On Tue, May 22, 2007 at 12:37:41PM -0500, Neil Gunton wrote: > Are we talking about desktop workstations here? Forgive my ignorance, > but what on earth requires that much RAM? Video processing? I have 1 GB > in my desktop at the moment, and that's useful for when I'm running > VMWare, but that's about it. Most of the time, it's just being used for > disk cache.
Ram is cheap, firefox leaks memory (or wastes it) like crazy. KDE doesn't seem much better. Until people start taking code quality seriously, it is simpler to throw more ram at it. > I think I must have gone to sleep for a couple of years. When I was last > looking at Intel vs AMD, "they" were saying that AMD's architecture was > much better than Intel's, because (I think this is right) for > communication between cores, the AMD doesn't have to go off-chip, but > Intel's architecture requires use of the external bus, and AMD's design > just plain scaled better. Or something. Then fast forward to today, > where apparently Intel's Core 2 Duo is apparently kicking the pants off > AMD... how did this happen? Is Intel really all that much better? The Core 2 Duo has an internal connection between the two cores (they are a single die) just as the Athlon 64 X2 does. The Core 2 Quad has two Core 2 Duo dies attached together using the front side bus. So for a quad design, the Core 2 is similar to the dual core design intel did with the Pentium 4 (aka Pentium D). The Core 2 is based on the Pentium-M core which goes back to the PPro (it is derived from the P6 core). The pipeline is in the low to mid teens, unlike the netburst which managed to go past 30 stages (great for clock frequency, bad for dealing with conditional branches). So in terms of design, the Core 2 has a lot more similarity with the Athlon than the Pentium 4, except it is a bit more modern and has some clever tricks, which makes it able to run faster than the Athlon 64 at the same clock speed. Hopefully those improvements AMD is promising in the next version of the Athlon 64 will in fact give them the same or hopefully better performance per clock than the Core 2 Duo. > Also, I only really hear comparisons between the Core 2 Duo and Athlon. > How about Opteron? Is the Opteron still a good choice for servers? Or > has Xeon leapt ahead there too? The Opteron is an Athlon 64, except it (usually) uses registered memory (allows more banks of memory in the server, at a slight speed penalty). Current Xeon's are Core 2 Duo or Core 2 Quads, with a different bus speed (I believe they tend to run 1333MHz effective bus rather than the 1066MHz of the Core 2 desktop chips). Xeon's also usually have more cache. Of course the opteron has the fast hypertransport link between cpus, and per cpu memory controllers, so the memory bandwidth is better on the opteron with lower latency, which is why the opteron still scales better than the xeon. For single or dual cpu the xeon is usually fastest, but for 4 or more cpus the opteron is better off since the xeon still has to share a single bus to the chipset for all the cpus while the opteron has the hypertransport links between cpus instead for memory accesses and only has to use the link to the chipset for accessing devices. Adding opterons and memory gives more overall memory bandwidth. Adding cpus to a xeon system doesn't add bandwidth, just processing power. Until intel some day gets an on chip memory controller. > Sorry for the ignorance. I don't pay much attention to hardware stuff in > between computer purchases. Last time I really looked was in 2005 or so. Lots has happened. It is nice to have some competition between AMD and intel to keep them both going, although I like to root for AMD being the underdog. -- Len Sorensen -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
