Replies to post #28592 on Intel Corp (INTC)

[smooth2o]

re: I hope AMD spends lots of their precious marketing resources on coprocessors.

If I were AMD, I would let others spend their precious engineering and marketing resources on elevating co-processing in the industry, something, it seems they are able to do and Intel isn't. You really need only a few high cost applications that "fit the bill" so to speak. Once these are identified (and there are some immediately apparent ones) and you can show that they provide an order of magnitude performance differences, you provide the capability to provide them in the *library* of modules in the AMD core architecture, something they say they are doing. It's then that you spend a lot of your own engineering resources to incorporate something that Intel, to date, is not offering. That done, the show could be over. Adding Virtualization, if they can, is icing on the cake.

For years, Intel's mantra was to incorporate MB functionality into the CPU, something that was very successful. The 8087 was one of them. They ceased the OBMC effort and they didn't rush to incorporate an external Hyperxport either, when that was an apparent direction in the Alpha architecture.

Somewhere along the line, Intel seems to have lost their paranoid behavior, and, as I stated before, the diffusion made possible by such AMD's proposed architecture could keep AMD in the perceived lead. Obviously, pulling this off, even if it's possible, doesn't seem like it will happen before Intel has similar features anyway.

At this point, I wouldn't call it crap. It may hang out there for 5 years or so if not only to point to the fact that AMD "opened up their internal architecture" to the customer and thereby "serving their needs better".

<marketing hat off>
<engineering hat on>

Smooth

[tecate]

It wouldn't surprise me at all if Intel weren't working on their own coprocessors :) now I have no knowledge of this, but I wouldn't be surprised.

[dbudde]

Coprocessing is just crap. If any of you are old enough to remember what it took to get the software industry to support the x87 when it was a "socket" you'll remember it was very heavy lifting. It wasn't until it was in the 486dx that it was starting to take off and even then the 486sx was offered for those who wouldn't pay the extra at the CPU level.

I'm old enough, and I don't remember it quite this way. I rememebr sitting in product develoment review meetings at Intel and occasionally Gordon Moore would attend. The most common question he asked when significant new features were discussed was: "Is there any way you can make this feature into a coprocessor?"

In fact, the x87 chips were extremely profitable and sold like hotcakes. Software issues aside, Intel wanted more products like these. It was quite controversial getting the FPU integrated into the 486.