videocardzExclusive: AMD Radeon R9 480 based on Polaris 10

[digitaldurandal]

Because a 390-level GPU for $230 (current 380X price) is boring. You can almost get there now if you sell the game that comes with a 390.

We all want a Fury for $300 instead.

I suspect the former is what we will get. I believe the focus was to bring a design for SoCs and mobile first. Almost everyone on this forum is not really the target market for the card.

I would like to see Vega this year please.

 

[Bacon1]

What has AMD stated about price/performance? Do you have an official statement? I know they have made comments about performance per watt, but I dont recall any official comments about specific price/performance gains. Personally, I think people are expecting way too much from this.

I posted it just a few posts above:

http://forums.anandtech.com/showpost.php?p=38243061&postcount=44

 

[tviceman]

I think Polaris 10 will have Fury (vanilla) performance at 1080p and closer to 390x perfomance at 4k for $299 with Pitcairn-level power consumption. Although AMD will undoubtedly have improved color compression, going from 384 gb/s bandwidth to 192 gb/s will show performance regression when moving up in resolution.

 

[krumme]

I suspect the former is what we will get. I believe the focus was to bring a design for SoCs and mobile first. Almost everyone on this forum is not really the target market for the card.

I would like to see Vega this year please.

I think its difficult for people to accept they are not the target group .
This product is mainstream desktop and mainstream enthusiast mobile.
I think many users in this forum is in the small Vega 1070 price segment. In my world thats enthusiast market.
I seriously dont know who is going to buy the big core from nv when 1080 is probably at the 3% outer edge of the market. I think they went a step to far this time but perhaps they can move the market because they are the market leader.
As rs showed as of now 85% market is below 300 usd.

 

[ZZZAAA]

This is completely off from what I have been reading on the web until today. I thought 300 for a 390/390x would make sense considering its still usually slower than a gtx 980 and the 1070 which is faster than a ti (or titan?) is going to retail for 379!

Also, I don't know if this is relevant but is it usually smart to crossfire a x8x/x card? The top radeon card which I think is the duo is a combination of the x9x/x card, am I right? No way they are two Fury's or X's put together...

 

[3DVagabond]

This is completely off from what I have been reading on the web until today. I thought 300 for a 390/390x would make sense considering its still usually slower than a gtx 980 and the 1070 which is faster than a ti (or titan?) is going to retail for 379!

Also, I don't know if this is relevant but is it usually smart to crossfire a x8x/x card? The top radeon card which I think is the duo is a combination of the x9x/x card, am I right? No way they are two Fury's or X's put together...

The 390X is not "usually" slower than the 980.

 

[sirmo]

I upgrade fairly frequently and I also have multiple rigs. At least two gaming machines in the house at the time.

I can afford any GPU pretty much. But I rarely ever buy one over $400. I will spring for a $500 GPU if it's exceptional value or getting a lower tier card results in a significant performance drop.

Funny thing is the one time I paid more than $500 for the card turned out to be the card with least amount of longevity (my gtx780). Should have went with AMD.

Now I can understand if you're on 4K, you kind of have limited choice there, you do need a top card to push that. Problem is a lot of games or software don't really support UI scaling and reading becomes an issue especially on large monitors you don't want to have close to you. So I stick to my 2k monitor.

 

[ZZZAAA]

LOL, from what I can recall, it has the AMD pulling ahead in most games at 1440p and/or 4k resolutions while the GTX would pull ahead by a slightly larger margin at most games on 1080p, no?

 

[sirmo]

Yeah my 780 hasn't aged well at all.

What sold me on it was the power efficiency. And it has been good in that regard. But it was way overpriced for what it is in terms of performance.

 

[krumme]

The 390X is not "usually" slower than the 980.

Well this graph also shows its utterly uninteresting if p10 is 390x level or fury level.
This 7% (percent points) difference is blown out of proportikns imo. Its like a psychological barrier for this forum.
If you want 39% (percent points) better perf as a 390x user go buy a 1080 for 700.
But besides from that minority what matters is this new tech gets far more affordable.
If anyone is still in doubt; 390x is not mainstream out there

 

[ZZZAAA]

Again, would I gain price/performance from crossfiring the Polaris (390/390x) or would it be better to just get a single vega or 1070/1080 gtx? Keep in mind I have no plans to upgrade my system after this.

 

[master_shake_]

Yeah my 780 hasn't aged well at all.

What sold me on it was the power efficiency. And it has been good in that regard. But it was way overpriced for what it is in terms of performance.

where is this from?

those numbers are staggering.

 

[krumme]

where is this from?

those numbers are staggering.

Its only getting started. In one year when we have clean sheet dx12 games 290x will reach above 980ti levels imo.

Rs have said this for a year and Zlatan warned us like 2.5 years back as i recall. Its serious stuff.

 

[sirmo]

where is this from?

those numbers are staggering.

Indeed. https://www.reddit.com/r/Amd/comments/4kfhhn/amd_cards_ageing_better_than_kepler_and_even/

 

[C@mM!]

Doubt we will have a launch soon. Reason being

A: GloFo is advertising 14nm FinFet availability for Q3 for other customers. Either, AMD is getting every bloody wafer out of GloFo, or the process isn't right yet.

B: AMD launched\rebadged a bunch of 28nm parts. This points to having to service GloFo guaranteed wafer buys if 14nm isn't working

C: AMD signed for wafers from Samsung. Either they are expecting a surge in buying (possible I guess, I'm much more enthused about AMD new architecture than Nvidia's maxwell refresh), or the fact is that GloFo can't deliver on 14nm.

 

[Mopetar]

Again, would I gain price/performance from crossfiring the Polaris (390/390x) or would it be better to just get a single vega or 1070/1080 gtx? Keep in mind I have no plans to upgrade my system after this.

DX12 is supposed to be better for multi-GPU, but recently neither CF or SLI has really been worth it, especially if you like to buy and play titles around launch.

Doubt we will have a launch soon. Reason being

A: GloFo is advertising 14nm FinFet availability for Q3 for other customers. Either, AMD is getting every bloody wafer out of GloFo, or the process isn't right yet.

B: AMD launched\rebadged a bunch of 28nm parts. This points to having to service GloFo guaranteed wafer buys if 14nm isn't working

C: AMD signed for wafers from Samsung. Either they are expecting a surge in buying (possible I guess, I'm much more enthused about AMD new architecture than Nvidia's maxwell refresh), or the fact is that GloFo can't deliver on 14nm.

A) Could also be explained by them having limited capacity (as they work out the kinks and ramp) which AMD is completely buying up. Even TSMC is still adding capacity.

B) Seems highly unlikely as those 28 nm parts were based on TSMCs process and it wouldn't be worth the extra cost of switching over for an old node. That AMD is launching their smallest dies first for better yields also suggests against this.

C) Another Global Floundaries disaster is a distinct possibility. I know Apple has been itching to get away from Samsung, so that may open some additional capacity for AMD in the future, but I imagine its tight for them right now as Samsung just launched new Galaxy phones and was likely using a lot of capacity to make their own SoCs.

 

[casiofx]

Well this graph also shows its utterly uninteresting if p10 is 390x level or fury level.
This 7% (percent points) difference is blown out of proportikns imo. Its like a psychological barrier for this forum.
If you want 39% (percent points) better perf as a 390x user go buy a 1080 for 700.
But besides from that minority what matters is this new tech gets far more affordable.
If anyone is still in doubt; 390x is not mainstream out there

From the graph, 1080 is actually 64% faster than 390X, not 39%.

 

[swilli89]

where is this from?

those numbers are staggering.

Yep, and people continue to say nvidia has better drivers

AMD's achievement with Hawaii and also Tahiti is truly staggering like you said, possibly never before seen value for your dollar over the course of several years.

 

[boozzer]

Yeah my 780 hasn't aged well at all.

What sold me on it was the power efficiency. And it has been good in that regard. But it was way overpriced for what it is in terms of performance.

this should be posted in every amd vs nv thread. it is that relevant. especially in threads where people who ask for help in what to buy.

 

[sirmo]

Doubt we will have a launch soon. Reason being

A: GloFo is advertising 14nm FinFet availability for Q3 for other customers. Either, AMD is getting every bloody wafer out of GloFo, or the process isn't right yet.

B: AMD launched\rebadged a bunch of 28nm parts. This points to having to service GloFo guaranteed wafer buys if 14nm isn't working

C: AMD signed for wafers from Samsung. Either they are expecting a surge in buying (possible I guess, I'm much more enthused about AMD new architecture than Nvidia's maxwell refresh), or the fact is that GloFo can't deliver on 14nm.

We've seen Polaris 11 chips demoed a while back. My theory on GloFo advertising Q3 for customers is because they are making AMD's wafers. Keep in mind the owners of GloFo have a 10% stake in AMD, and AMD has priority. Also AMD has announced Apple design win for P11 in Q1, and is likely producing Polaris 11 chips for them for the upcoming Apple Mac refreshes (which are long overdue).

A schedule like that is set well in advance, and Lisa Su wouldn't be lying to the investors this close to when the products are supposed to launch, if GloFo schedule made it impossible to accomplish.

Rebaging of lower models happens with every generation. It has no meaning.

Securing a second source from Samsung I think has another purpose entirely, but I would rather not speculate on it.

 

[3DVagabond]

Yeah my 780 hasn't aged well at all.

What sold me on it was the power efficiency. And it has been good in that regard. But it was way overpriced for what it is in terms of performance.

Remember all of the concern for Fury's 4GB of VRAM? This is far more of an issue, IMO.

The 980 ti and Fury X are dead even. The 390X is ahead of the 980. I truly believe as we get more DX12/Vulcan games in the mix and nVidia stops optimizing for new games Maxwell is going to nose dive badly. I think people who bought $1000 cards deserve better service and products.

 

[frozentundra123456]

Yeah my 780 hasn't aged well at all.

What sold me on it was the power efficiency. And it has been good in that regard. But it was way overpriced for what it is in terms of performance.

Those changes are somewhat misleading at best. For instance the first box says correctly that 780Ti lost the performance lead. However 290x performance did not increase 138% over 2.5 year. That is patently mis-stated. Furthermore the way the performance difference is calculated is technically correct, but hugely magnifies the relative change in performance. What it is really showing is the change in the difference.

Again looking at the 780Ti, over 2.5 years the 290x went from 90.3 percent of the performance of 780Ti to 103.7% of the performance. So calculating the true difference in performance, not just the change in the difference, gives only a 15% gain in absolute performance.

 

[master_shake_]

this should be posted in every amd vs nv thread. it is that relevant. especially in threads where people who ask for help in what to buy.

i tried to suggest an amd card in another thread...wrong place...

didn't to well.

i'm even running 2 980's..

Remember all of the concern for Fury's 4GB of VRAM? This is far more of an issue, IMO.

The 980 ti and Fury X are dead even. The 390X is ahead of the 980. I truly believe as we get more DX12/Vulcan games in the mix and nVidia stops optimizing for new games Maxwell is going to nose dive badly. I think people who bought $1000 cards deserve better service and products.

imagine how the people who bought in to the titan z feel.

3 months of support.

 

[Stuka87]

Those changes are somewhat misleading at best. For instance the first box says correctly that 780Ti lost the performance lead. However 290x performance did not increase 138% over 2.5 year. That is patently mis-stated. Furthermore the way the performance difference is calculated is technically correct, but hugely magnifies the relative change in performance. What it is really showing is the change in the difference.

Again looking at the 780Ti, over 2.5 years the 290x went from 90.3 percent of the performance of 780Ti to 103.7% of the performance. So calculating the true difference in performance, not just the change in the difference, gives only a 15% gain in absolute performance.

290X performance HAS increased though. Not at a hardware level, but on the software side. AMD's drivers continue to get better over time.

 

[escrow4]

A lot of people here bashed the 1070/1080 for their pricing. It seems that a lot of people care about price.

In the end price is all that matters. Along with decent quality whatever the price is. I'm not expecting a solid gold $300 GPU but one that is well built. But cheap. And proper HEVC decoding. AMD needs an updated UVD.