Support NeoGAF

[Adry9]

I own a 7950, so in the end, I WIN when I upgrade.

7950 here too, can't wait for the new cards to roll.

 

[daveo42]

At $300, that would help seal the deal on a full PC build as opposed to just upgrading my current card.

 

[DieH@rd]

If this is true (I doubt it is, though), it will be the card for VR.

Remember, the same Polaris 10 GPU chip [although lower clocked, and with 36 active CUs] is inside PS4K APU. This summer we will get tons of Polaris benchmarks, and when they come, we will know for sure what kind of boost will PS4K get over PS4.

Weeeeeell... If P10 is Hawaii on 14nm with 256 bit GDDR5X then technically it may somewhat reach Fury's performance if it'll be clocked high and the architectural optimizations will give it some 10-20% on average. Fury is some +10-15% of performance to 390X, not a gap that can't be closed by a better architecture with higher clocks.

Reaching 980Ti is a different matter however, especially if we're assuming that it's factory OC 980Ti cards which it needs to reach.

What I don't buy is the $300 price though. If it'll beat Fiji and they'll price it on $300 then what will they sell between $300 and $1500?

We know already that Polaris is not just "old GPU tech brought down to 14nm". AMD changed a bunch of things for this "GCN 4th gen" architecture, optimizing it even more for game rendering.

I own a 7950, so in the end, I WIN when I upgrade.

For more than half a year my PC is borked. When start up any game system freezes within minutes and I get all kinds of TDR timeouts all the time. When Polaris gets out, I will upgrade everything, and then... the true winner will be me.

 

[dgrdsv]

We know already that Polaris is not just "old GPU tech brought down to 14nm". AMD changed a bunch of things for this "GCN 4th gen" architecture, optimizing it even more for game rendering.

We don't know anything. These yellow splashes there don't provide any amount of information.

 

[DieH@rd]

We don't know anything. These yellow splashes there don't provide any amount of information.

They provide information that AMD has changed things. We are not getting any of the GCN architectures like on 28nm.

 

[chaosblade]

I'd definitely be in for that price and performance. Seriously doubt it's legit though.

 

[dgrdsv]

They provide information that AMD has changed things. We are not getting any of the GCN architectures like on 28nm.

AMD has been changing things on 28nm all the time - we've got three GCN revisions on 28nm. There's hope that Polaris will be a bigger change than what we've got between Tahiti and Fiji but that's just hope for now, with no solid reason to expect this.

 

[SolidSnakeUS]

AMD Polaris 10 GPU To Offer Near 980 Ti Performance For 299 USD?
Granted we've heard this before from them but it would cool if this will turn out to be true.

If it's true, that would be HUGE for AMD. I'd grab one over my GTX 970.

 

[Marlenus]

Weeeeeell... If P10 is Hawaii on 14nm with 256 bit GDDR5X then technically it may somewhat reach Fury's performance if it'll be clocked high and the architectural optimizations will give it some 10-20% on average. Fury is some +10-15% of performance to 390X, not a gap that can't be closed by a better architecture with higher clocks.

Reaching 980Ti is a different matter however, especially if we're assuming that it's factory OC 980Ti cards which it needs to reach.

What I don't buy is the $300 price though. If it'll beat Fiji and they'll price it on $300 then what will they sell between $300 and $1500?

We could also do the calculation based on perf/watt.

If 2.5x is true then a 150 watt P10 with 8GB GDDR5 would have similar performance to 375watt Hawaii with 8GB GDDR5. That is in excess of 390X performance and probably on par with Fury X so very near 980Ti performance.

As far as $299 is concerned it does not seem too far fetched. Fury Nano is around $500 and that is a bigger die and uses more expensive HBM. A smaller die with cheaper ram could easily be doable at $299.

This feels very similar to the 4xxx series launch. I remember the disbelief when the 800SP rumours dropped and then even more when the prices were leaked. It was such a bombshell that Nvidia dropped the 280 prices by $150 overnight.

Still I will wait and see. That level of performance seems feasible as does the price but feasible is very far from true.

 

[DjRalford]

I've really let me PC go for a while now, not upgraded for a couple of years, since i got my release date 4790K actually.

Not played much lately and still running on my old HD7970, so when the next big thing hits it's upgrade time methinks.

 

[kami_sama]

7950 here too, can't wait for the new cards to roll.

Count me in for the 7950 club! Great card it has been those three years and while some games do not run as fluid as they should, I can still play most new AAA games at high.

Remember, the same Polaris 10 GPU chip [although lower clocked, and with 36 active CUs] is inside PS4K APU. This summer we will get tons of Polaris benchmarks, and when they come, we will know for sure what kind of boost will PS4K get over PS4.



We know already that Polaris is not just "old GPU tech brought down to 14nm". AMD changed a bunch of things for this "GCN 4th gen" architecture, optimizing it even more for game rendering.

For more than half a year my PC is borked. When start up any game system freezes within minutes and I get all kinds of TDR timeouts all the time. When Polaris gets out, I will upgrade everything, and then... the true winner will be me.

I hope we get those benchmarks as soon as possible.

 

[Nachtmaer]

We could also do the calculation based on perf/watt.

If 2.5x is true then a 150 watt P10 with 8GB GDDR5 would have similar performance to 375watt Hawaii with 8GB GDDR5. That is in excess of 390X performance and probably on par with Fury X so very near 980Ti performance.

As far as $299 is concerned it does not seem too far fetched. Fury Nano is around $500 and that is a bigger die and uses more expensive HBM. A smaller die with cheaper ram could easily be doable at $299.

This feels very similar to the 4xxx series launch. I remember the disbelief when the 800SP rumours dropped and then even more when the prices were leaked. It was such a bombshell that Nvidia dropped the 280 prices by $150 overnight.

Still I will wait and see. That level of performance seems feasible as does the price but feasible is very far from true.

Not to discredit everything you've just said, but there's one problem with that 2.5x perf/W figure. We don't know what it applies to since all of their chips are different. Use it on let's say Tonga and you're going to get very different results than if you were to start with Fiji. There can even be a big difference between SKUs based on the same chip, like Nano vs Fury X.

I'm still not buying this rumor of P10 being close a 980 Ti yet. My conservative guess still is somewhere inbetween 390X and Fury, maybe Fury X since the gap between Hawaii and Fiji isn't THAT big. Perhaps in some edge cases where GCN4 shines it might come close to a reference 980 Ti.

And yeah, the fallout the 4000 series brought was beautiful.

 

[icecold1983]

We don't know anything. These yellow splashes there don't provide any amount of information.

it will at least be bigger than the changes 1.1 and 1.2 brought. Raja has stated its the biggest change to gcn yet.

 

[wachie]

I would temper expectations, 4000 series is maybe once in a decade type thing.

 

[RenegadeXV8]

This feels very similar to the 4xxx series launch. I remember the disbelief when the 800SP rumours dropped and then even more when the prices were leaked. It was such a bombshell that Nvidia dropped the 280 prices by $150 overnight.

That was legendary. I still distinctly remember forum posters in hardware sites analyzing the die space and concluding "THERE IS NO POSSIBLE WAY 800SP FIT IN THERE", adamant that it would be 480 at best.

 

[DieH@rd]

This feels very similar to the 4xxx series launch. I remember the disbelief when the 800SP rumours dropped and then even more when the prices were leaked. It was such a bombshell that Nvidia dropped the 280 prices by $150 overnight.

Haha, that was really awesome time. I remember it really shaking up the GPU market, Nvidia reacted immediately like mad and several of my friends purchased 4850/4870 cards.

I would temper expectations, 4000 series is maybe once in a decade type thing.

Radeon HD 4000 series - June 25, 2008
Radeon Polaris series - summer of 2016

 

[dgrdsv]

We could also do the calculation based on perf/watt.

If 2.5x is true then a 150 watt P10 with 8GB GDDR5 would have similar performance to 375watt Hawaii with 8GB GDDR5. That is in excess of 390X performance and probably on par with Fury X so very near 980Ti performance.

As far as $299 is concerned it does not seem too far fetched. Fury Nano is around $500 and that is a bigger die and uses more expensive HBM. A smaller die with cheaper ram could easily be doable at $299.

This feels very similar to the 4xxx series launch. I remember the disbelief when the 800SP rumours dropped and then even more when the prices were leaked. It was such a bombshell that Nvidia dropped the 280 prices by $150 overnight.

Still I will wait and see. That level of performance seems feasible as does the price but feasible is very far from true.

That's not how perf/watt works usually as the figure provided is for the chip of the same performance, you can't assume that it means that you'll get 2.5x more performance in the same power draw envelope. And the 2.5x figure itself is kinda misleading as they've been switching between 2x and 2.5x and it's not really clear if they're talking about desktop P10 chips or notebook P11 - which may obviously have a better perf/watt due to lower clocking and better sorting.

The best basis for any kind of estimation we have now is the 232mm^2 die size which should put that chip somewhere around Hawaii's complexity. And as I've said, with some enhancements and better clocks a chip of Hawaii complexity could reach Fury's performance levels.

4xxx series was a big architectural change. I'm not expecting anything like that from Polaris.

 

[icecold1983]

That's not how perf/watt works usually as the figure provided is for the chip of the same performance, you can't assume that it means that you'll get 2.5x more performance in the same power draw envelope. And the 2.5x figure itself is kinda misleading as they've been switching between 2x and 2.5x and it's not really clear if they're talking about desktop P10 chips or notebook P11 - which may obviously have a better perf/watt due to lower clocking and better sorting.

The best basis for any kind of estimation we have now is the 232mm^2 die size which should put that chip somewhere around Hawaii's complexity. And as I've said, with some enhancements and better clocks a chip of Hawaii complexity could reach Fury's performance levels.

4xxx series was a big architectural change. I'm not expecting anything like that from Polaris.

amds not going to sell 980ti performance for 299. its bad business when nvidias comparable performance chip will probably cost at least 200 more.

edit - oops, this was in response to the person you quoted

 

[Inspectah_Deck]

Roy Taylor:
“If you look at the total install base of a Radeon 290, or a GTX 970 or above [the minimum specs required for VR], it’s around 7.5 million units, but the issue is that if a publisher wants to sell a £40/$50 VR game, there’s not a big enough market to justify that yet. We’ve got to prime the pumps, which means somebody has got to start writing cheques to big games publishers. Or we’ve got to increase the install TAM [total addressable market].”

“The reason Polaris is a big deal, is because I believe we will be able to grow that TAM significantly. I don’t think Nvidia is going to do anything to increase the TAM, because according to everything we’ve seen around Pascal, it’s a high-end part. I don’t know what the price is gonna be, but let’s say it’s as low as £500/$600 and as high as £800/$1000. That price range is not going to expand the TAM for VR. We’re going on the record right now to say Polaris will expand the TAM. Full stop.”

http://videocardz.com/59445/amd-polaris-aiming-at-vr-capable-graphics-cards

So it's confirmed, what we speculated on the last pages.
Polaris will not compete with Nvidia in high-end.

 

[chaosblade]

Roy Taylor:
“If you look at the total install base of a Radeon 290, or a GTX 970 or above [the minimum specs required for VR], it’s around 7.5 million units, but the issue is that if a publisher wants to sell a £40/$50 VR game, there’s not a big enough market to justify that yet. We’ve got to prime the pumps, which means somebody has got to start writing cheques to big games publishers. Or we’ve got to increase the install TAM [total addressable market].”

“The reason Polaris is a big deal, is because I believe we will be able to grow that TAM significantly. I don’t think Nvidia is going to do anything to increase the TAM, because according to everything we’ve seen around Pascal, it’s a high-end part. I don’t know what the price is gonna be, but let’s say it’s as low as £500/$600 and as high as £800/$1000. That price range is not going to expand the TAM for VR. We’re going on the record right now to say Polaris will expand the TAM. Full stop.”

http://videocardz.com/59445/amd-polaris-aiming-at-vr-capable-graphics-cards

So it's confirmed, what we speculated on the last pages.
Polaris will not compete with Nvidia in high-end.

Didn't we already know this? Vega is the big chip that will be powering high end GPUs.

 

[DieH@rd]

Roy Taylor:

So it's confirmed, what we speculated on the last pages.
Polaris will not compete with Nvidia in high-end.

We knew this from the first moment chip sizes started being rumored.

This summer AMD will launch low [~50W for 1080p gaming] and mid [100-120W for 1440p60 gaming] cards, while Nvidia will immediately start with stronger/bigger chip [~350mm2].

AMDs high end will come out around new year.

 

[Inspectah_Deck]

Yes we knew this (as I wrote), but not officially.

 

[Gemüsepizza]

So Vega will compete with a GP100 (HBM2) variant? Welp, good luck with that AMD.

Edit: Seems like a really strange strategy to me. Nvidia will be able to compete with AMD in the mainstream area, because they have products in this range. They just have to price them accordingly. But now AMD can't compete with nvidia in the high-end area, because nvidia will have much faster cards. Also lol at their price predictions. $600 for the cheapest Pascal card? Does he really believe this?

 

[Inspectah_Deck]

The source article is btw a nice read:
http://arstechnica.co.uk/gadgets/2016/04/amd-focusing-on-vr-mid-range-polaris/

 

[THE:MILKMAN]

We knew this from the first moment chip sizes started being rumored.

This summer AMD will launch low [~50W for 1080p gaming] and mid [100-120W for 1440p60 gaming] cards, while Nvidia will immediately start with stronger/bigger chip [~350mm2].

AMDs high end will come out around new year.

I assume both are trying to avoid competing directly but you would think Nvidia especially would want a mainstream card seeing as that is where all those 960/970 sales are.

If the speculation is that the full 40CU Polaris 10 will be in the region of $350 as a 232mm2 chip I can't see how Nvidia could compete with a bigger/more expensive chip (I read somewhere it was measured at 316mm2?) unless they take a big hit on price?

I love this though. I just built a new PC sans GPU and would love to get more bang for my buck than I thought.

 

[FreezerGeezer]

I imagine this has been brought up;

https://youtu.be/aSYBO1BrB1I

Had it shown to me. Now I do not understand a lot of what he is saying in the technical sense, but it does sound to be a fairly solid argument towards his case.

Or is this something that we will not notice until Navi is detailed?

Note: Just ignore if this is too OT.

 

[VariantX]

I imagine this has been brought up;

https://youtu.be/aSYBO1BrB1I

Had it shown to me. Now I do not understand a lot of what he is saying in the technical sense, but it does sound to be a fairly solid argument towards his case.

Or is this something that we will not notice until Navi is detailed?

Note: Just ignore if this is too OT.

That had its own thread a few days ago.

 

[FreezerGeezer]

Oh fantastic. I did have a look, just apparently a sub par one!

I'll have to have another search and take it there, thanks!

 

[Frankfurter]

So Vega will compete with a GP100 (HBM2) variant? Welp, good luck with that AMD.

Edit: Seems like a really strange strategy to me. Nvidia will be able to compete with AMD in the mainstream area, because they have products in this range. They just have to price them accordingly. But now AMD can't compete with nvidia in the high-end area, because nvidia will have much faster cards. Also lol at their price predictions. $600 for the cheapest Pascal card? Does he really believe this?

AMD and Nvidia both will have their respective product range on their own at launch, i.e. Nvidia won't have a Polaris 10 (or 11) equivalent and AMD won't have a GP104 equivalent. The only more or less direct competition might be the fastest Polaris 10 GPU vs. the slowest GP104 GPU.
Nvidia will apparently release their own chip in the ~250mm² range earlier (~Oct.?) than AMD (1.Q 2017), but that's only gonna give them a couple of months.

 

[dgrdsv]

We knew this from the first moment chip sizes started being rumored.

This summer AMD will launch low [~50W for 1080p gaming] and mid [100-120W for 1440p60 gaming] cards, while Nvidia will immediately start with stronger/bigger chip [~350mm2].

AMDs high end will come out around new year.

I wouldn't read too much into what AMD's CVP is saying about NV's lineup. I mean, he's saying that GP104 cards will cost between $600 and $1000? I find it very unlikely.

They also seem to presume for some reason that GP106 which should be a direct P11/P10 competition won't be launched this year while latest rumors are saying that GP106 cards should come out around Aug/Sep.

 

[Mokujin]

So Vega will compete with a GP100 (HBM2) variant? Welp, good luck with that AMD.

Edit: Seems like a really strange strategy to me. Nvidia will be able to compete with AMD in the mainstream area, because they have products in this range. They just have to price them accordingly. But now AMD can't compete with nvidia in the high-end area, because nvidia will have much faster cards. Also lol at their price predictions. $600 for the cheapest Pascal card? Does he really believe this?

Do you have benchmarks of future products already?

 

[Durante]

Do you have benchmarks of future products already?

No, but if the rumoured die sizes for the initial products of each vendor are at all accurate, it would require a colossal blunder for NV not to dominate the high-end.

(The last time they made such a colossal blunder was with the FX series 13 years ago)

 

[spyshagg]

The source article is btw a nice read:
http://arstechnica.co.uk/gadgets/2016/04/amd-focusing-on-vr-mid-range-polaris/

some mistakes but a good read indeed

 

[Wollan]

"If you look at the total install base of a Radeon 290, or a GTX 970 or above [the minimum specs required for VR], it's around 7.5 million units,"

High-end PC market is smaller than I thought to be honest.

 

[Self_Destructive]

No, but if the rumoured die sizes for the initial products of each vendor are at all accurate, it would require a colossal blunder for NV not to dominate the high-end.

(The last time they made such a colossal blunder was with the FX series 13 years ago)

And seeing as Pascal is more of an evolution of Maxwell than a new architecture, I'd say the chances of that are slim.

 

[Turin Turambar]

No, but if the rumoured die sizes for the initial products of each vendor are at all accurate, it would require a colossal blunder for NV not to dominate the high-end.

(The last time they made such a colossal blunder was with the FX series 13 years ago)

It depends. Does releasing a big die, high end gpu at 1000 bucks count as "dominate"? Pretty much in the eye of the beholder.

 

[Bolivar687]

It depends. Does releasing a big die, high end gpu at 1000 bucks count as "dominate"? Pretty much in the eye of the beholder.

I have to think at least one GP104 will launch at the upper mid range slot of the 970/390. Theyre not going to have three GPUs above $550.

RTG needs something either significantly cheaper or definitively more powerful at that tier if they want any chance of clawing back market share.

 

[chaosblade]

It depends. Does releasing a big die, high end gpu at 1000 bucks count as "dominate"? Pretty much in the eye of the beholder.

That is what some people like to look at (X has the performance crown, regardless of price), but it isn't really the case here. Nvidia is just aiming at the bracket above AMD, but still not 500mm^2 behemoths.

 

[SneakyStephan]

So Vega will compete with a GP100 (HBM2) variant? Welp, good luck with that AMD.

Edit: Seems like a really strange strategy to me. Nvidia will be able to compete with AMD in the mainstream area, because they have products in this range. They just have to price them accordingly. But now AMD can't compete with nvidia in the high-end area, because nvidia will have much faster cards. Also lol at their price predictions. $600 for the cheapest Pascal card? Does he really believe this?

If vega is meant to compete with gp100 then there will be a massive gap in amd's lineup between polaris 10 and vega

It's pretty weird tbh that polaris10 is so small, maybe they'll have some heavily cut version of vega that competes in the midrange with gp10.

I'm hoping they'll deliver beyond our expectations with polaris because it's really bumming me out, if nvidia end up having the performance segment all to themselves then prices are going to suck horribly again.

It almost feels like a repeat of 2012 if this happens (well even worse, as right now it seems likely that gp104 might be as far ahead of polaris10 than gk110 was of tahiti, and it launches at the same time instead of 6 months later), when amd didn't have anything decent performance wise for ages and prices shot up massively. It took like 2 years for prices to go back down

 

[dgrdsv]

It depends. Does releasing a big die, high end gpu at 1000 bucks count as "dominate"? Pretty much in the eye of the beholder.

Yes, because we never had any cards for less than $1000 bucks on the same GPU.

If vega is meant to compete with gp100 then there will be a massive gap in amd's lineup between polaris 10 and vega

It's been confirmed already by AMD that there will be two Vega GPUs, just like there will be two Polarises - Vega 10 and Vega 11.

 

[MuchoMalo]

If vega is meant to compete with gp100 then there will be a massive gap in amd's lineup between polaris 10 and vega

It's pretty weird tbh that polaris10 is so small, maybe they'll have some heavily cut version of vega that competes in the midrange with gp10.

I'm hoping they'll deliver beyond our expectations with polaris because it's really bumming me out, if nvidia end up having the performance segment all to themselves then prices are going to suck horribly again.

It almost feels like a repeat of 2012 if this happens (well even worse, as right now it seems likely that gp104 might be as far ahead of polaris10 than gk110 was of tahiti), when amd didn't have anything decent performance wise for ages and prices shot up massively. It took like 2 years for prices to go back down

Or there's a Vega 11 releasing next year with HBM2 and AMD will have a lead. Getting disappointed before seeing benchmarks is not a good idea. You seem to have extremely high expectations of GP104.

 

[SneakyStephan]

Or there's a Vega 11 releasing next year with HBM2 and AMD will have a lead. Getting disappointed before seeing benchmarks is not a good idea. You seem to have extremely high expectations of GP104.

expecting a 330mm² die to significantly outperform a 230mm² die is outrageous, I know.

I'm totally out there with my expectations

 

[thelastword]

High-end PC market is smaller than I thought to be honest.

Yeah, thought this would be much higher too, especially if it's 970 and 290's combined installbase. In any case, this number only indicates the installbase for entry level GPU's for VR, the combined installbase of fury+980ti won't add too much to that tally, but perhaps take it past 10 million if we're being positive? I agree, still on the low end though. (for an installbase)

 

[callsignkrogan]

I imagine this has been brought up;

https://youtu.be/aSYBO1BrB1I

Had it shown to me. Now I do not understand a lot of what he is saying in the technical sense, but it does sound to be a fairly solid argument towards his case.

Or is this something that we will not notice until Navi is detailed?

Note: Just ignore if this is too OT.

wow that was a well made video. Definitely makes you think. I always wondered if we would see improvements on AMD's pc cards from the console market.

 

[dgrdsv]

AMD Radeon R9 M480 based on Polaris 11 GPU
Considering that the mobile parts naming is usually one step above the desktop AMD's lineup that likely means that P11 will be used for R9 470 cards on the destop.

 

[Irobot82]

AMD Radeon R9 M480 based on Polaris 11 GPU
Considering that the mobile parts naming is usually one step above the desktop AMD's lineup that likely means that P11 will be used for R9 470 cards on the destop.

We may very well end up seeing AMD hit the mainstream market to improve "TAM" like Roy said and Nvidia hitting enthusiast market.

So what if the 480X only costs $300. Let's assume for a moment that it can reach about Fury X levels.

Let's then assume the 1070 is $400 and 1080 is $500. What if these chips are +100 and at $500 and $600.

Who comes out on top? This is all pretty crazy.

 

[Bolivar687]

We may very well end up seeing AMD hit the mainstream market to improve "TAM" like Roy said and Nvidia hitting enthusiast market.

So what if the 480X only costs $300. Let's assume for a moment that it can reach about Fury X levels.

Let's then assume the 1070 is $400 and 1080 is $500. What if these chips are +100 and at $500 and $600.

Who comes out on top? This is all pretty crazy.

Depends on the 1070. There's no basis for pricing right now but if we assume both are FuryX/980Ti performance, then I would think $300, 125W 480X becomes the editor's choice in "Best Graphics Cards 2016" articles and AMD claws back some marketshare, no? Although not enough to dissuade a significant amount of Nvidia users to switch brands (and a few Gaffers will work very hard to assure that doesn't happen). It was always going to come down to the 1070 and full Polaris 10 either way. Iirc the 970 and 290 sold more than the rest of their respective lineup, possibly even combined.

There's some rumors AMD changes their naming scheme, which would make sense. The x80 has traditionally been the $250 and under tier. A $300 Fury X sounds more like an R9 490.

 

[Kezen]

(and a few Gaffers will work very hard to assure that doesn't happen)

Right, AMD are being plotted against I'm sure.

 

[Bolivar687]

Right, AMD are being plotted against I'm sure.

Its a joke, don't get so defensive!

 

[IMACOMPUTA]

Right, AMD are being plotted against I'm sure.

I definitely prefer Nvidia these last couple years, but don't act like people aren't fanboying.

Do we know the die size for Vega?