To: Bill Fischofer who wrote (89 ) 1/25/2000 11:48:00 AM From: Daniel Schuh Read Replies (2) | Respond to
One last post:I suspect the performance gap in real-world use will be a lot less than one might at first suppose. Clearly TM's emphasis on developing appropriate benchmarking awareness points to this. Most chips today spend much of their cycles bottlenecked on memory, I/O, busses, etc., so it is not inconceivable that Crusoe chips, which tend to have much larger caches since they don't need chip area for a lot of other hardware functions, will be very competitive in the performance department. transmeta.com , the one benchmark that shows any processor dependence in completion time is Office 2K, and the completion times listed are .133 hrs for Mobile PIII 500 mhz, .167 for the 6-months-out TM5400. Head scratching time there.transmeta.com , the TM5400 has 256k L2 cache, same as the PIII. More L1 for the TM5400, but it's still a little funny. The chip size is 73mm^2 for the TM5400 vs. 106 mm^2 for the mobile PIII, so the Transmeta chip is about 1/3 smaller. It's true they could have bigger caches if they make the chip as big as the PIII, but that would cost them in other ways.theregister.co.uk Sources tell The Register that behind the scenes, there are very low thermal K6-III microprocessors using copper interconnect technology being developed and perfected at AMD's Dresden part, in Germany.
