Whats Upcoming In The GPU Industry From AMD & NVIDIA? Do I Sell & Wait?

[UseItNow]

Hiya,

Anyone have any idea on what higher range GPUs are coming out this year if that is the case from AMD and Nvidia?

Would want an improvement on a GTX570 I'll be replacing soon.

It'll most likely go into a HTPC (which is used for gaming), so would need to have 24p output without issues.

Any updates as to what'll be coming possibly will be great, are there any leaks of performance increase the new ranges may bring?

Thanks.

[scaryjim]

Not sure. I'm pretty sure AMDs next line will be another VLIW hash - probably a full line of VLIW4 cards (althoguh that's me guessing wildly rather than based on any rumours I've picked up). After that they're hoping to move to an entirely new, more compute-oriented architecture which I believe they are calling Graphics Next or some such nonsense. That's a year or two away though.

I've not heard about nvidia's plans at all. One assumes they'll have an evolution of Fermi on the cards, but whether that will mean more shaders, faster clocks, or what, I really don't know.

It's unusually quiet on the graphics rumour-mill, at the minute...

[CAT-THE-FIFTH]

AMD might move over to a new architechure for its high end HD7000 or HD8000 series cards. They were talking about it at one of their events.

[kalniel]

Anyone have any idea on what higher range GPUs are coming out this year if that is the case from AMD and Nvidia?

Almost certainly none.

Both are moving to 28nm processes, and it doesn't look like the high end will be ready at the end of this year.

nVidia are working on the fermi replacement, which doesn't look so far to be anything revolutionary - just taking the efficiency improvements of the gtx 500 series line and doing more of it at 28nm.

Like above, AMD have a new thing coming, but it's going to take a long while - they're only talking about it now because they want developers onboard and need to warn them in advance. And that thing is basically moving towards are more nVidia-like compute model. In the meantime they have a very efficient graphics acceleration model - their latest VLIW4 architectures were very successful so I'd expect more of the same at 28nm as well, depending on how many things they held back from the 40nm line that was meant to be 32nm.

[shaithis]

I'm wondering if either of them are trying to license the 3D transistor tech from Intel.....

[kalniel]

I'm wondering if either of them are trying to license the 3D transistor tech from Intel.....

That's a fab tech really - neither AMD or nVida have fabs anymore.

[Pob255]

I wonder if there will be the production issues there where with the current 40nm?
(if you remember back there was a small delay on the 5xxx series from ati due to them having to make changes to the layout, although they caught them early so the delays pale compared to the issues nvidia had with the 4xx series)
If ati still go their standard route then we should expect to see a mid ranged card in 32nm, a sort of hybrid card from the current series, before they do the full shift over.
And normally ati embrace changes more than nvidia who tend to go more the refinement of an existing architecture.

[kalniel]

I wonder if there will be the production issues there where with the current 40nm?
(if you remember back there was a small delay on the 5xxx series from ati due to them having to make changes to the layout, although they caught them early so the delays pale compared to the issues nvidia had with the 4xx series)
If ati still go their standard route then we should expect to see a mid ranged card in 32nm, a sort of hybrid card from the current series, before they do the full shift over.
And normally ati embrace changes more than nvidia who tend to go more the refinement of an existing architecture.

There is no 32nm. TSMC cancelled it - AMD had to re-jig and release their 32nm designs as 40nm chips.

nVidia claim they've learned their lessons from the 40nm fisco:
http://www.xbitlabs.com/news/other/d...echnology.html

[Pob255]

sorry I meant 28nm

We were successful on so many different nodes for so long that we all collectively, as an industry, forgot how hard it is.

Umm no nvidia just you