Slashdot Mirror
NVIDIA To Buy AGEIA
The two companies announced today that NVIDIA will acquire PhysX maker AGEIA; terms were not disclosed. The Daily Tech is one of the few covering the news to go much beyond the press release, mentioning that AMD considered buying AGEIA last November but passed, and that the combination positions NVIDIA to compete with Intel on a second front, beyond the GPU — as Intel purchased AGEIA competitor Havok last September. While NVIDIA talked about supporting the PhysX engine on their GPUs, it's not clear whether AGEIA's hardware-based physics accelerator will play any part in that. AMD declared GPU physics dead last year, but NVIDIA at least presumably begs to differ. The coverage over at PC Perspectives goes into more depth on what the acquisition portends for the future of physics, on the GPU or elsewhere.
I always thought that GPU + physics engine would be a perfect combination. Ultimately, the AGEIA card is just a DSP + software driver for calculating physics. A GPU is... also a DSP + software driver for calculating graphics. It wouldn't be too hard to write a driver that does both: some of the pipelines could be allocated to graphics, and some to physics. Might even make a software-configurable to dedicate more/less units to physics.
___
If you think big enough, you'll never have to do it.
The interesting thing about the processor on an AGEIA card is it's similair in design to an IBM Cell processor. Just a fewer number of SPE's...
I can't seem to find the link to the paper that discussed it in detail, if I can find it I'll post it later...
I'm sick of following my dreams. I'm just going to ask where they're goin' and hook up with 'em later.
http://en.wikipedia.org/wiki/Physics_processing_unit#Cell_Processor_vs_PPUs
http://en.wikipedia.org/wiki/Physics_processing_unit#GPUs_vs_PPUs
There are differences. Otherwise Sony wouldn't have wet themselves when they announced Cell technology in the PS3 or Microsoft could of countered their ATI GPU was pretty much the same thing or more powerful or however the market types would of spun it if that was the case
I'm sick of following my dreams. I'm just going to ask where they're goin' and hook up with 'em later.
Unless something's changed in the past year or two it's been since I stopped using Nvidia, their drivers always tended to be quite good.
They were Binary-only, but they were good in that they were fast, stable, and supported all the major functions of their cards. Hardly half-assed if you ask me.
-- If you try to fail and succeed, which have you done? - Uli's moose
I completly disagree, so nvidia dont open source theyre driver, but at the end of the day they release good binaries, I see no advantage to open source drivers for videocards:
:'( )
*a community isn't going to develop video card drivers as well as the people who make the cards
*a community is much more likely to stall and slow down
*in most cases the fact software is open source doesn't mean much as one company or another has complete control over the product (look at OO)
the only arguments for it are that
*more people will find the bugs (this is a mu point, look at FF, plenty of eyes on code but still plenty of bugs)
*some genius could improve it (look at OO it needs serious work in some areas but nobody bothers)
*there could be spyware in a binary ( stop being paranoid)
the end effect of open sourcing the drivers will be similar to open sourcing secondlife, it just means that its easier to cheat with, no major work has been done on second life but a few people have figured out how to gain unfair advantages ( in the end it will either have to be closed (impossible) or they will waste CPU making sure your not cheating (a real pain) ).
Nvidia are fully committed to linux, they release public betas that are usable for linux, sure they might come a bit later than windows but they do come. Why are ATI open sourcing thier drivers, im guessing because linux users were switching to nvidia as thier drivers worked, either that or they cant be arsed to support linux anymore.
p.s titanic special effects were done on linux-nvidia clusters.
ANYWAY...my point was that this is great news because it means that linux will get fully supported physics cards, meaning some graphics effects can become physical or we can do some 3d physics on the desktop (not sure what we could do maybe throw windows inside the cube? meh i dont even have compiz
All we need after that are a few opensource to take full advantage, of it.
*Hell even for non linux users this is good news, if nvidia release seperate cards then, linux servers can start taking advantage of server side physics, and allow even physics card less users to benifit!
IranAir Flight 655 never forget!
except Intel doesn't think they need Nvidia... that's why they've got nearly all the notebook vendors pumping out crappy built-in graphics that just barely run Windows Vista. ATI saw the writing on the wall and got themselves bought by AMD. Now AMD battles on CPUs, integrated graphics, and high end graphics... Intel can never buy Nvidia because they'd be instantly sued. Nvidia overpriced themselves, even with all the work they did for AMD, and the matching logos... stock holders were just too rich for AMD.
This makes Nvidia the "odd man out" because they don't make processors. Both Intel and AMD have integrated solutions and obviously want physics processing on the CPU so that they can sell 7 core 3.21GHz processors. NVidia has to break the mold if they want sales... they got shunned the last round of consoles for IBM and ATI, and Microsoft pretty much let ATI write the book for DX10 this round. NVidia + Ageia only makes sense if they'll make an open source console that runs either AMD or Intel CPUs. Games would need to run flawlessly, without "installing" just like a console. There's a hole for PC gaming right now... Apple's not filling it (they think it's stupid) Wintel is not helping (Microsoft only wants Vista gaming, and Intel wants to sell integrated graphics) so a well done Linux console could help... but there's too much IP in the way to make it happen.
That's only true from a *very* large distance. The differences far exceed the similarities. Both chips have FMAC units, sure. But Cell is a moderately-pipelined single threaded RISC design while a GPU is massively-pipelined + hyperthreaded CISC. For instance, the Cell has no exp or sincos instruction, and the RSX can read a bilinearly-filtered texel with zero apparent latency by switching between O(1,000) threads once per cycle.
The memory systems of both are also radically different, with the SPU providing a single 256K area of unmapped memory and the GPU providing various caches at various levels with all manner of data reorganization and duplication going on under the covers.
What the hell is a "DSP" anyway, these days? The term was coined to cover architectures that (a) had a Harvard architecture and (b) could pipeline one FMAC per cycle. ALL architectures do those two things now.
The 56000 is ancient history, my friend, and the modern derived architectures are far, far more complex and varied than you give them credit for.
Ironically, in a few years, your comment will be correct, as GPU design is converging with designs like the Cell. But it ain't there yet, not by a long chalk.
Well, there's already Open Dynamics Engine.
Sigs are for losers
But Sun doesn't have an x86 processor, and this is the key.
Nvidia needs a an x86 processor to compete. Sure, Nvidia could just adapt their GPU architecture and expand the language to make a general-purpose VLIW processor. They could package it and sell it as an Itanium competitor. But nobody wants to use a non-x86 chip in mainstream markets, and that's where the long-term money is.
This is why Windows, Linux, Solaris, BSD, and now even OS X run on x86: if your OS has redeeming or unique qualities, more people will buy based on OS features alone if your hardware platform is agnostic.
And herein lies two problems: one, while you can make x86 processors without a license, you are constantly in danger of litigation from Intel's massive patent portfolio. In the last two decades, every x86 chipmaker has eventually negotiated a cross-license agreement with Intel. The other problem is, it is hard to build a new x86 processor from-scratch. Thus, a takeover bid for an x86 processor manufacturer is likely the best way to solve Nvidia's problem; they get a license to keep Intel at-bay, and a solid starting point.
I'm thinking Via, personally. Their sales have slumped in the last year, and they've stopped making Intel chipsets. In fact, Intel has been bullying poor Via for the last year, offering a new Intel chipset license if they just stop manufacturing CPUs. Either Nvidia will buy Via, or Via will spin-off their processor division for some cash. Thanks to the Intel cross-license Via purchased along with IDT, their processor arm is a goldmine in the long-run.
Man is the animal that laughs.
And occasionally whores for Karma.