Pace of technology slowing down?

[natty1]

In graphics, it's been 10 months since they've released a card with significant performance increase (5870). In CPU's, it's been even longer. And worst of all, the prices have not been dropping much in either category. 5870 is still selling for the same price it did last September. Intel CPU prices are completely stagnant. Why are things slowing down here?

 

[mwmorph]

It'ss been like that for a while now.

GPUs used to be on 12 month cycles where you could huge performance gains (2x then 1.5x a la 5800->6800 and then 6800->7800) but it's slowed down considerably as GPUs get bigger and more power hungry and complex.

 

[GaiaHunter]

In graphics, it's been 10 months since they've released a card with significant performance increase (5870). In CPU's, it's been even longer. And worst of all, the prices have not been dropping much in either category. 5870 is still selling for the same price it did last September. Intel CPU prices are completely stagnant. Why are things slowing down here?

Lack of competition in both situations.

Additionally 10 months isn't that much.

GF 8 November 2006.

Cards with significant performance boost over it were only launched in July 2008 in form of GTX280/260 and 4870. Almost 2 years later.

Cards with significant performance boost over GT200 and RV770 were released in the form of 5800 series in September 2009. Barely 1 year later.

Now the lack of price drops in GPUs have been caused by several factors - TSMC 40nm process problems and then NVIDIA own problems combined with TSMC.

Maybe the new GTX 460, that will be released on the 12th, will finally bring price pressure.

 

[thedosbox]

Lack of competition also plays a part. AMD don't really have anything that can compete with the higher-end CPU's from Intel, so why should Intel lower prices?

There's a similar situation playing out on the video card side where Nvidia don't (currently) have a serious competitor for most of the sub-$300 market, and the parts above $300 are not positioned directly against each other.

 

[nenforcer]

Try the worldwide global recession playing a big part in product timelines being lengthened. The next generation consoles are not suppose to arrive until 2012 now at the earliest with Microsoft releasing the 'Kinect' and Sony adding the 'Move' controllers as the only hardware the next few years.

 

[zephyrprime]

The real problem is we've hit the wall with Moores law. People keep talking about extending Moore's law but it's BS. When processors staled out at 3ghz, we were screwed. And traditionaly von neumann architechture is staling out too. Multicore just isn't scalable enough easily enough and superscaler is pretty much played out. GPUs don't have it so bad but they have some problems too. We really need some breakthrough technology. Something totally different.

 

[Dark4ng3l]

The real problem is not moores law but the thermal limit. Chips are too hot for how small they are. The result is that we don't get clock speed increases at the same rate as we used to because you can't ramp up the power consumption any higher than it is right now. We are still doubling the amount of logic on the chips but with almost no clock speed ramp unlike 10 years ago when both clock speed and the amount of logic were going up lightning fast.

 

[Ben90]

While I agree with Dark4ng3l for the most part, I think the reasons are different.

It is true that having no thermal limits would yield 5ghz screamers from both companies, however, eventually frequency scaling is limited by factors other than heat. I see clock speeds as a temporary solution to the advance of technology. IPC is king.

The only thing that will hold back Moore's law is power constraints. Yes eventually we will no longer be able to shrink transistors, but die size wise, it won't be a problem as stacked transistors will basically allow us to place many orders of magnitude more transistors using even the 32nm node.

Thermal management is going to be a huge factor in the design of 3d microprocessors. The amount of power possible per CM^3 is going to be insane. This is when I see thermal management to be much more of a problem then it currently is.

I am very interested in the development of High Temperature Super Conductors, as this would circumvent the thermal problems and yield some ridiculously powerful microprocessors.