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.

Technology Stocks : AMD:News, Press Releases and Information Only! -- Ignore unavailable to you. Want to Upgrade?

To: Elmer who wrote (6132)	5/13/1998 10:54:00 PM
From: Ed Sammons	Read Replies (2) \| Respond to of 6843

Re: <Clusters would not be effective with caches less than 1 meg as the bus would saturate at around 4 processors. > Elmer, I am still astounded by the price differntial. Of course an SMP board with 4 CPUs will approach saturation, and will undoubted saturate at 8 CPUs, 2MB cache no withstanding. I was comparing 1 or 2 450MHz-2MB CPU board w/ a 2 or 4 CPU 400MHz-2MB ($2700 each) or 450MHz-512KB ($1150 each). I am assuming the second and third examples will yield about 1.4-1.6 times the performance (Transactions/second) of the first, at similar or less cost. So if you were to design a system you could sell for the most money and spend the least to build (maximize gross margin), what would you build? Cache conflict is a nonissue in clusters since since each element (can be a multiprocessor element) has its own memory space, but memory sharing (e.g. message passing via gigabit ethernet) latency is. Software is available to balance the workload and also make make the cluster look like a single system image to the administrator.