Radeon RX 480 "Polaris" Launched at $199

[Jono8]

derp

 

[d_brennen]

This is all stuff and nonsense until we have some user benchmarks. Not even sure if review sites are worth a damn these days

 

[humbug]

I've said for a while, it's looking like the 232mm^2 AMD core is going to be significantly closer to the 314mm^2 core from Nvidia than their GP106 which is supposedly around the 200mm^2 mark, if so then there is nothing to believe GP106 will be competitive with RX480, in fact there isn't much to say it will be competitive with RX470 at the moment.

Thats the thing isn't it, AMD Polaris GPU's are tiny, Nvida's Pascal are much bigger.

AMD's smaller GPU's are slower, tho not clock for clock, we don't yet know if that is because AMD want to keep the 480 reference sub 100 Watts and let AIB's go mad on clocks, whatever it is they looked really good for the money on old drivers, a 15% boost on "Polaris Drivers" they are starting to look spectacular, the 480 is still a way behind the 1070 with much lower clocks, but if they have the clock rates which AIB's can exploit for under £300 there is no reason why they can't compete with the 1070 on performance.

 

[AlamoX]

The problem with AMD's strategy is is completely ignores the fact that they will still be competing with Nvidia for the mid and low-end segments. It would make sense if nvidia weren't releasing the the 1060 and 1050 for 6 months, AMD really would grab some market share, but since these mainstream cards are imminent form Nvidia all it means is AMD wont be competing in the high end and nvidia can enjoy a large market segment all to themselves.

Gibbo has pointed out that the 1080 is selling faster than the 970 which was the previous fastest selling card. That should send alarm bells ringing at AMD HQ.


I think AMD would want to have high end cards released around now but 1 or more things have conspired against them:
*) Limited R&D budgets means they simply could get the full product stack out on time, or are even having to provide a reduced product line up. I can imagine we get the 460-470-480 for the low end and Vega which takes on the 1080ti, but the 1070 and 108 will sit in a niche by themselves.
*) Global Foundries have technical issues for chip larger that 220mm^2, AMD have a 350mm chip that they just can't release due to yields.
*) AMD expected HBm2 to be ready much sooner than it is and didn't expect GDDR5X to be ready so soon. they have a mid-sized vega chip to use with HBM2 but can't release it until end of the year at the earliest.
*) They really do want people to go crossfire. Instead of doing all the R&D, and manufacturing of a 350mm^2 chip they rather sell you 2x480. AMd have released 2 official figures for the 480, 1 VR score with a single card which wasn't so exciting, and 1 Crossfire score with heavy marketing showing how it was aster and cheaper than a 1080. This relates to the first point - R&D budget constraints may come to play here.



Either way, Nvidia and AMD can have a price war of the 1060 vs 480 cards and NVidia can recoup plenty of profits form the higher margins 1070 and 1080 cards that sit without a competitor.

i do not believe nvidia have the 1060/1050 as close as you think, and i think the launch of 1070/1080 put a constraint on stock, and nvidia have to chose which one to prioritize, 104 or 106, but in the end one of them will suffer from shortage.
thinking that AMD is going from 200$ card to 800-1000$ card with nthing in between shows the delusion you live in, it's not even about competeting with nvidia, this just cant happen, they will have 2 GPU with Vega 10 & 11, the last one should and have to come as soon as possible with GDDR5X, to fill what you call 1080 nich
i love how you affirm that glofo have issues, source ? your wishing star ? the price of the 470/480 should be a good indicator, that they found the sweat size/spot, and yield are great.
there is a slide where AMD says they chosed the chip size because they know games will be tailored around it, i see the possibility of polaris being the chip used on Navi in 2018.

i think AMD's strategy is smart, they will be making small but steady profit, they might increase market share maybe by double digit, we will see at Q3 results anyway, and they know what performance the vega 11 should be at to beat 1080...

 

[TheRealDeal]

Are they sure the 16k run is a 470?

All the others say 84w tdp but the 16k score says 130w tdp (which is what I suspect the tdp of the 480 will be)

The code name is 67DF:C4 which suggests it is the 470. The 480 is c7.

 

[Jono8]

The code name is 67DF:C4 which suggests it is the 470. The 480 is c7.

I'm being silly and looking at the cpu tdp

Still odd though. There is a 13k score and a 16k score on the driver 16.200.0.0. The there are two 13k scores on 16.150.0.0

So is the 16k a heavily overclocked score or something?

I'm also probably being totally blind but where does it say the gpu codename in the 3dmark links? I know they have a picture of it but on the actual links it isn't there?

 

[sprite-]

I'm being silly and looking at the cpu tdp

I was about to say ....

 

[TheRealDeal]

This bit is interesting and i can't remember if this was ever confirmed.

However, there have been instances in the past where AMD has hinted that it might not be sticking with the 64 SP to 1 CU ratio this time around, so until the official numbers are out, the exact count remains suspect

 

[MyBrainz]

So is the 16k a heavily overclocked score or something?

The 16k score is 470 crossfire isnt it?

 

[TheRealDeal]

The 16k score is 470 crossfire isnt it?

The 16k p score is 2 x but the 16k Graphics score is from a single card.

 

[Jono8]

The 16k score is 470 crossfire isnt it?

No there is this

http://www.3dmark.com/3dm11/11167887

but also this on the same driver

http://www.3dmark.com/3dm11/11167781

 

[MyBrainz]

The 16k p score is 2 x but the 16k Graphics score is from a single card.

Ah ok.. well seems good as my OC'd 7950 gets 'Graphics Score 9689'.

 

[D.P.]

i do not believe nvidia have the 1060/1050 as close as you think

They have publicly shown die shots that predate GP104 retail chips. They ave publicly stated that Drive PX2 units with GP106 chips are shipping to auto OEms in April.

, and i think the launch of 1070/1080 put a constraint on stock, and nvidia have to chose which one to prioritize, 104 or 106, but in the end one of them will suffer from shortage.

A constraint on what stock?

thinking that AMD is going from 200$ card to 800-1000$ card with nothing in between shows the delusion you live in,

Where did I say that?
Polaris 10 will sell up to $300 as explained by AMD. Vega could come in at $400 for the lower end model. Where did you get $800-100- form, no one is going to pay that much for an AMD cards anytime soon.

i love how you affirm that glofo have issues, source ? your wishing star ?

I didn't confirm anything, is English not your native language? I merely pointed it out as 1 possibility. Its called speculation.

the price of the 470/480 should be a good indicator, that they found the sweat size/spot, and yield are great.

Or that they are desperate for sales and can cut prices of their GPUs because they are using GF for which they are contractually obligated to produce a certain number of wafers per year or get fined. In any case, a GPU's price has nothing to do with yiedls, its all down to the value the market is willing to pay for the card. I also don't see how yields on a 220mm chip have any bearing on a larger chip.

there is a slide where AMD says they chosed the chip size because they know games will be tailored around it, i see the possibility of polaris being the chip used on Navi in 2018.

AMD chose that chip size de to costs which is why the GP106 is similar.
I would dam well hope that AMD have a new architecture by 2018.

i think AMD's strategy is smart, they will be making small but steady profit, they might increase market share maybe by double digit, we will see at Q3 results anyway, and they know what performance the vega 11 should be at to beat 1080...

I didn't say AMD's strategy isn't smart, but it is only smart when considering there must be some external or internal factors that are preventing a complete lineup. It is very obvious not a choice AMD would want to make, let Nvidia have a free reign of the highly profitable high end for the next 5-8 months they may have had no choice in the mater, and thus the strategy to focus on mainstream cards is quite obvious.

Their entire strategy will depend on Polaris being significantly better than GP106 such that they can take market share away from nvidia in this segment.I don't see that really being the case based on the known performance and efficiency of Pascal. 470 and 480 cards might still have an edge in price-performance but it wont be so large as to create a massive surge in AMD market share. A lot will simply depend on how Nvidia ant to play it out, GP106 is gong to be slightly cheaper to make and Nvidia have enough positive cash-flow to let 1060 margins creep downwards. Whether they want to or not is a different mater. The last major price wars between AMD and Nvidia were great for consumers but bad for the companies

 

[meldarthx]

No there is this

http://www.3dmark.com/3dm11/11167887

but also this on the same driver

http://www.3dmark.com/3dm11/11167781

If that's with 470 - seems to match other benchmarks - blog from Newegg wrote his thoughts that only 480 would be VR......I had to counter - if the 470 is matching 290 to 290X speeds - it would also be VR ready

Card that can do VR for 150 - that's more game changing than the 480.......

 

[MyBrainz]

No there is this

http://www.3dmark.com/3dm11/11167887

but also this on the same driver

http://www.3dmark.com/3dm11/11167781

Weird as only major difference is 2 more cores from the 3960x CPU

LoL

 

[Jono8]

Weird as only major difference is 2 more cores from the 3960x CPU

LoL

Looks like the 470 has the same core clock of 1266mhz as the 480 as well.

 

[TheRealDeal]

Weird as only major difference is 2 more cores from the 3960x CPU

LoL

I think they might have had some cores disabled on the 3960 as it scored lower in the physics test. It certainly made no sense to me.

 

[Jono8]

Are wccftech sure the C4 part is the RX470 and not for example a mobile variant of the 480?

 

[MyBrainz]

I think they might have had some cores disabled on the 3960 as it scored lower in the physics test. It certainly made no sense to me.

Yeah these results have just confused the whole thing now.. and whats with the 1000Mhz core clock results vs the 1266Mhz? If they're all 470's?

Its all getting silly around here now

 

[drunkenmaster]

Thats the thing isn't it, AMD Polaris GPU's are tiny, Nvida's Pascal are much bigger.

AMD's smaller GPU's are slower, tho not clock for clock, we don't yet know if that is because AMD want to keep the 480 reference sub 100 Watts and let AIB's go mad on clocks, whatever it is they looked really good for the money on old drivers, a 15% boost on "Polaris Drivers" they are starting to look spectacular, the 480 is still a way behind the 1070 with much lower clocks, but if they have the clock rates which AIB's can exploit for under £300 there is no reason why they can't compete with the 1070 on performance.

No, Polaris is bigger. Polaris 11 is up against GP108, Polaris 10 is up against GP106, Vega 11(assuming it's the smaller one) is up against Gp104 and Vega 10 against GP102.

GP106 is under 200mm^2 where Polaris 10 is 232mm^2, because of the size of Polaris it's fairly unlikely that Vega will be less than 350mm^2 meaning 314mm^2 vs 350+mm^2 for Vega 11 and if Vega 11 is that big it stands to reason Vega 10 will be bigger than GP102 as well.