Discussion Delidded the Wii CPU and GPU

Joined
Mar 29, 2021
Messages
9
Likes
9
Was interested in getting a shot of the silicon underneath the heat spreaders, and potentially improving thermals, so I took a razor blade, an old credit card, and a huge leap of faith, and was able to pop both of them off and get a nice gander at the silicon. Admitted there is a bit of scratching on the substrates that could have posed a risk to the thing even functioning at all, but I did a test afterwords and everything works just fine. And in the words of my mother, “ye pays ya money and ye takes ya chances.”

Should anything go wrong with the trimming and build of this portable Wii, I have a backup that I can use. Granted it is the power hungry revision 01, but better than nothing I guess.

Look at that tiny PowerPC G3 core, I’m genuinely surprised something that tiny can give so much power to the Wii, but I guess that’s the glory of RISC computing. And now the architecture is open-source so anyone can probably reverse-engineer the Wii CPU and put it on an FPGA.

I’m guessing that little thing above the GPU die is an EPROM of some kind.
4D820338-8BF7-42EC-81D9-89D2EFE13220.jpeg

FD61DB53-5BA6-44CB-8D94-F87C13E81C15.jpeg


And now I have a couple of badges of honour. :)
 

Stitches

2 and a Half Dollarydoos
Staff member
.
.
Joined
Feb 5, 2017
Messages
1,826
Likes
1,560
Location
Banana Bender Land, Australia
Portables
6
And now the architecture is open-source so anyone can probably reverse-engineer the Wii CPU and put it on an FPGA
I dunno, it took two enterprise grade FPGAs worth like $2k each working in tandem to fully replicate a SNES. It'd take a chonky array to replicate a Wii, but I'd love to see it
 
Joined
Mar 29, 2021
Messages
9
Likes
9
I dunno, it took two enterprise grade FPGAs worth like $2k each working in tandem to fully replicate a SNES. It'd take a chonky array to replicate a Wii, but I'd love to see it
Well I heard of a GitHub project called MiSTer which uses an FPGA board to emulate old systems based on the Motorola 68000, primary Atari ST, Amiga, and several arcade boards. Hell there’s an FPGA-based Amiga called the Vampire4 whose “68000” processor outperforms any real 68k ever made. But granted we’re talking mid eighties here and not early 90s like the SNES so maybe I’m wrong.
 

Stitches

2 and a Half Dollarydoos
Staff member
.
.
Joined
Feb 5, 2017
Messages
1,826
Likes
1,560
Location
Banana Bender Land, Australia
Portables
6
The SNES was also a bit of an odd architecture to replicate, the guys who did it ended up needing to sign an NDA to get access to the original firmware code for the SNES to be able to pull it off. Perhaps other systems might be easier to do?
 
Joined
Mar 29, 2021
Messages
9
Likes
9
The SNES was also a bit of an odd architecture to replicate, the guys who did it ended up needing to sign an NDA to get access to the original firmware code for the SNES to be able to pull it off. Perhaps other systems might be easier to do?
Yeah the SNES uses a modified version of the 16 bit modification of the 6502 which is already a modification of the Motorola 6800 (Two zeros, not three) so it is a bit of an oddball, although a company called Analogue sells FPGA-based SNESes that they claim are engineered with maximum compatibility in mind.

Indeed other systems with more widely used CPUs will be easier to do but I think the cutoff point is around 1995 because then CPUs increase dramatically in terms of transistor count and complexity and optical microscopes are no longer able to see the tiny transistors beneath all that silicon.

I just learnt that it would take the resources of probably entire decades’ worth of the combined GDP of several small nations to even reverse engineer a PowerPC G3 CPU from the late 90s and even then it would likely never match what’s outlined in the original proprietary VHDL/Verilog files nor the proprietary programs that translate that code into CPU silicon, much less the doping process for the conductivity of said silicon... so yeah I can probably kiss my dreams of reverse engineering these chips goodbye, although I probably still will get a couple die shots because there’s a certain beauty in that.
 
Joined
Mar 29, 2021
Messages
9
Likes
9
Huh, check this out.
There seems to be a little pinkish discolouration on the CPU die.
Not sure if this is because I turned it on for about a minute or so to make sure everything works before I trim the PCB (which it does fortunately), but I’m curious as to what this is.
2B9714E5-DD72-4E5D-BD04-87224C6ABE6F.jpeg
 

Benge

.
Joined
May 18, 2020
Messages
100
Likes
219
Location
France
Portables
2- O-Wii / BS2
Huh, check this out.
There seems to be a little pinkish discolouration on the CPU die.
Not sure if this is because I turned it on for about a minute or so to make sure everything works before I trim the PCB (which it does fortunately), but I’m curious as to what this is.
Maybe it is just the air contact with eat ?
Can't wait to see if it dissipates better the heat, if it's better than with heat spreaders, I think I'm going to do the same !
Normally it should be better because it removes a thermal resistance because of the old thermal paste which has dried well since time
In addition, it could make portable Wii a few millimeters thinner !
 

Benge

.
Joined
May 18, 2020
Messages
100
Likes
219
Location
France
Portables
2- O-Wii / BS2
Is there thermal paste on the die?
Yes obviously, it is not directly in contact with the metal, thermal paste it is use to fill the nano hole in the metal to increase the thermal conductivity.
 

Shank

Moderator
Staff member
.
.
Joined
Jan 31, 2016
Messages
1,167
Likes
2,343
Portables
4
Delidding for thermal improvements is about removing the bottleneck in dissipating the heat from the die of the chips. With 5-8 watts spread across 2 chips, thermal dissipation from the die of the chips themselves has never really been a bottleneck: it's mostly getting the heat off the heatsink and out of the case. The volumes of heat we are dealing with are so much smaller than high-end computing that we don't really have to face the same engineering challenges they do.

As for running Wii on an FPGA, I've discussed that with a few FPGA nerds, including a mister dev. Short answer is we are a long, long way from having consumer-affordable FPGAs with enough LEs to handle it.
 
Joined
Mar 29, 2021
Messages
9
Likes
9
Delidding for thermal improvements is about removing the bottleneck in dissipating the heat from the die of the chips. With 5-8 watts spread across 2 chips, thermal dissipation from the die of the chips themselves has never really been a bottleneck: it's mostly getting the heat off the heatsink and out of the case. The volumes of heat we are dealing with are so much smaller than high-end computing that we don't really have to face the same engineering challenges they do.

As for running Wii on an FPGA, I've discussed that with a few FPGA nerds, including a mister dev. Short answer is we are a long, long way from having consumer-affordable FPGAs with enough LEs to handle it.
Eh I mostly did it because I wanted to look at the silicon under a microscope. I’m amazed at how they managed to fit essentially an iMac G3 in such a small package, but I guess it shouldn’t be a surprise now since we have smaller and more powerful ARM CPUs.

So no iMac G3 on an FPGA either? damn
 

Aurelio

ᕕ( ᐛ )ᕗ
.
.
.
.
Joined
Mar 3, 2016
Messages
2,082
Likes
2,738
Portables
2
Eh I mostly did it because I wanted to look at the silicon under a microscope. I’m amazed at how they managed to fit essentially an iMac G3 in such a small package, but I guess it shouldn’t be a surprise now since we have smaller and more powerful ARM CPUs.

So no iMac G3 on an FPGA either? damn
You won't be able to see much from that side. It's just a slab of silicon. I had them scanned with an SEM and nothing was visible from that side
 

Shank

Moderator
Staff member
.
.
Joined
Jan 31, 2016
Messages
1,167
Likes
2,343
Portables
4
Cpu 30 is different. I hit the little chip on the left because I thought it's right
On the 30, the CPU (smaller, bottom chip) has the smaller substrate chip with the smaller package: Broadway-1. The die of the broadway was shrunk prior to this (even as early as some 01 boards IIRC) without changing the package. To my knowledge, power consumption between 01 boards with 65nm and 90nm CPUs have not been compared.

However, on the 30, the GPU (bigger, top chip) is NOT a smaller, die shrunk chip. This means it will consume more power than a 4 layer board, all of which have die shrunk CPUs AND die shrunk GPUs.
 
Top