Power to Wikipedia project

[henriok]

Hi!
Let me introduce myself.
I'm new to this board, in respect to being an active member, but have been trawling for several years. My name is Henrik, am in my mid 30's and I live in Sweden. I've been involved in the Swedish Mac community since the mid 90s and that's where my affinity for PowerPC processors comes from. My intresset deepened when I, as a hobby technology reporter, tried to do what the mainstream media couldn't, namely write sanely about processors. In the heat of x86 vs PowerPC, CISC vs RISC, I felt that that was needed. Much was known about the x86 side, but there were a lot of myths about the other side so I tried to read up on the topic, do the research, query people in the know and present a balanced view. Something that often seems too hard for journalists.

Many years later, I discovered Wikipedia, and I've beed trying to curate the POWER, PowerPC and Power Architecture related articles ever since. With much of the same goal. My mission now is more like that of a historian, trying to patch up the missing pieces of information of technologies that's been seen as obsolete for years or even decades.

My primary tool for finding information is Google, but finding product descriptions of parts in the 1990s is hard and is getting harder. There are a lot of white spots on my map concerning the early days, like the POWER1 and POWER2 families and the RS64 series.

One aspect of this collecting and organizing of data is something I think all of you can appreciate, and that the visual apperance of the processors. I've always thought that a good article in any encyclopedia should have illustrations and that's where my current effort lies. It'd be great to have a visual representation of processors on Wikipedia so that you can just look at it to identify it. The size, shape, color, surface details, codes, arrangement of components can be enough even if there's no text that's easily to decipher. And when it comes to IBM's more expensive parts, there's usually just a FRU code that often doesn't tell you much.. 80L6710 anyone? That's an unlidded POWER3 in a purple ceramic singe chip module. Up until now, Google just rendered one hit on that, and it was not this.

Wikipedia has a policy of not accepting any media that's not under a very liberal copyleft license. I just can't "steal" any picture I find and publish it on Wikipedia, but I can be inspired by photographs and draw my own version of what I see. So that's what I've done.

You can see my collection at Wikipedia and Flickr.

I'd appreciate comments on accuracy and ways to better represent the appearance.
I try to depict one actual chip, but most cases the picture is a hybrid of several due to lack of information in any one picture. Most are incomplete, with missing FRU codes and other valuable information, but I think that all contains enough information to do a proper ocular identification and that's always been the goal. And, I'm not after photo realism. It's ok for the images to look illustrated.

So, what's missing? Lots!
But we can start with the main core of IBM's server class processors and leave the plethora of Motorola/Freescale parts and stuff from fringe manufacturers like LSI, C*Core, Xilinx and AMCC be for this time.

Does anyone have OK pictures of these parts?
* A10 Cobra, predecessor to RS64, and in the A35/30 Muskie family, a SCM on white substrate.
* RS64 Apache. Haven't seen it. Probably SCM on white.
* RSC - Risc Single Chip. The one-chip version of POWER1.
* P2SC, PS2C+ - POWER2 Singe Chip. The one-chip version of POWER2.
* Any unlidded POWER4+. I haven't found any such in any configuration during my years on Google. There should be a SCM and DCM on white substrate and a QCM on brown.

That's all for now. I'l leaving some samples of my work here so you can quickly get what I'm about.
And yes, the POWER3 is aberco's and the POWER6 is the one that ZseriesFreak sold to Marcis.

I'll be grateful for any help I can get!
And I'll leave this gem here for you. Found it today: An A25/30 Muskie.

[isa-d]

I can help with some pictures

* A10 Cobra, predecessor to RS64, and in the A35/30 Muskie family, a SCM on white substrate.

* RS64 Apache. Haven't seen it. Probably SCM on white. not completely sure about this chip, I bought it already desoldered

* RSC - Risc Single Chip. The one-chip version of POWER1.

* PS2C+

[isa-d]

Power2+

[cooperalp]

Hey Henrik,
Your renderings look awesome, I had seen them on Wikipedia while I was searching for chip photos. Welcome to the forum.

[hardwarecop]

[aberco]

That's a cool collection isa-d, still hunting for some of these, and it's good to put a picture on those early RS64

On your Power2+ series the 2 smaller are not part of the actual CPU chipset, the smallest last one is the clock chip and the one before is the I/O chip.

[Vlasta]

Great initiative - lovely photos. Keep up the good work.

[henriok]

Thanks for the kind replies and the awesome photos isa-d! I'll put them to good use immediately!

I'm making these entirely in Photoshop using a multitude of techniques, much layering and much use of layer styles. By now, I have archetypes of many details, like substrates, capacitors, textures, lettering and metal markings so some designs comes quickly, but some really takes time. From 10 minutes to a full work day if I have to do it all from scratch. The hardest part is probably to get decent pictures that shows enough detail and isn't warped too much. I've uploaded a sample for you to look at.

[CPUShack]

POWER5+ QCM if needed
http://www.cpushack.com/2014/02/18/cpu-of-the-day-ibm-power5-qcm/

[cooperalp]

The NEC MCM may be an interesting candidate:)
http://www.cpu-world.com/forum/viewtopic.php?t=18613&highlight=nec+mcm

[henriok]

My latest crop.

I'm not particularly sure about these. The look of the lid and the lidless 90 nm version is pretty straight forward, but there's a bit of a uncertainty of how the pin-out is of the different parts. The three Cell/B.E. sizes should be pin compatible but I've found at least two different configurations. There might be more. Stencils for re-balling them exist in even more configurations. And I've yet to figure out with certainty which corner is pin#1.

I found four different part numbers for the Cell/B.E. parts:
90 nm: CDX2964GB
65 nm: CXD2981GB, CXD2989GB, CXD2990GB
45 nm: CXD2992GB
As these doesn't look at all like IBM's FRU codes, I'm assuming this is Sony's part numbering. I guess that IBM has regular FRU codes for the parts they are using, but I haven't been able to find any with certainty. And Toshiba supposedly have branded parts too, but I haven't seen any of those at all.

I haven't seen a lidless PowerXCell 8i either, but the configuration of the capacitors can be found in research papers studying the effects of the new packaging due to new DDR2 interface. The die is a bit larger, and the packaging too. It's not pin compatible due to a completely different pin out. I haven't seen any PowerXCell 8i from below.

It'd be great to have documentation of these parts that include descriptions about package size, pinout and part numbering, but I haven't found any. This surprises me a bit since Cell/B.E. was offered a a commodity part by IBM for use by those who wanted. It did see some use in third party products, such as military computers, accelerator cards and custom made supercomputers.

[CPUShack]

CXD = Sony, Digital (usually an ASIC) (CXA is Analog, CXK is memory, CXP is Processor)

[Guest]

ufff, ufff, else very good and excellent work at end.
**

[aberco]

Use of original IBM CELL is in the BladeCenter QS20, 21 and 22.
Could not find any IBM FRUs on the SONY parts.

[CPUShack]

with any luck I will have a QS22 for pics