Intel goes out on a (rotten?) limb in efforts to catch AMD

[unixronin]

One of the reasons behind AMD's recent successes (read: taking away market share from Intel) in the server arena is its adoption of the HyperTransport high-speed interface for communication between multiple processors.  It is now reported that Intel, in an effort to reverse the loss of market share, plans to implement a similar feature on their own processors.

However, in what I personally consider a very ill-advised move, instead of adopting the tried, proven, and currently industry-standard¹ HyperTransport interface, Intel is inventing its own, which it calls CSI, for Common System Interface.  Intel makes quite a few showy-sounding statements about what CSI will do, but so far appears oddly silent on what it will do that HyperTransport doesn't -- apart, that is, from being a new, untested, untried, and undoubtedly initially buggy interface.

AMD, meanwhile, plans to move next year to HyperTransport 3.0, which will allow 41.6GB/s bandwidth, interconnects between up to 16 processors, and allows interconnects across multiple boards and even multiple chassis.

I don't know about you, but it sounds to me as though Intel's new Great White Hope is obsolete before it even gets off the drawing board.

[1]  HyperTransport, which grew out of technology developed by Digital Equipment Corporation for the late lamended Alpha processor, is an open standard maintained by the HyperTransport Consortium.  It is currently used by AMD, Apple, Cisco, IBM and Microsoft, among others.

Comments

[otherbill.livejournal.com]

...Intel's new Great White Hope...

...as opposed to their old Great White Hope, IA-64 and the Itanium family? :)

(Hey, if I can't laugh at this, nobody can. ;-)

[unixronin.livejournal.com]

...as opposed to their old Great White Hope, IA-64 and the Itanium family? :)

Also widely known as "Itanic" :)

[7leaguebootdisk.livejournal.com]

It is about time!

Shared bus SMP is dumb (but cheap), and it has been all Intel ever offered. It is about time they caught up with what AMD was shipping back in the K7 1200 days.

We shall see if intel can reverse the losses they have suffered, or if they are doomed long term. We can only hope they are.

[7leaguebootdisk.livejournal.com]

2009 to get the Xeon line to where AMD was with the K8 (the 64 bit chips) in what, 2003? How much more market share will they loose?

[ithildae.livejournal.com]

I think Intel is taking away the wrong lessons from it's partnership with Micro$oft.

Intel needs to suck it up and update the cross-license agreements with AMD. They have some cool stuff to offer. Intel has a problem in that they have always seen themselves and underdogs, on the verge of losing everything. Perhaps this will provide some long overdue focus on who their customers really are, and what those customers want.

[7leaguebootdisk.livejournal.com]

What, may the best processor win?

Seeing as Intel has been a performance winner only sparingly since the introduction of the K7 (Athlon) in what, 1999, I don't expect them to go for something that would alow you to switch cpu by just switching cpus, and not motherboards, and ideally, ram (remember Rambus?).

[ithildae.livejournal.com]

Yeah, those are the lessons the Intel absorbed from Micro$oft. Not helpful.

Processor performance is the outcome of many pieces of technology, it is not one single, monolithic patent. Being out of the raw performance lead does not mean that Intel does not offer anything of value. (Think Notebooks) AMD is only stealing Intel's lunch money for Servers, there are lots of other market niches. If Intel plays nice, they will likely continue to have much to offer. Otherwise, there are lawsuits pending. Another lesson from Micro$oft.