http://semiaccurate.com/2013/08/07/amd-to-launch-hawaii-in-hawaii/So now you know what AMD is launching, a 28nm new-generation GPU called Hawaii in late September. You also know what Nvidia is doing as a counter, basically nothing, and subscribers who follow the links will also get the reasons why, this being a key one. And lastly but somewhat anecdotally you have an understanding about the demographic changes behind Hawaii in Hawaii, it does actually make sense beyond the name.
It would show that they are serious about competiting with nvidia which I think many buyers would be swayed by a little fighting spirit from amd.
But they are not competing, they've been trying to keep up for quite some time now.
And following your logic confirms it. (release a bit updated cards just to match nVidia).
Competing would be releasing a significantly better cards.
What information are you privy to that says these upcoming cards will not be faster?
I just can't imagine how AMD can beat the highly successful 780/Titan if they stay on their current tech branch.
Top end isn't where the money is. There's far more revenue in sales from the mid-range market. Very few people are willing to drop $600 on a new video card every year.
The discrete desktop graphics market is slowly dying.
But they are not competing, they've been trying to keep up for quite some time now.
And following your logic confirms it. (release a bit updated cards just to match nVidia).
Competing would be releasing a significantly better cards.
I just can't imagine how AMD can beat the highly successful 780/Titan if they stay on their current tech branch.
um even when the next gen 7970 launched it was only around 35-40% faster than the 6970 so dont act like 30% is nothing for refresh. heck in the techpowerup 7970 launch review the 7970 was only 30% faster than the 6970. so if anything 30% will be a massive performance increase for a refresh on the same node.You dont think 30% is within reach of a refresh, on a tweaked architecture, mature node? Come on man, be real. 30% is NOTHING. If they dont get 30%, they failed.
and I am sure you know that we dont normally get refreshes on the same node that are anywhere near as fast as an architecture change on smaller node. the 7970 ghz is already a 250 watt tdp card so regardless of what you claim, a 30% real world performance increase will be massive in this case.You know full well the 7970 is a lot faster now, because GCN is matured. In newer games, its close to 100% faster.
The "refresh" is a tweaked GCN, obviously it wont suffer from the VLIW to GCN transition. Im sure someone of your background in GPU would know that. Coupled with a mature node, and heaps of clock speed headroom, if they cant deliver 30% they arent even trying.
Previous "refreshes" did not include a (rumoured) much bigger chip with more of everything, and this seems to me like a stopgap because TSMC cannot ramp up 20nm fast enough. It looks and smells like a compromise, something meant for 20nm that had to be cut down and squeezed into 28nm.
again the point is you act like 30% is no big deal. in reality its basically unheard of for a refresh especially on the same node. and also again the 7970 ghz is hardly a low power chip to begin with. it already uses as much power as the Titan.
the 7790 is not as efficient as people claim. its maybe about 5% more efficient. just a mild 75 factory oc makes it use the same max power as the 7850 even though the 7850 is 20% faster than the slightly oced 7790. http://www.techpowerup.com/reviews/ASUS/HD_7790_DirectCU_II_OC/24.html7970 was pushed beyond optimal clocks for that silicon (OC to compensate for new NV products). When Hawaii is "designed" to be run at a given speed, rather than pushed there like Tahiti, it should help power consumption. Look at Bonaire.
the 7790 is not as efficient as people claim. its maybe about 5% more efficient. just a mild 75 factory oc makes it use the same max power as the 7850 even though the 7850 is 20% faster than the slightly oced 7790. http://www.techpowerup.com/reviews/ASUS/HD_7790_DirectCU_II_OC/24.html
7970 was pushed beyond optimal clocks for that silicon (OC to compensate for new NV products). When Hawaii is "designed" to be run at a given speed, rather than pushed there like Tahiti, it should help power consumption. Look at Bonaire.
and I am sure you know that we dont normally get refreshes on the same node that are anywhere near as fast as an architecture change on smaller node. the 7970 ghz is already a 250 watt tdp card so regardless of what you claim, a 30% real world performance increase will be massive in this case.
Semiaccurate's charlie posted months back Volcanic Islands will be an great architectural improvement over the original GCN core.
Sea Islands is the barely tweaked GCN cores found on 7730/7790 and many of 8000m series GPUs.
Sea Islands is not going to be a 2012 product, it is looking like a March 2013 release for now. Things could of course change, but that is unlikely given what the card is. CI is a 28nm evolution of the current 7xxx line of cards, think optimizations rather than massively new features.
yeah..............still waiting!with more measured sources saying +15% is more likely. AMD is power bound and die size equals cost, so what you are looking at is process improvements and algorithm optimizations rather than any new big bangs.S|A
Yes, thats a lot of information for chips that never existed. What a load of ........lies! Of course I am sure ole Charlie will come up with a quick turn around, "the gk114 was UN-manufacturer-able" and his moles told him that the entire line was scrapped......but what about the performance he quotes so conveniently? Oh, uhm.....you know......it got scrapped right after they, they...ran a bunch of benchmark data to leak to charlie D personally. Right after that all the card blew up so no one else on gods green earth seen them..........yeah, thats how it happened.what’s new on the Nvidia side? Just like AMD, the short answer is not much. SemiAccurate’s sources have used two names for the ‘big’ gaming GPU, GK114 and GK114-GX. Given how similar the naming is, we wouldn’t count out GX114 either. For the sake of sanity, we will refer to it as GK114 until we get better documentation. Like AMD, it will be late, not much of an advance, and have everything dictated by process limits.
GK114 is currently slated for March “best case”, but given Nvidia’s track record of late that may be very optimistic. Sources who have been briefed on the GPU say that Nvidia is really hedging on mass availability, so April or May is much more likely.
Like AMD, Nvidia is also putting out a minor update to the GK104, not that it is a bad starting point. On the performance side, Nvidia is also power and size bound, but a bit less so than AMD, but the numbers line up pretty closely. They are promising those who have been briefed the same +15% increase that AMD is.