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 : CYRIX / NSM -- Ignore unavailable to you. Want to Upgrade?

To: Craig Freeman who wrote (24078)	2/25/1998 1:57:00 AM
From: Joe NYC	Read Replies (1) \| Respond to of 33344

Craig, if you look closely as the 90MHz benchmarks you will see that the speed of a processor does not improve linearly with the bus speed. I don't know what data on the benchmark you are looking at to arrive at your conclusion. The only thing that's anywhere close to comparing apples with apples is K6 at 2.5x90=225 and 3.5x66=233, where slower CPU on faster bus beats (by small margin) a faster CPU on slower bus. I would have run the tests differently from John Howland to see how increase of bus speed contributes to performance. With a special MB, heavy-duty heatsink and hand-selected memory what you can squeeze out of a CPU is, at best, about what you would expect from the next higher clock speed of the same CPU. As long as a significant percentage of the code can run from L1/L2 cache, you are right. But if the application is memory bandwidth intensive, the a fast CPU will just be waiting (very fast) for the memory. You will always get some improvement from a faster CPU, but for example a 1 GHz CPU Socket 7 CPU will not be 5 times faster than 200 MHz CPU. Of course, all depends on the application. There may be an application where you get almost nothing by increasing the CPU speed while memory speed is the same. I read a while ago about a company that is working on a database server running purely from memory. They claim 10 times greater performance compared to traditional database servers. But with everyone readying their 0.25u offerings at ever-increasing native speeds, the idea of trading cost and reliability for faster bus speeds and memory problems makes little sense to me. You don't have to be the first to go through the problems and price premium of faster memory. But in about 6 months all the problems will be ironed out, and there will be no reason not to go with 100 MHz bus. There will be an alternative approach. MXi will double the memory throughput by doubling the data-path to 128-bit Joe