AMD's next GPU uarch is called "Polaris"

[AnandThenMan]

The way AMD is presenting it, Polaris will just be another "generation" of GCN, implying only a few small tweaks yet again. This is really bad. Pascal is going to leave it in the dust...

Nope. Every time people make these proclamations they end up being dead wrong you will join that club. Until AMD or Nvidia gets a major leg up on a superior fab process the resulting GPUs will be very similar in performance.

 

[Silverforce11]

Surprised they denote no new changes to the Rasterizer or the ROPs as this is where Fury X is struggling against Maxwell.

Actually it's not, because if ROPs were the bottleneck, the performance will drop off as high resolution is used as well as MSAA or DSR/VSR.

We're seeing the opposite, where performance is poor at 1080p and improves at higher resolutions.

The bottleneck with Fury X is shader uptime, the front-end designed for Hawaii/Tonga is made to drive 4K shaders, way too many for it to feed.

@Techdog
All your gloom and doom goes against reality. Fury X is very competitive despite being a minor Tonga iteration, with the same front-end setup, single GPU 4K its close, and in multi-GPU it beats GM200.

AMD is doing a presentation apparently tomorrow for Polaris and the next-node tech. They are claiming a launch of the real next-gen in Summer, which is ~6 months time-frame.

Basically no point at all to launch Fury X2, is DOA. Both AMD and NV's real next-gen is about to land and make everything currently obsolete. Huge node jump + new uarch = POWER!

 

[mysticjbyrd]

Basically no point at all to launch Fury X2, is DOA. Both AMD and NV's real next-gen is about to land and make everything currently obsolete. Huge node jump + new uarch = POWER!

+ hbm2

 

[AtenRa]

I will agree with Silverforce11 that ROPs is not the problem at 1080p with Fiji, other wise at 4K performance would plummet.

Also since Hawaii, Tonga and Fiji, ROPs AND Rasterizers are now part of the Shader Engine and im sure the new 14/16nm cards will have more than 4 Shader Engines (6 or more) with 4 or even more Render Backends.

 

[beginner99]

Basically no point at all to launch Fury X2, is DOA. Both AMD and NV's real next-gen is about to land and make everything currently obsolete. Huge node jump + new uarch = POWER!

I disagree. The first 14 nm parts will be small die due to new process. Hence the will probably be somewhere in midrange at best. I guess max. 290x performance at lower cost and 2.5 less power and obviously much smaller boards, think Nano.

 

[Arachnotronic]

I disagree. The first 14 nm parts will be small die due to new process. Hence the will probably be somewhere in midrange at best. I guess max. 290x performance at lower cost and 2.5 less power and obviously much smaller boards, think Nano.

🙂

Expect AMD to focus on good performance at extremely high levels of efficiency for the first round of Polaris-based GPUs.

 

[Mondozei]

I disagree. The first 14 nm parts will be small die due to new process. Hence the will probably be somewhere in midrange at best. I guess max. 290x performance at lower cost and 2.5 less power and obviously much smaller boards, think Nano.

Yup. New node always means that the big die/fullfat GPUs come later, for consumers, as maturity increases and yields improve.

I still expect that we'll see full fat GPUs for server/HPC for the simple reason that the cost is worth the extra hassle of low yields/low maturity(designing a 600 mm^2 GPU is not the same as doing smartphone SoCs so don't even try to compare the two), at least from Nvidia's side, but since AMD is not a big player there, it remains to be seen if they'll even bother.

People who are gearing up for 70-100% performance increases are bound to be disappointed.

 

[desprado]

Nope. Every time people make these proclamations they end up being dead wrong you will join that club. Until AMD or Nvidia gets a major leg up on a superior fab process the resulting GPUs will be very similar in performance.

Totally wrong. Many people were right about predicting the performance of Titan X and GTX 980 Ti. No one here over-hyped GTX Titan X or GTX 980 Ti like some of member did with Fury X by predicting that it will be 20% to 40% faster than Titan X or GTX 980 Ti.

 

[AtenRa]

I disagree. The first 14 nm parts will be small die due to new process. Hence the will probably be somewhere in midrange at best. I guess max. 290x performance at lower cost and 2.5 less power and obviously much smaller boards, think Nano.

I believe a Fury X performance + 20-30% at close to 300mm2 is what we will see.
That will surely make the Fury X2 look outdated even if it is 50-60% faster but have more than 2x the power consumption.

 

[Seba]

We're seeing the opposite, where performance is poor at 1080p and improves at higher resolutions.

What? You get higher FPS as you increase the resolution? Because that would be "an improvement of performance at higher resolutions".

 

[AtenRa]

What? You get higher FPS as you increase the resolution? Because that would be "an improvement of performance at higher resolutions".

He means performance efficiency, where Fiji is barely faster than Hawaii at 1080p, at 4K it is as fast or faster than GTX 980Ti.

 

[desprado]

He means performance efficiency, where Fiji is barely faster than Hawaii at 1080p, at 4K it is as fast or faster than GTX 980Ti.

Wrong. GTX 980 Ti beat Fury X in 4k as well.

 

[desprado]

AMD PR is in such a desperation that they did not waited for pascal to see its performance per watt.

Lol AMD made comparison of Maxwell and GCN 4.0 in performance per watt.

Pascal is 2X more efficient than Maxwell + advantage of new node.

 

[3DVagabond]

He means performance efficiency, where Fiji is barely faster than Hawaii at 1080p, at 4K it is as fast or faster than GTX 980Ti.

I think the relative lack of performance at lower resolutions is caused by API overhead. The apparent increase at high res, on the other hand is Maxwell's drop off. Maybe due to less bandwidth than Fiji.

 

[AtenRa]

Wrong. GTX 980 Ti beat Fury X in 4k as well.

Only on some Gameworks titles and that by a few fps.

But it is way faster than Hawaii at that resolution.

 

[desprado]

only on some Gameworks titles

Totally wrong again. You said Fury X beats GTX 980 Ti in 4K yes but only with reference and if you talk about custom models than you are wrong.

https://www.techpowerup.com/reviews/MSI/GTX_980_Ti_Lightning/12.html
https://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_Waterforce/17.html

Fury X can never beat a proper GTX 980 TI OC even if Fury X OC and people need to get over it and face the reality.

 

[AtenRa]

Totally wrong again. You said Fury X beats GTX 980 Ti in 4K yes but only with reference and if you talk about custom models than you are wrong.

https://www.techpowerup.com/reviews/MSI/GTX_980_Ti_Lightning/12.html
https://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_Waterforce/17.html

Fury X can never beat a proper GTX 980 TI OC even if Fury X OC and people need to get over it and face the reality.

Thank you for proving my point with those links, oh and im talking at default clocks, not everyone OC.

Ryse Son Of Rome, Fury X faster than GTX 980ti

Shadow Of Mordor, Fury X faster than GTX 980ti

GTAV, Fury X and GTX 980Ti equals
And that is a GW title.

Witcher 3, Fury X faster than GTX980Ti.
Also a GW title.

 

[3DVagabond]

This thread is about AMD's new tech. It's not about prior cards or nVidia. Maybe we can try and keep the speculation and discussion on topic? We don't need to rehash Fiji vs. GM 200 vs stock vs overclocked vs frames per watt, etc...

 

[desprado]

Thank you for proving my point with those links, oh and im talking at default clocks, not everyone OC.

Ryse Son Of Rome, Fury X faster than GTX 980ti

Shadow Of Mordor, Fury X faster than GTX 980ti

GTAV, Fury X and GTX 980Ti equals
And that is a GW title.

Witcher 3, Fury X faster than GTX980Ti.
Also a GW title.

You really miss what i said? I said reference non OC GTX 980 Ti is beatable by Fury X on 4K ,however, their no one who keeps their GTX 980 Ti at stock.

 

[AtenRa]

I think the relative lack of performance at lower resolutions is caused by API overhead. The apparent increase at high res, on the other hand is Maxwell's drop off. Maybe due to less bandwidth than Fiji.

Well if you compare Hawaii vs Fiji at 1080p and then at 4K you will see that Fiji is scaling better at 4K than Hawaii and they both have the same 64 ROPs.

So Fiji cannot be ROP starved at 1080p. Perhaps its the API overhead as you say at 1080p but certainly not the ROPs.

 

[desprado]

This thread is about AMD's new tech. It's not about prior cards or nVidia. Maybe we can try and keep the speculation and discussion on topic? We don't need to rehash Fiji vs. GM 200 vs stock vs overclocked vs frames per watt, etc...

You think this pic is making sense?

Making a comparison of their GCN 4.0 comparison with Maxwell rather than pascal.

So GCN 4.0 is the competition for Maxwell?

Did Nvidia compare Pascal with GCN 1.1, 1.2?

 

[AtenRa]

You really miss what i said? I said reference non OC GTX 980 Ti is beatable by Fury X on 4K ,however, their no one who keeps their GTX 980 Ti at stock.

I will agree that the custom OC GTX 980 Ti models are faster than Fury X but they are also more expensive.

Anyway i was not trying to make it a battle between Fury and GTX980Ti but to make a point that Fury with the same number of ROPs as Hawaii is faster than any Hawaii card at 4K and reaches the same level of performance as reference GTX 980Ti.

 

[Techhog]

The problem is that nobody buys reference 980 Ti cards, and every single factory OC'd card stomps the Fury X. If AMD isn't making any major changes this time, then we'll still be in a situation where Nvidia cards have 10-15% factory overclock with another 15% or more of headroom while AMD will continue to top out at 15% MAX overclock after adding a ridiculous amount of voltage. Nothing will change.

 

[AtenRa]

You think this pic is making sense?

Making a comparison of their GCN 4.0 comparison with Maxwell rather than pascal.

So GCN 4.0 is the competition for Maxwell?

Did Nvidia compare Pascal with GCN 1.1, 1.2?

Ehm, Maxwell is what currently on the market TODAY, that is what they can compare with. Lets see NVIDIA compare Pascal to GTX950 at the same game same settings and Compare then 😉

 

[desprado]

The problem is that nobody buys reference 980 Ti cards, and every single factory OC'd card stomps the Fury X. If AMD isn't making any major changes this time, then we'll still be in a situation where Nvidia cards have 10-15% factory overclock with another 15% or more of headroom while AMD will continue to top out at 15% MAX overclock after adding a ridiculous amount of voltage. Nothing will change.

This......It is just a excuse for AMD and some review site to keep fury X competitive.