[Ars] AMD confirms high-end Polaris GPU will be released in 2016

[Silverforce11]

@JDG1980
There's no way we are getting 980Ti performance for $350. Even if next-gen's mid-range is very fast, it will be priced accordingly, aka expensive.

Best we can hope for, Polaris 11 ~390X/980 performance for $299, which is good but doesn't shift the perf/$ mark by much (I see it being very popular if its ~100W). I just don't see it going for cheaper, they need to make good margins.

Vega is the major unknown though.

 

[3DVagabond]

@JDG1980
There's no way we are getting 980Ti performance for $350. Even if next-gen's mid-range is very fast, it will be priced accordingly, aka expensive.

Best we can hope for, Polaris 11 ~390X/980 performance for $299, which is good but doesn't shift the perf/$ mark by much (I see it being very popular if its ~100W). I just don't see it going for cheaper, they need to make good margins.

Vega is the major unknown though.

We got ~GTX-580 for around that price from the 7870 on the last shrink. It wasn't like AMD low balled us with pricing with the 7000 series either.

 

[MrTeal]

We got ~GTX-580 for around that price from the 7870 on the last shrink. It wasn't like AMD low balled us with pricing with the 7000 series either.

The 7870 was about halfway between a 6970 and 580 at launch. It was actually closer to a 570 at launch than a 580, and 570s were selling under the $350 that the 7870 launched at.

 

[3DVagabond]

The 7870 was about halfway between a 6970 and 580 at launch. It was actually closer to a 570 at launch than a 580, and 570s were selling under the $350 that the 7870 launched at.

Depends on what site you use. TPU had the AMD cards closer to Fermi in performance than a lot of sites did. But even there the difference isn't great.

So, for the sake of argument, or avoiding any, let's say it could give 980ti performance for $300. Are you better with that? Does that really defeat the general point?

 

[MrTeal]

Depends on what site you use. TPU had the AMD cards closer to Fermi in performance than a lot of sites did. But even there the difference isn't great.

So, for the sake of argument, or avoiding any, let's say it could give 980ti performance for $300. Are you better with that? Does that really defeat the general point?

I would be ecstatic with that. It would be amazing.
AT found the 7870 relatively close to Fermi

At each launch AMD has undercut NVIDIA at critical points, allowing them to push NVIDIA out of the picture, and the launch of the Radeon HD 7800 series is no different. AMD’s decision to launch the 7870 and 7850 at roughly $25 to $50 over the GTX 570 and GTX 560 Ti respectively means that NVIDIA’s cards still have a niche between AMD’s price points for the time being, but this is effectively a temporary situation as NVIDIA starts drawing down inventory for the eventual Kepler launch.

Starting with the Radeon HD 7870 GHz Edition, AMD is effectively in the clear for the time being. At roughly 9% faster than the GTX 570 there’s little reason to get the GTX 570 even with the 7870’s price premium; it’s that much faster, cooler, and quieter. With the launch of Pitcairn and the 7870 in particular, GF110 has effectively been removed from competition after a nearly year and a half run.

The 7870 was a much better buy than the 570 due to its more modern uarch, the lower power, etc. In terms of perf/$ though, it launched at roughly the same as the outgoing 570. I would love to see AMD launch a card at $300 that matches a $600 980 Ti, but that kind of value is a fantasy. Especially if AMD launches much before nVidia, you really think they'll sell 980 Ti performance for $300 when they would sell every working chip they can make at $500?

 

[AtenRa]

$300 for 980Ti performance only after H2 2017.

 

[3DVagabond]

I would be ecstatic with that. It would be amazing.
AT found the 7870 relatively close to Fermi


The 7870 was a much better buy than the 570 due to its more modern uarch, the lower power, etc. In terms of perf/$ though, it launched at roughly the same as the outgoing 570. I would love to see AMD launch a card at $300 that matches a $600 980 Ti, but that kind of value is a fantasy. Especially if AMD launches much before nVidia, you really think they'll sell 980 Ti performance for $300 when they would sell every working chip they can make at $500?

I can see both sides of the debate. Why sell for less? Because they badly need market share and sell through.

I'm not convinced that we'll see that level of performance. At least not across the board. Maybe in DX12 it can match/beat the 980ti.

If it's launched as the 480 then it should be ~$300.

 

[ShintaiDK]

There's no way we are getting 980Ti performance for $350. Even if next-gen's mid-range is very fast, it will be priced accordingly, aka expensive.

Agree, a 125-150W 8GB GTX980TI/Fury X performance will pull ~500$. If not on performance, then performance/watt.

Memory alone I guess will cost ~100$ for GDDR5X and ~200$ for HBM2. Certainly not going for 350$ unless someone is very pressed.

Not to mention how long 14/16nm will have to last for GPUs.

 

[gamervivek]

Charlie said that the Polaris GPUs are GDDR5 only.

Charlie also said there are three new chips so I doubt that at least one of them is not using hbm2.

 

[3DVagabond]

Charlie also said there are three new chips so I doubt that at least one of them is not using hbm2.

The 3rd isn't Polaris though.

 

[beginner99]

Even so, that would still be a huge improvement if we got that kind of perf/$ across the board. It would mean GTX 980 Ti level performance at about $350. (The 980 Ti has about 40% better performance at 1080p than the 290X, and the 290X was being blown out at about $250 last year.)

And don't forget that all the new cards will have HDMI 2.0 output, hardware HEVC decoding support, and much lower power consumption. To many buyers, this makes a difference.

It doubt it will be priced at $350. If it research 980TI levels it will probably be more like $450+ at least for the first couple month and hence worse performance/$ than 290s. Why should they sell it for less? They will copy NV. Sell mid-range for flagship price and introduce the real flagship 6 month or more later for an even higher price. Bad for us, good for AMD. BTW I hope I'm wrong but underpricing makes 0 sense for AMD.

 

[guskline]

Yeah, AMD isn't going to offer better perf/$ than the custom 290s that were available. It will likely increase decently relative to some of the poor options in AMD's lineup like Fury X, but nothing is going to touch the $220 R9 290s or $250 290X.

Spoken by a man who loves his 290s in CF.

I also have 290s in CF (stk at 1000 core) in my 4790k rig below and they are fast, even at stock. The 290s produce heat and since I'm using custom water cooling that solves the problem.

I won't knock Fury X, Fury etc since AMD was striking out on a new path, HBM.

I opted for a single GTX980TI instead of a Fury X to replace my CF 290s in my water cooled 5960x rig and "repositioned" the 290s in a new 4790k build. I opted for the GTX980TI because it outperformed the Fury X slightly in all resolutions and games except a few in 4k, overclocked WAY better, was the same or lower price AND was available.

High end Polaris might be the new and improved version of Fury X; should be designed to go head to head with Pascal.

It has to overclock better AND be available.

 

[StereoPixel]

It's just my opinion...

 

[Glo.]

If you would have TWO graphics scheduler(Command Processors) on one GPU it would be recognized as two GPUs.

 

[StereoPixel]

If you would have TWO graphics scheduler(Command Processors) on one GPU it would be recognized as two GPUs.

OK, I agree, your thought is good

http://ranker.sisoftware.net/show_r...dbe6dfe9dcecdef88ab787a1c4a19cac8af9c4fc&l=en

 

[Glo.]

It makes sense for VR, and with Asynchronous Compute. Two Graphics Command Processors on one GPU with 64 CUs would make that GPU a true MONSTER. 8 TFLOPS of Compute Power.

 

[StereoPixel]

Polaris announcement?
http://insightfulvr.com/insightfulvr/

 

[xpea]

It doubt it will be priced at $350. If it research 980TI levels it will probably be more like $450+ at least for the first couple month and hence worse performance/$ than 290s. Why should they sell it for less? They will copy NV. Sell mid-range for flagship price and introduce the real flagship 6 month or more later for an even higher price. Bad for us, good for AMD. BTW I hope I'm wrong but underpricing makes 0 sense for AMD.

^^THIS
for all dreamers around already salivating on GTX980Ti at $200, better to wake up right now to avoid a big big disillusion...

 

[tential]

@JDG1980
There's no way we are getting 980Ti performance for $350. Even if next-gen's mid-range is very fast, it will be priced accordingly, aka expensive.

Best we can hope for, Polaris 11 ~390X/980 performance for $299, which is good but doesn't shift the perf/$ mark by much (I see it being very popular if its ~100W). I just don't see it going for cheaper, they need to make good margins.

Vega is the major unknown though.

You don't see it doing well, but the GTX 970 was like that and it did AMAZING. If they can provide great performance at 100W, then I'd upgrade my card to it. Sidegrade? Yes. But that's very little power consumption for crossfire along with the other platform improvements.

People will definitely buy it, just like they made the upgrade to the 970/980.

 

[StereoPixel]

Roy Taylor: "AMD has just completed the shrink to 14 nanometer [with Polaris Architecture]. What this means is, and this is where it comes home to everyone in this room, is that we can produce GPUs that will run the minimum spec of VR at a lower cost, in larger volume, consuming less power and running faster."

Raja Koduri: "Polaris is 2.5 times brighter today than when Ptolemy observed it in 169 A.D"

So 596mm2/2.5 = 238.4mm2 and Polaris 11 faster than Fiji

 

[Glo.]

Roy Taylor: "AMD has just completed the shrink to 14 nanometer [with Polaris Architecture]. What this means is, and this is where it comes home to everyone in this room, is that we can produce GPUs that will run the minimum spec of VR at a lower cost, in larger volume, consuming less power and running faster."

Raja Koduri: "Polaris is 2.5 times brighter today than when Ptolemy observed it in 169 A.D"

So 596mm2/2.5 = 238.4mm2 and Polaris 11 faster than Fiji

Lets genuinely hope that this is reality. It would be fantastic achievement.

 

[Silverforce11]

Roy Taylor: "AMD has just completed the shrink to 14 nanometer [with Polaris Architecture]. What this means is, and this is where it comes home to everyone in this room, is that we can produce GPUs that will run the minimum spec of VR at a lower cost, in larger volume, consuming less power and running faster."

Raja Koduri: "Polaris is 2.5 times brighter today than when Ptolemy observed it in 169 A.D"

So 596mm2/2.5 = 238.4mm2 and Polaris 11 faster than Fiji

That would be competitive and I will get 2, thanks!

 

[raghu78]

That would be competitive and I will get 2, thanks!

AMD mid range GPUs of the latest gen typically match their previous flagships or get very close.

HD 4870 -> HD 5770
HD 6970 -> HD 7870
HD 7970 -> R9 280X
Fury X -> Polaris mid range

AMD has mentioned that improving shading efficiency was a major goal of Polaris. That coupled with higher clocks can get the mid-range Polaris on par with last gen flagship (Fury X).

 

[Silverforce11]

AMD mid range GPUs of the latest gen typically match their previous flagships or get very close.

HD 4870 -> HD 5770
HD 6970 -> HD 7870
HD 7970 -> R9 280X
Fury X -> Polaris mid range

AMD has mentioned that improving shading efficiency was a major goal of Polaris. That coupled with higher clocks can get the mid-range Polaris on par with last gen flagship (Fury X).

7970 = 280X.

But I get what you mean.

On NV's side, it's similar too.

680 mid-range > 580 big-chip

Maybe I've been too jaded with 28nm performance crawl, I have forgotten how much of a leap it is for a full node transition.

Ok. I'm on the hype train!

Polaris 11 ~= Fury X at ~100W!

Where does this put Polaris 10? It's equivalent GCN 28nm part would be 2.5x 120mm2 = ~300mm2... slightly below 380X??

 

[raghu78]

7970 = 280X.

But I get what you mean.

surely I know that. just trying to prove a historical trend.

On NV's side, it's similar too.

680 mid-range > 580 big-chip

Actually 660 got very close to 580.

https://www.techpowerup.com/reviews/MSI/GTX_660_Gaming/26.html

Maybe I've been too jaded with 28nm performance crawl, I have forgotten how much of a leap it is for a full node transition.

Ok. I'm on the hype train!

Polaris 11 ~= Fury X at ~100W!

I am trying to go with Raja Koduri's statement of Polaris being 2.5 times brighter than Ptolemy. If we take that as Polaris perf/watt is 2.5 times that of last gen. Fury X was 275w. So mid range Polaris would come at 275/2.5 = 110w. I would say 110-120w range. My guess is Polaris mid range would be 2560 sp with 256 bit GDDR5 bus and improved memory compression, 64 ROPs. Combine that with significantly improved shading efficiency (perf/sp) and they could catch up to the Fury X.

Where does this put Polaris 10? It's equivalent GCN 28nm part would be 2.5x 120mm2 = ~300mm2... slightly below 380X??

My guess is R9 380 performance at 50W.