AMD announces fourth-generation graphics architecture 'Polaris'

Shawn Knight

Posts: 15,291   +192
Staff member

AMD on Monday unveiled its fourth-generation Graphics Core Next (GCN) architecture. Polaris is significant in that it moves from the tired 28nm manufacturing process to a revitalized 14nm FinFET process which, according to Radeon Technologies Group SVP and Chief Architect Raja Koduri, results in the largest performance-per-watt jump in the company's history (which also includes ATI).

As EuroGamer points out, the architecture includes a number of new features and improvements including a primitive discard accelerator, instruction pre-fetch, hardware scheduler, improved shader efficiency and better memory compression. It also introduces support for DisplayPort 1.3 and HDMI 2.0a

What sort of power-savings can we expect? Although it's early, the recent side-by-side comparison in Star Wars Battlefront is telling. As seen in the clip above, the system running AMD's new board was able to do so at nearly half the power consumption as the competition, an Nvidia GeForce GTX 950 (~86 watts versus ~140 watts). This will translate to major advancements in notebook GPUs.

What if you don't care all that much about power consumption? Is there still a reason to get excited? Absolutely.

Efficiency is only one part of the puzzle. More real estate to work with means AMD can cram far more transistors into the same amount of space used by today's top cards. What you'd then have is a card that has the same power requirements as today's top dog with far more transistors in play which of course translates to far more processing power.

Polaris-based GPUs are expected to go on sale by mid-2016. With any luck, we'll see some more details leak out over the course of CES.

Permalink to story.

 
I guess AMD and their customers are finding rebrands even more tiring than the old 28nm SKU.

It's enough to make you write the Consumer Protection Agency and ask for mandatory labeling that included generation numbering so we could easily figure out which was the newest. Of course, TigerSoft would hate that one!
 
Polaris vs Pascal, can't wait to see the numbers, and hopefully they'll both come out around the same time (summer 2016).
 
I guess AMD and their customers are finding rebrands even more tiring than the old 28nm SKU.

Amazing isn't it that "re-brands" are beating NV's products in every pricing segment except the $600+ flagship level:
http://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_Waterforce/23.html

Those same "re-brands" are ageing very well too.

R9 280X (aka HD7970Ghz) is now a whopping 21% faster than GTX770 at 1080p, and 29% faster at 1440p. $399 R9 290 aka 390 is just 4% behind $699 GTX780Ti at 1440p. 390X > 980 at 1440p. Re-brands and long-term driver support is doing great for anyone who isn't brainwashed by brand names.

----

Interesting how you also ignored the major architectural changes of Polaris architecture, and also ignored the presentation where an i7 4790K + Polaris with 87-88W achieved 60 fps in the same scenario where a 950 was hitting 150W. The more competitive AMD's Polaris is, the more pressure it puts on NV to deliver even faster products and price them more competitively. I suppose such basic logic is beyond the comprehension of NV sheep though.
 
Amazing isn't it that "re-brands" are beating NV's products in every pricing segment except the $600+ flagship level:
Ignoring power efficiency again are we? That's the only way AMD is beating nVidia (if in fact they are). HBM memory may turn the tides, but we are not there yet.
 
Amazing isn't it that "re-brands" are beating NV's products in every pricing segment except the $600+ flagship level:
http://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_Waterforce/23.html

Those same "re-brands" are ageing very well too.

R9 280X (aka HD7970Ghz) is now a whopping 21% faster than GTX770 at 1080p, and 29% faster at 1440p. $399 R9 290 aka 390 is just 4% behind $699 GTX780Ti at 1440p. 390X > 980 at 1440p. Re-brands and long-term driver support is doing great for anyone who isn't brainwashed by brand names.

----

Interesting how you also ignored the major architectural changes of Polaris architecture, and also ignored the presentation where an i7 4790K + Polaris with 87-88W achieved 60 fps in the same scenario where a 950 was hitting 150W. The more competitive AMD's Polaris is, the more pressure it puts on NV to deliver even faster products and price them more competitively. I suppose such basic logic is beyond the comprehension of NV sheep though.
They've stretched that old silicon to breaking point now and milked it for all it's worth, even they've woken up to the fact that it's time they chuck something new out of the door.
 
I guess AMD and their customers are finding rebrands even more tiring than the old 28nm SKU.

Amazing isn't it that "re-brands" are beating NV's products in every pricing segment except the $600+ flagship level:
http://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_Waterforce/23.html

Those same "re-brands" are ageing very well too.

R9 280X (aka HD7970Ghz) is now a whopping 21% faster than GTX770 at 1080p, and 29% faster at 1440p. $399 R9 290 aka 390 is just 4% behind $699 GTX780Ti at 1440p. 390X > 980 at 1440p. Re-brands and long-term driver support is doing great for anyone who isn't brainwashed by brand names.

----

Interesting how you also ignored the major architectural changes of Polaris architecture, and also ignored the presentation where an i7 4790K + Polaris with 87-88W achieved 60 fps in the same scenario where a 950 was hitting 150W. The more competitive AMD's Polaris is, the more pressure it puts on NV to deliver even faster products and price them more competitively. I suppose such basic logic is beyond the comprehension of NV sheep though.

It is great to see some great logic applied to this.

AMD look to be aiming to push the limits with NV
 
This makes me really happy, I was genuinely worried AMD was going to do the same thing they did with their CPU's and just sort of, not improve them.

Nvidia would then charge a fortune, This news makes me very happy if indeed any of it is true. This means Nvidia won't be able to charge a small fortune which is good as I've been waiting for Pascal for a while now.

I don't know if I'm blind but did anyone read anywhere in this article if AMD's new architecture uses HBM2 instead of HBM?
 
It is great to see some great logic applied to this.

AMD look to be aiming to push the limits with NV
nVidia is the one doing the pushing. AMD is the one that can't get a grip on their thermals, and you say AMD is the one pushing. There is no logic in that at all, no matter how many years they continue re-branding.

What if next year the same 7970 was rebranded as an R5 430 to compete with nVidia's budget card at the time. Would you still think AMD is pushing nVidia's limits, even though the rebrand R5 430 requires a 500W PSU? *sarcasm* Yeah that's logical all right! *sarcasm*
 
I really enjoy these discussions, despite the fact the most "advanced" video card I own, is a 9500-GT. Oh, don't laugh buster, it's the one with the GDDR-3 :cool:

It's always gratifying to hear the AMD fanbois soldier on, as if it was still the "Athlon 3800" versus "Pentium 4" decade.

"Polaris" makes a wonderful rallying cry! In fact, I think it would also make a great name for a porn star or movie.

"Make way for Paul Polaris in, "The Ballistic Missile Cruiser...." :cool: Maybe, "Scuds are Duds, Meet Paul Polaris". How about possibly, "Pablo Polaris" in, "Surfacing to Fire"...;)
Polaris_CCAFS2009RK_4.jpg

Up your first person shooter with that, sucka!

(Now that I think about it some more, it would even make a good name for a woman).

For example, "Introducing Stella Polaris in, "The Pole Star".
 
Last edited:
For some reason I feel that buying a GTX 970 was a mistake in the long term, this news from AMD sounds great since I want to come back to red team.
 
It is great to see some great logic applied to this.

AMD look to be aiming to push the limits with NV
nVidia is the one doing the pushing. AMD is the one that can't get a grip on their thermals, and you say AMD is the one pushing. There is no logic in that at all, no matter how many years they continue re-branding.

What if next year the same 7970 was rebranded as an R5 430 to compete with nVidia's budget card at the time. Would you still think AMD is pushing nVidia's limits, even though the rebrand R5 430 requires a 500W PSU? *sarcasm* Yeah that's logical all right! *sarcasm*

Should NV be embarrassed that AMD beat them with old gen then?

I think the "rebrand" of the AMD cards (300 Series), new drivers and the HBM cards of 2015 show a new direction for AMD that represents them wanting to push the limits with NV.

If you re-read the comment I didn't say they were currently pushing the limits but they look to be setting themselves up to be doing this in the next 12 months.

I think it is a stretch to say that NV are the ones doing the pushing when AMD were the first ones to release a HBM card in 2015.

AMD managed to beat alot of the current gen NV cards in performance with there rebaged cards and have zero issues with thermals.

NV seem to be the ones in the news last year for issues with there cards and drivers while AMD had many articles around how they have a great new direction and the 300 series were fantastic cards.

Had memory issues on GTX970

NV flopped at CES 2016

But I NV do have good power efficiency
 
It is great to see some great logic applied to this.

AMD look to be aiming to push the limits with NV
nVidia is the one doing the pushing. AMD is the one that can't get a grip on their thermals, and you say AMD is the one pushing. There is no logic in that at all, no matter how many years they continue re-branding.

What if next year the same 7970 was rebranded as an R5 430 to compete with nVidia's budget card at the time. Would you still think AMD is pushing nVidia's limits, even though the rebrand R5 430 requires a 500W PSU? *sarcasm* Yeah that's logical all right! *sarcasm*
eerm , everyone rebrands, remember GTX 680 > GTX 770?
the gcn architecture is revolutionary, why? even nvidia tries to mimick it, some gens ago they used fewer cores higher clock

and 1-2 generations ago, radeon had the best power efficiently
if you look at 290X vs fury X, they have around the same power usage, yet the fury X is almost twice as fast
 
Back