Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: jspeed who wrote (213665)	10/16/2006 9:28:14 AM
From: Plissken	Respond to of 275872

I think he's talking about a gradual shift, not something that happens by year's end 2007. Torrenza is the key to that, moving the GPU closer to the CPU at first before on-die integration becomes feasible. PCI-E is set to go the way of the FSB. Also keep in mind that any SLI-capable system should be able to upgrade onboard graphics with add-on cards, so you can just plug in extra power if you need it, sail along at 2D resolution with the integrated part (while putting the add-in to deep sleep, saving energy) and add computing/GPU power on demand from the plug-in. I also see the possibility to shift between stream computing (i.e. for physics simulation) and 3D output seamlessly, using a flexible system of GPU "cores" (pipelines) in near-future generations of 3D chips. That's for Ati, of course, Nvidia will have a bit of a harder time competing unless they keep close ties with AMD.

To: jspeed who wrote (213665)	10/16/2006 10:20:31 AM
From: dhellman	Read Replies (1) \| Respond to of 275872

If AMD lets ATi discrete gpu die it wont be long before their integrated offerings will be as slow as Intel's. And shortly after that the ATi division will become a sucking wound on the financials of the company. -dh-

To: jspeed who wrote (213665)	10/16/2006 11:20:58 AM
From: pgerassi	Respond to of 275872

Dear Jspeed: AMD/ATI will not let discrete die slowly on the vine. Like in cars where the OEMs use racing to try new things, AMD/ATI use discrete to push the envelope. Doing just integrated gets you to where Intel is now. Just enough of a bottom feeder, but no good for mainstream 3D uses. You need the power of discrete to drive performance so that developers will find uses for the performance. That in turn drives demand up and allows you to sell more product. The high margin business is in the discrete end of the market. Just like servers are to the CPU market. Sure it takes R&D to keep up, but the punishment for doing nothing is a painfull death. The strategy of just doing integrated like JO wants AMD/ATI to do, will stop the waterfall of performance enhancements from the discrete end to the integrated end. Intel has tried that method for the last few years and by the time they get their GPU out, the performance bar has pushed even further away then when their previous version came out. Like AMD with creating server end CPUs and downsizing them into the desktop and notebook territories as time goes by, so ATI and nVidia did with discrete GPUs into mainstream and integrated territories. Besides, one miscomprehension is that although ATI and nVidia have 6 month product cycles, the cores only have major changes every 18 months. The first cycle shows the new core at the very high end. The second cycle just moves that new core both higher to stay on pace with the competition and lower into the mainstream. The third cycle lowers power and adds it into the intgrated realm while waterfalling the high end into the mainstream. The fourth cycle starts it all over again. This is also true of the CPU business. K7 had a number of cycles, but the core stayed relatively constant, but it was tweaked a number of times, Classic with L2 die on slot, Tbird with on die L2, Palomino on 130nm and Tbred AXP. K8 with Clawhammer, Newcastle, Toledo, Windsor, Brisbane and now, Barcelona. Every six months a new revision of the core comes out with more tweaks. Some times they add cores, change cache amounts or simply go to a new process. But each seemed to be an increment to a previous core. Going from K7 to K8 was a major core revision. Only Barcelona may be considered a major core change (dual core Toledo was already designed for in Clawhammer) over the original Clawhammer. So GPU hype aside, the CPU business also has a 6 month new product cycle. As for discrete chipsets not being moneymakers, then where does he put the server chipsets? Even Centrino is a discrete chipsetand that is a money maker. So is NForce and Crossfire. I think its JOs way of justifying Intel's past practices and hopes that AMD screws up with ATI like Intel did with their GPU. Too bad JO, AMD will make ATI chipsets into Centrino Crushers and Duo Shredders. He forgets that AMD is into "Have it your way!" Pete