Replies to post #1627 on Advanced Micro Devices Inc (AMD)

[Elmer Phud]

YB -

That system will beat all 4-way servers in the world, including Itanium and Power4.

Once again you are comparing tomorrows AMD vapor with what will very soon be yesterday's Intel systems. So easy to shoot off your mouth but you have no proof whatsoever.

[Tenchu]

YourBankruptcy, <So 4-way Opteron has only 17% overhead over single Xeon. Practically 4 cpu's on a single core.>

OK, if I use the same assumptions AMD uses in their slides (i.e. idle latency), that means four Xeons on a single bus have the same overhead as one Xeon on that bus. That means, by your logic, a 4-way Xeon system is "practically four CPUs on a single core."

Now wait a minute, you say. A 4-way Xeon on a shared multiprocessor bus can't possibly be compared to a quad-core CPU? Well, neither can a 4-way Opteron system. The fact is that NOTHING matches the scalability of a quad-core CPU. All four cores in such a CPU share a single on-die cache. Four Opterons can't, and four Xeons can't either. A quad-core CPU can also have all four cores coordinate its memory accesses closely with each other. Opteron must broadcast its memory accesses across the HT links, while Xeon must broadcast its accesses on the shared bus. (Taking this aspect alone, i.e. out of context, the Xeon bus has lower latency than Opteron's HT links.)

No offense, but you really shouldn't jump to conclusions on subjects you know little about. The notion that 4-way Opteron is like a quad-core CPU is ridiculous.

Tenchu