• Updated 2023-07-12: Hello, Guest! Welcome back, and be sure to check out this follow-up post about our outage a week or so ago.

Have you seen this chip? A minor CUDA mystery


Well-known member
TL;DR: Have you got a chip labelled 343S0150-B or CS4217-KL on your Power Macintosh's logic board?

Over the last week or so I have been at war with my SuperMac S900, which died an unceremonious death last year. The machine's current status is that it either, A: emits 50-100% of a chime but still does not boot, or B: emits pure static noise from the speaker.

I was warned that these machines tend to have bad CPU connectors so I cleaned it thoroughly and checked the continuity of every pin (this took a long, long time). Everything was connected properly with low resistance and no shorts so I am confident that the socket is not the root cause of this issue.

I set about trying to track down the source of the static noise I was hearing from the speaker. After a painstaking process I arrived at a chip that many of us dread, the CUDA... at least I am fairly sure that it is the CUDA. The chip either produces normal digital signals or descends into "analogue" chaos, triggering the staticky sounds of doom. I think I should swap out this chip.

For the uninitiated, the S900 shares most of its major components with the Power Macintosh 9500. The ROM, Bandit chipset, Grand Central I/O controller, and many more components of varying importance are identical and interchangeable between these logic boards. I was hoping that I could graft the CUDA from the 9500 board into the S900 but this is where I stumbled upon a minor Power Macintosh mystery: these Macs seemingly have different CUDAs. Or do they?

The CUDA in the S900 is a chip fabricated by National Semiconductor, labelled 343S0150-B.


The CUDA in the 9500 is a chip fabricated by CRYSTAL, labelled CS4217-KL.


It struck me as strange that two machines that are otherwise so similar would have different CUDAs so I set about figuring out which Macs did actually use the 343S0150-B. At first I really struggled. It is surprisingly difficult to find good photos of random PowerPC Mac logic boards. Then, as is often the way, a piece fell into the place that shed some light on the mystery.

The CS4217-KL chip's name was ringing a bell and I couldn't figure out why, until I realised I'd seen it before. It's the same chip found on the Performa 5400's logic board. Then it occurred to me that I'd actually seen it a second time on the "Typhoon" logic board from the SuperMac C500. A hunch followed by some web searching confirmed that this chip was also used in the PM 7X00 series.

This revelation fueled some further web searches and soon I stumbled upon the first instance of the 343S0150-B outside of the S900.


This is a Power Macintosh 6500 logic board. In the top right you can just about make out the CUDA's markings.


That's a match!

This discovery brings about a cascade of follow-up questions.

Why is this chip found on the 6500 and the S900, but none of those machines' close relatives (6400 and 9500 respectively)?

Why is the CS4217-KL found in a similarly random array of machines that are seemingly related to machines that have the 343S0150-B instead?

Finally: are these actually the same chip but produced by different manufacturers?

This is the point at which I realised I need some help from the community. If we can establish that these chips are interchangeable then it opens up an avenue for repair that may not have previously been possible. PowerMac 5400/6400 logic boards are common and cheap so being able to use them as donors for the much rarer S900 would be a huge help in the future. This leads me to my "big ask":

If you are reading this and you own one of the machines mentioned in this post (or any mid-90s PowerMac), please can you check which of these two chips is on the logic board?

I am hoping that we will discover that these chips were used "either/or" and are in fact electronically the same. All we need to confirm this hunch is two separate machines of the same model that have "mismatched" CUDAs.

This is a long shot but I am optimistic that we'll glean some useful knowledge regardless of the outcome. Thanks in advance to all who contribute.

It is also possible that I have identified these chips entirely incorrectly, in which case at least I will have learned something.