Polaris 10 benchmarks...

[AtenRa]

I asked this in the other thread before it got locked but do you think polaris pricing could effect GTX 1070 and 1060 pricing?

AMD Polaris already made NVIDIA lower their GTX1070 to sub $400, otherwise NVIDIA would sell GTX1070 closer to GTX 1080.

 

[ShintaiDK]

AMD Polaris already made NVIDIA lower their GTX1070 to sub $400, otherwise NVIDIA would sell GTX1070 closer to GTX 1080.

Nothing points to that.

Polaris sells for what it does because its performance and perf/watt is what it is. Its a GP106 competitor at best. Far from GP104.

 

[Qwertilot]

Definitely. Look, NV are incredibly predictable when there's no competition interfering. ~35% up/annum on the top card, performance dropping a commensurate amount down the pricing tiers on more of a two year cycle.

It doesn't seem that likely that the 1060 is going to be badly affected by P10 price wise. The 960 launched at $199 too, and NV can typically hold a slight premium vs AMD without pain. Straight up competition seems likely.

 

[Riek]

Nothing points to that.

Polaris sells for what it does because its performance and perf/watt is what it is. Its a GP106 competitor at best. Far from GP104.

From what we know now the 1070 seems to be less than 15% faster (3dmark). Not so sure if that is 'far away' to warant a >50% price uplift

Assuming both use exactly 150W.. The perf/wat advantage is not really big if we assume the metrics from the 'leaked' 3dmark. So saying polaris sells for that due to perf/watt??
it sells for that because it was designed for that price slot. (perf/watt is a bogus metric for this).

 

[exar333]

AMD Polaris already made NVIDIA lower their GTX1070 to sub $400, otherwise NVIDIA would sell GTX1070 closer to GTX 1080.

That could be true (not sure about that) but I wouldn't be surprised if we see the 1070 get price cut at or around when Vega is released to maybe $329 or so, similar to where the 970 sat for a long time.

Unless AMD releases a 'full p10' that is close to $300 and gets close to 1070 performance, I don't see that price changing much until Vega.

 

[Qwertilot]

That's quite likely, if perhaps as likely to be triggered by big Pascal appearing. The 1080 would logically come down somewhat in price at the same time.

Happened to the 6xx stuff, never to the 9xx's but then hopefully Vega will be a better response to the 10xx stuff than Fury was.

 

[provost]

Yes, NV has been predictable (30% -35% performance, regardless of arch or node changes, and optimizations riding off into the sunset at the outset of subsequent newer releases) and that fact hasn't been lost on AMD either, thus, AMD's pricing surprise for Nvidia...lol

For the potential addressable market, "$199 for 980 performance" has a much nicer ring to it than "$699 for TitanX performance"....lol

 

[seitur]

290/290x/970 performance has been around for almost 2 years now, and you could snag those deals for close to $250. This being $200, with some updated connections (which 4K and HDR likely won't mater much in this performance class and/price range anyway) just isn't compelling.

Yeah, but like RussianSensation said, this is not true outside of USA.

Prices in rest of the world are diffrent, including second hand used market and this is result of more than just added value tax.

So imho if 480 will have at least moderately good characteristics (noise, OC, availability, etc) then it will sell nicely.

Remmeber that most people are still on 1080p and while 1440 and 4k monitors are coming down in price heavily, with economic uncertainity 199-230$ 480 will be really good offering and will allow many gamers to finally experience smooth high fps 1080p AAA gameplay on high/ultra settings.

 

[tviceman]

AMD Polaris already made NVIDIA lower their GTX1070 to sub $400, otherwise NVIDIA would sell GTX1070 closer to GTX 1080.

No one, not you not I, can prove any of this conjecture but I believe Nvidia was (is) aware of P10's capabilities before 1080's launch and is the reason why GTX 1070 is as cut down as it is. What Nvidia didn't know, and couldn't have predicted, is AMD's rock bottom price out of the gate.

Anyone with GTX 770 or HD 7950 performance (or less) wanting to upgrade is going to have a hard time considering GTX 1070 at $379-399 when $219-239 will get 80% of that performance. The big wild card is how GP106 will end up in price and performance as well as its release date. AMD incited a price war out of the gate and if Nvidia decides to play ball I expect Nvidia to paper launch GP106 within a week or two (assuming a launch no later than early August) of RX 480's release.

 

[Vesku]

Quite on the nose regarding this pricing for those with first gen 28nm and older. Thought I'd ride my 7950 until large die Finfet but at $199 I might as well get a 480 to play with. Interested to see how it OCs.

 

[exar333]

Yeah, but like RussianSensation said, this is not true outside of USA.

Prices in rest of the world are diffrent, including second hand used market and this is result of more than just added value tax.

So imho if 480 will have at least moderately good characteristics (noise, OC, availability, etc) then it will sell nicely.

Remmeber that most people are still on 1080p and while 1440 and 4k monitors are coming down in price heavily, with economic uncertainity 199-230$ 480 will be really good offering and will allow many gamers to finally experience smooth high fps 1080p AAA gameplay on high/ultra settings.

I can get on board with that. As I have said, for a user here in the US and as an enthusiast, its not 'exciting' but IS a fantastic product.

I don't like including VAT in the pricing discussion because (not to get off topic) that cost is in place for everyone. It just may not be 'VAT', but those services are paid for somehow. I do recognize though that international 'base' prices can be very different. Outside of exchange rate issues, I cannot claim to understand why those may be there.

 

[tviceman]

Quite on the nose regarding this pricing for those with first gen 28nm and older. Thought I'd ride my 7950 until large die Finfet but at $199 I might as well get a 480 to play with. Interested to see how it OCs.

I sold my entire system a month ago with plans to rebuild this December. I was dead set on getting a used GTX 970/980 or 390/390X until AMD announced its RX 480 pricing. 480's pricing is going to murder after-market sales worse than the GTX 1070.

 

[antihelten]

Anyone with GTX 770 or HD 7950 performance (or less) wanting to upgrade is going to have a hard time considering GTX 1070 at $379-399 when $219-239 will get 80% of that performance. The big wild card is how GP106 will end up in price and performance as well as its release date. AMD incited a price war out of the gate and if Nvidia decides to play ball i expect Nvidia to paper launch GP106 within a week or two of RX 480's release.

If GP106 is to GP104 as GM206 was to GM204, then GP106 won't stand a chance against RX 480 (based on the current vague performance leaks for RX 480, i.e 80% of a 1070 and 65% of a 1080).

Rumours currently indicate that GP106 will basically be half a GP104 (2 GPCs instead of 4), which is exactly the same as GM206 to GM204. The top GM206 (GTX 960), was basically half as fast as the top GM204 (GTX 980).

If the top GP106 (1060) is half as fast as a 1080, then it would be about 25% slower than RX 480 and about 30% slower than a theoretical RX 480X.

As far as I can tell, the only real option Nvidia has against Polaris 10, is to make a third SKU off of the GP104 die. In other words a 1060 Ti basically.

 

[tviceman]

Rumours currently indicate that GP106 will basically be half a GP104 (2 GPCs instead of 4), which is exactly the same as GM206 to GM204. The top GM206 (GTX 960), was basically half as fast as the top GM204 (GTX 980).

If the top GP106 (1060) is half as fast as a 1080, then it would be about 25% slower than RX 480 and about 30% slower than a theoretical RX 480X.

You and I are reading different rumors then. Die shots have made the rounds with a 200mm2 estimate, putting GP106 firmly above the 1/2 GP104 rumor you are talking about.

 

[antihelten]

You and I are reading different rumors then. Die shots have made the rounds with a 200mm2 estimate, putting GP106 firmly above the 1/2 GP104 rumor you are talking about.

The die shots I have seen estimate the GP106 at 170-200mm2

GM206 was 57% the size of GM204 (227mm2 versus 398mm2). 57% of the GP104 die would be 180mm2, which is right in line with the aforementioned die shot.

 

[exar333]

The die shots I have seen estimate the GP106 at 170-200mm2

GM206 was 57% the size of GM204 (227mm2 versus 398mm2). 57% of the GP104 die would be 180mm2, which is right in line with the aforementioned die shot.

And coincidentally that ~57% is just about where the GX480 sits in terms of performance, based on the data from AMD's Computex CF FPS numbers. We may have a close one here...

 

[showb1z]

1070 is already 3/4 of GP104. So GP106 being half of GP104 is the only thing that makes sense no?

 

[tviceman]

The die shots I have seen estimate the GP106 at 170-200mm2

GM206 was 57% the size of GM204 (227mm2 versus 398mm2). 57% of the GP104 die would be 180mm2, which is right in line with the aforementioned die shot.

Looks legit, thanks for that link.

 

[antihelten]

And coincidentally that ~57% is just about where the GX580 sits in terms of performance, based on the data from AMD's Computex CF FPS numbers. We may have a close one here...

True, but it's worth noting that even though GM206 is 57% the size of GM204 it is only about 52% of the performance

So RX 480 would be about 10-15% faster, and a fully unlocked RX 480X (if such a thing exists) would be about 25% faster.

So not necessarily that close of a race (of course the aforementioned 480X would cost significantly more than $200).

1070 is already 3/4 of GP104. So GP106 being half of GP104 is the only thing that makes sense no?

This is of course far and away the best argument against the GP106 being 3 GPCs. Which would then mean that 2 GPCs is really the only option (GP107 then being 1 GPC)

 

[exar333]

True, but it's worth noting that even though GM206 is 57% the size of GM204 it is only about 52% of the performance

So RX 480 would be about 10-15% faster, and a fully unlocked RX 480X (if such a thing exists) would be about 25% faster.

So not necessarily that close of a race (of course the aforementioned 480X, would cost significantly more than $200).



This is of course far and away the best argument against the GP106 being 3 GPCs. Which would then mean that 2 GPCs is really the only option (GP107 then being 1 GPC)

Well said.

NV has gotten away with the last few gens on the x60 SKU being not as desirable with per/$ compared to AMD choices. It will be interesting to see if they stick to that, or get more aggressive with pricing this time around. 1/2 1080 performance would be rather limiting, unless it clocks insanely high.

 

[Qwertilot]

I wouldn't say the 960 got away with it, more that it was very well targeted for its intended audience vs the competition.

If they've still got a non trivial power efficiency advantage this time, they might well price it in again. Not enormously, but it is a definite plus for a decent chunk of that bit of the market.

One thing is for sure - with nothing above the 480 there is very much more motivation for AMD/AIB's to really push the performance of that chip in ways that no one is likely to try with GP106.

 

[therealnickdanger]

The success of the 960 was a stroke of marketing wizardry on Nvidia's part. I think the main difference between the 960 and 480 from a marketing perspective is that MOBAs (targeted by the 960) were and still are extremely popular and command a massive audience. VR, on the other hand, is still a very small niche. It's great that we can soon buy a VR-ready GPU for $199, but the barrier of entry is still $600-800 for a headset and controllers. It's not the F2P target audience the 960 had.

 

[tential]

The success of the 960 was a stroke of marketing wizardry on Nvidia's part. I think the main difference between the 960 and 480 from a marketing perspective is that MOBAs (targeted by the 960) were and still are extremely popular and command a massive audience. VR, on the other hand, is still a very small niche. It's great that we can soon buy a VR-ready GPU for $199, but the barrier of entry is still $600-800 for a headset and controllers. It's not the F2P target audience the 960 had.

See how I would have done is show the GTX 960 vs the RX 480 and show the generational leap in performance AMD has brought at the $200 pricepoint.

A lot of people know people with GTX 950s and 960s. Those are the most accessible Nvidia cards. Comparing to these cards and showing just how much performance has moved would be beneficial.

 

[lilltesaito]

See how I would have done is show the GTX 960 vs the RX 480 and show the generational leap in performance AMD has brought at the $200 pricepoint.

A lot of people know people with GTX 950s and 960s. Those are the most accessible Nvidia cards. Comparing to these cards and showing just how much performance has moved would be beneficial.

I agree with you, that would have been a smart move. I think people in here would of given AMD a harder time comparing it to an "Old Model". Thing is AMD can not win right now, no matter what they do. People are going to say something bad no matter how they do something, sure they could of showed the card better then they did this time around(that xfire slide just confused a lot of people).

 

[tential]

I agree with you, that would have been a smart move. I think people in here would of given AMD a harder time comparing it to an "Old Model". Thing is AMD can not win right now, no matter what they do. People are going to say something bad no matter how they do something, sure they could of showed the card better then they did this time around(that xfire slide just confused a lot of people).

I don't think it would have mattered.

I think no matter the criticism, people ALWAYS want to 1 up their friends.

People who saw that massive performance boost at the $200 level, and didn't upgrade to a GTX 960 last gen, would be far more inclined to do so at the $200 level.

I think AMD showing these confusing benches just turns off casual gamers who don't want to get an answer about GPU utilization/efficiency/frame limiters/etc. when seeing some charts.

Comparing to AMD's current offerings would also be useful, but to be honest, most peopl edon't own them or know people who do. They should focus on showing the benefits of a $200 GPU vs the competition, and the benefits of the freesync + GPU combo coming in FAR cheaper than an Nvidia Gsync + GPU combo. Although all Nvidia Gsync monitors are 144hz.