ATI & Nvidia to slowdown competiton (inquirer "speculation")

[harl]

http://www.theinquirer.net/?article=16111

Maybe this is only Fuad's paronia...

But, anyone belive this could be posible in the next generation?
(or even in this generation)
that's are the beneficts of the oligopoly ...

[Rytr]

Retap? I did not think the Nv40 was a re-tap? I thought it was a complete new design. Anyway, I don't think a purposely inacted technology slow down will happen as that would potentially open the door for other IHVs. Suppressing implementation of that technology is possible, if both could come to an agreement. Something akin to price fixing, IMO. Hopefully, the 'die-hards' maintain the upper hand without over-taxing resources of either company.

[Nutty]

I believe it. Its getting harder and harder to produce top end cards. You cant just keep shrinking the frabriction process forever. There are physical limits that we will soon reach in a few years, and then we'll have to try something different. When the silicon paths get too small for an electron to move along, what do we do then ?

Then theres also power requirements. In about 2 years, we'll be nearing 500M transistors, then 2 years after than 1billion ?? All for what ? For gamers to play games on. Thats where most of NV's money comes from. Can you imagine the cost in produce a 1Billion transistor chip ?

[harl]

Can you imagine the cost in produce a 1Billion transistor chip ?

Montecito (Itanium II) has 1.7 Billon transitors
(I think in more than 1 chip, 1 for cores, 2 or 4 for cache?)

[coldpower27]

If they reach the 65 nm process in time for a 1 Billion Transistors chip they might be ok, since it will only be slightly bigger then 6800 Series is now :clap:

[Nutty]

Montecito (Itanium II) has 1.7 Billon transitors
(I think in more than 1 chip, 1 for cores, 2 or 4 for cache?)

And how much does of those cost ?

[Gator]

it makes perfect sense, I can't beleive how quickly video cards are being upgraded these days. Can anyone recall how long the Rage128 series was out before Radeon came on the scene? I think it was a pretty long time.

If they slow down, then when they do come out with the new stuff it'll be worth having and fewer bugs and more features.

Furthermore, it's annoying for us consumers to buy top of the line, only to have it be yesterday's crap before long.

[digitalwanderer]

Didn't ATi & nVidia say the same thing about a year, year and a half ago?

[Evildeus]

I think NV4* will last till DX10/Next so 18 months approximatively i think.

[harl]

Well it will be nice if they decide to make longer th life of
their card, if they make also biggers jumps in performance
(like skipping generations)

[nutball]

Montecito (Itanium II) has 1.7 Billon transitors
(I think in more than 1 chip, 1 for cores, 2 or 4 for cache?)

Montecito is a very different beast. The vast bulk of those transistors are in the (24MB?) cache, the cores are a very small fraction. The caches are designed to be self-healing, divided into small blocks so that any blocks containing defects can be mapped out. AFAIK cache uses nowhere near as much power as core logic; GPUs are pretty much all core logic.

The GPU guys do a similar thing with the disabling of faulty pipelines in the top-end chips and selling them as lower models, but there's only so far they can go with this scheme.

My bet for future GPUs is the multi-chip solution. That doesn't solve the overall power problem at the card level, but might make building individual GPUs that actually work in sensible quantities a little easier.

[borntosoul]

i cant see them agreeing to slow down at all. look at what theyve done just recently to stay ahead of each other(all the stunts theyve pulled). only thing will slow them down is internal choice to do so for more profit. i also think they will be going the same road as AMD/intel with dual chip as well.

[harl]

Dual core seems meaningless in grafics,
since they are already parelel archtictures
(X pipes, X shaders)

[nutball]

Dual core seems meaningless in grafics,
since they are already parelel archtictures
(X pipes, X shaders)

That makes it even more appealing. It's already been done don't forget (3dfx to name but one).

[harl]

Dual core it's different from Dual chip

Dual core it's made to take advantage of high
transistros count on chip

Dual chip is made to improve performance
without making expensive chips (but mading the PCB more expensive)
(2 chips with 2 pipes instead of 1 chip with 4 pipes)

[nForceMan]

ATI has already slowed down. No SM3.0, no floating point blending, no FP32, just the 2 years old crap on steroids. :rofl

[harl]

But it seems to be enough to beat Nvidia :fkr: (daz) (duel)
(at least in Quality/performance AA)

[smeagol]

ATI has already slowed down. No SM3.0, no floating point blending, no FP32, just the 2 years old crap on steroids. :rofl

But it's still beating the crap out of nextgen hardware.

[Nutty]

Hardly beating the crap imo.

[nForceMan]

But it seems to be enough to beat Nvidia :fkr: (daz) (duel)
(at least in Quality/performance AA)
I agree. ATI is definitely better in cheating! :lol: