All of the speculation directly contradicting the stated information is futile and frankly misleading. Working within the revealed limits however is useful.I think it depends on what kind of clock speeds they can get out of the chips they have. AMD isn't at the top of the market with Polaris 10 so they price their cards relative to the rest of the market and past offerings from the previous generation.
If they had a card that could perform for $350 of performance, that's what it would sell for, but if it can only justify $300, that's where it would be. For $350 I'd expect something that trades blows with the 1070, because eventually it will drop close to MSRP and at $350 you can't sell a card that's going to have more than a ~5% performance gap or most people will just pay the extra $30 for the added bump.
We'll eventually see a full die, but I expect that AMD might be waiting until they can get better clock speeds out of their chips. An extra ~10% performance from a full die isn't worth $100, so it likely needs a good bump from clock speeds as well, because for a 50% increase in price (or perhaps closer to 30% given that a 480X wouldn't ship with less than 8 GB) you'd expect at least a 25% increase in performance.
The real question is who's going to get to that magical $300 price point first as right now with the 480 at $200 and the 1070 at double the cost, there's a gaping hole in the middle and as the 970 proved, that's a sweet spot to own in the market.
It's just common sense that they want to target the $300 price point given the success of the 970.
http://www.guru3d.com/news-story/amd-radeon-rx-480-3dmark-11-performance-benchmark-surfaces.html
Fury/980 for $200-250 under 150w?
Nice. Now it remains to be seen how GCN4 on 14nm overclocks. Driver version seems to be old.
That fits the previous leak, ~Fury performance. Above 390X/980.
Polaris 10 (67DF:C7): 18.060
http://www.3dmark.com/3dm11/11263084
R9 390X: 18.865
http://www.3dmark.com/3dm11/10039529
Let's face it guys, whether you dislike AMD or not, if they deliver amazing deals like the RX 480 and this new Polaris series, it forces NVIDIA to compete. When the 1060Ti and 1060 launch, it won't come with a premium tax on it if Polaris is a better product.
Don't be such a downer (all that heated argument on an unreleased product..) on an AMD product thread. A competitive AMD is good for gamers.
All of the speculation directly contradicting the stated information is futile and frankly misleading. Working within the revealed limits however is useful.
Let's face it guys, whether you dislike AMD or not, if they deliver amazing deals like the RX 480 and this new Polaris series, it forces NVIDIA to compete. When the 1060Ti and 1060 launch, it won't come with a premium tax on it if Polaris is a better product.
Don't be such a downer (all that heated argument on an unreleased product..) on an AMD product thread. A competitive AMD is good for gamers.
I think the point is that the cut die/480 was meant to be in the $300 range but they reduced the price because the performance of the 1070 turned out to be faster than expected because of the high clocks and AMD couldn't really justify the price given the performance difference. Why they won't release a full die model is still a mystery although I guess we will find out eventually, although remember AMD did the same thing with Tonga.
Let's face it guys, whether you dislike AMD or not, if they deliver amazing deals like the RX 480 and this new Polaris series, it forces NVIDIA to compete. When the 1060Ti and 1060 launch, it won't come with a premium tax on it if Polaris is a better product.
Don't be such a downer (all that heated argument on an unreleased product..) on an AMD product thread. A competitive AMD is good for gamers.
Right now AMD and Nvidia are in completely different market segments so there's no competition at all.
Only a fool would claim that competition is not good,
However, at $300 the card will perform relative to its price to the 1070,
I think the point is that the cut die/480 was meant to be in the $300 range but they reduced the price because the performance of the 1070 turned out to be faster than expected because of the high clocks and AMD couldn't really justify the price given the performance difference. Why they won't release a full die model is still a mystery although I guess we will find out eventually, although remember AMD did the same thing with Tonga.
That fits the previous leak, ~Fury performance. Above 390X/980.
For a $199/$229 SKU, it's a great bang for buck. I expect the gaming power usage to be close to 100-120W, AMD's TDP is a max board power, only reached in compute workloads.
This means there's a good chance custom boards with higher clocks = Fury X.
Though I expect the major thing to look out for, the results won't be like that across the board, in less complex scene games, it will be slower. But in more complex games, it will be faster. It's down to how effective the Discard Accelerator can be on any given scene.
The RX 480 is actually equivalent to the 4850, a cut down chip at $199. The bigger brother is coming at $299, as per Lisa Su's statements about the RX Polaris series $100 to $300.
ps. For you guys who have been around for awhile, did you remember what happen to the 260/280 prices after the 4850 and 4870 was released? Yeah that's right. Tell me competition isn't good for us gamers. -_-
The GTX1070 is 33% faster than a Fury in the same benchmark. But what is 33% today, right?
The GTX1070 is 33% faster than a Fury in the same benchmark. But what is 33% today, right?
with the dripfed performance gains of the last 3 years, and looking to continue from now on, it just might be.480 soon the card of a decade
It's also quite good to see AMD getting this kind of performance from 2304sp and 100Mhz over 28nm GCN in what is effectively Pitcairn's replacement, that's a healthy reduction from Hawaii's 2816sp and more importantly Fiji's 4096sp that here show how underutilized they are. GCN4 finally seems to bring improvements to the most critical areas in GCN1-3 that were bottlenecking performance, relative to the amount of SPs in each chip.
It certainly makes Vega and its supposed Fiji amount of SPs promising, that could end up being a monster chip.
I'm hoping RX480 and its variants shake the market just as the 4850 and 4870 did back then. We all need prices to come down to saner levels that don't feel like a rip off.
Unlike Maxwell -> Pascal where they regressed in IPC, I don't expect GCN 3 -> GCN 4 to regress.
Given the focus on shader efficiency and utilization, new pre-fetch and cache, it should be an increase in IPC.
You made a good point on overclocking. Early Pitcairn and Tahiti was very low clocked for perf/w. My 7850 OC 40% and 7950 an amazing 50% (800mhz -> 1.2ghz).
Could you see a situation where the 480 is like the 970? I.e. Gangbusters at 1080p but falls behind at 1440p?
I don't share your optimism, vodka. Nvidias new 1920 shader card beats their old 2816 shader card at stock. At Max OC it's a wash.
AMD's new 2304SP card should handily be faster than their old 2816SP card then. They merely need to match the gains Nvidia made.
Even look at the 1080. New 2560 easily beats old 3072, stock and max OC both. 83.3% of the shaders, kills it. 2304SP to 2816SP Hawaii is 81.8% of the shaders. It should beat the 390X easily then, just slightly less than the lead 1080 has over Titan X.
What a failure if it does not. Pricing wise they can price it a winner, but from a technology perspective? How does this give you any confidence in Vega? If RX 480 is a wash with the 390X, then you better hope it is memory bandwidth starved and that Vega with HBM2 will show similar gains to Nvidia. Otherwise AMD is very far behind, and indeed Nvidia is dominating.
We'll see on June 29 what happens in the overclocking front, that's the only main question left. I'd like to believe AMD is pulling another 78xx/79xx style of launch here, conservatively clocked cards that excel in perf/w, yet have lots of overclocking headroom if you want pure performance.
I think that's what you meant.
NV shareholders and NV marketers would want no competition though. Why do you think it's the "usuals" that are trying to downplay RX 480 and hype up GTX 1070?
I hope it can clock well. Indeed that is the main reason for Pascal's gains. And that is also why it would be a failure if AMD did not find a way after all these years to change GCN enough in either clockrates OR large architectural changes for IPC gains.
I don't share your optimism, vodka. Nvidias new 1920 shader card beats their old 2816 shader card at stock. At Max OC it's a wash.
AMD's new 2304SP card should handily be faster than their old 2816SP card then. They merely need to match the gains Nvidia made.
Even look at the 1080. New 2560 easily beats old 3072, stock and max OC both. 83.3% of the shaders, kills it. 2304SP to 2816SP Hawaii is 81.8% of the shaders. It should beat the 390X easily then, just slightly less than the lead 1080 has over Titan X.
What a failure if it does not. Pricing wise they can price it a winner, but from a technology perspective? How does this give you any confidence in Vega? If RX 480 is a wash with the 390X, then you better hope it is memory bandwidth starved and that Vega with HBM2 will show similar gains to Nvidia. Otherwise AMD is very far behind, and indeed Nvidia is dominating.