Slashdot Mirror
Nvidia Is Trying To Make an x86 Chip
Slatterz writes with a story from PC Authority which says that "Word has reached us that Nvidia is definitely
working on an x86 chip and the firm is heavily recruiting x86 engineers all over Silicon Valley. The history behind this can be summarised by saying they bought an x86 team, and don't have a licence to make the parts. Given that the firm burned about every bridge imaginable with the two companies who can give them licences, Nvidia has about a zero chance of getting one."
Except that Intel and AMD hold vital patents to the set of technologies that are part of the x86 architeture. They have to cross license because they depend on each other, but they have no obligation to license to NVidia.
how does pcauthority.com.au get away with re-posting others articles without even linking back to the original source (yes, I know that they credit theinquirer.net at the top, however it just links to all articles stolen from theinquirer.net).
Why does a firm wishing to enter the x86 market need to buy licenses
They're probably alluding to possible patents held. Of course, NVidia has them in the graphics part and could leverage that anyway. Just another reason why patents need to be scrapped and replaced with a non-exclusive system of financial incentive, if we need one at all.
however did AMD come to own any
Ancient history. AMD got into the x86 market in the 80's when the USG required multiple sources for many components, so Intel was more or less forced to let them in if they wanted USG business. Once they were established they've worked on improvements themselves which they license to Intel, etc.
The PC Authority site got slashdotted, but this sounds terribly like Charlie Demerijan's article from 2 days ago.
And while Charlie's articles are terribly fun to read, they don't quite qualify as news. Call them rants, speculation, whatever you wish, but not news. At least unless they get picked up blindly by other publications...
Intel licensed a bunch of stuff to VIA after a legal battle some years ago.
here: http://arstechnica.com/old/content/2003/04/397.ars
What ? Me, worry ?
Some will be expired, but the technology employed on the current chips (state of the art and previous generations) are covered by more recent patents, and if NVidia wants to produce anything more advanced that the good old 8086, they will have to negotiate.
Check this and this articles. That shows the heavy politics involved between the big processor companies in order to be able to produce our beloved processors.
Perhaps they could be making GPGPU that with a translation layer for x86 instructions, like the Transmeta Crusoe did in VLIW, or maybe they are enhancing a Via Nano CPU design with on die GPU (rather like they did with the Tergra ARM11 chip). Either way this won't be a desktop CPU, and it won't be serious competition for Intel, but could be targeted at the growing netbook market.
Intel could step in and try to block them, but they have lost against Via and Transmeta in the past, and they would also put themselves in a difficult situation, since they are being watched in the US, EU and Asia for antitrust violations. This would look quite bad for them.
... sortof. NVIDIA has a 386(!) SoC from the acquisition of ULI.
I'm skeptical about a new entrant like NVIDIA gaining any traction in the x86 market, they would have better luck pushing out their ARM chips.
No, the author is Charlie Demerjian from The Inquirer. Some years ago Charlie broke a NDA, so nVidia has removed him from the pool of journalists given notice of new releases. Since then Charlie writes only negative things ("they are broke", "they produce only faulty chips", "ATI is much faster", "CUDA stinks", "3D glasses are no good", etc. etc.) about nVidia. I've a spam filter about "news" about nVidia by Charlie (it's a pity slashdot reports this junk...)
They may have the base architecture available, but not any of the fancy simd or 64-bit instruction sets.
First appearances (not necessarily patent dates):
MMX - 1997
3DNow! - 1998
SSE - 1999
SSE2 - 2001
AMD 64 - 2003
Intel 64 - 2004
SSE3 - 2004
SSE4 - 2006
Of course, most software doesn't use any of these extensions, but Intel and AMD can use this as a weapon in a possible FUD campaign.
Back in the day, many purchasers demanded that manufacturers of electronics had a secound source of components so you wouldn't get stuck with a product line you could no longer build. AMD was Intel's second source provider. This agreement went to court http://findarticles.com/p/articles/mi_m0EKF/is_n1961_v39/ai_13734404 and the result was a forced agreement that meant AMD had access to Intel intel.
There are places where the networks are not touching,and there are places where they are-Boeing's Lori Gunter
C allows for things that just don't make sense on GPUs. Arbitrary branching, pointer aliasing, etc. are poisonous for GPU performance.
GPUs excel at tasks that map N input values to one output value, with a minimum amount of unpredictable branches. If a task fits in this well, it is likely being accelerated already, via CUDA, Stream, CTM. If it doesn't fit, forcing it on the GPU is a waste of time.
What you want to look at are things like C++ DSELs, which create expression templates out of compile-time defined language specifications. This way, you can have a "shader language" that is evaluated at compile-time, either to a "real" shading language, or to plain old C++ code for the CPU.
This sig does not contain any SCO code.
So why not just buy Via? They have the license to make x86, and more importantly they have low power CPUs that are ready to go, and with Netbooks and Notebooks taking a big chunk out of the market this would give them a BIG advantage in the market.
3 words: Ownership Transfer Clause
Intel is already waving that sword at the offsprings of their soon-to-be-late AMD competitor (namely, the question whether The Foundry Company will be covered by the x-licenses or not). Usually licensing agreements are set to be terminated if ownership of the licensee passes to a third party, so NVidia might even get a total of zero licenses if it buys Via.
No, the above post really overstates what goes on inside today's x86 chips.
It is true that Intel and AMD internally break up x86 into simpler "micro-ops" to simplify the internals of the chip. However, the specific micro-ops uses are tailored explicitly for x86 instructions, and many match up with x86 instructions one-to-one. The mapping really isn't that programmable, either. Most of the mapping is hard-coded and highly optimized. It would not be trivial to support another ISA such as PowerPC, even for just user-mode instructions. If you then consider all the privileged instructions, virtual memory, and virtualization stuff, you have a real mess. It would likely be easier to start from scratch rather than try to retrofit a current x86 to be anything other than an x86. Sure, you could reuse some of the arithmetic units and memory controllers perhaps, but the core would have to change pretty dramatically.
That said, Transmeta (RIP) did have technology that would likely make it easier to run non-x86 code on its processor, and the translation was done in software. But even its internal instructions were likely closely match to specifics of the x86 ISA.
VIA Technologies Inc.'s Market Cap is 461.0M according to: http://investing.businessweek.com/research/stocks/snapshot/snapshot.asp?capId=679305