Editors Note: From time to time, SemiAccurate will be republishing some older articles by its authors, some with additional commentary, updates and information. We are mainly reprinting some of the oft referenced articles that originally appeared on the Inquirer. Some will have added content, but all will be re-edited from the originals as per contractual obligations. You may see some slight differences between the two versions.
This article has had some of the original links removed, and was published on Tuesday, October 24, 2006 at 12:52 AM.
NVIDIA IS MAKING a CPU, the only questions are what kind of CPU, and how the hell are they going to do it. Making an x86 based CPU is not a trivial venture, and there are enough problems to make even a company with the engineering bandwidth of Nvidia cringe. Those problems are mainly called lawyers.
There is no doubt that given engineers, time and money, Nvidia is capable of making an x86 CPU. The Stexar people have done it before, several times in some cases, so it is even less of a problem. It can be financed, and while it may not happen soon enough, it will happen before NV runs out of cash.
The overwhelming problem is patents and other evil things often used to bludgeon the bright eyed and bushy tailed into submission. Any x86 implementation would have to negotiate a minefield of patents and deals just to make it out of the fab. This won’t be easy.
To step aside for a moment, lets ask the question of what if NV does not make an x86 CPU, but an ARM, PPC or some other ISA? Well, they are again more than capable, but that would mean giving up the market they have a pretty solid lock on now. Buy the new NV CPU, it doesn’t run Windows and won’t run your software or games, but the theoretical graphics power is astounding! I can see the lines forming to buy one as soon at they are released. Not. Linux is very cool, but basing a mainstream CPU only on it may not be a bright business move.
Back to reality, what do you do if you want to make an x86 chip? You get a license from the people who have one, in this case most likely AMD, Intel and a few other people. You could also buy a company that has a license, but those licenses may not be transferable if a company changes control.
There are currently four companies that can make an x86 CPU from what we are told, AMD, IBM, Intel and NatSemi. Those guys are in the clear, and the others like Via and Transmeta are in a greyer area, at least according to my sources. If you want to go x86, at the very least you need to have the blessing of AMD and Intel.
The four above companies have cross licensed each other to the point of sillyness and beyond, so they are all OK with whatever they want to do. Related to this, there are four companies with a license to the Intel bus; Nvidia, ATI, SIS and one other, possibly VIA but we weren’t able to pin it down. These may allow you to make chipsets, but it most emphatically does not allow you to make CPUs.
If you want to make an x86 CPU, you have to either have a patent that one of the big four is stepping on and can’t live without, usually in the form of something they have already shipped by the tens of millions, or pay a lot of money. Or both. Lets boil it down to blackmail or wads of cash.
For Nvidia, the blackmail option probably isn’t going to work, at least not on Intel. If that would have flown, they would have made an ‘Intel edition’ chipset a long time ago, but that mess is another story in itself. If Nvidia had that card to play, they would have done so long ago.
This leaves them with the wads of cash option. They could, and probably will try this one, but if someone shows up at your door and asks to be your fiercest competitor in exchange for short term money, do you say yes? I would guess Intel and AMD are smarter than your average flagstone and will not jump at any generous offer NV makes. They will either point and laugh or more likely set the number so high no one in their right mind would consider making a product with that much IP tax. Of the two, pointing and laughing would be far less cruel.
There is a dark horse option, having IBM fab the parts, something that almost worked with the Cyrix based IBM Blue Lighting CPUs. Given the acrimonious relationship between NV and IBM nowadays, I kind of doubt this will happen. The relationship was a miserable failure when they tried to fab GPUs, so something orders of magnitude more difficult to make does not seem like it is worth a serious look, much less an attempt.
So, what is a green GPU maker to do? If I had to guess, I would say that NV is doing the classic thing and ignoring the patents for now. If they wait to sign whatever agreements they are going to sign, well, that delays them by a year or two. If they start designing now while negotiating, as long as the ink is dry by the time they fab parts, all is well, it is better to beg for forgiveness than to ask for permission. I would guess that NV is going ahead and will worry about details later.
One side note, from the sound of it, NV hired the Stexar people but does not appear to have purchased the company or any IP they might have had. We have no idea if this is because of baggage, money, or there was nothing there.
In the end, it all boils down to mistakes. AMD let Stexar slip away for a multitude of reasons. Intel let them slip away for politics. Nvidia picked them up and earned the wrath of both their dear friends. Nvidia might end up crushed like a bug by the companies it needs to survive short term. This could possibly be a lose/lose/lose.
Then again, Nvidia might just end up with an astoundingly good team that will easily beat the best of what is out there, negotiate the maze of patents, and end up winning. The future of high performance computing may be a lighter shade of green than AMD, and certainly not blue, dropped e or no.
Stexar might have been rebuffed because they were not all they were cracked up to be, and AMD and Intel will laugh all the way to the bank. Maybe they were right to say no, far stranger things have happened.
Whatever the case, Stexar will have played a pivotal role in the future of x86. They were involved in the plans of the combined CPU/GPU for all three major players, and could have ended up with any of them. If NV pulls it off, Stexar will have been their greatest achievement. If not, they may have been poison for all three companies in an indirect fashion. Oh how I do love these soap operas.S|A
Note: Part 1 can be found here.
Latest posts by Charlie Demerjian (see all)
- You can now buy bare Snapdragon SoCs with 410E and 600E - Sep 28, 2016
- Spin Transfer Technologies talks about their ST-MRAMs - Sep 27, 2016
- ARM adds CMN-600 interconnect and DMC-620 memory controllers - Sep 27, 2016
- Globalfoundries 7nm process isn’t even close to the name - Sep 26, 2016
- ARM upgrades realtime offerings to v8-R and adds Cortex-R52 - Sep 21, 2016