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

[Artorius]

More powerful than Polaris? Yes.

My guess is that small Vega will come with GDDR5/X to compete with GP104 while big Vega will be HBM2 and compete with GP100/102.

 

[krumme]

More powerful than Polaris? Yes.

Top vega according to specs will be aprox 20% faster than the 1080 you paid in excess of 900 usd for. For probably 10%-20% less $.
You then buy the next titan or big nv gpu. And so on.

 

[sirmo]

Top vega according to specs will be aprox 20% faster than the 1080 you paid in excess of 900 usd for. For probably 10%-20% less $.
You then buy the next titan or big nv gpu. And so on.

10-20% faster on release and 40-50% faster in 2 years. D:

 

[zentan]

Top vega according to specs will be aprox 20% faster than the 1080 you paid in excess of 900 usd for. For probably 10%-20% less $.
You then buy the next titan or big nv gpu. And so on.

Yea, extra Tax on GPUs and other components is something that most of the EU and Asian countries pay along with many other nations. While the performance in yet unknown for Big Vega but I hope it's extremely good. Not to forget many 1080FE owner would have gotten the card 6-7 months(maybe more) before big Vega shows up.

 

[Yakk]

$199 looks like the best sweetspot to me for mass distribution and increasing user base.

 

[ShintaiDK]

Top vega according to specs will be aprox 20% faster than the 1080 you paid in excess of 900 usd for. For probably 10%-20% less $.
You then buy the next titan or big nv gpu. And so on.

Im sure the top GP102 will be 40-50% faster than my 1080. GV104 being 70% faster and so on.

But good thing I can play and enjoy now, rather than sitting and doing the wait forever game

Life is short. And Vega is far away at this point. Just like GP102.

 

[boozzer]

10-20% faster on release and 40-50% faster in 2 years. D:

just like the 1080 oc vs 980 ti oc. 1070 oc isn't even in the same ball park

prepare for minimum of 2 more generations of 20% increases as they milk the node shrink.

if the next node shrink is harder still, we could even see 4th or 5th gens on this current node. 20% increases. 3-5 years on the same node with 20% increases. argh

 

[SelenaGomez]

Top vega according to specs will be aprox 20% faster than the 1080 you paid in excess of 900 usd for. For probably 10%-20% less $.
You then buy the next titan or big nv gpu. And so on.

According to what specs? Where was this data released.

 

[crisium]

Isn't that part of the cost assessment of Apple more than AMD? If Apple prefers a bin that only needs eight BGA memory packages over one that needs twelve BGA packages and cut down Tonga is the best fit, then why wouldn't you as a company sell this cut down bin to Apple?

The better question is why it never hit the retail dGPU market. 380X 4GB was introduced at $230, and the 390 8GB was $330. AMD said they couldn't find room to fit a 380X 384-bit 6GB card. Can you find the room?

It could be a clickbait header. "9/10 AMD executives fail this quiz on where to price full Tonga. Can you pass?"

 

[R0H1T]

The better question is why it never hit the retail dGPU market. 380X 4GB was introduced at $230, and the 390 8GB was $330. AMD said they couldn't find room to fit a 380X 384-bit 6GB card. Can you find the room?

It could be a clickbait header. "9/10 AMD executives fail this quiz on where to price full Tonga. Can you pass?"

The 290/x were still selling at that time, weren't they? I doubt they could squeeze anything while the original Hawaii cards were still available for sale & relatively cheap as compared to the 390/x

 

[Mopetar]

Yea $250 for a 8gb 480 seems like a very good way for them to sell. If this is the case I will get one myself. But I'm guessing close to $300 is where it would land. Still though, $400 for a gtx 1070 or $100 less for the 480 will only matter if performance benches of each card make it worth that much to someone.

$300 seems like it will be the full die P10, probably with higher clocks. With the 480 at $200, the 8 GB version will be no more than $250. This is AMD not Apple.

 

[Vesku]

The better question is why it never hit the retail dGPU market. 380X 4GB was introduced at $230, and the 390 8GB was $330. AMD said they couldn't find room to fit a 380X 384-bit 6GB card. Can you find the room?

It could be a clickbait header. "9/10 AMD executives fail this quiz on where to price full Tonga. Can you pass?"

Without access to the performance numbers of a 380X 384 bit the question can't be answered well. Just as we need more performance info to better judge 480 price/perf.

 

[sze5003]

$300 seems like it will be the full die P10, probably with higher clocks. With the 480 at $200, the 8 GB version will be no more than $250. This is AMD not Apple.

I hope you're right but we shall see. I hope at that price we can pick a non reference model too.

 

[zlejedi]

10-20% faster on release and 40-50% faster in 2 years. D:

And 200% faster in 20 years when AMD is still on GCN architecture

 

[3DVagabond]

Still confused as to what may go up against the 1070. Full Polaris chip or cut down Vega?

Well, the 480 is $199. It's also a much smaller chip. If it's able to compete with a card that's going to be released at $450 then the 1070 will be the biggest failed rip off attempt ever.

 

[3DVagabond]

And 200% faster in 20 years when AMD is still on GCN architecture

You don't think just because nVidia slaps a new name on something it makes the design more advanced, do you? It's all marketing.

 

[el etro]

I think we don't need more than one Polaris thread. Too much of the same info, would be better to have a unique Polaris thread.

 

[maddie]

Even if that's true it's not going to matter that much. Nvidia has a much stronger brand and can flex more marketing muscle if necessary. AMD is going to be able to regain share, and I think it makes more sense for them not to but heads with NV at the high-end when each can soak up their own market segments. If the 8 GB 480 comes in at below $250 for third party chips with good coolers, I think it will do extremely well.

Also, I still don't think we'll see P10 be branded as a 490. Sure adding an additional X to what would then be RX 480X is a bit ridiculous, but no one calls the cards by their full names. It'll just be a 480X, just like there was a 380X. No one bothers to call it an R9 380X.

The shocking $199 price has a lot of gamers interested in AMD, many for the first time. I wouldn't count on the much stronger brand name being as effective as it has been lately. If RX480 and others models bench well, it will further erode such blind faith

You have to realize that Vega is not just Polaris with HBM2. It is a new design family with additional innovations on top of the use of HBM2. They are claiming a substantial increase in perf/W over Polaris. 2.5X with Polaris and 3.5X with Vega. Vega is 40% more perf/W than Polaris. You don't get this with HBM2 alone.

Silverforce11 posted this in another thread.
Quite clear in JPMorgan's interview, AMD claims this is their fastest ever roll out of a follow up new architecture after Polaris. Time-frames between new generations are normally at least a year or more. They are aiming for 6 months.

I really do see them using a new naming scheme probably based on the Fury name to differentiate the GPU families.

 

[zlejedi]

You have to realize that Vega is not just Polaris with HBM2. It is a new design family with additional innovations on top of the use of HBM2. They are claiming a substantial increase in perf/W over Polaris. 2.5X with Polaris and 3.5X with Vega. Vega is 40% more perf/W than Polaris. You don't get this with HBM2 alone.

Sure you do if you produce radeon 390 power gpu on 150w tdp then hbm2 and bit more conservative clocking would be enough.

Just look at Nano vs Hawaii.

 

[guskline]

Since this is the Radeon RX480 thread, were any RX480s given out, loaned etc to reviewers who flew over to Taipei? Is this info under NDA?

I don't want to start a flame war but Nvidia had a stack of GTX 1080s to hand out to reviewers at release and I understand that AMD is probably in a tighter financial bind but I did not see any mention of handing out RX 480s. I wonder if they will be sent to the persons that attended later in June after the GTX 1070 is released?

 

[Mopetar]

The shocking $199 price has a lot of gamers interested in AMD, many for the first time. I wouldn't count on the much stronger brand name being as effective as it has been lately. If RX480 and others models bench well, it will further erode such blind faith

There are a lot of consumers who don't look at benchmarks or do much research. A few might read some engadget or ars article about the new card launching, but that's the limit for non-enthusiast purchases.

The $199 price tag is probably the best thing AMD has going for it right now, and depending on where the 8 GB 480 lands at on the price spectrum, it's going to make for a compelling purchase for anyone on a budget.

 

[maddie]

Sure you do if you produce radeon 390 power gpu on 150w tdp then hbm2 and bit more conservative clocking would be enough.

Just look at Nano vs Hawaii.

Nano is 37% bigger. I really don't think they will go that route when they claim Vega is a new family. You don't make a much larger, more costly GPU to throw away performance if you can help it.

 

[RussianSensation]

What I see on this forum is people who are for the most part not buyers in the mainstream market complaining. This gpu is not for them.

The smart ones see an incredible value that will vastly increase the lowest common denominator in pc gaming performance. Gtx980/390x performance at $199-229 is going to no only improve the VR market, but also make the viable pc gaming market itself much bigger.

All those people playing old games/moba games on onboard graphics now have a cheap way to get their pc running the latest triple a titles and could even step up to vr if they desire.

Polaris 10 is a boon to pc gaming in general so everyone should just stop with the veiled personal attacks and thread crapping and just wait for reviews. More game studios will cater to pc gamers if the market for the games grows in size enough to rival consoles.

Excellent post; but instead trivial metrics that consumers don't care about such as perf/mm2, die size. The only reason we even discussed P10 and GP104's die size was to gauge their standing in their next Ben lineups relative to Vega and Big Pascal. Even if 1070 was a 120mm2 did, unless NV's is going to sell it for $299, who cares if P10 has worse perf/mm2.

AMD is also attacked for bringing this level of GPU power to $199/$229 price levels as if we should cheer paying more for graphics cards?! It's amusing people trying to crap on P10 by using useless engineering metrics like perf/mm2 when the same people bought GTX780/780Ti over 290/290X.

Right now as it stands $379 1070 costs 90% more than $199 RX 480 4GB, 66% more than the RX 480 8GB, and still 52% more than AIB $249 480 8GB variants. As hard as some people are trying to prove that 1070 is better value, it simply isn't. It's also amazing seeing the elephant in room being ignored over and over -- 84-85% of PC gamers buy GPUs in the $100-300 range. That means it practically doesn't even matter even if 1070 was 52% faster than the 480 8GB because most of the target market isn't going to increase their budget from $229-249 to $379-449.

Looking at 1080's launch, market pricing outside of the US suggests 1070 will cost substantially more as well. Russian stores are reporting at least $500-530 USD for a 1070. Mainstream gamers who bought 750/750Ti/360/370/950/960 clearly don't play games maxes out and are willing to compromise. For this target market using 4xMSAA benchmarks on Ultra (with GWs features on) is meaningless. Someone buying 480 will find it a perfect budget 1080p 60Hz card that's looking like it will obsolete every single videocard in the $150-300 space. We will surely expect a 1060/Ti response from NV but until then the comparisons to a 1070 are odd to say the least. It's like as if 7850 cost $199-229 and GTX670 cost $379-449 back then. Almost no one cross shops like that. Besides, chances are the type of customer who is aiming to get a 1070 probably already has a pretty good card, from which 480 wouldn't be a good upgrade in the first place.

The biggest FAIL of all is that the people promoting 1070 and downplaying 480 are mostly the same ones who never recommended the more powerful 270/270X over 750/750Ti, or a 380 2GB over a 950 or a 280X/380X/290 over a 960. The only way this is possible is if someone is 100% NV brand loyal. You cannot objectively have ignored the entire R9 200/300 series for smashing every NV card in the $100-300 space but then turn around and start promoting that gamers are better off spending more for a 1070.

Simply said, 480 hits the sweet spot and NV has no response and its fans have run out of arguments other than starting to pull out 4GB VRAM limitation at 1080p 60Hz without even any proof of a faster Fury X being bottlenecked by the same VRAM limit at 1080.

Smart mainstream buyers either buys a $199 480 4GB, and upgrades again in 2018, or he/she can spend a bit more for the 8GB version.

Another point: Engineer's perf/watt TPU uses are useless for PC gamers. Trying to claim that 1060 is world's more efficient if it uses 90-100W is highly misleading since in an actual gaming rig, it'll be 170W PC vs. a 230W PC. Who cares!!! This is literally engineer's water cooler talk. You cannot use a videocard in a vacuum. What matters is the Consumer's perf/watt and that is a function of Total System Power Usage.

 

[Silverforce11]

There's some talk/rumor from some EU tech press, AMD is officially launching at E3 PC Gaming Show (June 14th), review samples sent out to the press, NDA for reviews likely 19th.

 

[tviceman]

Excellent post; but instead trivial metrics that consumers don't care about such as perf/mm2, die size. The only reason we even discussed P10 and GP104's die size was to gauge their standing in their next Ben lineups relative to Vega and Big Pascal. Even if 1070 was a 120mm2 did, unless NV's is going to sell it for $299, who cares if P10 has worse perf/mm2.

You don't care; and that's completely fine. Other people like to discuss the technical merits of a chip, including transistor count, die size, available headroom, driver features, peripherals, etc. with the cost to the consumer not withstanding. What is wrong with that? This is a "Video card and graphics forum" and not " video card and graphics hot deals sales forum." Why can't we talk about the technical merits of a GPU without being assaulted about prices?

AMD is also attacked for bringing this level of GPU power to $199/$229 price levels as if we should cheer paying more for graphics cards?! It's amusing people trying to crap on P10 by using useless engineering metrics like perf/mm2 when the same people bought GTX780/780Ti over 290/290X.

Anyone who is ridiculing AMD for bringing great perf/$ is an idiot. Anyone that wants to argue WHY AMD is bringing great perf/$ is not an idiot but is equally attacked the same as the ones you suggest are attacking AMD for low pricing.

Right now as it stands $379 1070 costs 90% more than $199 RX 480 4GB, 66% more than the RX 480 8GB, and still 52% more than AIB $249 480 8GB variants. As hard as some people are trying to prove that 1070 is better value, it simply isn't.

USUALLY (not always), as graphics cards climb the ladder in performance, value goes down. The 1070 is obviously not a better value at $379. It's still probably not a better value at $329, but if Nvidia were to drop the price to $329 upon RX 480's release, I'd sure as heck have no qualms saying it'd be a good investment to step up even if it's still worse perf/$. Would you, or is having the best perf/$ the only metric that matters? If GP107 or P11 comes out, lets say at $129 and offers 25-30% better perf/$ than RX 480, are you suddenly going to start telling people that RX 480 is overpriced and not to buy it? And what is a 1440p or 1080p 144hz gamer to do when the gap in price for 25% more performance exceeds the best perf/$ card that is borderline adequate or inadequate for the games he/she plays? Suck it up and by the cheaper card anyways even if it doesn't deliver the performance desired?

It's also amazing seeing the elephant in room being ignored over and over -- 84-85% of PC gamers buy GPUs in the $100-300 range. That means it practically doesn't even matter even if 1070 was 52% faster than the 480 8GB because most of the target market isn't going to increase their budget from $229-249 to $379-449.

You're probably right, but at the same token Nvidia sold more GTX 970's than GTX 960's according to Steam and the 970 spent most of it's life above $300.

Simply said, 480 hits the sweet spot and NV has no response and its fans have run out of arguments other than starting to pull out 4GB VRAM limitation at 1080p 60Hz without even any proof of a faster Fury X being bottlenecked by the same VRAM limit at 1080.

Isn't it a bit premature to say Nvidia has no response? RX 480 won't be out for another 27 days, GP106 die shots have been on the web for two months now, and Nvidia still can release a third GP104 SKU that can sell for ~$279 or so and still be ~15-20% faster than RX 480 if need be. I'd go ahead and say that with GP106 and GP107 coming in July/August and giving them a complete top-to-bottom refresh (sans the ultra high end) means AMD has no answer for Nvidia. But that's just me.

Smart mainstream buyers either buys a $199 480 4GB, and upgrades again in 2018, or he/she can spend a bit more for the 8GB version.

Unfortunately, the "smart" solution isn't always the be-end-all best solution for enjoyment. Put a price on 60fps for us all... or not, because I'll tell you right now it doesn't matter what card I have; if I can't play with minimum frame rates rarely dropping below 60fps (sans turning down a very select few bells and whistles), I just wait until graphics drivers fix possible performance issues or until I upgrade again.

Another point: Engineer's perf/watt TPU uses are useless for PC gamers. Trying to claim that 1060 is world's more efficient if it uses 90-100W is highly misleading since in an actual gaming rig, it'll be 170W PC vs. a 230W PC. Who cares!!! This is literally engineer's water cooler talk. You cannot use a videocard in a vacuum. What matters is the Consumer's perf/watt and that is a function of Total System Power Usage.

So again, we're not allowed to discuss the merits of a particular chip's attributes? Perhaps we should have split the video card sub-forum into "graphics cards prices" section and "technical discussion" section. And then, from there, further divide each section into an AMD and Nvidia section. I wouldn't want to get caught red handed talking about die sizes or transistor counts in the face of consumer dollar signs! How dare I discuss my own interests and fascinations!