AMD Polaris Thread: Radeon RX 480, RX 470 & RX 460 launching June 29th

[nevdawg]

*Assuming 480 performs right between a 970 and a 980.

Pretty big win for gamers who don't already have 390 and 970's and above

 

[coercitiv]

The other thing is what people said about not being sure what the power draw really is until it gets tested. If it is a genuine 150w though? Not a good place to be. Nil laptop sales, won't be near competitive at the top end etc.

You can't have 150W TDP for 390X performance levels and claim up to 2.8x increase in perf/watt over last gen chips. It simply doesn't add up: 2.8x over 280W TDP equals 100W TDP for same performance. Either performance is significantly higher, power is significantly lower, or AMD is lying about the perf/watt improvements with a straight face.

 

[Keysplayr]

Is it at all possible that AMD had no intention of competing against 1070/1080 with Polaris10, but perhaps against what could be 1060/1060Ti GP106? Ya know, whenever that comes round?

 

[boozzer]

*Assuming 480 performs right between a 970 and a 980.

Pretty big win for gamers who don't already have 390 and 970's and above

it is basically the default upgrade for anyone who doesn't have a 290 or 970 right now with a budget. now nv needs to answer back with 1060 if it can.

 

[crisium]

If it is actually close to 150W there could be OCing problems.

AMD wants the power efficiency image for these, but having a solid OC is also an image worth having.

Was there a big OC difference between 270 (1x6pin) and 270X/7870 (2x6pin) even though all 3 cards were fully enabled Pitcairn?

 

[Stuka87]

You can't have 150W TDP for 390X performance levels and claim up to 2.8x increase in perf/watt over last gen chips. It simply doesn't add up: 2.8x over 280W TDP equals 100W TDP for same performance. Either performance is significantly higher, power is significantly lower, or AMD is lying about the perf/watt improvements with a straight face.

You are confusing TDP with power consumption, they are not the same.

 

[Stuka87]

Is it at all possible that AMD had no intention of competing against 1070/1080 with Polaris10, but perhaps against what could be 1060/1060Ti GP106? Ya know, whenever that comes round?

This is what all rumors have been saying, and what AMD has basically been saying since day one. This is not a surprise. Polaris is a main stream chip.

Vega is AMD's big chip, it will fill the roll of what 390X/Fury used to fill last gen.

 

[exar333]

If it is actually close to 150W there could be OCing problems.

AMD wants the power efficiency image for these, but having a solid OC is also an image worth having.

Was there a big OC difference between 270 (1x6pin) and 270X/7870 (2x6pin) even though all 3 cards were fully enabled Pitcairn?

Seems like the 270 OC'd decently with power target at 120%. Didn't seem to limit it much at all. Look's like both the 270 and 270x overclocked similarly from what I can see:

http://www.bit-tech.net/hardware/graphics/2013/11/13/amd-radeon-r9-270-review/10

http://www.hardocp.com/article/2014/08/11/asus_radeon_r9_270x_directcu_ii_top_video_card_review/3#.V08j1ZErL-g

 

[Keysplayr]

This is what all rumors have been saying, and what AMD has basically been saying since day one. This is not a surprise. Polaris is a main stream chip.

Vega is AMD's big chip, it will fill the roll of what 390X/Fury used to fill last gen.

I mean, not really a surprise to me either looking at performance and price bracket. I really wasn't keeping up with what AMD was saying from the beginning nor paying attention to rumors. But thanks for the thoughts on this. It makes the most sense, it seems.

 

[Daedalus685]

Is it at all possible that AMD had no intention of competing against 1070/1080 with Polaris10, but perhaps against what could be 1060/1060Ti GP106? Ya know, whenever that comes round?

Yes it is possible and is certainly the case.

There is one more pascal chip to be released to gamers (edit: one more "larger" chip) (1080Ti/Titan). AMD has 2 more chips to release bigger than this Pascal (Hawaii and Fury replacement - Rx 490 and Fury 2 if you will).

For the few that insist on comparing them, it is either disingenuous or simply ignorant to force a comparison between the rx 480 and the gtx 1070/80. Or there are people that believe AMD will segment the market with no cut down SKUs but a new chip at each price point.

Now, that is not to say the Nvidia marketing machine won't make this strategy explode in AMD's face... but that is neither here nor there.

I'd assume that AMD planned HBM2 chips for their high end and enthusiast. Thus there is another factor in when they can release. That may prove detrimental to their share of the high end but won't know for some time.

 

[exar333]

Yes it is possible and is certainly the case.

There is one more pascal chip to be released to gamers (1080Ti/Titan). AMD has 2 more chips to release bigger than this Pascal (Hawaii and Fury replacement - Rx 490 and Fury 2 if you will).

It is either disingenuous or simply ignorant to force a comparison between the rx 480 and the gtx 1070/80. Or there are people that believe AMD will segment the market with no cut down SKUs but a new chip at each price point.

Now, that is not to say the Nvidia marketing machine won't make this strategy explode in AMD's face... but that is neither here nor there.

I'd assume that AMD planned HBM2 chips for their high end and enthusiast. Thus there is another factor in when they can release. That may prove detrimental to their share of the high end but won't know for some time.

Good post.

For better or worse, AMD hitched their higher-end options to HBM/HBM2. Unfortunately that meant a limit of 4GB on their Fury line (although it performed pretty well regardless) and is a prerequisite for Vega. On the positive, it gave some nice efficiency gains and worked well.

NV went a different route and mitigated somewhat by sticking with high-speed GDDR5, using GDDRX and only using HBM2 for P100 (so far). GP102 likely will have the option for HMB2.

Because of this, NV likely released mid/mid-high first and will follow with lower-end and the high, high end later. AMD went with mid/mid-low and is waiting to add the high end with HBM2 (most likely).

Different strategies, but both can work. It is important to not focus too much on the technology solution, but rather performance and constraints (if applicable). Sometimes I think the forum gets so set that 'only one company can be right'. It is not always that way...every decision has it's pluses and minuses.

 

[topmounter]

AMD is certainly getting a lot of good press with the $199 price point. The only negative comments are from folks that weren't paying attention (or are trolling) and creating the fiction that Polaris10 is somehow supposed to be a direct, card for card competitor to 1070/80. I just wonder if NV and AMD are overplaying the VR aspect since all that hullabaloo seems to have disappeared (at least until a AAA game or movie shows up).

 

[coercitiv]

You are confusing TDP with power consumption, they are not the same.

I'm not confusing anything, all this started with people claiming it's a 150W TDP card because AMD slide said Card Power: 150W. It was convenient for some to interpret this as "card power consumption".

 

[Concillian]

Was there a big OC difference between 270 (1x6pin) and 270X/7870 (2x6pin) even though all 3 cards were fully enabled Pitcairn?

Not really. I don't know about the 270/270x era, but I had a 7850 and it OCed essentially the same as all the 7870s did.

You're essentially "overclocking" the connector specs in terms of the current flowing through the pins, but like everything else, there's some margin built in.

 

[Zstream]

Not really. I don't know about the 270/270x era, but I had a 7850 and it OCed essentially the same as all the 7870s did.



You're essentially "overclocking" the connector specs in terms of the current flowing through the pins, but like everything else, there's some margin built in.

Yep, myself and a few friends had purchased 270's and we had no issues at all over clocking to 270x specs or beyond.

 

[VulgarDisplay]

If these things can overclock 15-25% they will go from being great cards for the money to probably the most ridiculous deal in gpu history.

That would put then within spitting distance of a stock 1070 wouldn't it?

Sent from my XT1575 using Tapatalk

 

[Head1985]

If these things can overclock 15-25% they will go from being great cards for the money to probably the most ridiculous deal in gpu history.

That would put then within spitting distance of a stock 1070 wouldn't it?

Sent from my XT1575 using Tapatalk

Around GTX980TI stock(if rx 480 is at GTX980 level or faster)

 

[Concillian]

AMD is certainly getting a lot of good press with the $199 price point. The only negative comments are from folks that weren't paying attention (or are trolling) and creating the fiction that Polaris10 is somehow supposed to be a direct, card for card competitor to 1070/80. I just wonder if NV and AMD are overplaying the VR aspect since all that hullabaloo seems to have disappeared (at least until a AAA game or movie shows up).

I actually think that segment of people who aren't paying attention is pretty large... like most of the market. I'm very happy with a $200 390 --> 390x, it's the performance I was expecting, but honestly a little cheaper than I expected. However, I feel like most people in that price range pay little to no attention to the details of what's coming.

As far as VR, to me it's 3DTVs all over again. It's cool, but it's very unlikely that enough people will spend the necessary coin to have a critical mass of people with the headsets to really justify enough developer time and energy spent on it.

 

[antihelten]

Is it at all possible that AMD had no intention of competing against 1070/1080 with Polaris10, but perhaps against what could be 1060/1060Ti GP106? Ya know, whenever that comes round?

The main target for Polaris 10 and 11 would appear to be GP106 and GP107, but even so AMD may still intend to compete to some degree with the GTX 1070. It all depends upon whether or not a higher binned version of Polaris 10 exists (Lisa Su's comment about covering the entire segment from $100-300 would seem to suggest so), and how said version performs.

I don't expect a $300 Fully unlocked Polaris 10 to beat or even match a GTX 1070, but if it can get close enough (say within 10%), then they would effectively be competing with each other.

The GTX 1080 is of course out of the question and will probably have free reign until GP102 and Vega.

 

[exar333]

I actually think that segment of people who aren't paying attention is pretty large... like most of the market. I'm very happy with a $200 390 --> 390x, it's the performance I was expecting, but honestly a little cheaper than I expected. However, I feel like most people in that price range pay little to no attention to the details of what's coming.

As far as VR, to me it's 3DTVs all over again. It's cool, but it's very unlikely that enough people will spend the necessary coin to have a critical mass of people with the headsets to really justify enough developer time and energy spent on it.

I know its a bit off topic, but I do think VR has some legs. 3DTV was more gimmicky and tried to take a group activity (like watching a movie) and force everyone to use goggles. VR is pretty engrossing, and software is really being built to take full and unique advantage of it. Where VR and AR go is definitely up in the air, but I think it is here to stay in some form or another. It is a LOT of fun.

 

[Headfoot]

Remember of course that the target for a $200 midrange GPU are people still using 7850's, 7870s or their rebranded equivalents, 660s, 660tis, 760s. Old $200 to $250 gpus. Maybe even folks still holding onto 7950s and 7970s.

The jump from a 7850 to a RX 480 even at just matching a 390x would be quite substantial for $200.

And the folks still on 40nm ala 560 Ti or 6950s will see a colossal upgrade.

 

[Concillian]

If these things can overclock 15-25% they will go from being great cards for the money to probably the most ridiculous deal in gpu history.

That would put then within spitting distance of a stock 1070 wouldn't it?

It's really not all that different from 28nm launch, where 7850s were around $200 and OCed to equal OC'ed GTX580. Definitely not the most ridiculous deal in gpu history, but certainly not a bad deal.

 

[Qwertilot]

I'm not confusing anything, all this started with people claiming it's a 150W TDP card because AMD slide said Card Power: 150W. It was convenient for some to interpret this as "card power consumption".

To be fair, what else would it be rationally taken to mean? It was in a marketing presentation for pities sake and they're hardly shy of being generous

It seems very possible that - for instance - the efficiency gains are seen on lower clocked mobile chips, while clocking the SKU they were describing high enough meant that they blew the power up a bit more than you'd like.

On a different topic - one thing I wouldn't take as absolutely read is that AMD are going to fill every niche in the market with cards. They were only really managing that at 28nm by rebranding all over the place. Could definitely imagine them being quite selective with where they target.

This low end up stuff ties in with their console deals etc and Vega is probably going to be after some professional compute markets as well as gaming.

 

[Stuka87]

Maybe even folks still holding onto 7950s and 7970s.

The jump from a 7850 to a RX 480 even at just matching a 390x would be quite substantial for $200.

And the folks still on 40nm ala 560 Ti or 6950s will see a colossal upgrade.

This is me. I just need to determine if the 480X will be enough for me, or if I want to wait for Vega. My 7950 is still running fine, but it not longer runs things at higher settings @1080P.

 

[exar333]

To be fair, what else would it be rationally taken to mean? It was in a marketing presentation for pities sake and they're hardly shy of being generous

It seems very possible that - for instance - the efficiency gains are seen on lower clocked mobile chips, while clocking the SKU they were describing high enough meant that they blew the power up a bit more than you'd like.

On a different topic - one thing I wouldn't take as absolutely read is that AMD are going to fill every niche in the market with cards. They were only really managing that at 28nm by rebranding all over the place. Could definitely imagine them being quite selective with where they target.

This low end up stuff ties in with their console deals etc and Vega is probably going to be after some professional compute markets as well as gaming.

Power levels playing a DX11 may be different than a DX12 title w/ async and also may be different that doing a compute task. TDP can be defined as the 'max' or the 'typical'. Its not always equal for the same task...