And that benefits AMD how exactly? Do they produce the PCB or buy the memory chips?
Many people thought the GPU would get less complex and smaller - the contrary is true. Why? Hidden CPU on board? :hmm:
It benefits AMD because their AIB partners would be able to sell it for cheap under fierce competition from Maxwell soon. If its still Tahiti, they would not be able to afford price drops without going bankrupt. If AIBs can't sell, AMD make no money. It's all connected... connect the dots.
i wonder , we did see something similar with kaveri (although most of that could be attributed to a node change) when they added xdma , the dsp and co. how much space is the uncore taking?
And that benefits AMD how exactly? Do they produce the PCB or buy the memory chips?
Many people thought the GPU would get less complex and smaller - the contrary is true. Why? Hidden CPU on board? :hmm:
Depends on how much NV charges for GM204 "mid-range" GPU, no?
That doesn't matter. Radeon is now just as terrible as FX. I forsee a 350W 20nm single-GPU that falls significantly behind a 200W 20nm Maxwell next year.
so. you are buying graphics cards based on power consumption? troll much?:thumbsdown:
Why would anybody buy a graphics which uses twice the power for the same performance?
It's like the CPU market: Power consumption is king.
It benefits AMD because it lowers the total cost of the final product. Lower end cost = better sales = more GPUs sold by AMD. Pretty straightforward.
No, I'm talking about performance per watt and what all of that extra power will mean. Basically, Maxwell will OC like crazy, while GCN 2.0 will require a custom loop to get a 10-15% overclock. After factoring in the cost of watercooling, Nvidia will offer the cheaper option. I don't need all of that heat dumped into my room/case.
"The same" is being awfully generous.
That doesn't matter. Radeon is now just as terrible as FX. I forsee a 350W 20nm single-GPU that falls significantly behind a 200W 20nm Maxwell next year.
do you know or are you just speculating? also who buys a high end gpu to worry about power consumption, especially when one even mentions unrealistic theoretical overclocking? just seems silly, something a troll would do.:thumbsdown:
I don't understand how you release a card that is supposed to be great for Mantle but Mantle isn't ready for it. Look at the [H]OCP and Tom's reviews... Mantle is broken for this card. You are telling your customers: "Hey, trust us, we're going to get you Mantle for this card down the road." The hardware/software dynamic between GCN "iterations" and Mantle has always been, and continues to be, "trust us, this is coming down the road."
I don't know why this bothers me. I simply think that AMD should do better about having its hardware launches and Mantle ready at the same time. You'd think AMD would have worked with the few software developers that have games in the market that use Mantle to have their games GCN 1.3 ready at launch.
You aren't looking at the issue correctly. Back in the days when GPU manufacturers could simply increase the power usage of top-end cards to achieve more performance, people didn't care about power consumption.
That isn't the reality anymore. New chips must be more efficient because there isn't any room for power consumption to grow (on top end cards). Thus, if a card, like Maxwell, can achieve the same amount of performance using less power, the manufacturer can then scale up performance using that extra power.
If AMD cannot reduce power consumption at a set level of performance, AMD cannot scale performance any higher (at least without process shrinks, which are obviously becoming fewer and farther between).
The reason people should be disappointed in Tonga is because it isn't much more power efficient than its predecessor which launched almost 3 years ago. Taking that to its logical conclusion, AMD doesn't have an architecture that can scale power down to open up room for additional performance at the high end . . . and Nvidia may not have as much pricing pressure as we would all like, assuming Maxwell 2nd gen delivers the same efficiency gains as Maxwell 1st gen.
You don't know that, this is not GCN 2.0. This is just Hawaii, made cheaper to secure the mid-range segment.
Also, my R290s are still damn fine compared to any high-end competition for perf/$ and perf/w. It would be delusional to class Radeons as terrible as FX.
But it certainly seems AMD is late with their next-gen and Maxwell may have a good time by itself.
you are making too many assumptions, having a low power uarch doesnt mean it can scale well. There are many factors in performance scaling for tdp headroom to be the only one.lets see how maxwell big gpus scale before the usual doom and gloom for amd sets in.
The fact of the matter is that AMD can't even come remotely close to Kelper efficiency on 28nm. I don't see them suddenly making a huge improvement in a year. Hawaii can't overclock much at all. AMD is content with just throwing more power at problems, and I can't see that changing soon. Even my 7950 makes my room a bit more toasty...
I've heard doom and gloom about AMD for years. Yet to see anything arising from that.
On a related note, does it seem odd that AT doesn't have a review up yet but everyone else does? Instead, AT has posted front page stories about AMD's FX series and a MIPS 64-bit chip.