videocardzAMD Radeon R9 490X and R9 490 launches in June/Pro Duo launches on April

[Eymar]

If AMD want keep fury name then:
Polaris11-470x/470
polaris10-480x/480
Vega11-490x/490
vega10-Furyx/fury

If they dont want keep fury name:
polaris11-460x/460
Polaris10-470x/470
vega11-480x/480
Vega10-490x/490

Vega11-Rage Fury
Vega10-Rage Fury Maxx

Make it happen ATI, I mean AMD.

 

[Adored]

There are two possibities of making the VR bottom limit cheaply.

Polaris 11 at 1500MHz will be close to a 970 around 125W or so. This one looks more like fantasy than reality but you never know.

The cut down Polaris 10 should be well ahead of the 290 or 970 at $250. I don't like that one either but it seems more likely than the first choice.

Incidentally I believe we'll be really surprised at how cheap all of these cards are. This is all about grabbing mindshare back and damaging Nvidia's profit. The only chip that matters to AMD in monetary terms is Zen, and if it fails the company is dead anyway regardless of how good Polaris is.

 

[xpea]

There are two possibities of making the VR bottom limit cheaply.

Polaris 11 at 1500MHz will be close to a 970 around 125W or so. This one looks more like fantasy that reality but you never know.

The cut down Polaris 10 should be well ahead of the 290 or 970 at $250. I don't like that one either but it seems more likely than the first choice.

Incidentally I believe we'll be really surprised at how cheap all of these cards are. This is all about grabbing mindshare back and damaging Nvidia's profit. The only chip that matters to AMD in monetary terms is Zen, and if it fails the company is dead anyway regardless of how good Polaris is.

and by same time put even more pressure on AMD revenue ! with their current financial situation, not sure Lisa Su agrees with this strategy...

 

[topmounter]

Don't do it. I'll give you 2 solid reasons. The first is that Fury's price/performance relative to an after-market 390 is terrible. If you want a stop-gap card, maybe find an after-market 290/290X or wait for a deal on a 390. It's possible to find an after-market 390 for $275 from time to time (with MIR) and then sell Hitman for $30. That makes it at $245 stop-gap. The second reason is that after selling the Division for $30, you can end up with a 980Ti for about $540. I'd say it's easily worth the premium over a $470 Fury. Personally, right now I'd only go after-market used 290/290X or 390 or bust. Otherwise, I would just wait since you already waited this long.

I'm trying I've been working through my Steam back-catalog. My R9 270 is still serviceable for 1080p, but my ideal target for a new card is no compromises 1440p and playable non-twitch 4k.

 

[Adored]

and by same time put even more pressure on AMD revenue ! with their current financial situation, not sure Lisa Su agrees with this strategy...

But the money from discrete GPU is nothing compared to what they'll get from a good Zen.

AMD aren't stupid, they know they'd need to be massively ahead for anything to change wrt Nvidia. Even when they were massively ahead at 40nm, people ended up buying Nvidia anyway.

They need to comfortably beat the 970 at a knockout price and then beat the 980 Ti at a really good price with the Polaris cards. Nothing else will make any difference.

 

[RussianSensation]

I'm trying I've been working through my Steam back-catalog. My R9 270 is still serviceable for 1080p, but my ideal target for a new card is no compromises 1440p and playable non-twitch 4k.

What are the rest of your system specs? PSU and CPU + CPU overclock? If you have an R9 270, I sure hope you are at least rocking an i5 3570K OC or similar before considering a $500 GPU upgrade.

They need to comfortably beat the 970 at a knockout price and then beat the 980 Ti at a really good price with the Polaris cards. Nothing else will make any difference.

AMD noted that since R9 290/290X launched, the entire market of 290/970 up until recent months is only 7.5M cards.

Now consider the monumental damage NV did to AMD's market share with 750/750Ti, and then 950/960/970. If you look at it from a volume, market share and mind-share perspective, AMD needs <$300 dGPU market MUCH more than >$300 dGPU market. This is especially so considering how weak NV's sub-$300 dGPUs are right now as far as performance and price/performance goes. This means there is a lot of opportunity for AMD to strike and take away more market share in the $100-300 segments than there is in the $350-650 segments.

The problem is a lot of gamers don't build their own rigs and when they buy a an OEM system with a 300-400W junk PSU, well it comes with a 750/750Ti/950/960. The 450W version will probably have a 970. That means AMD is eliminated from those OEM design wins by virtue of power usage, no other metric matters. That means what AMD really needs more than anything are $50-300 OEM desktop dGPU and mobile dGPU design wins with Polaris 10/11. AMD must focus on sub-75W and sub-150W TDP markets and release cards that are as fast as possible within those 2 power segments. As far as worrying about 980Ti's replacement, they have plenty of time as GP102/100 doesn't seem to be in the cards for 2016.

 

[Kris194]

There is no way that Polaris 10 will be 390x successor, it's simply too small.

 

[xpea]

But the money from discrete GPU is nothing compared to what they'll get from a good Zen.

AMD aren't stupid, they know they'd need to be massively ahead for anything to change wrt Nvidia. Even when they were massively ahead at 40nm, people ended up buying Nvidia anyway.

They need to comfortably beat the 970 at a knockout price and then beat the 980 Ti at a really good price with the Polaris cards. Nothing else will make any difference.

well, for long, RTG has been the money maker at AMD, only took over recently by the semi-custom business.
Problem is that Zen, as good as it will be, has very small chance to succeed. May I remember you that they are fighting Intel that will do whatever it will take to keep their market share. And as history has proven, Intel is not shy of using unethical behavior when needed. With unlimited money, they won't have any issue to counter whatever Zen will bring to the table. This battle is already lost...
On the other side, graphics is more open. They still can make money with it as Nvidia is a more accessible competitor.
So (acting as AMD CEO armchair) I will put more effort in graphics and make sure Polaris will have a successful launch (please no stupid reference design cooler and stable drivers). Then make sure price is adequate with right balance between sales QTY and profit.
my 2cts

 

[Glo.]

There is no way that Polaris 10 will be 390x successor, it's simply too small.

AMD demoed P10 in Hitman DX12 in 1440p Ultra settings and it was locked at 60 FPS. So minimums were higher than R9 390X which is not able to average that framerate in the same settings.

Everything points that GPU that was running that game was the one from SiSoft leak with 2304 GCN cores clocked at 800 MHz, and 8 GB of GDDR5.

Now compare it to Grenada: 1050 MHz, 2816 GCN cores. Looks like Mahigan may be right.

But all of this is only my personal opinion.

 

[Face2Face]

AMD demoed P10 in Hitman DX12 in 1440p Ultra settings and it was locked at 60 FPS. So minimums were higher than R9 390X which is not able to average that framerate in the same settings.

Everything points that GPU that was running that game was the one from SiSoft leak with 2304 GCN cores clocked at 800 MHz, and 8 GB of GDDR5.

Now compare it to Grenada: 1050 MHz, 2816 GCN cores. Looks like Mahigan may be right.

But all of this is only my personal opinion.

Let's hope that 800MHz core clock is as overclockable like Tahiti.

 

[Pinstripe]

I think Polaris 10 will be ~Hawaii level performance and really low TDP, while GP104 will be ~GM200 performance with GM204 level TDP. Now if both come out at the same time, I'm afraid AMD will be humiliated (again).

 

[topmounter]

What are the rest of your system specs? PSU and CPU + CPU overclock? If you have an R9 270, I sure hope you are at least rocking an i5 3570K OC or similar before considering a $500 GPU upgrade.

I took care of the rest a few months ago: i5-6600K, Asus Z170-a Mobo, 16GB DDR4, 650W Corsair p/s

 

[IllogicalGlory]

There is no way that Polaris 10 will be 390x successor, it's simply too small.

It's not the 390X successor, but it is likely that level of performance. Really, it should be called the 470/X.

 

[xpea]

I think Polaris 10 will be ~Hawaii level performance and really low TDP, while GP104 will be ~GM200 performance with GM204 level TDP. Now if both come out at the same time, I'm afraid AMD will be humiliated (again).

I have no idea yet what performance Polaris and consumer Pascal will bring, but one thing is sure, because AMD went public on their Polaris strategy long time ago, Nvidia had plenty of time to prepare (secretly) their counter attack. Crazy to think that some people here believe that AMD will take easily the market by storm with Polaris and Nvidia will watch doing nothing

 

[Adored]

There is no way that Polaris 10 will be 390x successor, it's simply too small.

Assuming Polaris 10 is the 232mm2 GPU mentioned on LinkedIn, it's nowhere near too small to beat the 390X.

Just do a dumb shrink of Hawaii/Grenada at 438mm2. At even 50% it would be around 219mm2. Hawaii doesn't even have delta color compression meaning Polaris wouldn't need the 512-bit MC to get near the required bandwidth. Or just look at the 980, it can match the 390X with a 256-bit bus. Why wouldn't Polaris 10?

If Polaris 10 doesn't beat the 390X then AMD needs to just quit altogether. If Nvidia can get 1480MHz out of big Pascal this early, AMD really ought to smash the 390X at 1250MHz or so.

I try not to get carried away with these things because you can only end up looking stupid, but I expect Polaris 10 to beat Fury X, even with a 256-bit GDDR5 bus.

 

[Adored]

Now consider the monumental damage NV did to AMD's market share with 750/750Ti, and then 950/960/970. If you look at it from a volume, market share and mind-share perspective, AMD needs <$300 dGPU market MUCH more than >$300 dGPU market. This is especially so considering how weak NV's sub-$300 dGPUs are right now as far as performance and price/performance goes. This means there is a lot of opportunity for AMD to strike and take away more market share in the $100-300 segments than there is in the $350-650 segments.

The problem is a lot of gamers don't build their own rigs and when they buy a an OEM system with a 300-400W junk PSU, well it comes with a 750/750Ti/950/960. The 450W version will probably have a 970. That means AMD is eliminated from those OEM design wins by virtue of power usage, no other metric matters. That means what AMD really needs more than anything are $50-300 OEM desktop dGPU and mobile dGPU design wins with Polaris 10/11. AMD must focus on sub-75W and sub-150W TDP markets and release cards that are as fast as possible within those 2 power segments. As far as worrying about 980Ti's replacement, they have plenty of time as GP102/100 doesn't seem to be in the cards for 2016.

I agree that in terms of market share, the lower end stuff is what counts.

What I'm talking about is mindshare though. AMD needs to make people with 970's feel like they need an upgrade. That should be #1 priority for both companies but for AMD it's a must.

Ask yourself what it would take for you to upgrade a 970 to Polaris 10. That's what AMD needs to achieve, or close.

 

[RussianSensation]

I think Polaris 10 will be ~Hawaii level performance and really low TDP, while GP104 will be ~GM200 performance with GM204 level TDP. Now if both come out at the same time, I'm afraid AMD will be humiliated (again).

1. What makes you think they are going to be priced to compete against one another based on your above performance expectations? That's not even logical.

2. If the above is true, that's the most underwhelming 2016 in the cards. Who the hell wants a 120W TDP Polaris with 390X performance for $429? Not happening and that contradicts everything AMD said. Either Polaris has various performance tiers with large disparity aka 4850-4870-4890, or it will be priced well below R9 390X, or it will cost > $329 but perform faster than a 390X.

If both NV and AMD just give us lower power usage at similar prices, who is going to upgrade to that? That would be the most lackluster generation.

I took care of the rest a few months ago: i5-6600K, Asus Z170-a Mobo, 16GB DDR4, 650W Corsair p/s

:thumbsup: Ya, you are good to go then for the most part.

It's not the 390X successor, but it is likely that level of performance. Really, it should be called the 470/X.

That sounds reasonable. 120TDP, 232mm2 die size with ~ 390X performance but priced much more aggressively. Right now 390X is $385+ and 980 is $450+. Bring that level of performance for $279 before rebates and discounts and that's not a bad start. Still I cannot reconcile Polaris 10 running Hitman faster than Fury X.

 

[Head1985]

Well if we look at latest techpower UP review FuryX is only 16% faster than 390x and 6% faster than GTX980 in 1080P.

If samsung 14nm brings same clock increase like TSMC 16nm where 610mm2 tesla have 1500bost thats 25% clock increase vs GM200.

390x-1050Mhz
polaris 10 with 2560SP at 1300mhz(25%more than 390x) + better architecture = FuryX in 1080P

Polaris should Match FuryX in 1080P.But to do that polaris will need better delta compression atleast like NV maxwell level.Because polaris will be 256bit with DDR5
Fury X is just bad in 1080.In higher resolution it will be slower than FuryX.
Furyx is like worst AMD card ever build for 1080p...

 

[Kris194]

AMD demoed P10 in Hitman DX12 in 1440p Ultra settings and it was locked at 60 FPS.

You don't know that. "Ultra settings" rumor comes from the...comment

http://videocardz.com/58621/amd-polaris-10-engineering-sample-pictured#comment-2574370429

 

[RussianSensation]

Well if we look at latest techpower UP review FuryX is only 16% faster than 390x and 6% faster than GTX980 in 1080P.
...
Fury X is just bad in 1080.In higher resolution it will be slower than FuryX.
Furyx is like worst AMD card ever build for 1080p...

Almost all modern NV/AMD GPUs are bottlenecked at 1080p. Compare anything you want from 980 vs.770, 980Ti vs. 780Ti, Fury X vs. 380X, etc. I have long moved on from gaming at 1080P 60 fps. I suppose if they start doing 120-144Hz 1080P benchmarks OR start using SSAA/8xMSAA/DSR/VSR for 1080P, I'll consider those benches more relevant for GPU testing. I cannot imagine spending $400+ on a brand new Pascal/GCN 4.0 GPU and gaming at 1080p 60 fps. Literally a $200 after-market 290 can do that. Even though 1080p is the most popular resolution Steam, that's because most people in the world have GPUs below 970/290 and 1440P and above monitors are expensive in 3rd world/developing countries relative to their income levels.

I think for developed countries, testing $400+ GPUs in 2016 at 1080p 60 fps without DSR/VSR/SSAA or moving the testing to 120-144Hz monitor is almost meaningless now. Of course that's my opinion but I would never upgrade to modern $400+ GPUs and still be gaming on a $150-$200 1080P monitor, not a chance. Ya, I get it there are people with $500-650 GPUs and $200 1080p monitors but those gamers would benefit much more going with a monitor upgrade then. 1080P is the new 1280x1024 in the era where enthusiasts/high-end gamers have moved on to 1440p/1600p/4K. It's time to move 1080p to the lowest 1600x900 tier at TPU and re-introduce 3440x1440 or triple monitor gaming instead.

I've been saying for a while now that TPU needs to either introduce SSAA/VSR/DSR 1080p benches or specifically start benching 1080P 120-144Hz. 1080p @ 60Hz for expensive modern GPUs is like buying a Porsche 911 and putting Honda Accord Coupe tires on it.

 

[Head1985]

Almost all modern NV/AMD GPUs are bottlenecked at 1080p. Compare anything you want from 980 vs.770, 980Ti vs. 780Ti, Fury X vs. 380X, etc. I have long moved on from gaming at 1080P 60 fps. I suppose if they start doing 120-144Hz 1080P benchmarks OR start using SSAA/8xMSAA/DSR/VSR for 1080P, I'll consider those benches more relevant for GPU testing. I cannot imagine spending $400+ on a brand new Pascal/GCN 4.0 GPU and gaming at 1080p 60 fps. Literally a $200 after-market 290 can do that. Even though 1080p is the most popular resolution Steam, that's because most people in the world have GPUs below 970/290 and 1440P and above monitors are expensive in 3rd world/developing countries relative to their income levels.

I think for developed countries, testing $400+ GPUs in 2016 at 1080p 60 fps without DSR/VSR/SSAA or moving the testing to 120-144Hz monitor is almost meaningless now. Of course that's my opinion but I would never upgrade to modern $400+ GPUs and still be gaming on a $150-$200 1080P monitor, not a chance. Ya, I get it there are people with $500-650 GPUs and $200 1080p monitors but those gamers would benefit much more going with a monitor upgrade then. 1080P is the new 1280x1024 in the era where enthusiasts/high-end gamers have moved on to 1440p/1600p/4K. It's time to move 1080p to the lowest 1600x900 tier at TPU and re-introduce 3440x1440 or triple monitor gaming instead.

I've been saying for a while now that TPU needs to either introduce SSAA/VSR/DSR 1080p benches or specifically start benching 1080P 120-144Hz. 1080p @ 60Hz for expensive modern GPUs is like buying a Porsche 911 and putting Honda Accord Coupe tires on it.

So why is non reference 980TI 26% faster than FURYX in 1080p?
GPUs are not bottlenect in 1080 its just FURYX is that bad in 1080p.

 

[Adored]

It's the front-end of GCN holding them back at 1080p and not just on Fury X. AMD seems to be constantly struggling with unbalanced GPUs but Polaris will fix it.

 

[RussianSensation]

So why is non reference 980TI 26% faster than FURYX in 1080p?
GPUs are not bottlenect in 1080 its just FURYX is that bad in 1080p.

DX11 driver bottleneck. Also, it's 11% not 26%. You are comparing a 1.4+Ghz 980Ti to a stock Fury X? My point is even if you compare NV GPUs to each other, their true GPU raw power comes out at 1440P and 4K. I mean if it makes some people feel better that 980Ti is 26% faster than Fury X at 1080P, they can knock themselves out but at least compare apples-to-apples (stock vs. stock or OC vs. OC). If 980Ti was 30% faster than Fury X at 1080P and lost by 15% at 1440P/4K, it only matters if you play at 1080p. I don't. When I buy a $300 GPU now, I don't look at 1920x1200 and below benches. If those matter, go NV until more games are DX12. Also, I tend to keep my GPUs longer than 12 months so long-term performance, as is often gauged by a card's ability to perform better at higher resolution, matters far more to me. Even since I started gauging the card's longevity by high resolution gaming benches, it hasn't once let me down.

If you care about 1080p benches, but you also have a 970 OC, I wouldn't waste $ upgrading to 2016 cards. 970 OC ~ 980/390X OC is probably good enough for another 1.5 years at 1080p, maybe longer.

BTW, stock vs. stock

900p
980Ti beats 980 by 14%
980Ti beats 970 by 27%

1080p
980Ti beats 980 by 19%
980Ti beats 970 by 35%

1440p
980Ti beats 980 by 24%
980Ti beats 970 by 43%

4K
980Ti beats 980 by 27%
980Ti beats 970 by 47%
http://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_XtremeGaming/23.html

It's pretty clear the point I am illustrating. If someone buys a high-end card for CPU limited resolutions, well your GPU is waiting for that CPU. The benches aren't showing true GPU vs. GPU comparison since so many of them are instead showing CPU limited gaming scenarios.

The reverse of that is showing 900p CPU benchmarks of i7 6700K OC vs. i7 4790K OC but the gamer uses 1440p/4K gaming resolution. Ya, not really relevant for CPU upgrade path for the Devil's Canyon user, is it?

 

[Head1985]

well 90% people still game in 1080P and i am also playing in 1080p.I really dont care how cards are fast in 4K when FuryX have 33Fps and 980Ti have 30Fps.Its unplayble on both of them.
If poaris fix this problem in 1080P its HUGE win for AMD.
If 232mm2 2560SP 256bit DDR5 polaris 10 can match 600mm2 4096SP with HBM FIJI it only shows how bad is FIJI in 1080p.

 

[jpiniero]

Even though 1080p is the most popular resolution Steam, that's because most people in the world have GPUs below 970/290 and 1440P and above monitors are expensive in 3rd world/developing countries relative to their income levels.

On the Steam Survey, above 1920x1200 is a whopping 2.2%. Maybe that will go higher at some point but until then 1080p is going to be the main resolution.