Free Capitalist Network - Community Archive
Mises Community Archive
An online community for fans of Austrian economics and libertarianism, featuring forums, user blogs, and more.

barriers to GPU making?

rated by 0 users
This post has 8 Replies | 2 Followers

Top 100 Contributor
Posts 814
Points 16,290
No2statism Posted: Sun, Jul 1 2012 3:19 PM

Besides patents, are there any other state barriers to entry?  Do you think the fact that there only 2 (nvidia and AMD) is due to market forces or due to the state?

  • | Post Points: 20
Top 75 Contributor
Posts 1,612
Points 29,515

intel is actually the #1 GPU manufacturer.  intel makes onboard graphics for their motherboards and they move more of those than nVidian and AMD do their gfx expansions.  There are many more than two, also; Matrox, Via, SIS, PowerVR.  The real barriers are the resources needed to produce them.  Yes, ATi (now AMD) and nVidia work hand in hand with M$ (DirectX API for almost every new video game) and Apple, this is why they are larger than Matrox.  I used to have a Power VR Kyro 64 and a 3dfx Voodoo Banshee (now pretty much nvidia) and they worked just as well as my Radeons and GeForces have.  Matrox and SIS make onboard graphics for motherboards.  The smaller GPU comapines just don't get contracts for the Xbox or gamecube, etc.

The GPU manufacturers work with many companies to produce the cards as well.  To be honest with you, concerning your OP, I don't think there are many state barriers to entry in the computer chip market.  Most CPUs and GPUs are defective and sold based on their level of defective-ness.  It costs a lot of money to make ten different versions of a chip and since the production process isn't that great (the machines and lasers make many many mistakes) they only try to make the top of the line models.  Every chip in a particular line of gfx cards is intended to be the same.  It is just that when they are produced they cut features and sell it for less due to the deficiencies in the process.  There is a lot of wasted money, so the defective units are dwarfed and sold as a lower end part of the line.  ATi used to cut rendering pipes off with a laser.  If you bought a medium line (always has been my price range) you can just look at the copper wires going to/fro the GPU itself and see.  The box on the medium range card might say 12 pipes where the top end model says 16 pipes and when you look at the medium range card you'd be able to see that you have 16 copper wires leading in/out of the GPU, but 4 of them had been severed by a laser.  Your card was defective and the defective process was disabled and sold as a lesser chip.

The above paragraph was a long winded way of saying that there are not many GPU manufacturers because it is expensive and because the ones that exist now are in bed with the OS compaines. (Not necessarily bad)

Hardware IP is very different that software IP.  Software can be reproduced by anyone, hardware requires resources not all are going to have available to them without big money funding.

And if you made this post because of the linux, nvidia, chinese government's custom OS, know that Linux linus was being a baby.  nVidia is not going to allow an open source market to craft their drivers out of their control.  If the linux community could make nVidia chips perform better than nvidia themselves, which is a real possibility, then they will lose tons and tons of money.  Not to mention that nVidia may not want the chinese government to have their hardware architechture knowing how they counterfeit things over there...

"The Fed does not make predictions. It makes forecasts..." - Mustang19
  • | Post Points: 20
Top 100 Contributor
Posts 814
Points 16,290

That's true, but intel doesn't make anything geared towards hardcore gamers and Matrox hasn't either since like 1999 (the G400 MAX, which was actually the very best at the time other than that many games ran best under glide back then).  nvidia's hardware is actually pretty good IMO, but their drivers are so terrible I don't know how they could still be in business.

Thanks for answering: )

  • | Post Points: 35
Top 10 Contributor
Male
Posts 6,885
Points 121,845
Clayton replied on Sun, Jul 1 2012 4:39 PM

I work for one of the above-mentioned companies. It is IP law which is the driving force behind the scale of the major tech giants. The fact that hardware cannot be copied easily is immaterial. It is the patenting of more-or-less "obvious" electrical circuits that stalls innovation if for no other reason than that it is very costly just to be sure your chip doesn't contain anyone else's patented electrical circuit or, if it does, that you are reasonably sure that you will win in the event of a court challenge. You have to employ a small army of patent lawyers to do this.

Clayton -

http://voluntaryistreader.wordpress.com
  • | Post Points: 20
Top 75 Contributor
Posts 1,612
Points 29,515

Matrox hasn't either since like 1999

The last matrox attempt at the ATi and nVidia market that I remember was this - Matrox Parhelia.

nvidia's hardware is actually pretty good IMO, but their drivers are so terrible I don't know how they could still be in business.

I like nVidia.  I'm running some beta drivers now on Win 7 64 with a GeF 460 and have no problems other than I cannot run the DX11 version of my pirated CivV.  But I don't think that has to do with the gfx drivers.

"The Fed does not make predictions. It makes forecasts..." - Mustang19
  • | Post Points: 5
Top 75 Contributor
Posts 1,612
Points 29,515

The fact that hardware cannot be copied easily is immaterial.

Well, for this point, but that is a significant reason why the market for computer chips in general is smaller than say...bottled water.

It is the patenting of more-or-less "obvious" electrical circuits that stalls innovation if for no other reason than that it is very costly just to be sure your chip doesn't contain anyone else's patented electrical circuit or, if it does, that you are reasonably sure that you will win in the event of a court challenge. You have to employ a small army of patent lawyers to do this.

Are you referring to the GPU architecture?  I noticed a few years ago that ATI and nVidia went down pretty different development patterns (so much so that they are difficult to compare in terms of power/ability with the given information) and thought to myself, they must not be worried about the fact that their "chip doesn't contain anyone else's patented electrical circuit" anymore. 

Also, at what level does API come into this?  The simple fact that AMD and nVidia develop their architecture hand in hand with M$ API seems like it is still software IP that limits the developmental capabilities of hardware.  It is simply a developmental resource that is not available to all chip manufacturers (contracts with Sony, Nintendo, Apple, and M$).  Those contracts help strengthen market confidence in the ability and sales (almost necessity) in way s that SiS and Matrox cannot anymore

 

"The Fed does not make predictions. It makes forecasts..." - Mustang19
  • | Post Points: 20
Top 10 Contributor
Male
Posts 6,885
Points 121,845
Clayton replied on Sun, Jul 1 2012 5:10 PM

at what level does API come into this?

It just creates an interlocking SW/HW IP barrier because APIs themselves can be patented. The idea of patenting APIs is no less absurd than any software or hardware patent but its effects are particularly vicious because it's basically saying "not only can you not create HW/SW that copies or mimics ours, but you cannot even create HW/SW that talks to ours without our permission."

This creates a monopoly/monopsony relationship between the respective hardware and software companies that are producing the given solution (which necessarily requires both hardware and software components to actually work). The consumer technology market is a huge, interlocking network of such sweetheart relationships.

The trust-busting types would want to focus on the sweetheart relationships themselves as the problem but they are just symptoms of the problem. The problem is IP law. That is what has created this monster and until we kill IP, it will never die.

Clayton -

http://voluntaryistreader.wordpress.com
  • | Post Points: 20
Top 75 Contributor
Posts 1,612
Points 29,515

This is pretty much what I thought.

In the real world where IP laws do exist.  I can understand nVidias reticence to turn over their driver and architecture information to the CCP due to the loss of money that is guaranteed by the IP.

This creates a monopoly/monopsony relationship between the respective hardware and software companies that are producing the given solution (which necessarily requires both hardware and software components to actually work). The consumer technology market is a huge, interlocking network of such sweetheart relationships.

I cannot help but think that without IP, companies could still agree to this...I don't think Rothbard would contest nVidias decision to work with M$ to promise each other sales through working together and preventing competition from, like you said "even talking" to their systems.

"The Fed does not make predictions. It makes forecasts..." - Mustang19
  • | Post Points: 20
Top 100 Contributor
Posts 814
Points 16,290

@ Aristophanes:  It's true that they'd be allowed to work with each other, but remember that MS got as big as it has because of IP.   Programmers would just be programming to the metal if there were no IP.   DX really isn't that wonderful, people just think it is because they haven't seen anything different.

  • | Post Points: 5
Page 1 of 1 (9 items) | RSS