AMD Polaris architecture to succeed Graphics Core Next

What are the chances that Polaris will hit GTX 980 power levels (hopefully at a lower TDP)? Or should I expect to wait until Vega?

I guess no one could possibly know and I'm just impatiently waiting for benchmarks.
 
Supposed benchmarks for the Polaris 10:

You have indentified it as cut down Polaris 10, but techpowerup identified it as cut down Polaris 11 [small 50W chip]
http://www.techpowerup.com/222156/amd-radeon-r9-480-non-x-performs-close-to-r9-390x

If this bench is correct [and who knows if it is], I think we are in for a great ride this summer. Cut down version of small Polaris 11 [480 non-X] being around 390X for ~$200? Sign me up.

IMO, we badly need some official news. All these rumors are pointing in so many directions. :D
 
Where's the rush? They will be on Computex, and they can still launch at the same time as Nvidias paper launch happens.

They need those benchmarks out before Nvidia starts shipping out Pascals. I ain't gonna wait if they're coming in later but I hope they can even the market a little.
 
I predict weaker than Pascal at much cheaper prices. It should be a good fit. But I'd like to know Polaris before buying Pascal.
 
AMD better show up at Computex.

I have a Freesync monitor now, so I'd prefer a nice AMD card to tide me over until Vega to get the advantages of Freesync. I can probably pretty easily survive on my 970 until next year. But I'd like to see a good offering from AMD regardless as I'm hoping it gives me enough incentive to switch.
 
So we are targeting Polaris 10 to be somewhere between "barely enough for good vr" to "almost beating 980Ti".

:D

Are 390X or Fury not good for VR?

VR is supposed to be good with 970 or 390X if Oculus and HTC's recommended specs are anything to go by. The launch titles for Rift and Vive have generally conformed to these specs, and I expect that both companies will be pretty strict on enforcing specs to get certified in the early going since you can make people vomit if your game stutters or drops too many frames.
 
Can't wait to see the Polaris specs. Building a first-time PC and I'm thinking about going with AMD now, as the price I can save on a FreeSync monitor can go towards better parts elsewhere. Now that I've heard AMD is stressing efficiency, that negates my concerns over their previous cards. Thinking about starting with their rumoured $300 card and moving onto the big Vega upgrade when that happens. But we'll see with the unveiling and benchmarks.

When is their next showing?
 
AMD needs to have this out in early 2017 before 1080Ti.

ipZ4INe.jpg
b3S8PpE.jpg
 
You have indentified it as cut down Polaris 10, but techpowerup identified it as cut down Polaris 11 [small 50W chip]
http://www.techpowerup.com/222156/amd-radeon-r9-480-non-x-performs-close-to-r9-390x

If this bench is correct [and who knows if it is], I think we are in for a great ride this summer. Cut down version of small Polaris 11 [480 non-X] being around 390X for ~$200? Sign me up.

IMO, we badly need some official news. All these rumors are pointing in so many directions. :D

The most interesting result in those benches is the texturing one as that is half of the 390X result. Either it is a full P10 running at around 600 Mhz, a P11 clocked at around 1,375 Mhz or something in between.*

The texturing result makes it possible it is a cut down laptop variant of P10 with reduced clock speeds and if that is the case then the P10 being around the level of Fury X / 980 Ti seems entirely possible if bandwidth is not an issue.

The more I look at the rumours/leaks the more it looks like we are going to get a P10 that has a similar performance uplift over the 390x as the 7870 had over the 6970 which as I have said before puts it right into the Fury X and 980Ti performance window.

* 390X has 176 TMUs @ 1050Mhz. to get half of that result from P10 would need the 160 TMUs (based on the 40 CU rumour) to be clocked at around 600 Mhz. To get it from P11 would need the 64 TMUs (based on the 16 CU rumour) to be clocked at around 1,375 Mhz.
 
1070 should be around 980 Ti at 379$. There is no reason to think AMD can not match more or less that. Then, it will be the usual marketshare/marketing battle.

Then, since AMD will be pressured, they will lower their prices even more like they did with the previous line.
 
1070 should be around 980 Ti at 379$. There is no reason to think AMD can not match more or less that. Then, it will be the usual marketshare/marketing battle.

Then, since AMD will be pressured, they will lower their prices even more like they did with the previous line.

Will be interesting to see if amd talks more about DX12 performance than NV, the closest NV came to it was saying they had improvements in async over Maxwell right?
 
If it's true that the 480 will perform close to a 390X at half consumption and gets positioned in the 380 price point, it will be a hell of a card for 200$ and 1080p. (though realistically I guess it will be more like 249$)
 
Will be interesting to see if amd talks more about DX12 performance than NV, the closest NV came to it was saying they had improvements in async over Maxwell right?
DX12 performance aren't in any way different from DX11 performance or OpenGL performance or Vulkan performance, and NV did talk about performance quite a lot. AMD likes to talk about DX12 performance because their DX11 performance still isn't as goog as NV's, that's all.
 
DX12 performance aren't in any way different from DX11 performance or OpenGL performance or Vulkan performance, and NV did talk about performance quite a lot. AMD likes to talk about DX12 performance because their DX11 performance still isn't as goog as NV's, that's all.

I wonder if they'll do something about that cpu overhead in dx11 with the polaris release.

As long as that's a thing I don't really want to get an amd card again.
 
what hardware issue is causing higher cpu overhead for amd gpus under dx11?

I wasn't talking about just higher CPU overhead in DX11 as there are other reasons why AMD's h/w is performing worse there than in DX12 but you may argue that a higher cost of driver's dispatch is related to how the h/w frontend is made as well.
 
I predict weaker than Pascal at much cheaper prices. It should be a good fit. But I'd like to know Polaris before buying Pascal.

This is about what I expect, but I wonder if Pascal is going to impact their pricing at all. If they were planning to release the full P10 at $300 I wonder if it might end up a bit less than that due to the 1070. Assuming the 1070 is about 980Ti level and P10 is 390X level, $80 cheaper would be okay but if they could dip lower it would probably be better for them.
 
Hoping for a 300€ card with the same value for money as the 1070.

It will be interesting to see what AMD does. Nvidia doesn't have any Pascal cards to replace 960/950. They could rebrand 980/970 and move them down. AMD might just be looking at competing at that level with Polaris.
 
It will be interesting to see what AMD does. Nvidia doesn't have any Pascal cards to replace 960/950. They could rebrand 980/970 and move them down. AMD might just be looking at competing at that level with Polaris.
GP106 cards should be out for back to school.
980 and 970 are EOL.
It's possible that there will be a third card on GP104 as well.
 
This is about what I expect, but I wonder if Pascal is going to impact their pricing at all. If they were planning to release the full P10 at $300 I wonder if it might end up a bit less than that due to the 1070. Assuming the 1070 is about 980Ti level and P10 is 390X level, $80 cheaper would be okay but if they could dip lower it would probably be better for them.

Until we see production drivers and numbers from people other than Nvidia, I think a 1070 reaching 980ti performance is a possibility but not something people should take to the bank.
 
GP106 cards should be out for back to school.
980 and 970 are EOL.
It's possible that there will be a third card on GP104 as well.

Oh I hadn't heard anything about those. Still really interested in AMD cards. The the multiple display pane things Nvidia are doing are super intriguing though. I wonder if AMD has anything similar or anything as surprising.
 
1070/1080 will sell great, so chances for another cutdown of GP104 are IMO possible only if they have serious yield issues.

1070 is 3/4 of 1080 with full memory bus. It's possible that they'll make another SKU which will be the same 3/4 of GP104 but with 3/4 of MCs as well granting it 192 bit bus and 6GB of memory - would fit nicely under 1070 as a 1060Ti.
 
1070 is 3/4 of 1080 with full memory bus. It's possible that they'll make another SKU which will be the same 3/4 of GP104 but with 3/4 of MCs as well granting it 192 bit bus and 6GB of memory - would fit nicely under 1070 as a 1060Ti.

Just like the 660 ti, and that GPU launched at $100 cheaper than the one it was cut down from (and I think with the same amount of VRAM).
 
Don't fuck up, AMD. I bought an R9 390 just to support you (and I've been happy with it. Amazing GPU.) But, I will go back to Nvidia if you have nothing to compete with.
 
People keep saying this, but where is the it coming from? I'll believe after the rumored late May announcement. Also what is midrange? I would put 1070 and 1060 as mid cards.

And wasn't AMD ahead of Nvidia last year in DX12 benches? Even their 290x was beating 900 series in some benches in DX12.

So if they have their driver front end in line with these 14nm cards and they are cheaper, I can see it putting lots of pressure on nvidia. Especially if they show games running on regular current drivers and not some custom Vulkan api that may not be what we see on final drivers for 1080.
 
People keep saying this, but where is the it coming from? I'll believe after the rumored late May announcement. Also what is midrange? I would put 1070 and 1060 as mid cards.

1080/1070 are a 330mm^2 die. Polaris 10 is a 230mm^2 die. Some expect more out of them since they are on a slightly smaller node than Nvidia and typically have more dense designs. Rumors are pegging it at up to 390X tier performance at $300 and under.
 
Especially if they show games running on regular current drivers and not some custom Vulkan api that may not be what we see on final drivers for 1080.
What do you mean? Nvidia had Vulkan support in their regular drivers since very shortly after its release, and it's not really "custom" in any way.
 
People keep saying this, but where is the it coming from? I'll believe after the rumored late May announcement. Also what is midrange? I would put 1070 and 1060 as mid cards.

Speculation based on conflicting statements made by AMD.

The benchmarks, demonstrations, specifications, and actual insider rumors have suggested the opposite.
 
1080/1070 are a 330mm^2 die. Polaris 10 is a 230mm^2 die. Some expect more out of them since they are on a slightly smaller node than Nvidia and typically have more dense designs. Rumors are pegging it at up to 390X tier performance at $300 and under.

AMD die sizes right now are pure speculation. The only thing we know is that 2306 cores or whatever from a benchmark leak.
 
And wasn't AMD ahead of Nvidia last year in DX12 benches? Even their 290x was beating 900 series in some benches in DX12.

So if they have their driver front end in line with these 14nm cards and they are cheaper, I can see it putting lots of pressure on nvidia. Especially if they show games running on regular current drivers and not some custom Vulkan api that may not be what we see on final drivers for 1080.

Funny, Vulkan is basically an evolution of AMD's Mantle but it's bad now because NV is using it in their demos 8)
 
Funny, Vulkan is basically an evolution of AMD's Mantle but it's bad now because NV is using it in their demos 8)

I really hope Vulkan becomes viable. The fact that it stretched Windows versions and works in Linux it could take off for once.
 
I really hope Vulkan becomes viable. The fact that it stretched Windows versions and works in Linux it could take off for once.

Yeah, well, even though I hope so too I don't think that VK will be used very often on Windows. On Linux, Android - that's another story. Apple could've helped here but alas they want to become MS as much as MS wants to become Apple it seems.
 
Top Bottom