Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Elmer who wrote (73743)	10/2/1999 10:39:00 PM
From: grok	Respond to of 1572953

RE: <Ahhh... the ASIC mentality! Of course that approach will work, if you can live with 5K-20K DPM. Nobody will accept that with RamBus. EP> If it were only a DPM issue then people could learn to live with it in manufacturing by providing lots of system testing and rework. But what about people who want to upgrade in the field? I guess that an escape plan for Rambus might be to drop RIMMs and require that Rdrams be soldered on mobos and no upgrading is allowed. The lack of connectors might allow the systems to work. Of course, reworking the ones that don't would be much worse.

To: Elmer who wrote (73743)	10/3/1999 1:11:00 AM
From: kash johal	Read Replies (1) \| Respond to of 1572953

Elmer, Now you've done it!!!! Us ASIC boys have nothing like 5-20K DPM. Seriously the validation of critical paths by simulation is fine and works day in and day out- AS LONG as the SIMULATIONS show that 99.9999% of parts easily meet the critical path. With Rambus clearly they have to test the critical paths and as has been suggested guardband it as well. I agree that testing it and ensuring that several different types of RIMMS work interchangeably seems like a tough task. When data is flowing at 800Mhz not much leeway for mismatches. I can even see shlocky RIMM connectors or lousy board manufacturing causing lots of problems even if everything is perfect. I recall that Intel went to the cartridge approach of implementing L2 cache directly with the CPU die and that cache was running slower than todays RDRAM. I sure hope they can fix the problems but they are certainly Non-Trivial. regards, Kash