[B3D News]

Ouch, only two stories in nearly 3 weeks! Good thing the industry has been rather slow so we didn't miss anything too interesting. And GT200/RV770 rumours from people who obviously have no idea what they're talking about don't count, sorry. Either way, let's quickly catch up (link-only)...

Read the full news item

[digitalwanderer]

Looks like I'm not the only one who keeps forgetting B3D has got a frontpage...

/rimshot

[AlphaWolf]

Originally Posted by digitalwanderer

Looks like I'm not the only one who keeps forgetting B3D has got a frontpage...

/rimshot

ouch.

[Arun]

No, no; I merely forgot that other sites had a frontpage!

[MulciberXP]

Big companies love to think that new org. charts will solve all their problems. To me it just signifies that they're still focusing on their management structures and can't yet get their eye on the ball.

Oh and unless talks have already fallen through...I still think nVidia could by VIA. They have to buy someone to grow their staff of engineers. Not everyone can be like Nintendo and just sit on billions of dollars.

[Arun]

Well, I personally tend to think of those things in very... practical terms. Buying VIA costs a lot of money and makes Intel even more paranoid about you; so you sure as hell better have a plan to get a great combined product out of the door *fast*. Taking 3 years like AMD did with Fusion just doesn't cut it for NV.

That means you need to think in terms of aligning the process technology of the GPU and the CPU. The problem is that right now VIA/Centaur is using Fujitsu as a foundry, while NVIDIA is using TSMC. I suspect, sadly for NV, that they're going to stick for Fujitsu for 45nm; the latter's process is very competitive and, I suspect, still has higher performance than TSMC's.

Unless VIA does go to TSMC at 40nm, I can't see the fabs & process variants converging for Centaur CPUs and NVIDIA GPUs/Northbridges until 28nm in early 2012. That's really quite ridiculous, and what should have been the priority is making sure that doesn't happen. Acquiring VIA early would have been a good way to guarantee that... Alternatively, they could just have pushed TSMC really hard to significantly undercut Fujitsu, potentially paying themselves for part of the difference.

Yes, they could port their GPUs to the same process VIA is on; but that means the integration time would be quite long, which as I said is highly undesirable. So the problem is that now they risk not having much flexibility in the short & mid-term. For their discrete GPU business, it's not a big deal at all; but for their chipset business, it puts them at very real risk.

Another factor I like to point out is the feedback effect between the quality of the ultra-low-end solutions and the CPU ASPs, and the impact that has on GPU ASPs. If I can buy an awesome computer with the entire semiconductor value around $25... Why would I want to pay $200 for a CPU, especially when a GPU adds a lot more value? This kind of thing shouldn't just be thought in terms direct business opportunities, but also the industry-wide consequences of the different scenarios.