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 : Intel Corporation (INTC) -- Ignore unavailable to you. Want to Upgrade?

To: puborectalis who wrote (160014)	2/25/2002 8:22:44 AM
From: Dan3	Read Replies (1) \| Respond to of 186894

Re: As a result, Intel says the Xeons can operate up to 30% to 80% faster than earlier chips. What's up with that? Up to 30% faster or up to 80% faster? You can't have two maximums. In a single processor configuration, on typical code, it will probably run about 5% slower than a non-SMT chip. This isn't due to the chip, it's because most code is single threaded, and Microsoft's SMP kernel (required to run multiple threads) runs about 5% slower than the uniprocessor kernel. You can't switch kernels without running (at minimum) the utomp utility and re-booting the machine. That will take you from the uniprocessor kernel to the kernel that supports the dispatch of multiple threads. I am not aware of any way to go from the SMP kernel to the uni kernel. Some of the difference between the two kernels is due to concurrency checking that would be needed to keep the OS from getting corrupted by processes returning in an order that differs from that in which they were scheduled (which should be just as possible with SMT as with SMP). So SMT will require the slower kernel. This is probably why SMT is being activated only for servers (that are more likely to run concurrent threads), or SMP workstations (that already have to load the slower kernel to support the second processor).