I'm rushing this one out because the news is timely and I'm pushing myself to become a better writer. As I start this I have thirty minutes before I need to be at work. Luckily I have a five minute commute, so at least I get to use most of that time.
I've been expecting some sort of settlement between NVidia & Intel for awhile now. I was fairly sure it would involve Intel paying NVidia a large chunk of money (it did) but the rest of this deal is a surprise to me, but not in a bad way.
In addition to $1.5 billion over the next six years NVidia & Intel have also entered into a cross licensing agreement that gives Intel access to IP that NVidia has developed for their GPU's while in turn giving NVidia access to IP that Intel has created in the processor space. This might seem like a one sided deal in Intel's favor if not for the fact that NVidia is working on integrated ARM/GPU based chips. With this settlement behind them NVidia should have much smoother sailing in implementing both their discreet graphics solutions and project Denver, the already mentioned ARM/GPU project.
It's no accident that the duration of this agreement is six years. In August of last year Intel settled with NVidia & AMD over numerous other complaints. One of the technical requirements was that Intel provide a high speed PCI based solution for the next six years. NVidia needed this to keep their discreet GPU business viable. What has essentially happened with these two agreements is that Intel has provided itself six years of reduced legal wrangling and access to patents that should help it's always lagging GPU development efforts. AMD is the big loser here as their main edge has been the fact that they have both a very competitive GPU and processor and were thus ahead of Intel in the race to integrate these two components into a package that is both fast and cost effective. In fact, Intel could potentially partner with NVidia to integrate an NVidia GPU & Intel CPU onto the same packaging, and perhaps even the same wafer in the next few years. That would be a killer combination as NVidia GPU's have double precision floating point and error correcting memory to go along with excellent performance. AMD has double precision floating point but no Error correcting memory. Essentially you would have the best of both Intel & NVidia in that scenario and AMD would be sorely tested.
One little factoid I missed when talking about CES last time was the fact that Microsoft has announced that they'll be supporting the low power ARM CPU architecture in Windows 8. This should make NVidia happy as it means they'll be able to provide their partners with chips(assuming Denver is a success) that run Windows. I'm not a Windows fan boy and I tend to think that the importance of platforms, particularly Windows is going to continue to decrease in the future but it's hard to see this announcement as anything other than a win for ARM. What this means is that NVidia & Intel will likely be competing across the entire spectrum of devices in about three years. Intel has the advantage of size and money. NVidia has the advantage of starting with a CPU architecture that is much more power efficient than Intel and a clear head start on the GPU side of things. Plus Intel will be making annual payments of $250 million to them for the next six years.
OK, ten minutes to go and I need to fix my links so I'll wrap things up here by saying that Intel & NVidia both appear to have done well for themselves here. It's going to come down to which side executes best on the technical, operational and marketing fronts. Great technology by itself has never been enough. History is littered with companies that came out with technically superior products that failed the test of the market and/or their competition.
At the end of this six year window we'll either see Intel & NVidia working very closely together or at each others throats. I don't see a lot of room for anything in between.
No comments:
Post a Comment