Intel Settles With NVIDIA: More Money, Fewer Problems, No x86by Ryan Smith on January 10, 2011 9:20 PM EST
As was the case with the AMD/Intel settlement, today’s settlement with NVIDIA paints Intel as being the loser in the proceedings. Officially both sides are settling their differences and dropping their suits, however the terms of the settlement look to be in NVIDIA’s favor versus Intel’s favor.
The biggest (or at least, least ignorable) component of the settlement is a cash settlement between Intel and NVIDIA. Intel will be paying NVIDIA a total of 1.5 billion dollars as part of the terms of the settlement. This is larger than the EU fine ($1.45bil) and larger than AMD’s cash payout ($1.25bil). Notably, this is not a lump sum but rather will be paid in installments. As per the new six year cross-licensing agreement between the companies, Intel will be paying a portion of the settlement for each of those six years - $300mil on January 18th of this year, another $300mil on 2012 and 2013, followed by $200mil in 2014-2016. As a result this doesn’t give NVIDIA an immediate and large cash infusion, but it will add to their bottom line for the next several years; NVIDIA has never had a yearly loss larger than the Intel payments, so it’s safe to say that they are likely going to be in the black for the next six years even if their operations generate a loss.
The cash settlement goes hand-in-hand with the rest of the settlement, which settles the outstanding legal ambiguity about the previous Intel/NVIDIA chipset licensing agreement, along with establishing a new six year agreement between the companies that largely extends the previous chipset agreement.
The most notable bit here is that the chipset license agreement will now formally define that NVIDIA does not gain rights to DMI/QPI, which the agreement defines as being Intel processors with an on-chip/on-die memory controller. So while the company can continue to produce C2D chipsets, they will not be able to produce a Nehalem or Sandy Bridge chipset. This seems to be quite alright with NVIDIA, who claims they are done making chipsets – as far as we know they wound-down their chipset operations some time ago, and the GeForce 320M chipset (seen in Apple’s 13” and 11” notebooks) was the final chipset for the company. This also recognizes the long-term problem with producing a chipset for these processors, as with an on-die memory controller there’s little for NVIDIA to do on DMI-based CPUs beyond adding a south bridge (although we would like USB 3 support…). One way or another the 3rd party chipset market is dead.
|NVIDIA Gets||NVIDIA Doesn't Get|
|1.5 Billion Dollars, Over six Years|
|6 Year Extension of C2D/AGTL+ Bus License||DMI/QPI Bus License; Nehalem/Sandy Bridge Chipsets|
|Access To Unspecified Intel Microprocessor Patents. Denver?||x86 License, Including Rights To Make an x86 Emulator|
NVIDIA also does not get an x86 license. x86 is among an umbrella group of what’s being called “Intel proprietary products” which NVIDIA is not getting access to. Intel’s flash memory holdings and other chipset holdings are also a part of this. Interestingly the agreement also classifies an “Intel Architecture Emulator” as being a proprietary product. At first glance this would seem to disallow NVIDIA from making an x86 emulator for any of their products, be it their GPU holdings or the newly announced Project Denver ARM CPU. Being officially prohibited from emulating x86 could be a huge deal for Denver down the road depending on where NVIDIA goes with it.
So what does NVIDIA get out of this? On top of the 1.5 billion dollars, much of it is a continuation of the status quo: the six year chipset agreement (amended to explicitly forbid QPI/DMI) will be extended another six years. However there’s one item that sticks out in our minds based on the NVIDIA conference call this afternoon: Denver.
Last week’s announcement of Project Denver firmly established that NVIDIA is making its bet on ARM and not x86, and today they reiterated that not having an x86 license is not a problem for the company because they don’t intend to make an x86 processor. However based on what little we know about Denver it certainly is going to compete with Intel’s CPUs at some level. To this end, NVIDIA specifically mentions that they are getting access to Intel’s “microprocessor” patents, excluding the x86 (and XScale) technologies we previously mentioned.
Although this is not laid out in the settlement (because the settlement refers to the original agreement, which is confidential), NVIDIA has made it clear that the agreement gives them the right to “take advantage” of Intel’s patents for the “types of processors” they’re building. Our best guess is that as a result this agreement includes at least a partial preemptive settlement over Project Denver. Just as NVIDIA has many GPU patents Intel has many CPU patents, and it may be difficult to build a desktop/server CPU like Denver without infringing on those patents. If this is the case then today’s agreement implies that Intel and NVIDIA are cross-licensing to the point that Denver is mostly safe from Intel. While Intel’s approval isn’t necessarily essential for Denver like it would be for an x86 CPU, it clearly is easier to build Denver without the risk of Intel suing the pants off of NVIDIA again.
|Intel Gets||Intel Doesn't Get|
|Continued Access To NVIDIA's Graphics Patents||?|
|No NVIDIA Nehalem/Sandy Bridge Chipsets|
|No sharing x86 With NVIDIA|
So we’ve established what NVIDIA gets, but how about Intel? The Intel situation looks to be much more straightforward. As we mentioned previously, NVIDIA and Intel originally cross-licensed in 2004 so that Intel could build IGPs using NVIDIA patented technologies and methods. That agreement was set to expire this year, which would have been a massive problem for a company whose CPUs almost always include a GPU. Today’s agreement with NVIDIA renews and extends that original agreement: Intel continues to cross-license with NVIDIA, allowing them to produce IGPs that use/infringe on NVIDIA patents. To be clear we believe this is a continuation of existing practices, and not any kind of agreement to integrate actual NVIDIA GPUs into future Intel CPUs as others have claimed elsewhere.
The rest of what Intel gets would appear to be gaining a market advantage through not having to give anything up. Intel doesn’t have to license x86 to NVIDIA, Intel doesn’t have to license DMI/QPI to NVIDIA, and if our reading is right Intel won’t have to face direct competition from NVIDIA using an x86-to-ARM emulator. This may not be an “exciting” outcome, but keep in mind that Intel already has some of the best gross margins in the chip industry, so to maintain status quo for the company is a big deal for them.