Date: Fri, 20 Apr 2001 00:10:14 -0700 From: "Charles Burns" <burnscharlesn@hotmail.com> To: vince@oahu.WURLDLINK.NET, jgowdy@home.com Cc: seanp@loudcloud.com, lplist@closedsrc.org, kris@obsecurity.org, mwlist@lanfear.com, freebsd@sysmach.com?, questions@FreeBSD.ORG Subject: Re: the AMD factor in FreeBSD Message-ID: <F936u1Wy9eabpAprqFp000049ce@hotmail.com>
next in thread | raw e-mail | index | archive | help
This is all getting highly offtopic, but it is very interesting. > > > and are available with up to a 200MHZ bus. > > > > They both have a 200mhz bus. Not up to, but 200mhz. > > Wouldn't that factor alone make it smoke the Intels? NO. No one factor EVER determines a winner in a processor war. There are so depressingly many factors that determine system performance, every one of them very important for certain types of software and unimportant for others and varying between the two at uneven levels for every other type of software. Three things seem to be widely believe to be either all that matters or all that is really important. CPU clockspeed, FSB clockspeed, memory clockspeed. Even if these THREE things all have very high numbers, there are just too damn many other factors. Intel's marketing/engineering team has built for us the perfect example. The Pentium 4 runs at 1.7GHz recently, has a 400MHz bus, and has 800MHz memory, yet it is the slowest modern architecture of the big three that you can buy. Here are some of the many, many reasons why: CPU@1700MHz: Does very little per clock (very low IPC), has very little cache memory FSB@400 MHZ (actually 100QDR--whatever). Umm, the FSB is fast and I can;t think of any disadvantages, but it doesn't make up for other weaknesses. RAM@800MHz. (Yes, I know it's actually 200QDR--whatever). That's all find and dandy until you use a 16-bit bus instead of a 64-bit bus, and increase memory latency with every RDRAM module that you add on top of an already high latency. They fixed the wrong problem and broke the right one. Bandwidth VS. Latency. It's all marketing. The Athlon is overall faster but the much faster FSB is only one factor. Take away any one advantage that the Athlon has and remove maybe 1%-3% of its performance. Its when all of those little touches are combined that the whole becomes greater than the sum of its parts. (Actually they end up exactly equal, but that sounded better ;-) The P3 tried to get the edge on performance by taking a single factor--the L2 cache bandwidth--and going for extreme values on that. Now the P3 has the fastest L2 cache in the world, but because most or all of the many factors have to compliment each-other, it's potential performance improvement was killed by many other bottlenecks, like the ancient core, the GTL+ bus, the slower and lower number of PUs, the shallower pipelining of the FPUs, etc. > Atleast it sounds better at AMD than it does at Intel. A Celeron >II is a Pentium III with 1/2 the cache, higher latency cache and 4way >cache instead of 8 ways. Setting cache to be more associative doesn't necessarily make it faster. 2-way set associative cache has a lesser miss penalty than 4-way or 8-way. It just so happens that in most cases the cache is slower because of the setting and the way that it is used. The Celeron 2 somewhat irritates me because it shows Intel's gross profit margins. It is the exact same core as the P3, so the P3 can clearly be sold for much cheaper. Good thing for AMD to shave their margins down. At least AMD is decent enough to make the Duron a separate core. _________________________________________________________________ Get your FREE download of MSN Explorer at http://explorer.msn.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?F936u1Wy9eabpAprqFp000049ce>