![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
and it goes both ways. But if we just look at AMD's information they have stated so far and shown, we can't get any info on power consumption and performance from this chip. Out side of 30% performance increase with the same power consumption or 50-60% (made a mistake before thought they stated 70%) at the same performance from the new node. Which is quite underwhelming, but this statement you have to look into what is the ideal power consumption for the die size and for the new architecture. There are too many if's and buts about what they showed to even presume anything. There are ideal die sizes, because the 950ti isn't the most power efficient chip out of the 9x0 family, the gtx 970 and gtx 980 are, and actually the 750ti is better too. It's really just too early and too loose a test to garner much more than it's not a space heater compared to what we presently have. It may well be a space heater relative to 14/16nm offerings.Hot shader clock wasn't Fermi,it was dropped for Fermi, unless you were reading Charlie's rumor mill blog (yeah the wood screw thing was a fuck up) which was so of base it wasn't funny, g80 and its derivatives had hot clocks and only that one generation. The hot clock thing wasn't about die space savings either.. Actually if you look at the architecture Keplar, Fermi, g80 they were all at the reticle limit of the process they were on. All of them around the 600mm2 if not I'm not mistaken, so the ones with hot clocks vs the ones without hot clocks didn't shift the size of the gpu (using more transistors vs less, hard to say). The G80 and the g200's may have had to drop transistor density because of the hot clocks, which we don't really know, but I can see why would have too since the higher clocks will lead to more leakage)Also, the reason nVidia had the hotclock shader clock on Fermi and Tesla (can't remember if Tesla was the first or if some of the earlier designs had it too) was a die size/power trade off. You have half as many shaders running twice as fast, although this costs in power because you have to raise the voltage to raise the clock frequency and that doesnt scale linearly, power wise, as we all know.
With kepler, nVidia was able to make the opposite tradeoff, as they could put twice as many shaders going ~slower and use more die space, but the 28nm process let them do that and stay within their target die-size, thus making the chip more efficient.
So I take that back., they both use the same Fermi architecture, well gen to gen tesla uses that gens architecture.Oh don't tell me about crap from Charlie, that is him talking right there. Again those things are so off base it wasn't funny.
Tesla was an arch as well, GTxx chips.Edit
Sorry it was Kepler they dropped the hot clocks, forgot, its been a while and acutally thats when I stepped away from looking at graphics card 7 years ago
BTW Tesla is a line Fermi is an architecture, two different things, don't want to confuse those two up
Oh don't tell me about crap from Charlie, that is him talking right there. Again those things are so off base it wasn't funny.
Ah ok, I didn't mean that, sorry if you read it that way. Yes now I agree with what you are saying lol . yeah its not a full node jump its a half node.
Its less than the same node without finfet is what I'm saying not from previous nodes.
EDIT: I'm not intending to come off as argumentative here, so sorry if that's how it seems.
And yeah, I am not sure why some articles are saying it's a full 2 node jump. I mean yeah you get a big benefit from the FinFet's themselves, but I would say closer to 1.5 -- and we need it after 5 years on 28nm.Arguments are ok, as long as we can see where they are coming from and get to a point of agreement
The purpose of the demo was twofold. First, it proved that AMD already has functioning 14nm GPUs. Secondly, it showcased the power efficiency of the new 14nm process and AMD's new Polaris video cards:
Given the fact that power efficiency has suddenly become the most important metric for purchasing a video card, I expect a lot of current Nvidia owners to jump ship to AMD once Polaris cards hit the store shelves.
No, chime in peacefully.