Fresh marchitecture skirmish, on borders
http://[img]http://netcomm.spinb...rMPU[/img]
JUST BECAUSE Intel followed AMD in saying clock numbers were a thing of the past doesn't mean that the embedded hostility which constantly invites us to compare them to Tweedledee and Tweedledum is* over.
Far from it. Now instead of competing over whose chip has the most megahertz, there's a new "war of the willies" nicely shaping up over dual and multiple cores and who can get the most in one package and quickest.
Anyone who believes that it's a coincidence that senior souls at both firms upped the ante on dual and multiple cores during the same week is living in a Looking Glass World. As we've pointed out many times before, Intel and AMD go together like love and marriage or like the kneebone connected to the legbone.
Despite endlessly telling us that they "listen to their customers", both firms have decided, just like the protagonists in Through the Looking Glass, to go to battle over dual and multiple cores. That's without any thought as to whether "their customers" - who aren't, by the way, consumers - but people far further up the food chain want this stuff or whether it will bring any additional performance to the party. Neither firm will disclose the performance benefits. We are expected to take it all on trust.
Never let us forget, as the good old Anglo-Saxon line goes: A pint of wine to a vintner is as a pippin to a costermonger.
Intel recently made available a PDF -
http://www.intel.com/intel/finance/p...s/2004_fam.pdf - of its autumn presentation to analysts, surely a flock of sheep endlessly led astray by bellwethers. Or like lambs to the slaughter.
Between pages 40 and 49 of this presentation, soon to retire CEO Dr Craig Barrett did a remarkable thing. He described the RISC and the IA-43 architectures as complementary, pitting the Itanium as the architecture of choice against RISC, and IA-32 as pushing the lead on price-performance per watt. This presentation confirmed what we've been able to pick from the bones of what the chip giant has said before. In 2005 on the Xeon front we'll see EM64T everywhere, larger 2MB cache, faster front side buses, power management and PCI Express IO. On the Itanium front we'll see dual core multithreading with Peliston and Foxton.
It won't be until 2006 that we'll see dual core Xeons with lower power cores, "enhanced memory", virtualisation, iAMT, IO packed acceleration and advanced storage controllers. Beyond 2006 we'll see multicore chips for both IA-32 and Itanium, Barrett posited.
On the AMD front, CTO Fred Weber has also started talking seriously about multicore technology, about DDR-3 and Hypertransport 3. This is all before we know what Microsoft proposes to do apart from chat way too vaguely about Longhorn. We wonder if either AMD or Intel knows any more about Microsoft's directions than we outsiders do? We somehow suspect maybe they don't. In the meantime, we're all going to have to hear a heck of a lot more about dual core and multiple core chips from both Intel and AMD, possibly meaning that by the end of 2005 we'll all be so heartily sick of the subject that we'll give in and neglect to keep asking what they actually bring to the PC party.
Source:
http://www.theinquirer.net/?article=20207
Linear Mode
