PS3 Frankenstein PHAT PS3: CECHA with 40nm RSX

@major, @squeept, you guys seeing this? More in depth pics. @LeoTW630 maybe when major responds here you two can exchange some info to get a better understanding of how this is done. I srill don't think anyone figured this out yet.
 
Hi
My English is not very good so use google translation
Share the PS3 purchased by my friend
(Replace PS3 with 40nm RSX)

I hope everyone can understand

I own a cecha00 (Made in China) because of the YLOD problem, so I need to replace the TOKIN/NEC capacitor.
I think it is the problem of TOKIN/NEC capacitors, and now I want to buy the capacitors.
After I get it, I have to replace my PS3.

However, because of fear of RSX also having problems.
So I also bought the re-balled RSX to replace it.

In addition, is there a way to make a simple measurement to determine whether RSX is good or is there a problem?

A few days ago, my friend bought a cecha00 (Made in Japan)
When disassembling to replace the thermal paste, he found that RSX is 40nm (CXD5300A1GB)
He informed me about this problem
I also took pictures to show me
His cecha00 motherboard is COK-001 just like mine

I saw the picture
RSX was replaced with a 40nm package
I also found that there are other differences besides the 45-degree resistance

Then I found this discussion on this forum
So also provide the photos he gave me for reference

The model of cecha00 is also very precious
Hope to find out the correct way to replace 40nm RSX
After all, 40nm has less heat than 90nm


Attached below is a photo of my friend's PS3 motherboard
That sticker looks identical to the one on @lcferrum's console. Unless someone can think up another reason as to what that might mean we now know they were doing this as early as 2011.
 
You could do with high resolution scans of the motherboard really to compare. At this kind of quality:

View attachment 23799

Unfortunately I only have a COK002 here. Maybe someone with a decent scanner has a COK001 lying around. Linked a COK002 image anyway in case it might be useful to someone. It a 20MB composite made from 2 scans, over 14000x11000 pixels. I cant really get a good scan of the other side.

https://imgur.com/a/9meGXek


Edit: Imgur has scaled it down so I've attached a full res version here too:
http://xmbmods.com/temp/COK002_1.jpg
Does this require some special hardware? I tried using a high res flatbed scanner I have to do some PCB scans the results were less than stellar.
 
Does this require some special hardware? I tried using a high res flatbed scanner I have to do some PCB scans the results were less than stellar.
Not exactly, I did use a high end scanner (Epson V800) but I do not think it makes much difference.

I think the main issue that causes blurry image is there are a few components that lift the board up off the scanner bed too much, For my scan of the back of the COK-002 I had to unclip the power pins and pop those out through the other side (mostly) so I could take that scan with the board flat on the bed. I did not have to damage the board to do this but it was a little risky so would be better to do with a completely dead board.

Ideally someone would have an A3 size scanner, and a DEAD motherboard, then they could remove the components that raise the board up too much on the front too, like the USB ports and the sata ports. Then it would sit fairly flat and could be scanned on both sides.

The COK-002 image is still hosted here if anyone needs it. You will see the black hole where the power pins were on the right in the middle.

http://devil303.com/temp/COK002_1.jpg
 
Last edited:
Not exactly, I did use a high end scanner (Epson V800) but I do not think it makes much difference.

I think the main issue that causes blurry image is there are a few components that lift the board up off the scanner bed too much, For my scan of the back of the COK-002 I had to unclip the power pins and pop those out through the other side (mostly) so I could take that scan with the board flat on the bed. I did not have to damage the board to do this but it was a little risky so would be better to do with a completely dead board.

Ideally someone would have an A3 size scanner, and a DEAD motherboard, then they could remove the components that raise the board up too much on the front too, like the USB ports and the sata ports. Then it would sit fairly flat and could be scanned on both sides.

The COK-002 image is still hosted here if anyone needs it. You will see the black hole where the power pins were on the right in the middle.

http://devil303.com/temp/COK002_1.jpg
I did it on just REX-001 board and it still looked terrible as you can see. As far as I can tell, none of the components are nearly as tall as the ones on the COK-001 board yet mine still looks far worse than yours.
You can see that some components are razor sharp like the 470uF caps near the RSX/CELL, but the brown capacitors right next to them are out of focus.
 

Attachments

You are in the US yes? If it were in Europe
I would send you 1 or 2 European C models with artifacting/GLOD and a working L model board, which is what I have. If nobody else popped their head out quickly, which I very much doubt, I wouldn't even rule out international shipping.
Cheers and good luck
 
Just noticed there was some new activity here. @sandungas @M4j0r @littlebalup @DeViL303 or anyone else with some credentials that wants to take a crack at figuring out a 65nm CECHA01, just send me the two boards needed (make sure the 65nm is known working, but the COK-001 can be YLOD) and I'll swap the RSX and send back for free.
Thanks for the offer, but I wouldn't have the relevant skills unfortunately.

Hopefully someone like @littlebalup or @M4j0r can take you up on that if they have the HW knowledge and soldering skills. I would be great to see this upgrade available as service someday.
 
I cant help with this research either, i dont have the hardware and also im not so sure if is going to be required to reprogram syscon
The hardware side looks doable for someone like @squeept that have good reballing equipment and experience, the resistor and any other small hardware modifications should be easy (incase we are missing some other small hardware detail it should be easy to find it and replicate it)
But the real challenge could be reprogramming syscon, we are lucky the syscon keys was published, this open doors to this kind of modifications, but is still a dark area that needs more research
 
The syscon needs to be changed in order to support the newer RSX. You'll need a 302GB or newer.
First you have to dump the eeprom of the original syscon.
Then:
- Change the layout (Cookie old to Cookie new: https://pbs.twimg.com/media/ENs1zGGXUAIwnZl?format=png&name=large)
- Delete the complete patch (0x2800-0x2C00 and 0x4400-0x5000)
- Change the RSX revision byte (set 0x3912 to 0x21)
and flash it to the new syscon.
If you do it via hardware you don't even need any key.
 
The syscon needs to be changed in order to support the newer RSX. You'll need a 302GB or newer.
First you have to dump the eeprom of the original syscon.
Then:
- Change the layout (Cookie old to Cookie new: https://pbs.twimg.com/media/ENs1zGGXUAIwnZl?format=png&name=large)
- Delete the complete patch (0x2800-0x2C00 and 0x4400-0x5000)
- Change the RSX revision byte (set 0x3912 to 0x21)
and flash it to the new syscon.
If you do it via hardware you don't even need any key.
Nice info, so the software modifications are easy :encouragement:
 
- Change the RSX revision byte (set 0x3912 to 0x21)
Wait, im just thinking... after changing that byte it would be needed to update the checksum of the "board config" area, right ? (last couple of bytes in the area)
How is made that in PC ?, there is some tool to calculate the checksum in PC ?

I know we can access syscon by serial, and the syscon itself is able to calculate the required checksum (pretty cool feature btw) so is relativelly easy to fix it

But... is a lot better to do it in PC because that way what we are writing is 100% valid data

---------
I was curious about this also because, if at some point someone starts distributing syscon patches for the factory fan settings (intended to be applyed to the "thermal config" area of the syscon eeprom) that files should contain the correct checksum
 
Last edited:
Have not to much time for investigation this days but soon or later after this period of year I will for sure.
Few questions :
Syscon on cok002 models can be complete erased?
Tools for Renesas I have tnm5000 /teensy 2 which can do some generic 78RL(R5F100xxx) .Need to test also ps4 syscon.
I may have help from friends with car keys programmers if something similar would detect id of Renesas.
I have some friends help c++ /assembler.
If syscon from another console can be completely clear I may be able to write dump from another working console/patched?
I would like to test slim gpu on cok002 model. Seen many people trying to fix this type of units. I have about 6 all waiting for tests.
Is that resistor 10k?
Sorry not reading full thread.
 
I gambled and bought a Chinese 304GB syscon chip. Just gonna set it aside with a 40nm GPU for the next time I suspect I have a dead GPU on a CECHA01.
 
I gambled and bought a Chinese 304GB syscon chip. Just gonna set it aside with a 40nm GPU for the next time I suspect I have a dead GPU on a CECHA01.

I hope you got a real one, I already tried all sources and only got fake syscons (still waiting for one order to arrive). Both the flash variant and the 303GB and 304GB weren't even made by Sony, the markings didn't match at all. But I'm a bit more confident on the last order which hasn't arrived yet.
The risk of putting a fake syscon in a system is high, because you can almost break all components by that (or at least the SPI configuration ports).
 
Last edited:
I was thinking about refurbishing those 90 nm gpu ram by clearing with cnc and add new ram on gpu would not be much easier then hunting some sort of syscons hard to get? I will try this as well. Fixing what factory released. Ram could work rather than programming. Drilling by cnc as they done with icloud.
 
There's no option in getting one from a dead mobo? Or it has to be a specific serial number?

I was thinking about refurbishing those 90 nm gpu ram by clearing with cnc and add new ram on gpu would not be much easier then hunting some sort of syscons hard to get? I will try this as well. Fixing what factory released. Ram could work rather than programming. Drilling by cnc as they done with icloud.
Isn't the main cause of all of this the bgas under the die itself? If you want to reball those rams, wouldn't the die suffer the consequences?
 
I mean desolder gpu and remove with cnc memories and epoxy on the minimum edge of solder points. Add new 4 rams and reball back hole gpu. In most cases if I reball with manual profile on Jovi RE8500 it has to be slow and more than 10 minutes will get boad quite burned. If I use manual profile for slim or any lead free profile, board it looks as factory but ram epoxy crashes and split solder out. Only on cok002 and few sem001. Tried many time backing boards. Same results. Mostly any other cpu/gpu must be desoldered in maximum 8 minutes to get board as new.
Not that I can control well temperatures but done about 420 slim units since 2018. Report from my brother who sold them. I am working only on worksop not selling.
Mostly after desolder even if not bleeding I see ram points shorted. And 2 ohms on ic die power line. On 6 units cok002 bought one is starting and running fine in games, one crashing in games (XMB will stay forever on 68 with webman) rest didn't touch them for about 2 years, nor that I want. I do full reball on cpu/gpu because is an tandem process (experts know better) on every single unit. Process to complete full assembled unit 8 hours without software side which is done by my brother.
After this time of the year will open thread about reball with videos on every board. People will think twice before putting pressure on bga, delid on board without jig, reflow, etc.
 
Last edited:
Back
Top