Power to Wikipedia project

[Guest]

Tried and failed on the green power6

[henriok]

I'm sorry, but thanks anyway. This picture and Aberco's will give me a complete picture of all the markings on the substrate.

[isa-d]

On the left a Power5 (or Power5+?) with an alternative layout


caricare immagini



free image upload

[aberco]

These are both Power5.

It seems that capacitors and dispensing of epoxy underfill is very variable on the Power5/5+

Here are my Power5, with an interesting one on a thinner substrate.

Then I was successful at opening a Power6 OLGA, CLGA and a Cell BE (opening the OLGA was very challenging!).

[Guest]

Can you tell me the method you use to successfully remove the top from the green PCB POWER6?

[aberco]

I used a dremel and a sawing disk.

It's not easy, a lot of care should be taken not to damage the organic substrate.

1. Make deep grooves on all of the red lines.
2. Use a flat blade screwdriver to pry the copper heatspreader off. Slide the blade in the groove you made and move it sideways. Gently pry the sides off first, then the part that cover the CPU die. Don't touch the part that has the L3.
3. Use the dremel again to mill the copper on the top of the L3 (orange zone). Mill untill only 2-3mm are left.
4. Gently lift one side of the heatspreader with a screwdriver. Don't use the organic substrate as a lever as you will leave marks on it.
5; Use a pair of plier to lift and roll the copper on top of the L3 die, like when opening a sardine tin box.
6. Repeat on the other side if needed.
7. Scrape off the grey silicone on top of L3 gently with your nail or very soft material. The die will get very easily scratched.
8. Use a wood stick to scrape off the black adhesive that remain on the organic substrate. I use a pointed bamboo stick under microscope, it does not leave any scrape marks.

[Guest]

Wow nice info, hopefully I find another chip sometime soon to try again.

[aberco]

Got a large lot of Wii, so here are some Broadway and Hollywood goodness

[henriok]

Excellent! You are getting pretty good at opening these!
Can you do a side by side comparison so we can get a good picture of how the size of the packaging of Broadway has changed?

Aaaand.. how cool is it that the Broadway B is named LUIGI!?!

[aberco]

Broadway and Hollywood are easy to open with a scalpel.

Henriok, if you want super highrez scans, I can place them on my server.

I now have an entire set to show!

BROADWAY 39X6735 (Canada)
BROADWAY A 43E5048 (Canada, Taiwan)
BROADWAY B 43E5070 (Canada, Taiwan)
BROADWAY-1 48J2662 (Canada, Taiwan)

HOLLYWOOD D10046F5 (Canada, Taiwan)
HOLLYWOOD AA C10073F5 (Canada, Taiwan, Korea)
HOLLYWOOD-1 D811301F5 (Canada, Taiwan)

Unusual for IBM, the last line of digits on both Broadway and Hollywood start with a datecode.

[aberco]

Well, I got a AS/400 9406-170 planar board, expecting to have something similar to the board seen on the left.

It is however different, I think I got RS64-III and than the left one sports RS64-IV.

[aberco]

I have been browsing to get info on the RAD750 hardened PowerPC 750.

Here's a nice photo I found of it, it is in CCBGA, with columns, and made by IBM (there's a part number on the substrate).

On boards, it seems to be used in conjunction with another BAE chip, larger and with white edges. This could be a radhard CPC700 or similar function.

Henriok, I think you used the latter to draw your model of RAD750.

[henriok]

Hello everybody! Long time no see!

But this is my latest crop: This is what I can gather a complete collection of the PowerPC G4 family, including the Core e600 series of Freescale, i.e. the continuation of this family after Apple abandoned PowerPC.

One very curious addition to this family is the version of MPC7400 that IBM fabbed for Apple when Motorola's yields were too low for Apple's liking. Motorola promised Apple 400, 450 and 500 MHz, but they couldn't deliver on that promise so Apple had to dial back to 350, 400 and 450 MHz until yields rose about a year later and IBM picked up the slack. IBM's version doesn't have a proper name, just IBM's internal FRU code: 06K5319. And since the die is smaller than Motorola's I would guess it uses IBM 0.20 µm fab, instead of Motorola's 0.22 µm. And it's consistent with the further die shrink of the MPC7410 at 0.18 µm.

The light green ceramic substrate is from when Motorola spun of its semiconductor business into Freescale. So they changed from purple to green ceramics.

Also.. the color of the dies are more or less what they look like in certain angles. The pink dies really are pink, and they reflect the copper CMOS fab that Motorola had at the time. When they moved to a copper SOI fab, the color changed to a neutral blackish tint.

[aberco]

Awesome as always!!!

[henriok]

I did a size comparison of the Broadway B and Broadway-1 and they are exactly the same (4×4 mm), and the original "Broadway" is somewhat larger (4.4×4.4 mm).

Aberco, would you mind remove the lid of one of the Broadway As? It might reveal a codename (just like "Luigi" for Broadway B) and will confirm that the As are of the same die size as the original.

There's a assumption (NeoGAF and Wikipedia, and therefore the rest on the Internet) that Broadway went through a die reduction due to a new fabrication process in 2009 (90 nm to 65 nm). If the B's size reduction is representative of this new process then IBM did the change in 2007 (which would fit since that's when Cell, POWER6, XCPU and z10 came in 65 nm variants).

The misconception is that they switched fab when they reduced the size of the package in the Broadway-1, but they didn't see the die shrink in the previous Broadway B. And they missed the date on Broadway-1 too. Aberco's images shows that this part begun production in 2008.