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 (30315)	11/19/1998 7:30:00 PM
From: Joe NYC	Read Replies (1) \| Respond to of 33344

Craig, Where is the bottleneck? Good question. I don't have an answer. I can speculate, though. I think a lot has to do with Windows and it's memory management. You would expect that your hard disk should be almost unused when you have a lot of memory, and you think all you need is loaded, but a lot of times it is not. I notice some delays in disk activity when switching applications. Also, this whole registry thing is a nightmare in my opinion. I don't think it is cached at all. A lot of important stuff gets stored in registry and it is possible that when applications need some value, the computer may have to perform a disk activity. Just take the browser (IE. I am not sure Netscape had this problem.) When I start to type the url address, even though the computer has some MRU list, it never seems to be in memory. The same thing with the bookmarks in IE. A lot of times I get a delay and I hear the hard disk. It takes some time to open sub-menus. When I go back to the same sub-menu immediately, sometimes it pops up instantaneously. Other times it doesn't for no reason. I tend to have a lot of browser windows open at the same time. When I go to the bookmarks menu of another window, always seems to re-read them from the hard disk. Another big reason why you get only tiny increases in performance with increasing clock speed is memory latency. Anytime you have a L1 and L2 miss, you have to go to main memory which takes a set amount of time. So a 200 MHz CPU will wait for example 25 clocks. 400 MHz CPU will wait 50 clocks. I don't know how much Microsoft took this into consideration. Probably not at all. Why bother. It's not like people will choose BE OS over Windows because of this. According to one of the Cyrix presentations, the CPU spends 60 to 80% of time in the operating system. If the operating system produces a lot of cache misses, your high priced CPU is not going to help you much. (which brings me back to our earlier conversation. One thing that Cyrix PC had over PII was 1MB L2 as opposed to 512K of PII) Going from 66 MHz bus to 100MHz will help a tiny bit but not much. Maybe Scumbria can fill in: How much will the latency be reduced when I switch from say 66 MHz SDRAM to 100 MHz SDRAM? It will not do anything about time wasted in L1 and L2 lookup, but the main memory read should be a little quicker. Joe

To: Craig Freeman who wrote (30315)	11/19/1998 9:28:00 PM
From: Scumbria	Read Replies (2) \| Respond to of 33344

Craig, If the hangup is memory, a PII/350 on a 100MHz bus should be a great deal faster than a PII/333 (it is not). High-end video cards do a lot for video benchmarks but very little for business applications. And when I upgraded to a faster hard drive, I barely noticed any difference. Where is the bottleneck? The bottleneck is memory latency. A faster bus does nothing to fix this problem, because the latency is inherent to the dram. The fastest drams have a latency of 25ns on a page hit and 100ns on a page miss. Compare this to a clock period of 2ns on a 500 MHz CPU. A dram page miss will cost 100 clocks on a 1GHz CPU. All the fancy CPU architecture in the world will do nothing to resolve this issue. Scumbria