Nvidia Waiting In the Wings In FTC-Intel Dispute

The NY Times has a Bits Blog piece speculating on some of the fallout if the FTC prevails in its anti-competition lawsuit against Intel. The Times picks out two among the 26 remedies proposed by the regulator, and concludes that they add up to Nvidia being able to license x86 technology. This could open up 3-way competition in the market for combined CPU-graphics chips. There is a good deal of circumstantial evidence pointing to the possibility that Nvidia has been working on x86 technology since 2007, including the presence on its employment rolls of more than 70 former Transmeta workers.

Competition is a Good Thing

[xzvf]

In the long run getting multiple competitors in the CPU space is good. The problem is trust busting worked when the competitors were slow moving oil companies or railroads, by the time this gets through the court system the market will be significantly different. What computer were you using at the turn of the century?

Intel's ill-gotten-gains

[distantbody]

Intel effectively defrauded AMD of many billions of dollars in revenue. Intel should be forced to return those ill-gotten-gains to AMD and THEN be fined.

In the near future if AMD goes bankrupt (possible given their current uncertain situation) and Intel's unlawful actions could reasonably be considered to have led to the demise of their main competitor (AMD), Intel shouldn't be allowed to live with the benefits of their wrong-doing, namely a monopoly, and instead be forced to establish an equivalent competitor. The FTC may indeed be acting along these lines as Nvidia could possibly be a capable CPU producer.

Is x86 shit?

[some_guy_88]

We've been using this instruction set for years and years now. There's gotta be something better around by now. Is it ARM? Cell?

Are Microsoft and Windows the only reasons we haven't moved on? How hard would it be for them to target a different architecture? Linux seems to manage fine in this regard. Rewrite a bit of assembly and choose a different c compiler. Shouldn't be too hard right?

Re:Is x86 shit?

[Rockoon]

No, there doesnt really have to be something better. There are many ways to view things, and certainly x86 is one of the ugliest instruction sets still in use.

But the modern x86 architecture has almost all the key features that make processors faster, and x64 has the one thing that x86 lacked (gratuitous amounts of registers)

Re:Intel's ill-gotten-gains

[SpazmodeusG]

Intel will be screwed if AMD goes bankrupt and the patents on a large part of the x86 tech fall into the hands of someone who has no desire to make x86 chips.
Currently they cross license to avoid a patent war. AMD going bankrupt will screw Intel over big time.

I'm still going my original guess for NVIDIA

[Akir]

I still think they're using Transmeta's engineers to run x86 code on their GPUs so they can get Windows to run on systems with other ISAs for their CPU. ARM and POWER, anyone? It sounds much cheaper and simpler than doing the insane amount of testing needed to roll out a new chip, and you'd get the added benefit of accellerating your everyday applications without needing to recompile them for CUDA. Plus NVIDIA will have the advantage of being the first ones out there with SSE5. So BAM!

Dumb Blog, And Not At All Correct

[rsmith-mac]

I suppose the NYT could be right, in the sense that they see NVIDIA getting an x86 license out of this in the same way that conspiracy theorists see that the Apollo 11 landings were filmed on a soundstage.

There's nothing about remedy 17 or remedy 18 that would lead to NVIDIA getting an x86 license directly from Intel. In short:

17: Intel has to license its chipset buses to other companies (e.g. NVIDIA) so that they can make chipsets for Intel's newest CPUs. NVIDIA only has an AGTL+ license for older Core 2 CPUs, they don't have one for DMI (low-end and mid-range Core i3/i5/i7) or QPI (high-end Core i7).

18: Intel can't get in the way of AMD's efforts to spin off their fabs in to Global Foundries. Up until AMD and Intel inked their own settlement, Intel intended to enforce provisions of AMD's x86 license that required them to do the vast majority of production in-house, which wasn't going to be possible if they spun-off their fabs.

The only way NVIDIA could end up with an x86 license out of this is that remedy 18 would allow VIA to transfer their x86 license, and in reality Intel has never fully acknowledged them having one. VIA only gets away with it because they have a couple of patents that are critical to Itanium, and those patents should be expiring soon.

So I don't know why the NYT is claiming that NVIDIA is going to get an x86 license out of this. This seems to be wild dreaming, or an attempt to generate traffic with ridiculous claims.

Re:Not necessarily.

[servognome]

Not only that, but what if Intel tries to leverage their monopoly to get Nvidia out of their graphics offerings, and instead tries to bundle their processors with their own integrated graphics chipsets? One of the FTC's complaints was that Intel was doing something pretty close to this on their netbook/atom platform. If they tried it on the higher end, I could see that backfiring for them.

You mean the same way Nvidia has integrated PhysX into their hardware and gone so far as to disable such acceleration if any additional cards made by a competitor are present.
The move to system on a chip is not an anti-competitive practice, it's the way the entire industry works. Third party hardware solutions have long been incorporated into mainstream designs as their silicon requirements decrease. Discrete math coprocessors and memory controllers were devoured by the CPU, video decoding and physics acceleration have been integrated into GPUs.
Why would SoC's from intel be considered anti-competitive, while AMD fusion and Nvidia Tegra, which are essentially the same, be considered innovative?
The FTC needs to consider whether the consumer would really benefit by forcing chipmakers to keep various pieces seperate for the sake of competition. The continued decline in average selling price, combined with the increasing capability of each new generation of microprocessor indicates that consumers are not negatively impacted by such design integration.

Re:Wow.

[drinkypoo]

When I recently went to price intel and AMD solutions, I ended up getting MB+CPU for less than intel's CPU with similar performance alone. And we're talking retail "black edition" overclockable CPU, not OEM, and a motherboard with every port I could want (well, OK, there's no fw800 on it) and support for overclocking, which I haven't even messed with yet. AMD's big problem is that they are fighting the perceptual technology leader — not the real one, but the one the public perceives as being there. It's much the same problem that's led to this as that which led to the perception that PowerPC chips were slow back when they were faster than intel chips: there were software problems (keep in mind, I'm talking about the K6 days) which mostly equate to problems with vendors.

Re:Wow.

[drinkypoo]

The intel part is superior only where the intel part costs twice as much or more. AMD is by far the leader in price:performance, regardless of what interconnect technology intel is using. An intel CPU twice as fast costs at least four times as much money, that's not a win!

Comment removed

[account_deleted]

Comment removed based on user account deletion