The Long Standing Issues Between Nvidia and Intel
Intel has tolerated Nvidia’s products for years, tacitly accepting that the Green Team’s chipsets and GPUs made Intel’s CPUs more attractive to many users, even though it cost Intel in their own chipset sales. This time last year, things weren’t exactly a love-in, but no one had said the other’s product was doomed to occupy less and less of the computing landscape.
The main bone of contention back then was SLI licensing: if you wanted to run multiple Nvidia graphics cards, you needed an Nvidia chipset on your motherboard. Nvidia said Intel wasn’t offering them a fair licensing rate; Intel, unsurprisingly, said Nvidia was asking for too much. In all fairness, it was a question that didn’t affect all that many users, since not that many people run multiple graphics cards.
Then again, for those that do, this SLI/Crossfire schism was important, since this was before ATI released their 48xx series, and ATI graphics had fallen behind Nvidia’s offerings significantly. For a time, the top of the line in graphics required an Nvidia chipset on the motherboard.
Words, Surprisingly, Matter, But Not as Soon as We Think
In April of last year, Intel and Nvidia exchanged some FUD. Specifically, Intel claimed, with some accuracy, that they ship more graphics processors than anyone else. That’s a big deal: everyone knows Intel is on top in CPUs, but people associate GPUs with Nvidia and AMD/ATI because they are who make discrete graphics. And discrete graphics get the lion’s share of the attention.
People who use integrated graphics may not even know what kind they have. Intel wanted to point out that all those integrated GPUs makes them the number one GPU maker by volume. Nvidia took exception and, referring to a study by an investment firm, pointed out that 73 million of Intel’s GPUs might be out there, but they aren’t being used.
In those cases, the computer has the integrated Intel graphics, but their operation is superseded by a discrete Nvidia card. Factor that in, and Intel sells more GPUs, but Nvidia GPUs are used by more people. So they both say they are number one, and they are both right, but they are measuring different things.
There was no agreeing to disagree, and both sides included a lot of rhetoric to strengthen their position. Actually, they did agree on one thing: that the move to parallel processing means that eventually one type of processor would do most or all of the work in a computer. Of course, Nvidia felt that as things got more parallel and computing becomes more visual, the GPU would win out, while Intel felt that as things got more parallel, CPUs could imitate GPU functionality well enough. Driving this point home is their Larrabee project, which uses a bunch of modified CPUs to act like the many stream processors in a GPU.
SLI Gets Licensed, and All Appears Forgiven
Telling a rival that their business is doomed to become a subset of your own is pretty aggressive. And, perhaps in a unrelated exchange, Intel stated that Nvidia couldn’t make chipsets for Nehalem, and Nvidia said they could. Despite their other differences, they both claimed they were still negotiating.
And back then, there were still months before Nehalem came to market. It was assumed they would work something out in the interim. More importantly, at the time at least, we found out that the two had worked out an agreement to include SLI on X58 Nehalem motherboards.
At first, Nvidia was going to insist these boards use a special Nvidia chip, which would have not only cost money on its own but have to be worked into the motherboard. Eventually, they settled for licensing it without the extra chip, for $5 bucks a motherboard.
Sure, Intel was still working on Larrabee, and Nvidia was hyping the heck out of CUDA (a way to program applications to run on GPUs), but no one was saying it was the way of the future anymore. Well, they weren’t saying it was the exclusive way anymore.
The Calm before the Storm
Everyone was too impressed with the X58 based boards that came out with Core i7 in November to worry about when the Nvidia ones would show up. And Nvidia seemed to be walking hand and hand with Intel at the other end of the market.
Their 9400 integrated graphics chipsets were as good as and often better than AMD’s 780G. Without the 9400, the choices in platforms with decent integrated graphics were decidedly in AMD’s favor. And AMD obviously doesn’t make chipsets for Intel processors.
Furthermore, Nvidia was moving ahead with the Ion platform. By pairing an Atom CPU with the 9400 chipset (actually, the whole chipset, both bridges and graphics, are on a single chip), Nvidia was about to completely rejuvenate the still burgeoning netbook market. Not to mention huge opportunities in the smaller, but still worthwhile, ultra small form factor, all in one, HTPC, and low cost laptop markets.
And they were using an Intel CPU to do it. So what would Intel say to this chance to sell Lord knows how many tens of millions of Atoms?
Then, Out of the Blue (Team)
Last month, we were surprised to hear that Intel had taken Nvidia to court to prevent them from making motherboards for Nehalem. After all, shouldn’t they have figured out that out when they negotiated the SLI license? Both sides say they have been in negotiation for a year, but the other won’t play ball.
Then, a document marked as internal and confidential was leaked from Intel. In it, a scathing list of points for their sales team to use in bashing the Ion. Nvidia, for their part, has said that Ion 2 will use a Via CPU, instead of Intel’s Atom. They then said they are looking into making their own CPUs.
The Nehlem and Ion questions could be based around Intel’s market segmentaion attempts. We discuss the how the Ion interfears with Intel’s segmentation of the low-end market here. A subsequent article looks at how Nvidia chipsets for Nehalem could hamper Intel’s segmentation attempts of that market.