voltage mod, but haven't made the time yet to investigate and see what is necessary to perform it. If it is actually better, then that is the way I should be doing it!! Ill have to bother the boys or find the guide/process and try it out on my next frankster.

 

[truemaster]

sony change the 8 pin power regulator. i think find a regulator do the same thing is possible. but they also change come resistors and capacitors. those have been mesured but in circuit so there real values are different

 

[Workz_777]

With all the stuff I have been doing, I've heard of the "Sony" way of doing the voltage mod, but haven't made the time yet to investigate and see what is necessary to perform it. If it is actually better, then that is the way I should be doing it!! Ill have to bother the boys or find the guide/process and try it out on my next frankster.

Would be great @Computer Booter ...to compare and see if it helps the success rate. Here is Felix's guide from page 76:- https://www.psx-place.com/threads/f...cecha-with-40nm-rsx.28069/page-76#post-324546

Click on... [Spoiler: (Optional) VDDR Voltage MOD (Only for 40nm RSX on COK-00X)]

Then open... [Spoiler: (Optional) SONY's Method]

It has all the components and where they go with photos. The size of the resistors would be 0402 i'd imagine, and although Felix put 1% tolerance resistors, maybe get 0.5% tolerance ones, because looking at the Sony COK service manual pdf they tend to mainly use 0.5% tolerance resistors.

You will also need to find some of the voltage regulator IC Sony used, here's the part number:- BD3504FVM ... @RIP-Felix has some he ordered from before, maybe he can let you know where he got them from, or send you a few by mail for test purposes.

 

[RIP-Felix]

...the Sony way allows for more ripple tolerance because it has a variable low and high voltage, compared to the fixed 0.95v coming from the "dirty" version.

Not quite how it works. You are confusing 2 different voltages entirely.

VDDR is the one affected by the voltage mod. Sony's method was to replace the OG MOSFET driver IC with one that allows the voltage output to be selected with external resistors. The OG was fixed. The "hacky" way replaces the MOSFET with a linear voltage regulator. The MOSFET driver does nothing now. It has not MOSFET to drive anymore. Basically we are just using the MOSFET footprint for power and output pads. Thats why we lift pins 2-4. And a MLCC is needed for the regulator to "regulate" which is why it needs the jupmer to that cap. It will now deliver a fixed 0.95v the moment SYSCON enabels 12v. Which just so happens to be fine, because VDDR is swithed on shortly thereafter anyway. So it doesn't even remove much of the SYSCON's control.

The voltage thresholds are for VDDC not VDDR. They determine how much ripple is permissible without triggering a YLOD. The upper limit is always the same. The lower limit is the one that can be changed. Sony made a very hacky solution to the tokin issue by simply allowing more through. It makes glitches more likely. The processores may have to work harder to resend data that was lost between them. That could make the run hotter and shorten their life.

What they should have done is replace the tokins.

Because I don't want to encourage people to choose the same lazy way out, I deliberately chose not to provide resistor numbers and a tutorial for exactly what to do. What they shoud do is attenuate the noise. That's what the tantalums are for. It's a good fix for this issue. SONY's method for it should not be co sidered the "official" or "correct" way.

And yes @NewRetroRepair attempted it the other day, unsucessfully. I suspect one of those changes is causing his issue. He didn't even make it to BitTraining.

 

[vyktormvmpay25]

I'll test that on my first test unit. I just need to read all and watch again his video. My first Frankie I didn't get time to change vddr vreg by now. I'll do this blind as I already did full recap and no scope to tell exactly. If you want to prove is working I can risk that unit, is cok002, having here many is not big deal.
Let me know Felix.

 

[Workz_777]

Not quite how it works. You are confusing 2 different voltages entirely.

VDDR is the one affected by the voltage mod. Sony's method was to replace the OG MOSFET driver IC with one that allows the voltage output to be selected with external resistors. The OG was fixed. The "hacky" way replaces the MOSFET with a linear voltage regulator. The MOSFET driver does nothing now. It has not MOSFET to drive anymore. Basically we are just using the MOSFET footprint for power and output pads. Thats why we lift pins 2-4. And a MLCC is needed for the regulator to "regulate" which is why it needs the jupmer to that cap. It will now deliver a fixed 0.95v the moment SYSCON enabels 12v. Which just so happens to be fine, because VDDR is swithed on shortly thereafter anyway. So it doesn't even remove much of the SYSCON's control.

The voltage thresholds are for VDDC not VDDR. They determine how much ripple is permissible without triggering a YLOD. The upper limit is always the same. The lower limit is the one that can be changed. Sony made a very hacky solution to the tokin issue by simply allowing more through. It makes glitches more likely. The processores may have to work harder to resend data that was lost between them. That could make the run hotter and shorten their life.

What they should have done is replace the tokins.

Because I don't want to encourage people to choose the same lazy way out, I deliberately chose not to provide resistor numbers and a tutorial for exactly what to do. What they shoud do is attenuate the noise. That's what the tantalums are for. It's a good fix for this issue. SONY's method for it should not be co sidered the "official" or "correct" way.

And yes @NewRetroRepair attempted it the other day, unsucessfully. I suspect one of those changes is causing his issue. He didn't even make it to BitTraining.

So if the Tokins are replaced with Tantalums, then it's better to not do the voltage mod the Sony way? (because it was just a quick fix by Sony?) But what about the units / Frankies where the Tokins are left on the board, like some of @Computer Booter ones, then is it better to do the Sony method if they allow more permissible ripple without causing a YLOD?

Or just o-scope the Tokins to see if they need a Tantalum replacement? And never do the Sony voltage mod?

Just one thing, (maybe it's splitting hairs) when you say:- "SONY's method for it should not be considered the 'official' way", that confuses me because the definition of "official" reads like this:- "official = relating to an authority and its activities and responsibilities."

Surely Sony is the authority when it comes to their own Sony PS3, and are responsible for it under warranty, and so the 40nm Frankies done by Sony (under warranty) also counts as their activity. So truly defined, the Sony voltage mod is the actual official method.

 

[DeadEnd]

The voltage thresholds are for VDDC not VDDR. They determine how much ripple is permissible without triggering a YLOD. The upper limit is always the same. The lower limit is the one that can be changed. Sony made a very hacky solution to the tokin issue by simply allowing more through. It makes glitches more likely. The processores may have to work harder to resend data that was lost between them. That could make the run hotter and shorten their life.

What they should have done is replace the tokins.

Because I don't want to encourage people to choose the same lazy way out, I deliberately chose not to provide resistor numbers and a tutorial for exactly what to do. What they shoud do is attenuate the noise. That's what the tantalums are for. It's a good fix for this issue. SONY's method for it should not be co sidered the "official" or "correct" way.

And yes @NewRetroRepair attempted it the other day, unsucessfully. I suspect one of those changes is causing his issue. He didn't even make it to BitTraining.

Well, that is interesting take on it. But sounds like we need to test it more in real life use to see if glitches are in fact more likely or not.

I see your point , but I'm still leaning towards Sony's way. At least until it's proven to be harmful.

NewRetroRepair is still learning, he may have had other issues so it doesn't have to be related to the Sony's mod in any way.

So if the Tokins are replaced with Tantalums, then it's better to not do the voltage mod the Sony way? (because it was just a quick fix by Sony?) But what about the units / Frankies where the Tokins are left on the board, like some of @Computer Booter ones, then is it better to do the Sony method if they allow more permissible ripple without causing a YLOD?

Or just o-scope the Tokins to see if they need a Tantalum replacement? And never do the Sony voltage mod?

Just one thing, (maybe it's splitting hairs) when you say:- "SONY's method for it should not be considered the 'official' way", that confuses me because the definition of "official" reads like this:- "official = relating to an authority and its activities and responsibilities."

Surely Sony is the authority when it comes to their own Sony PS3, and are responsible for it under warranty, and so the 40nm Frankies done by Sony (under warranty) also counts as their activity. So truly defined, the Sony voltage mod is the actual official method.

Yea, I feel like this too. You could say they have been "lazy" about not wanting to replace the Tokins, but surely modifying the lower threshold had to be approved by a Sony Technician/EE who understands the circuit. Would they do it knowing that it could cause more glitches?

 

[NewRetroRepair]

Not quite how it works. You are confusing 2 different voltages entirely.
And yes @NewRetroRepair attempted it the other day, unsucessfully. I suspect one of those changes is causing his issue. He didn't even make it to BitTraining.

For posterity I'll add a small update on that. I replaced the first RSX with a different one. The first chip was slightly bent in one corner and nearly touching the board. I replaced it twice more with two separate RSX chips, both exhibiting the same problem as each other but different to the first I tried.

I undid the POWERGOOD modification and put the 15k and 20k resistors back in place.

The errors I'm seeing are now

[ERROR]: 0xa0202120
[ERROR]: 0xa0213013

RSX #2 and #3 are both showing this error so it's related to /something/ I did. I checked F6302 and have continuity. I checked for shorts around C6320 and that area is fine. I even tried replacing TH2501.

Presumably I've messed something up related to Cell. Cell VDDC resistances test fine at least. I'm in half a mind to reball it.


tldr; I don't think the "official" sony powergood modification is the problem.

 

[squeept]

For posterity I'll add a small update on that. I replaced the first RSX with a different one. The first chip was slightly bent in one corner and nearly touching the board. I replaced it twice more with two separate RSX chips, both exhibiting the same problem as each other but different to the first I tried.

I undid the POWERGOOD modification and put the 15k and 20k resistors back in place.

The errors I'm seeing are now

[ERROR]: 0xa0202120
[ERROR]: 0xa0213013

RSX #2 and #3 are both showing this error so it's related to /something/ I did. I checked F6302 and have continuity. I checked for shorts around C6320 and that area is fine. I even tried replacing TH2501.

Presumably I've messed something up related to Cell. Cell VDDC resistances test fine at least. I'm in half a mind to reball it.


tldr; I don't think the "official" sony powergood modification is the problem.

This is literally what I'm dealing with right now on #7. A0403034 3 second YLOD -> swapped in known good 40nm RSX -> 1 second YLOD with no error codes -> TOKINs were failing and rework pushed the CPU set over the edge, replaced all TOKINS -> A0202120 1 second YLOD -> replaced HDMI encoder -> A0202120 with A0213013 1 second YLOD.

CPU is off and everything is in the drying oven again for a little bit then I'll report back. I did notice before I removed the CELL that there was a tiny little spot on the top of the die that looks like it might be discoloration from a microscopic burn. We'll see, but I lost a lot of hope when I noticed that mark. I suspect the CPU shit the bed unfortunately.

 

[RIP-Felix]

For posterity I'll add a small update on that. I replaced the first RSX with a different one. The first chip was slightly bent in one corner and nearly touching the board. I replaced it twice more with two separate RSX chips, both exhibiting the same problem as each other but different to the first I tried.

I undid the POWERGOOD modification and put the 15k and 20k resistors back in place.

The errors I'm seeing are now

[ERROR]: 0xa0202120
[ERROR]: 0xa0213013

RSX #2 and #3 are both showing this error so it's related to /something/ I did. I checked F6302 and have continuity. I checked for shorts around C6320 and that area is fine. I even tried replacing TH2501.

Presumably I've messed something up related to Cell. Cell VDDC resistances test fine at least. I'm in half a mind to reball it.


tldr; I don't think the "official" sony powergood modification is the problem.

My guess would be FBVDDQ (RSX VRAM). How many 2120 does each error generate? 10x 2120 and 1x 3013? Cuz booter and I triggered that combo sabotaging FBVDDQ.

 

[squeept]

@NewRetroRepair @RIP-Felix no change with CPU reball, but upon the nudge from Felix (mine has the 10x errors), I'm re-examing some power circuitry. No shorts or opens sticking out, but I am getting close to isolating an anomaly on FBVDDQ (now that I'm REALLY looking close at it) that I don't believe to be caused by the RSX. I'll update if I find anything useful.

edit: I found the anomaly - me. I'm a dumbass. Left the power supply plugged in to the board (not the wall, I'm still not that dumb) off to the side and it was wildly affecting my readings while I poked around. So, plans for this one now are to sit on a shelf until I have time to swap the 40nm back to the original board and see if it's still good.

 

[squeept]

Looks like ^ was probably due to RSX destruction during delid. Those 40nm IHS are REALLY on there, and I must have cracked some RAM when pulling it. I tested it on the KTE board after delid but before removal. Heat from rework must have propagated some damage.