I love this story, hahaha...
IBM makes 1 of the 160 million transistors...unfortunately IBM has been having yield problems and has given precedence to AMD so the lines that do not exhibit these problems are not available to any other client other than Sony and AMD. This has delayed the core of the X800XT s TSMC is making the R420 without problems at a lower clockspeed with 1 quad disabled.
Emphasis mine
That's some seriously cool crap that ATi has going on: they're somehow able to add a single transistor to the other 159,999,999 transistors that are built into a single silicon wafer! Talk about kick-ass soldering skills -- at the nanometer level! w00t! Actually, that's even cooler crap that IBM has going on -- making ONE transistor at a time, and only selling those single transistors to Sony and AMD. Bwahahaha! Talk about hard-to-handle cargo in shipping!
FedEx: "Dude, why are you shipping a bucket of silt insured for $3 Billion?"
IBM Rep: "Um, that's not silt, that's four trillion individual 130 nanometer transistors! BE CAREFUL, you know how long it takes to cut a single 130 nanometer transistor out of a 12 inch wafer? The packaging is a beyotch too!"
No wonder their yield on those single transistors is so low
I also love that TSMC is having absolutely no issues making cores at lower speeds with one quad disabled. The only problem is, they don't BUILD cores with one quad disabled, they disable a selected quad after it's complete and it goes thru testing -- hence the bridge joining. TSMC
must be having fabrication problems if they're consistantly churning out cores with a completely disabled quad
FUD is FUN
Since both NVIDIA and ATI are having a LOT of issues shipping their top-of-the-line product in any quantities, and both are relying on Samsung's 1.6ns GDDR3 ram modules, I'm still taking the safe bet and assuming that memory shortages is what's causing the issues.
That, or maybe even PCB issues. The signalling noise of 1.2ghz memory might be hard to build a PCB around, ya know?
