(GURU3D RUMOR) AMD Polaris 10 GPU To Offer Near 980 Ti Performance For 299 USD?

[moonbogg]

I don't play Firestrike.

LOL. You just said that. HAHA

 

[RussianSensation]

You're literally doing the exact thing by saying Polaris 10 will be slower than gpus it aims to replace utilizing 0 facts or data to arrive at that conclusion.....

For some reason almost no one on this forum wants to entertain the idea that AMD could bring R9 290/390/390X level of performance at much lower price levels, rather than try to dethrone Fury X cards at $299-349 levels.

Right now for AMD, it's $329 R9 390 and $429 R9 390X. Now imagine Polaris 10 at $199 with R9 290 (say +5%) performance and Polaris 10 $249 with R9 390X (say +5%) level of performance.

This is what we have now on Newegg:

GTX960 2GB = $170-$230

GTX960 4GB = $180-$250+

R9 380 2GB = $170-$220

R9 380 4GB = $170-$240

R9 380X = $220-$250

These low end cards are selling like hot cakes at $170-250 prices, especially 960 variety. This is because most PC gamers aren't buying GPUs above $300.

What's more impressive for the average PC gamer? ~R9 290 performance at $199 and ~R9 390X at $249 or Fury X/980Ti performance at $349?

The former because most PC gamers don't buy $349+ GPUs. Even based on the most realistic estimates, at best there are 10 million+ R9 290/970 and faster cards sold up until now. That's nothing considering R9 290/780Ti were around in 2013!

If AMD wants to gain market share, they need to target the market segments where the unit volume sales are the highest and where competition is the weakest. That is hands down the $179-199 and $199-249 market segments.

Using 1080p resolution that these gamers are most likely using in this price range, we get:

R9 290 is 76% faster than a GTX950, 45% faster than a GTX960. Bring this level of performance for $199 and it will sell like hot cakes with good marketing. R9 390X is 2X faster than a GTX950, and 67% faster than GTX960. In contrast to that, the 960 barely beats a driver and architecture gimped 760 by 19% on the same chart.

Now think about how well the 960 2GB and 960 4GB turds sold on the market while barely delivering adequate next gen performance over a 1.5 year older 760 they replaced?

What AMD should do is target $199 and $249 price levels and worry about 1080 GP104 later. This is because NV's competitive standing under $280 is atrocious and this is where AMD can deliver the biggest blow.

The most ironic part of all, is that 1080p 60Hz (non DSR/VSR) gaming with Fury X or 980Ti is CPU limited in many modern titles even with an i7-4790K @ 4.9Ghz:
https://www.youtube.com/watch?v=f5lfMogcrPU

Therefore, given how the vast majority of PC gamers do not have anywhere close to an i7-4790K even at stock speeds, until next gen games become even more GPU demanding, I think it's actually better for the mainstream PC gamer to have more GPU performance at $199 and $249 price brackets rather than Fury X performance at $299-349. Probably more than 95% of gamers buying Fury X level of performance in 2016 for 1080p gaming will not have the CPU required to max this GPU out.

Even a stock i7 4790K still bottlenecks 980Ti at 1080p:
http://www.purepc.pl/pamieci_ram/te...pamieci_ram_wybrac_do_intel_skylake?page=0,11

That's why I find it hilarious how so many people keep talking about 1080p resolution. Ironically, it is 1920x1200 and below where you require the fastest CPU possible to not bottleneck your GPU. So if you plan on gaming at 1080p with a 980Ti level card without DSR/VSR, you better be packing an i7 6700K @ 4.8Ghz and at least DDR4-3000. What are the chances the Polaris 10 customer will have this level of PC platform and yet be looking to purchase a $299 GPU? Almost 0%. When many of us discuss how i7 2600K/i5 3570K OC are still good gaming CPUs, that's assuming you are gaming at 1440p and above and you have them overclocked to 4.5-4.8Ghz. If you are using something like a stock i5-6400 or worse a stock i5-2400/2500K, cards like Fury X and 980Ti will be severely CPU bottlenecked at 1080p.

The idea that stock SB/IVB i5s are good enough in 2016 is simply false. Digital Foundry, Techno-Kitchen, PurePC have also shown this to be true.
https://www.youtube.com/watch?v=2ZxZiksWtRQ

Now think about how many people are using CPUs even slower than an i5-2500k for 1080p gaming?

Also, if AMD brings Fury X level of performance at $299, that's only $50 cheaper than their VR Ready chart that shows R9 290X at $349. If AMD is serious about revolutionizing the VR Ready Total Addressable Market, they need to bring R9 390X level of performance to $199-249 inside 110-130W TDP. For the short-term, they can then drop the price of Fury/Nano to $329 and leave Fury X at $399 once GP104 launches. Now, if AMD decides to bring out at 150-175W Polaris 10 with much higher clocks or if Polaris 10 is actually a 2816 SP card not a 2304 SP card, that's another story entirely.

I'd love to be wrong and see Polaris 10 full chip bring Fury X performance at $299 but after the hype behind Fury X, I am staying cautious on the hype.

 

[tential]

If it's R9 290 performance at $200, it's a failure and I'd have thought you'd be one of the first inline to say so.

 

[RussianSensation]

If it's R9 290 performance at $200, it's a failure and I'd have thought you'd be one of the first inline to say so.

You haven't been reading my posts carefully then. It would be a failure for gamers such as myself or yourself who keep track of deals or were able to find Hawaii cards for dirt cheap on the used market. For the rest of the world, current market prices highlight a significant performance gap between cards that today sell for $180-250 and an after-market R9 290/290X/390X. This is especially true for modern titles or DX12 titles where the highly popular 960 is left in the dust by Hawaii cards.

You do realize that R9 390X level of performance is what NV sells right now for $430+ US? Bringing 290X/390X level of performance to $199-249 is a huge improvement for the rest of the world.

Remember those $200 Sapphire Tri-X 290 and $255 PowerColor PCS+ R9 290X prices November 2014 were artificially low. Now in May 2016, you cannot get this level of GPU performance new without some deal like Jet.com, etc.

Think about it, the average PC gamer is now paying $180-200 for a GTX960 2-4GB. Look at how much faster a 290X is in modern titles today. The performance increase is simply staggering, 70%, sometimes 80-90%.

That doesn't mean there won't be some Polaris 10 card faster than a 390X, but I am simply stating that bringing 290X level of performance at $199 MSRP inside a 110-120W TDP would be HUGE for most of the world's GPU market. Don't forget, most people purchased 2GB versions of 960/380 based on the anecdotal defense on forums online. That means suddenly they may get an 8GB card at $199-249, and 6GB with 1060Ti. Finally, there will be rebates over time bringing this level of performance below $199.

Again, look at the math. Fury X is a 280W TDP card with an AIO CLC. AMD is showing 2.5X perf/watt increase over 2014 GCN, not 2015 Fiji. Polaris 10's performance is too hard to estimate since we first heard rumours of the card targetting 110-120W TDP (or even lower based on AMD's demos) vs. latest rumours of Polaris 10 scaling up to 175W TDP. That's a massive variance that makes it to hard to accurately estimate its performance. That doesn't change my opinion that AMD should target $179-199 and $200-$249 price brackets as that's where NV is the weakest.

 

[maddie]

For some reason almost no one on this forum wants to entertain the idea that AMD could bring R9 290/390/390X level of performance at much lower price levels, rather than try to dethrone Fury X cards at $299-349 levels.

....................................................................................................

Also, if AMD brings Fury X level of performance at $299, that's only $50 cheaper than their VR Ready chart that shows R9 290X at $349. If AMD is serious about revolutionizing the VR Ready Total Addressable Market, they need to bring R9 390X level of performance to $199-249 inside 110-130W TDP. For the short-term, they can then drop the price of Fury/Nano to $329 and leave Fury X at $399 once GP104 launches. Now, if AMD decides to bring out at 150-175W Polaris 10 with much higher clocks or if Polaris 10 is actually a 2816 SP card not a 2304 SP card, that's another story entirely.

I'd love to be wrong and see Polaris 10 full chip bring Fury X performance at $299 but after the hype behind Fury X, I am staying cautious on the hype.

Am I detecting a rethink of your earlier position on Polaris performance and pricing?

 

[tential]

I guess we'll see what the launch prices are. It's all about price/performance ratio.

I guess I should be selling my R9 290 soon huh?

Edit: Also, if we go by Steam, the 970 is more popular than the 960. So the $350 price range is the KEY price range to combat in.

Are you telling me AMD is going to completely give up the $350+ price range at launch vs Pascal, and let Pascal command 350+ while AMD goes for the 250 and below price range?

Because if so, that's a massive blunder....

The 1060Ti is going to be at $350, and if AMD doesn't have competition to that card? That's a massive problem.

Fiji commanding the high end for AMD until Vega is just silly too. I'm not buying Fiji after Polaris..... NO ONE SHOULD.

 

[SimianR]

A cool, quiet, low power card offering R9 290 performance for $200 would be great, not a failure. But it also depends on how it stacks up against NVIDIA's new cards.

 

[RussianSensation]

Am I detecting a rethink of your earlier position on Polaris performance and pricing?

It's too hard to predict for me given the TDP data of 110-175W and Polaris 10 leaked specs ranging from 850mhz-1000mhz and 2304-2816 stream processors. It's too much all over the place. I am just going off AMD's own statements that $349 is simply too expensive for wide adoption for VR Ready GPUs. If you bring Fury X performance at $299, even though the performance is good, that's still too expensive. Think about what would sell better 390X level of performance for $199-249 or Fury X level for $349? What else is contradictory is we haven't seen any credible rumours of AMD ending production of Fiji products and replacing them with Polaris 10. OTOH, we have consistently heard this being stated regarding 970-980Ti being replaced by GP104 stack.

"AMD's recent statements are seemingly contrary to those made by its graphics head, Raja Koduri, in January of this year. In an interview with VentureBeat, Koduri explained that one of the Polaris GPUs was a larger, high-performance GPU designed to take back the premium graphics card market currently dominated by rival Nvidia. It now appears that he was simply referring to bringing high-performance down to a more reasonable price point."
http://arstechnica.com/gadgets/2016/04/amd-polaris-will-be-a-mainstream-gpu/

That's why the projections of Polaris 10 competing with GP104 1070/1080 sound like wild fantasies. AT already stated that AMD intends to launch Vega 10 and Vega 11. This just means AMD is prioritizing high-end laptops and mainstream desktop products. I don't consider $299-349 mainstream pricing.

The 1060Ti is going to be at $350, and if AMD doesn't have competition to that card? That's a massive problem.

What makes you think 1060Ti will be $350 not less?

I think you are not keeping track of how fast R9 390X is. At 1440P, it outperforms 960 by 83% and is only a hair behind the 980.

If AMD can bring R9 390X to $199-249, that's ground-breaking given where the market currently sits. You honestly think GTX1060Ti will be much better than $199-249 with 390X level of performance at 1440p? If 1060Ti is 2X faster than a 960, that's only 9% faster than a 390X (122%x2 / 223%). If NV prices that at $350, who is going to buy that if AMD has Polaris 10 ~ 390X for $199-249?

Fiji commanding the high end for AMD until Vega is just silly too. I'm not buying Fiji after Polaris..... NO ONE SHOULD.

Ya and who said anything about Polaris 10 being a spiritual successor to Hawaii? Polaris 10 isn't supposed to be an upgrade path for $399-549 Hawaii GPU owners.

Some of you are living in the clouds if you think a 232-234mm2 Polaris 10 with 110-130W TDP and GDDR5 stands a chance competing with 300-330mm2 GP104 1080 GDDR5X that's likely to have 165W TDP, 180-185W after-market versions akin to after-market 980 cards.

Polaris 10 = 8Gbps GDDR5 256-bit = 256GB/sec
GP104 = if 10Gbps GDDR5X 256-bit = 320GB/sec
GP104 = if 12Gbps GDDR5X 256-bit = 384GB/sec

Sorry, I don't believe in magic.

Just because AMD cherry-picked a random Hitman 2016 scene where Polaris 10 was providing better performance than Fury X doesn't mean anything. They told us Fury X would be overclocker's dream and how did that turn out?

 

[maddie]

It's too hard to predict for me given the TDP data of 110-175W and Polaris 10 leaked specs ranging from 850mhz-1000mhz and 2304-2816 stream processors. It's too much all over the place. I am just going off AMD's own statements that $349 is simply too expensive for wide adoption for VR Ready GPUs. If you bring Fury X performance at $299, even though the performance is good, that's still too expensive. Think about what would sell better 390X level of performance for $199-249 or Fury X level for $349? What else is contradictory is we haven't seen any credible rumours of AMD ending production of Fiji products and replacing them with Polaris 10. OTOH, we have consistently heard this being stated regarding 970-980Ti being replaced by GP104 stack.

"AMD's recent statements are seemingly contrary to those made by its graphics head, Raja Koduri, in January of this year. In an interview with VentureBeat, Koduri explained that one of the Polaris GPUs was a larger, high-performance GPU designed to take back the premium graphics card market currently dominated by rival Nvidia. It now appears that he was simply referring to bringing high-performance down to a more reasonable price point."
http://arstechnica.com/gadgets/2016/04/amd-polaris-will-be-a-mainstream-gpu/

That's why the projections of Polaris 10 competing with GP104 1070/1080 sound like wild fantasies. AT already stated that AMD intends to launch Vega 10 and Vega 11. This just means AMD is prioritizing high-end laptops and mainstream desktop products. I don't consider $299-349 mainstream pricing.

I agree with the low pricing structure and mainstream performance.

Never saw anyone claim Polaris as GP104 competitor however, except when I once mentioned that the highest Polaris card might compete with the lowest GP104 model.

 

[tential]

Nano at $330 still gets stomped out by the 1060ti at $350. It needs to be even cheaper than that going up against a newer architecture like Pascal, with the unbalanced mess Fiji is.

 

[RussianSensation]

I agree with the low pricing structure and mainstream performance.

Never saw anyone claim Polaris as GP104 competitor however, except when I once mentioned that the highest Polaris card might compete with the lowest GP104 model.

The lowest GP104 model, if it has 980 level of performance is nearly 2X faster than a 960. If Polaris 10 and 1060Ti come in around 390X/980 level of performance, they'd be direct competitors at $199-249. Unless you think 1060Ti will be ~ 980Ti and cost $299-349? Like I said, I'd rather keep expectations realistic rather than over-hype yet another unreleased AMD card by setting expectations in the top 10th percentile.

A cool, quiet, low power card offering R9 290 performance for $200 would be great, not a failure. But it also depends on how it stacks up against NVIDIA's new cards.

Ya, especially if AMD brings 110-120W TDP R9 290X/390X level to $199. For the mainstream GPU market, that's FAR more impressive than $299-349 Fury X. The unknown is if there will be 150-175W Polaris 10 chips with much higher shader count than 2304.

 

[tential]

I think 1060Ti will be around 980Ti for $350.
Polaris 10 being at $200-250 and around 390x just means they are aiming at different markets. It's like AMD got the low end and Nvidia the High end. That's just not good...

Edit: With Vega 10 AND Vega 11 coming out afterwards though, that will probably break down into 4 cards (fury fury x r9 490x r9 490)

So it is possible AMD is just giving up the high end to Nvidia for now.

 

[moonbogg]

Based on what Nvidia did with the 8800GT, maybe we should be wondering if PASCAL will top 980ti for $300? BAHAHA! Just kidding. That sounds so incredibly stupid lol. Nvidia giving us a good deal? AHAHAHA!

OK, here are my predictions.

Polaris 10 = 980 for $250 or less. Why? Because AMD said they wanted to increase people with VR capable GPUs. The 970 is already around $300 and is said to be a VR card. It would have to be cheaper than a 970 in order to do a better job than the 970 is already doing. It simply must be cheaper than a 970 and perform better than it.

Polaris 10 will cost $250 or less and perform like a 980.

Take it to the bank. Bet the house on it (your house, not mine). This word is golden.

I stand by my prediction more secure than ever. I'm uh genius.

 

[maddie]

The lowest GP104 model, if it has 980 level of performance is nearly 2X faster than a 960. If Polaris 10 and 1060Ti come in around 390X/980 level of performance, they'd be direct competitors at $199-249. Unless you think 1060Ti will be ~ 980Ti and cost $299-349? Like I said, I'd rather keep expectations realistic rather than over-hype yet another unreleased AMD card by setting expectations in the top 10th percentile.



Ya, especially if AMD brings 110-120W TDP R9 290X/390X level to $199. For the mainstream GPU market, that's FAR more impressive than $299-349 Fury X. The unknown is if there will be 150-175W Polaris 10 chips with much higher shader count than 2304.

Some thing I've been thinking about.

1] Nvidia has 1060Ti, 1070 and 1080 models from GP104

2] Sale numbers skewed to lower end of price range meaning 1060Ti then 1070 then 1080.

3] Using a die yield calculator, a die of 320mm^2 and defect of 0.2/cm^2 [claimed by Samsung and applied to TSMC], we get 55% full die/wafer.

4] I can't see them getting enough defective die to fill 1060Ti orders as at best they can harvest 45% of all produced.


Nvidia will have to cut perfectly good 320mm^2 Gp104 dies to compete with a 232mm^2 top Polaris model. How will that affect pricing power?

 

[raghu78]

Again, look at the math. Fury X is a 280W TDP card with an AIO CLC. AMD is showing 2.5X perf/watt increase over 2014 GCN, not 2015 Fiji. Polaris 10's performance is too hard to estimate since we first heard rumours of the card targetting 110-120W TDP (or even lower based on AMD's demos) vs. latest rumours of Polaris 10 scaling up to 175W TDP. That's a massive variance that makes it to hard to accurately estimate its performance. That doesn't change my opinion that AMD should target $179-199 and $200-$249 price brackets as that's where NV is the weakest.

Again thats an assumption. In the image the 28nm GPUs are placed just above 2015 and not 2014. There are no GPU names mentioned in 28nm GPUs. So don't argue assuming your interpretation is the right one.

http://www.anandtech.com/show/10145/amd-unveils-gpu-architecture-roadmap-after-polaris-comes-vega

btw Polaris 10 seems to have around 4 SKUs with a high chance of a flagship with GDDR5X. That
could allow a wide range of performance and pricing. So there could be a Polaris 10 SKU with 980 Ti perf at USD 299-USD 349 and probably the slowest SKU (lots of sp disabled) with GTX 980/R9 390X perf at USD 199.

http://wccftech.com/amd-gcn-4-0-c99-flagship-polaris-rra-certification/

"The certification number is “MSIP-REM-ATI-102-C99301” and it’s the last six numbers that are of particular interest to us. The certificate covers 4 sub-variants of the product as well (C99302, C99303, C99304, C99305)"

http://vrworld.com/2016/04/13/amd-polaris-10-gpu-beat-competing-pascal/

AMD is generally an early adopter of graphics memory technologies such as GDDR5, HBM. Why do you think AMD cannot have a Polaris 10 with support for GDDR5/GDDR5X when Nvidia Pascal can have ?

 

[Silverforce11]

I think 1060Ti will be around 980Ti for $350.
Polaris 10 being at $200-250 and around 390x just means they are aiming at different markets. It's like AMD got the low end and Nvidia the High end. That's just not good...

Edit: With Vega 10 AND Vega 11 coming out afterwards though, that will probably break down into 4 cards (fury fury x r9 490x r9 490)

So it is possible AMD is just giving up the high end to Nvidia for now.

Ofc they are giving up mid-range and high-end. They have no chips ready for those segments to compete.

Polaris 11 ~110mm2 (1280 SP)
Polaris 10 ~232mm2 (2560 SP)

The next chips up, just follow this logic...

Vega 11 ~340mm2 (3840 SP)
Vega 10 ~460mm2 (5,120 SP)

Cut down Vega 11 = low mid-range (think 380), full Vega 11 = upper mid-range (380X).

Same applies for Vega 10 and it will complete AMD's stack.

Since they aren't releasing Vega chips until Q4 2016 or Q1 2017, they have nothing to compete with GP104 directly. They are different segments.

AMD directly calling it mainstream should be very obvious for people. No, it won't be expensive. No, it won't be super fast. Expect normal mainstream prices, ie. <$299 down to $199 for Polaris 10 SKUs.

 

[parvadomus]

Polaris 10 will match 980Ti performance with some SKU mark my words. RV770 will be back.

 

[Ken g6]

I would say if AMD can't find at least one benchmark to fit this description, Polaris 10 will be a disappointing product. Take [thread=2471634]the Hitman 2016 thread[/thread]. You can buy an AMD GPU now for ~$299 USD that will perform near - or even better than - 980 Ti levels in that game.

 

[Mopetar]

Cut down Vega 11 = low mid-range (think 380), full Vega 11 = upper mid-range (380X).

Same applies for Vega 10 and it will complete AMD's stack.

I'm thinking that Polaris 10 will be the 480, and the cut version will be the 470. Vega 11 is the 490 and Vega 10 will be the Fury replacement.

Vega 11 is probably at the size where they won't get many full chips per wafer and I suspect the ones they do will get saved up until they can release a 490X or something like that.

Demand for new GPUs is going to be massive so I suspect that they'll harvest pretty aggressively.

 

[escrow4]

When are these cards coming out? Doom 4 is less than a fortnight. Was considering a 390 non X, which is close to $100 more than the US price (390 is around $480 ish here local), what is that replacement?

 

[Silverforce11]

When are these cards coming out? Doom 4 is less than a fortnight. Was considering a 390 non X, which is close to $100 more than the US price (390 is around $480 ish here local), what is that replacement?

Still no other info besides after Computex, sometime in June. Which if true, Bits&Chips.it called it months ago, ie, paper launch with availability after.

 

[Riek]

Polaris 10 = 8Gbps GDDR5 256-bit = 256GB/sec
GP104 = if 10Gbps GDDR5X 256-bit = 320GB/sec
GP104 = if 12Gbps GDDR5X 256-bit = 384GB/sec

Sorry, I don't believe in magic.

I agree, but the numbers above are true for the 1080.
If 1070 and 1060 come with regular GDDR5 as the rumours go, the bandwidth is just the same as polaris 10 and by definition a lot less than 980ti.

So if we assume that polaris 10 doesn't have the bw to compete with 980ti, we can say the same for 1070 and for 1060 right?

 

[Timorous]

For some reason almost no one on this forum wants to entertain the idea that AMD could bring R9 290/390/390X level of performance at much lower price levels, rather than try to dethrone Fury X cards at $299-349 levels.

Right now for AMD, it's $329 R9 390 and $429 R9 390X. Now imagine Polaris 10 at $199 with R9 290 (say +5%) performance and Polaris 10 $249 with R9 390X (say +5%) level of performance.

This is what we have now on Newegg:

GTX960 2GB = $170-$230

GTX960 4GB = $180-$250+

R9 380 2GB = $170-$220

R9 380 4GB = $170-$240

R9 380X = $220-$250

These low end cards are selling like hot cakes at $170-250 prices, especially 960 variety. This is because most PC gamers aren't buying GPUs above $300.

What's more impressive for the average PC gamer? ~R9 290 performance at $199 and ~R9 390X at $249 or Fury X/980Ti performance at $349?

The former because most PC gamers don't buy $349+ GPUs. Even based on the most realistic estimates, at best there are 10 million+ R9 290/970 and faster cards sold up until now. That's nothing considering R9 290/780Ti were around in 2013!

If AMD wants to gain market share, they need to target the market segments where the unit volume sales are the highest and where competition is the weakest. That is hands down the $179-199 and $199-249 market segments.

Using 1080p resolution that these gamers are most likely using in this price range, we get:

R9 290 is 76% faster than a GTX950, 45% faster than a GTX960. Bring this level of performance for $199 and it will sell like hot cakes with good marketing. R9 390X is 2X faster than a GTX950, and 67% faster than GTX960. In contrast to that, the 960 barely beats a driver and architecture gimped 760 by 19% on the same chart.

Now think about how well the 960 2GB and 960 4GB turds sold on the market while barely delivering adequate next gen performance over a 1.5 year older 760 they replaced?

What AMD should do is target $199 and $249 price levels and worry about 1080 GP104 later. This is because NV's competitive standing under $280 is atrocious and this is where AMD can deliver the biggest blow.

The most ironic part of all, is that 1080p 60Hz (non DSR/VSR) gaming with Fury X or 980Ti is CPU limited in many modern titles even with an i7-4790K @ 4.9Ghz:
https://www.youtube.com/watch?v=f5lfMogcrPU

Therefore, given how the vast majority of PC gamers do not have anywhere close to an i7-4790K even at stock speeds, until next gen games become even more GPU demanding, I think it's actually better for the mainstream PC gamer to have more GPU performance at $199 and $249 price brackets rather than Fury X performance at $299-349. Probably more than 95% of gamers buying Fury X level of performance in 2016 for 1080p gaming will not have the CPU required to max this GPU out.

Even a stock i7 4790K still bottlenecks 980Ti at 1080p:
http://www.purepc.pl/pamieci_ram/te...pamieci_ram_wybrac_do_intel_skylake?page=0,11

That's why I find it hilarious how so many people keep talking about 1080p resolution. Ironically, it is 1920x1200 and below where you require the fastest CPU possible to not bottleneck your GPU. So if you plan on gaming at 1080p with a 980Ti level card without DSR/VSR, you better be packing an i7 6700K @ 4.8Ghz and at least DDR4-3000. What are the chances the Polaris 10 customer will have this level of PC platform and yet be looking to purchase a $299 GPU? Almost 0%. When many of us discuss how i7 2600K/i5 3570K OC are still good gaming CPUs, that's assuming you are gaming at 1440p and above and you have them overclocked to 4.5-4.8Ghz. If you are using something like a stock i5-6400 or worse a stock i5-2400/2500K, cards like Fury X and 980Ti will be severely CPU bottlenecked at 1080p.

The idea that stock SB/IVB i5s are good enough in 2016 is simply false. Digital Foundry, Techno-Kitchen, PurePC have also shown this to be true.
https://www.youtube.com/watch?v=2ZxZiksWtRQ

Now think about how many people are using CPUs even slower than an i5-2500k for 1080p gaming?

Also, if AMD brings Fury X level of performance at $299, that's only $50 cheaper than their VR Ready chart that shows R9 290X at $349. If AMD is serious about revolutionizing the VR Ready Total Addressable Market, they need to bring R9 390X level of performance to $199-249 inside 110-130W TDP. For the short-term, they can then drop the price of Fury/Nano to $329 and leave Fury X at $399 once GP104 launches. Now, if AMD decides to bring out at 150-175W Polaris 10 with much higher clocks or if Polaris 10 is actually a 2816 SP card not a 2304 SP card, that's another story entirely.

I'd love to be wrong and see Polaris 10 full chip bring Fury X performance at $299 but after the hype behind Fury X, I am staying cautious on the hype.

Last time we had a node shrink AMD launched the 7870/7850 that performed at the same level as the previous gen top end cards and Pitcairn was 212mm^2. It also had fewer shaders than Cayman but even with the poorly optimised for GCN launch drivers it still out performed the 6970 by a good margin.

I do not see any reason to think that the performance profile for P10 will be much different to the last node shrink which would put it in the 390X - Fury X performance category for the cut down and full die respectively.

The big question will be prices and they could do $249 and $349 again like with 7870/7850 or they could be more aggressive and do something like $229/$299.

 

[3DVagabond]

I'm curious to see how the newest gen copes with the bandwidth limitation from current available memory.

 

[RussianSensation]

Last time we had a node shrink AMD launched the 7870/7850 that performed at the same level as the previous gen top end cards and Pitcairn was 212mm^2. It also had fewer shaders than Cayman but even with the poorly optimised for GCN launch drivers it still out performed the 6970 by a good margin.

I do not see any reason to think that the performance profile for P10 will be much different to the last node shrink which would put it in the 390X - Fury X performance category for the cut down and full die respectively.

The big question will be prices and they could do $249 and $349 again like with 7870/7850 or they could be more aggressive and do something like $229/$299.

HD7870 vs. 6970 comparison isn't valid:

1) 6970 used far inferior IPC architecture with VLIW-4. 6970 was not a compute architecture. Fury X has the fastest Async compute horsepower out of any consumer GPU today. In DX12 games, Fury X starts to pull away from 290X and beats the 980Ti. Those 4000+ shaders are better utilized than under DX11. A lot more DX12 games will come out, further allowing Fury X to extend its lead over the 290X.

2) 6970 was not a 280W GPU with an AIO CLC (effectively the Fury X on air would likely be pushing close to 300W)

3) 6970 only had a die size or 389mm2, Fury X has 596mm2. Asking 232mm2 Polaris 10 to pull off a similar feat is much harder without a 30-40% GPU clock uplift over Fury X.

4) 6970 didn't have the most advanced per transistor/die space memory controller, Fury X does. HBM1 Fiji memory controller is more efficient than any AMD memory controller in the entire stack. It's why the even more efficient Vega uses HBM2, not GDDR5X.

5) 6970 didn't have almost 2X the memory bandwidth of 7870, Fury X appears to have that over 256-bit Polaris 10 GDDR5.

6) 6970 had lower clocks than 7870 too, while leaks for Polaris 10 show 850-1000mhz. Bad sign. Not 1 leak so far indicating Polaris 10 could have 1400-1500mhz clocks.

7) AMD decoupled the ROPs from the memory controllers in GCN which allowed a huge throughput in real world ROP performance. With the same # of ROPs, 7970 had > 50% higher pixel Fillrate.
http://www.anandtech.com/show/5261/amd-radeon-hd-7970-review/26

I don't expect Polaris 10's ROPs to be 50% faster than Fury X's.

Using 7870 vs. 6970 actually shows how much the cards are stacked against Polaris 10 to match the Fury X at 1440-4K. GCN was the biggest leap in a decade for AMD, surely far more than the leap GCN 4.0 will bring over Fury X.