AMD Polaris architecture for Arctic Islands / Radeon 400 series GPUs

ParallaxScroll2

New member
http://i.imgur.com/a4K5k9g.jpg?2

http://videocardz.com/58021/amds-polaris-to-succeed-graphics-core-next-architecture

AMD Radeon 400 series to utilize Polaris architecture

The picture you see above was released by HardwareBattle. It supposedly confirms AMD plans for major technological upgrade next year. Graphics Core Next, a name that accompanied each GPU since 2011, will soon be replaced by Polaris.

We have heard this name before. It was before RTG Summit in December. Tweets from Chris Hook and Raja Koduri made no sense, but some believed they were referring to new GPU codename, which was not exactly correct. Polaris is in fact entirely new architecture, that will be used in new Arctic Islands GPUs (and if rumors are correct one of them would be Greenland).

Introducing the Polaris Architecture

Our guiding lights is to power every pixel on every device efficiently. Stars are the most efficient photon generators of our universe. Their efficiency is the inspiration for every pixel we generate.

Companies like AMD are not known for sharing too much information about their future products. However since the ‘paper-split’ into two entities, Radeon Technology Group has made some major changes in their marketing. We have already learned a lot about ZEN architecture from AMD’s CPU departement, it is time to learn more about RTG’s graphics breakthroughs planned for next year. So far we have heard about GPUOpen initiative, major FreeSync update, and focus on image quality, but the most important question is: how efficient will new GPUs be on smaller fabrication node? Does the tweet from Koduri gives us a hint? One thing is certain, the future for RTG is bright, and it becomes brighter every day.

So it seems GCN 2.0 is now called Polaris.
 
Interesting that Raja specifically mentions the 2.5x figure for the Polaris star.....Could it be that GCN 2.0 which hopefully is what Greenland will use, gets 2.5x the amount of work done compared to the existing Fiji GPU, using the same amount of power overall?



Mind you I am expecting a 2X increase simply because of the change from 28nm to 14nm and that allows a huge increase in transistor budget while using the same amount of power, but to get an additional 50% more because the architecture itself is more efficient is a very nice bonus indeed....:)
 
The real test of the efficiency of the Polaris architecture would be to eliminate the fab process shrink altogether, and the fact that it will also use HBM 2.0 which provides twice the memory bandwidth over HBM 1.0.



Basically, built it at the same die size as Fiji (600 mm^), using the same transistor budget as Fiji (8.9 billion I believe), and also using HBM 1.0 and running at the same clocks as Fiji so that all the hardware variables are a non issue....Any difference in performance would then really be just from improvements in the architecture design alone and nothing else.



Then we'd get a clear picture of what Polaris is all about over the current GCN architectures efficiency wise, but I won't mind the die shrink to 14nm and the (potential) doubling of the transistor budget either, or the use of HBM 2.0 with 1TB/sec of memory bandwidth and having 8GB of ram on the card just the same.....Speed is good, no matter how it's delivered :D
 
Getting tired with triple screens myself, and hoping for a nice single 5K monitor (5120*2880 native resolution), at least in the 32" inch size......Got to put the pressure on those Greenland GPU's somehow right?.....:p :D
 
Polaris vs Pascal
seems they are going hard on 4k/3440x1440 HDR also with the new generation. with the 1.3 DP added I am upgrading my card and likely to a 3550x1440 144hz HDR screen along the way. Unless they charge my arm for it..

Not just your arm but a leg as well :lol: If they do bring 2.5x performance why are people surprised that they may cost 2x+ the cost? Simple economics really. Just remember how much the Titan's cost to see where the price is going.

I'm intrigued by the 2.5x performance because it doesn't say against what. The Fury Nano was 2x the 290X at the same TDP but wasn't as fast as a Fury Pro. I have this horrible feeling AMD are heading for another massive PR disaster.
 
Not just your arm but a leg as well :lol: If they do bring 2.5x performance why are people surprised that they may cost 2x+ the cost? Simple economics really. Just remember how much the Titan's cost to see where the price is going.

I'm intrigued by the 2.5x performance because it doesn't say against what. The Fury Nano was 2x the 290X at the same TDP but wasn't as fast as a Fury Pro. I have this horrible feeling AMD are heading for another massive PR disaster.


Difference being that the titan was never marketed as a gaming card to begin with, as not even me using triple screens could make use of 12 GB of ram on that card, with any game.....:lol:


That extra memory required a custom PCB where there was ram placed on both sides of the card (not just on the side where the cooler is located), so that adds up to additional costs that will inflate the final retail price, as there's 24 memory modules on the titan, versus 12 on the GTX980TI.



A large part of the performance increase is simply due to the transition to 14nm, and likely doubling the transistor budget over the fury, and HBM 2.0 to again double the memory bandwidth as well, to keep all those additional pixel shaders, rops, texture units and more that come with the 14nm transition, actually busy with something.



I don't doubt that Polaris has improvements in efficiency, but they're not that dramatic on their own without 14nm process + and HBM 2.0 being a factor.
 
Last edited:
Won't that annoy current Xbone owners?

I think it would probably annoy devs as well. That's one more platform they have to deal with. I just don't see them releasing an upgraded platform anytime soon. I think an xbone slim with the same performance is more likely.
 
The Xbox One's code name is and has always been Polaris. It showing up in Netflix is nothing new. Its been there since day one.

The same article shows Polaris as the xbone since fall 2014.

The first user to notice "Xbox One Polaris" was "BLNJ" which he posted on twitter. This was all the way back in October in 2014

So maybe just coincidence...
 
Really looking forward to this. Let's hope they release this + Zen before I turn 35 this November. Looking forward to put together an epic AMD build. :lol:
 
Still plenty of CPU firepower for the next several years here......I'm running a pair of Xeon's that can go all the way to 24 cores / 48 threads on this motherboard, and it supports up to 512 GB of ram across 8 channels....


So going for the other side with Zen + a new motherboard + DDR4 ram, just isn't in the cards anytime soon.....
 
Still plenty of CPU firepower for the next several years here......I'm running a pair of Xeon's that can go all the way to 24 cores / 48 threads on this motherboard, and it supports up to 512 GB of ram across 8 channels....


So going for the other side with Zen + a new motherboard + DDR4 ram, just isn't in the cards anytime soon.....

Damn bro you one rich mofo. :lol:
 
Still plenty of CPU firepower for the next several years here......I'm running a pair of Xeon's that can go all the way to 24 cores / 48 threads on this motherboard, and it supports up to 512 GB of ram across 8 channels....


So going for the other side with Zen + a new motherboard + DDR4 ram, just isn't in the cards anytime soon.....

You're getting paid to run it, right?:bleh:
 
"Backed by the first major node shrink for GPUs in over 4 years, RTG has laid out an aggressive plan for Polaris in 2016. At this point RTG needs to catch up and close the market share gap with NVIDIA - of this RTG is quite aware - and Polaris will be the means to do that. What needs to happen now is for RTG to fully execute on the plans they've laid out, and if they can do so then 2016 should turn out to be an interesting (and competitive) year in the GPU industry."

maybe i will get back on red camp much sooner than expected.
 
they also need to release first and hope nvidia hits a delay and is only on 16nm. AMD needs to hard launch the high and midrange cards.
 
Back
Top