Intel’s Response

Intel wasted no time in responding to the FTC’s suit. In their response, Intel has denied the accusations, and insisted that they have competed “fairly and lawfully,” noting that prices (ASPs) have been declining at a rate faster than any other industry.

Of particular note in Intel’s response is their claim that the FTC’s case “is based largely on claims that the FTC added at the last minute and has not investigated.” This we’re assuming means all of the GPU-related claims. You can read into this several ways, but our take is that Intel is more concerned with the GPU-related claims than the CPU-related claims at this moment.

Finally, Intel is understandably annoyed that this even reached the lawsuit stage. As we mentioned before Intel was already in settlement talks with the FTC, and believes that the issue should have been resolved there. In the failure of these talks, Intel has cited that the FTC “insisted on unprecedented remedies – including the restrictions on lawful price competition and enforcement of intellectual property rights set forth in the complaint.” As the FTC would not back down on their requested remedies and Intel would not accept them, this is what has lead to the case moving on to a lawsuit.

What the FTC Wants

So what does the FTC want? For one thing, not money. In their claims and requested remedies, they have not asked for any kind of fine, but rather are focusing exclusively on enforcing changes in the way Intel conducts business. This means their remedies are entirely corrective, rather than being a mix of corrective and punitive remedies such as what the EU has gone for.


What the FTC wants for Christmas: For Intel to license DMI

For their corrective remedies, here is what the FTC wants:

  1. For Intel to stop doing all of the things mentioned above.
  2. Intel cannot require OEMs to purchase only Intel CPUs and GPUs, purchase them in specific quantities, or to not purchase competitors GPUs and CPUs. This is effectively a stab at the rebates Intel has been offering for bulk purchasing, and the advertising help Intel has been offering to bulk purchasers.
  3. To stop prioritizing CPU shipments to loyal OEMs.
  4. To stop withholding technical support from disloyal OEMs.
  5. For Intel to be disallowed from producing/distributing any software or hardware that unreasonably excludes or inhibits the performance of competitors’ GPUs and CPUs.
  6. To stop selling things below cost. The FTC is defining this as being the average variable cost plus a “contribution to Intel’s fixed sunk costs in an appropriate multiple of that average variable cost.”
  7. For Intel to do a few different things about the versions of their compiler that put AMD at a disadvantage (which the FTC is calling the Defective Compiler): offer a substitute compiler to customers for free that is not a Defective Compiler, or to compensate customers in switching to another compiler, to provide notice to software buyers of products compiled using the Defective Compiler that they may need to replace their software.
  8. To stop Intel from making misleading statements.
  9. To prevent Intel from coercing benchmark organizations into adopting misleading benchmarks.
  10. For Intel to license the QPI and DMI buses to 3rd party chipset manufacturers.
  11. For Intel to not block the Global Foundries deal (AMD and Intel already settled this) or any similar deal that VIA might make.
  12. For Intel to stop badmouthing competing products unless they have solid scientific evidence.
  13. For Intel to foot the bill for the independent organization that will monitor this.

It’s a long list, but there’s nothing in it that’s particularly surprising. The FTC’s ultimate goal is to get Intel to stop engaging in all of the anti-competitive actions they have been engaging in for the last decade, and to have them directly monitored for an indefinite period of time in the future to make sure they do not resume these actions.

On an interesting note, the FTC chose an unusual way to go about this suit. Without getting into the nitty-gritty of anti-trust laws, the FTC has multiple sections of the FTC act to charge violators under. Normally when they engage in a lawsuit, they charge them under Section 2, which allows for harmed consumers to sue violators in private for triple damages. The FTC has decided that it’s in the best interest of everyone to not open Intel to that kind of liability or to take the risk that they’ll lose the suit based on that kind of liability, and instead charge them under Section 5. The big difference between the sections is that a Section 5 violation only leaves Intel open to the damages caused by their actions, and not to triple that price tag. Overall the use of Section 5 is very rare compared to Section 2, but the use of it is growing according to the FTC.

At this point we do not have any idea what the price tag would be on damages for Intel if they were to lose this suit, but it’s a reasonable expectation that it won’t be cheap, going into the billions of dollars. Triple damages would make that even higher. As triple damages were established as a punitive solution, this is consistent with the FTC’s position that they are not trying to enact a punitive remedy upon Intel.

Conclusion

So when does this battle royale kick-off? Not for a while, it seems. The case is currently scheduled to go before a judge on Wednesday, September 15th of 2010, which is 9 months from now. Even if it were to start on time (it likely won’t), a ruling would take an equally long time. It may be 2012 before the case is ruled on, later if the case starts late.

In the meantime, there are the positions of NVIDIA, Intel, and AMD to consider. Despite the FTC’s immediate concerns, with the recent cancelation of Larrabee Prime, Intel probably isn’t the risk to the GPU market that the FTC believes they are. The question will be what Intel will be announcing in 2010 as the successor to Larrabee Prime, and what actions they may be taking. It’s not in their best interests with this case to engage in anything that might be seen as disparaging of AMD or NVIDIA GPUs, which in turn may influence Intel’s actions here.

As for AMD, for them this entire matter is largely settled when it comes to CPUs. Their exposure on the GPU side is a bit more nebulous – they aren’t going to make integrated GPUs for Intel processors, so their exposure is in the smaller discrete GPU market. Without a better idea of what the FTC is accusing Intel of when it comes to discrete GPUs, it’s hard to say what the impact of this is. If this stops Intel’s anti-GPGPU efforts however, then it’s going to be good news for AMD’s efforts in that field.


Fermi: Intel's greatest fear?

And finally there’s NVIDIA. NVIDIA has been on a crash-course with Intel for some time now, and they would have it no other way. For NVIDIA this has been a very good month: first Larrabee Prime gets canceled, and now the FTC is going to fight Intel in court over several issues that effectively has the FTC fighting Intel on NVIDIA’s behalf. This could go a very long way in boosting NVIDIA’s GPGPU efforts with Fermi, not to mention the fact that the IGP chipset business has been quite good to NVIDIA lately and is something they would like to continue. This suit could come quite close to defanging Intel from NVIDIA’s perspective.

On a long-term perspective, we’re left wondering where this is going to leave the entire market when it comes to GPU/CPU integration. Both AMD and Intel have been pushing it, with Intel preparing CPUs with both on-chip and on-die GPUs. Could a successful FTC suit put a stop to this Fusion for Intel? Will this slow down or stop GPU/CPU integration for the entire market, and greatly benefit CPU-less NVIDIA in the process? The outcome of this case could very well have an impact greater than just stopping any anti-competitive actions Intel is engaging in, so it’s going to be something we’ll be keeping a very close eye on.

Index
Comments Locked

114 Comments

View All Comments

  • yuhong - Wednesday, December 16, 2009 - link

    "This CPU GPU convergence is a convenient excuse for Intel to force "bundle" its crappy graphics with its CPU. "
    And to be honest, AMD will be doing the same thing later in 2011 with Fusion, and of course the graphics will be better.
  • prodystopian - Wednesday, December 16, 2009 - link

    "Intel reworked their compiler to put AMD CPUs at a disadvantage. For a time Intel’s compiler would not enable SSE/SSE2 codepaths on non-Intel CPUs, our assumption is that this the specific complaint. To our knowledge this has been resolved for quite some time now."

    Can someone explain this? What compiler that Intel designs would an AMD CPU have to run? Does AMD not make their own compiler?
  • Ryan Smith - Wednesday, December 16, 2009 - link

    There are a number of x86 C/C++ compilers out there. The 3 most popular ones are the GCC compiler (free and open source), the Microsoft Visual Studio compiler, and the Intel C++ compiler. The Intel compiler in particular is rather popular as Intel has proven to be quite adept at extracting meaningful SSE optimizations out of code, and hence it's used in a lot of performance-critical situations. AMD does not make their own compiler (it's very, very hard work) but they do contribute to the GCC compiler.

    When a compiler is using advanced code and wants to generate a binary that is backwards compatible, it will create multiple codepaths, with the program itself deciding which one to use depending on what feature set the codepath needs versus what the current CPU offers. For a time, Intel's compiler was simply checking for Intel processors, and ignoring AMD processors regardless of the feature set. As a result programs compiled with the Intel compiler would not take SSE/SSE2 codepaths on AMD processors, even if the processor did support that feature.
  • JumpingJack - Thursday, December 17, 2009 - link

    Ironically, one example is Lame MP3 compilations -- the Intel compiler compiles better code for an AMD CPU than does the MS compiler.
  • prodystopian - Wednesday, December 16, 2009 - link

    Thank you very much. I am familiar with GCC and VS but didn't know Intel had their own C++ compiler.
  • ltcommanderdata - Wednesday, December 16, 2009 - link

    Could Intel argue that they didn't want to spend time verifying and supporting the functionality of the SSE/SSE2 code path on AMD processors and so disabled it? To avoid liability if something did go wrong and say data got corrupted. Or can it be assumed the AMD's SSE/SSE2 implementation is 100% compatible with that of Intel's implementation with no need to verify compiler compatibility?
  • Ryan Smith - Wednesday, December 16, 2009 - link

    It can be assumed that AMD's SSE/SSE2 implementation is 100% compatible. There was no technical reason to take a different codepath on any processor that identified itself as SSE2-capable.
  • jensend - Wednesday, December 16, 2009 - link

    The most crucial part of all of this is the demand that Intel license QPI and DMI. Ever since coming out with the Pentium M and Centrino Intel has been doing a lot of odd things to crush the 3rd-party chipset market.
  • DigitalFreak - Wednesday, December 16, 2009 - link

    Hell no. If Nvidia gets a DMI / QPI license, there goes SLI on Intel chipsets. Besides, Nvidia chipsets are buggy as hell.
  • jensend - Thursday, December 17, 2009 - link

    Intel would be perfectly within its rights to make the continuation of the SLI licensing agreements a condition for their licensing QPI to nVidia. I can't imagine the FTC having a problem with that; that's just sound business.

    Don't know where you're getting your beef with nV chipsets from. nV and VIA showed years ago that they have the resources to do just as well with chipsets as Intel and AMD, but they've been squeezed out by anticompetitive practices. The death of the third-party chipset market is really bad news for consumers.

Log in

Don't have an account? Sign up now