quaint little rutted bucket

Thursday, April 21, 2005

The power of two: AMD launches dual core processors

Gee, AMD's newest and greatest dual core processors thus far have shown some amazing scability and performance. Taking a look at AnandTech's recent tech preview of the Athlon 64 X2 and Opteron xx5's, these babies show amazing potential when compared alongside Intel's own dual core CPUs. AMD simply dominates.

The Potomac Quad (4 way) Xeons were just saved by its gargantuan (read: 8 megabytes!) amounts of L3 cache. Without those, Intel's server processors would be creamed right off the bat. Even if they were running above 3 GHz, clock speed weren't enough to win the day. As some in the Anandtech forums suggested, a four-socket AMD board loaded with dual core Opterons would be an entirely different beast.

True, Intel may have had a slight advantage before when it introduced Hype[d]rThreading, but remember that it was at best, just a stop-gap measure to keep Intel's wasteful NetBurst architecture busy at speeds well over 3.0GHz. Now that AMD built a processor that also takes advantage of multi-threading, Intel should be wise to stop, look and listen. (I just hope Intel's newer dual core Pentium-M based desktop processors, slated for a late 2006 release, should be enough to provide performance in the league of AMD processors that will be available at that time.)

In multithreaded/multitasking environments the Athlon 64 X2 is even more impressive; video encoding is no longer an issue on AMD platforms - you no longer have to make a performance decision between great overall performance or great media encoding performance, AMD delivers both with the Athlon 64 X2.

As some of you may know, Intel's HyperThreading "emulates" a logical processor within the same CPU to maximize the processor's use of resources. AMD has publicly acknowledged, through Fred Weber of DEC Alpha fame, that the current Athlon (K8) architecture is not suitable for simultaneous multi-threading (SMT). (Simple answer: AMD's K8 processors have less than half of the 30 or so stages that Intel's current NetBurst architecture processors have, thus less branch prediction errors and more instructions done per clock vis-a-vis NetBurst) By adding another core to the physical processor itself and activating the same bit used by Intel to identify a HT-enabled processor, AMD hits two birds with a single shot: it brings to the multi-threaded applications the same power and execution performance the current K8 architecture provides.

Note that this was just done with DDR1 and loose RAM timings. With any luck, a future high bandwidth memory platform (either DDR3 or XDR, not DDR2), with tight timings, and of course, PCI Express and SATA-NCQ enabled drives should make for interesting performance figures in the near future.

It's strange how tables have turned, making Intel look like the value CPU manufacturer in the dual core race.

That is, if AMD manages to bring in Fab36 in time to cope up with the demand; from a economics point of view, Intel wins simply because of its sheer manufacturing capacity compared to AMD. Intel would be able to bring huge amounts of its products, especially in the low-end compared to AMD. What is good is that, in the end, we end users all win. :)


Post a Comment

<< Home