Dell to use AMD Chips in Desktop PCs

bain writes "MarketWatch reports that Dell has decided to use AMD chips in its Dimension desktops due next month. The move to use AMD chips signals a break from its long standing reliance on Intel chips. The information slipped out of Dell's quarterly earnings report." From the article: "Before the announcement, which had been speculated in the financial community and the press, Morgan Stanley analyst Mark Edelstone wrote in a research note: 'It should have a negative impact on Intel and it could be a large offset to the expected benefits from Intel's restructuring efforts.' AMD, which has become a more formidable competitor to Intel, has been expanding its manufacturing capacity, a sign that it expects to be shipping more chips. Its chief goal is to put itself in position to supply 33% of the global microprocessor market by 2008. "

AMD owns server market, and 65nm coming

[Brian Stretch]

Intel's new chips won't stop AMD's share gains in the server market (no integrated memory controller, no HyperTransport, no 4P/8P options) and if you're standardizing on Opterons for your high-end x86 servers why not run AMD all the way down the line? How many corporate customers has Dell lost to HP because they had no Opteron option?

Plus AMD hasn't done their 65nm trasition yet (shipments start end of this year). That should be enough to leapfrog Intel, depending on how many architectural tweaks they do while they're at it. AMD doesn't switch process nodes until they've figured out how to get mature yields (which they say they have), then they do a rapid changeover.

Intel's C2D chips have got to be expensive to produce, what with their 2MB and 4MB L2 caches. I wonder what their yield rates are? Dell was probably worried about getting enough supply, especially with Apple getting first dibs now. Intel's strategy of throwing capacity at problems has to be becoming unsustainable, looking at their deteriorating balance sheet. (Ignore their income statement, that's much easier to manipulate. Cash is tougher to fake.)

Worst case, Dell has seen what's coming at wants to get on AMD's good side now.

Re:AMD owns server market, and 65nm coming

[MrNemesis]

Remember that (currently) Intel has much higher cache density than AMD, so can fit more cache in the same amount of silicon. AMD are meant to be adopting some new cache spec (sorry, I forget the buzzword involved, ZRAM maybe?) which will allow them a more competitive cache/silicon ratio, but I odn't think this is due out until K8L.

Granted though, I'm amazed how well AMD's chips compare to the C2D's considering their aging design and lith process. A shift to 65nm will make them even more competitive, although C2D will still retain the performance king I think, at least until K8L.

In short - yay for competition! I don't think the CPU market has ever had so many great chips on offer for such low prices. But then I guess you can say that about pretty much everything in computing...

Hyperthreading and multicore are different idea

[DrYak]

Hyperthreading was a precursor to multicore processors, idea-wise

While in the end, both hyperthreading and multicore enable you to run more task concurrently without buying extra chips, they don't have anything in common. HT isn't Multi-cores precursor, it's completly different idea.

The basic idea of HT is to fill-in the hole that happen in the pipe. Very often, the CPU waits a few cycle, while instruction are comming through the pipeline stage. The basic idea of HT is instead of a given stage stay idle, wainting on the previous to complete, we can feed it with data from another thread. 1 logical unit, but 2 threads run in parallel, the first one as usual, the second only serve to avoid staying idle each time a prediction turned out wrong. Over-all speed : almost the same, but background task "feel" more responsive.

The basic idea behind multicore is to try to takae the advantage of 2 CPU, but sharing some part : 1 packaging, 1 interface, 1 socket on a single-socket motherboard some times even 1 of the lowest level cache (and some times it is just two chip packaged together and using 1 interface), except from that sharing, it behaves mostly like two CPU. Over-all speed : doubled.

So the idea are basically different : HT is "try to keep the CPU busy even in case of pipe-line stall (and thus avoid wasting time)", Dual-Core is "try to make SMP by making two-processors-on-a-chip (and thus increasing theoretical max speed)".

Re:Strange timing?

[adam31]

the only thing that really would have made Dell change their tune is customer demand.

Well, that has been there for several years now. Perhaps because of the AMD lawsuit, Intel has had to tone down their vendor-bribing, and the revenue from AMD demand now outweighs the check Intel cuts to Dell.

This again?

[tashanna]

Not [slashdot.org] this [slashdot.org] again [slashdot.org]. Come [slashdot.org] on [slashdot.org] editors [slashdot.org], really [slashdot.org]!

- Tash [tashcorp.net]

Re:Price

[Wdomburg]

The Core 2 Duo architecture uses two cores (obviously) and shares the L2 cache (revolutionary, usually each core has a seperate L2).

Revolutionary? The POWER4 was sharing L2 cache between cores way back in 2001, followed by the UltraSparc IV+ in 2005.

Re:It's opposite land! - RTFA

[Anonymous Coward]

RTFA - It's for shipment next month, not just another AMD/Dell rumor. How did this cluelessness get modded up to 5?