[Ars] AMD confirms high-end Polaris GPU will be released in 2016

[railven]

AMD is going to most likely sell every chip they can. They'd have no competition from Nvidia. Without a performance metric, I can't really say how it will be, but if you're telling me they're going to offer a premium part at discount prices, you haven't been listening to AMD.

AMD wants to increase profits per unit sold. More so if they have a leg up on the market. You launch it high, get the early buyers, make bank, when the competitor comes out, you drop price since you've already made some bank, and thus put extra pressure on your competitor.

If AMD is going to go for volume versus margins, this is why they are facing bankruptcy.

 

[railven]

It absolutely blows my mind that there are people on these forums advocating high pricing. I don't think I've ever read anything more ridiculously stupid in my life. Literally, "Take my money" talk. Unfreakinbelievable!

Here you got an AMD fan basically telling AMD to repeat what basically led them to their current state.

Don't use your advantage, squander it. You're fans will love you for it! Oh, but they won't buy your next premium product well because it's too expensive. Sorry AMD.

 

[Glo.]

Why do you think the new uarch will be much more dense?

I do not think it will be MUCH more dense. I think it will be "optimized" in this manner. But enough to allow bigger amounts of cores in smaller packages.

 

[DooKey]

Here you got an AMD fan basically telling AMD to repeat what basically led them to their current state.

Don't use your advantage, squander it. You're fans will love you for it! Oh, but they won't buy your next premium product well because it's too expensive. Sorry AMD.

Yep, pretty much this in a nut shell. Some of the loudest squawkers on price only spend about $200-$250 on a video card and expect AMD to give them something top of the line for that amount. Unfortunately AMD has screwed the pooch and have had too many fire sales and now this vocal minority thinks those are somehow normal prices. Unless the supposed AMD fans change their thought process they aren't going to have AMD around any more.

 

[AtenRa]

I wouldn't take AtenRa's post as fact; it's opinion the same as anyone else here. For one, his scaling assumption is removing two 128-bit memory controllers from Tonga to save 66mm². Even outside the fact that it's questionable you would save that much die area, Tonga is already slightly gimped vs the 280X with its 256-bit bus. A 128-bit part with 2048 shaders would be ridiculous. Even then, looking at the Tonga die removing 256-bits worth of MC likely wouldn't even get you to 66mm² saved.

yea after a better look, seams to be closer to 30-40mm2 for those 4x 64bit controllers in Tonga.

As for the 128bit for 2048 shaders, dont forget that Pollaris has a new memory controller AND L2 Cache.

 

[MrTeal]

yea after a better look, seams to be closer to 30-40mm2 for those 4x 64bit controllers in Tonga.

As for the 128bit for 2048 shaders, dont forget that Pollaris has a new memory controller AND L2 Cache.

Unless they're shipping it with GDDR5X (and that seems unlikely in a budget product ready for back to school) even a new memory controller and L2 is going to have a big mountain to climb to make up for a 128-bit bus. Even at 7GT/s that's only 112GB/s bandwidth and a far cry from the 288GB/s of the 280X or 182GB/s of the 380X.

 

[AtenRa]

Unless they're shipping it with GDDR5X (and that seems unlikely in a budget product ready for back to school) even a new memory controller and L2 is going to have a big mountain to climb to make up on a 128-bit bus. Even at 7GT/s that's only 112GB/s bandwidth and a far cry from the 288GB/s of the 280X or 182GB/s of the 380X.

Well NVIDIA did it with Maxwell GM206,
GTX960 with only 1024 Shaders and 128bit memory is as fast as GTX680 with 1536 Shaders and 256bit memory.

 

[antihelten]

Well NVIDIA did it with Maxwell GM206,
GTX960 with only 1024 Shaders and 128bit memory is as fast as GTX680 with 1536 Shaders and 256bit memory.

But both the GTX 960 and the GTX 680 are significantly slower (20%-ish) than the 380X with 2048 shaders and 256 bit bus.

 

[maddie]

AMD is going to most likely sell every chip they can. They'd have no competition from Nvidia. Without a performance metric, I can't really say how it will be, but if you're telling me they're going to offer a premium part at discount prices, you haven't been listening to AMD.

AMD wants to increase profits per unit sold. More so if they have a leg up on the market. You launch it high, get the early buyers, make bank, when the competitor comes out, you drop price since you've already made some bank, and thus put extra pressure on your competitor.

If AMD is going to go for volume versus margins, this is why they are facing bankruptcy.

Yep, pretty much this in a nut shell. Some of the loudest squawkers on price only spend about $200-$250 on a video card and expect AMD to give them something top of the line for that amount. Unfortunately AMD has screwed the pooch and have had too many fire sales and now this vocal minority thinks those are somehow normal prices. Unless the supposed AMD fans change their thought process they aren't going to have AMD around any more.

Can't both of you see that overall profit is more than just unit profit. Sale volume is the other part of the equation. Simply put, total profit = unit profit X # units sold. Both of you keep getting stuck on maximizing unit profit, when the company is more interested in overall profits. The problem with the sale volume / unit pricing formula is that they are co-dependent variables and so an optimal solution is not as obvious as some may assume.

 

[MrTeal]

Well NVIDIA did it with Maxwell GM206,
GTX960 with only 1024 Shaders and 128bit memory is as fast as GTX680 with 1536 Shaders and 256bit memory.

The difference there is that nVidia introduced color compression with Maxwell 2.

Tonga already has DCC, which is probably part of the reason it wasn't hurt too badly dropping from a 384-bit bus to a 256-bit bus. They would need some pretty extreme additional changes to make up for halving the bus width again though.

 

[maddie]

Hope you get your system. I know a lot of people are downplaying MGPU, but I think it could be fun just messing with it.



How good would it be for us if AMD pulled another Evergreen release? If they can get their CPU business rolling again with Zen they wouldn't need the large profit margins on GPU's that nVidia shoots for.

It absolutely blows my mind that there are people on these forums advocating high pricing. I don't think I've ever read anything more ridiculously stupid in my life. Literally, "Take my money" talk. Unfreakinbelievable!

Someone finally said it. If my other speculation is realized then we're looking at a 1 yr + lead probably.

 

[AtenRa]

Someone finally said it. If my other speculation is realized then we're looking at a 1 yr + lead probably.

I will say 3 to 6 months top. Probably something like AMD will release in June and NV in September-October. But this time availability may be on the AMD side, NVIDIA will have to share the same fab with Apple and more companies than AMD manufacturing on the Samsung Fab.

 

[tential]

Basically what Intel has been doing. No need to really bring anything huge to the market if you're only fighting yourself.

Why I get this weird feeling AMD (and all of their outspoken supporters) are going to find out both Intel and NV have been quiet about things because well they aren't threatened.

Zen is going to come out swing, hit (throwing them a slice of pie) Skylake IPC numbers, Intel is going to laugh and unveil what they've been working on while AMD was playing catch. Same with Pascal. It's not like AMD is deterring ar from GCN so what they've had available has been on the table since 2012.


I will be pleasantly surprised if AMD can catch BOTH Intel and Nvidia with their pants down. But just reading AT's recent Carrizo artile makes me realize more so AMD has no idea what the hell they are doing. This Carrizo fiasco reminds me of when AMD sat approvingly and nodding of a Freesync vs Gsync competition while the organizers basically sabotaged it.

The CPU is MUCH harder for AMD to come back on. Because they have to integrate with a lot of partners. The GPU on the other hand, it's not like AMD is very far behind in GPU. It is ahead of Nvidia in some parts, and behind in others. I mean look how long Hawaii lasted? Up against Kepler, and now beating out even Maxwell. I'd say Fiji is a certain bust though, and you know how much I'm against Fiji.

I definitely think AMD is capable of doing this to Nvidia on a node switch.
Amd and Nvidia aren't that far apart and it's been AMD to push to new technologies quicker.

The issue isn't the hardware for AMD on the GPU side. It's the publisher support/driver support.

On the CPU side.... there are a lot of things wrong, and a lot of unknowns about Zen. My guess is that at this point, due to manufacturing process, all devices with Zen/Intel will have decent power consumption. Even if Zen loses out, I bet it will sell in products, just like other slightly higher power consumption socs in tablets sell.

As you noted though, it's how AMD handles it. Making it super easy for your OEM's to release products at the expense of user experience is insane. AMD's problem will be convincing people it's not the same old crappy stuff that it used to be in the past. People prefer to make safe choices, so unless Zen is actually great in some aspect, it will really not have much of an impact for AMD.

I think Zen's going to have a corecount advantage over Intel, be slower. Power consumption, who knows. But if it does of a corecount advantage on Intel at the same price as intel, and the speed is close to sandybridge, I'll probably pick it up for my server. It's the reason I haven't bought any CPU for my server yet actually. Waiting to see if Zen has any usage scenarios for me.

 

[tential]

An AMD Corporate VP said this? Then you know it's the wrong thing to do.

"Guy's we're hemorrhaging left and right, let's offer up out newest GPU family at consumer friendly prices! That should stop the bleeding!"

You're thinking way too short term though. If AMD sells at higher prices and continues to sell a similar amount of GPUs/marketshare and makes slightly more money because AMD fans continue to stay there then great.

AMD may however double down, invest more into advertising, maybe they'll find a competent marketing scheme, and utilize being first to market, at good prices to upgrade to, and gain actual market share traction.

Look, Nvidia had ok GPUs, but Maxwell's BIGGEST selling point, was it was an upgrade option, with no alternative for a LONG time. Being first to market is hard to ignore. Many people couldn't even wait from the 980Ti release to the Fury X release to buy the 980Ti.

If Polaris holds the GPU crown for a couple months, while we wait for Nvidia, I mean it won't even take me more than 1 month. I'll buy it if it's the fastest thing out and an upgrade option and it's available first. I'm not waiting around. If there is something better, I'll sell it, or figure it out later, but I'm enjoying this gpu gen asap and if the gains of node shrink GPUs/new architecture GPUs are significant over the 980Ti/Fury X, I doubt people will care to wait much longer either if AMD prices aggressively.

To suggest though that if AMD released first, their only option would be to use high prices is really just a low level look at the situation.....

You're literally suggesting that this GPU release is the same as the last 2 GPU releases....l which I mean, that's just a low intelligence look at the situation there is no other way to put it.

 

[railven]

Can't both of you see that overall profit is more than just unit profit. Sale volume is the other part of the equation. Simply put, total profit = unit profit X # units sold. Both of you keep getting stuck on maximizing unit profit, when the company is more interested in overall profits. The problem with the sale volume / unit pricing formula is that they are co-dependent variables and so an optimal solution is not as obvious as some may assume.

They went that route. They've been trying to sell a faster/cheaper card since Nov 2013. How did that work? Pretty bad.

When did you see AMD gain back market share? You think it was Fiji selling strong? It was Grenada. A higher priced Hawaii with more RAM.

It was received better, thus it sold better. To us "intelligent forum hopper" we knew it was a dolled up bargain bin <$200 290X. Joe Schmoe didn't. But AMD saw GPU growth. Meanwhile Fiji got it's teeth kicked in by GM200 on a pegged leg.

Someone finally said it. If my other speculation is realized then we're looking at a 1 yr + lead probably.

If the lead is that great, more of a reason for AMD to milk this for all they can. AMD needs money, perhaps if they had more money they could actually coax more devs on the PC side to give them a shot. They controll 100% of the console market. What good does that do them if they can't even keep a previous customer (Tomb Raider/Square-Enix) on the catalog list?


AMD needs to make some quick cash. So they can fund other aspects of their corporation. If you're solution is to sell superior GPUs at discount prices for probably a year or longer, that's just losing money.


Again, why didn't AMD milk the bitmining scene? Radeon Bitming Edition $1000 MSRP, you don't think it would have sold? Meanwhile NV is selling $1000 Titans. Then down the road (ala Nvidia) they release a proper 290X.

Rinse repeat (NV sure did, and they're laughing dragging bags of money to the bank, meanwhile AMD still couldn't sell <$200 290X).

Build your brand then marketshare will come. CPU wise they have the best bargain CPUs. Isn't doing jack to save them.

 

[maddie]

You're thinking way too short term though. If AMD sells at higher prices and continues to sell a similar amount of GPUs/marketshare and makes slightly more money because AMD fans continue to stay there then great.

AMD may however double down, invest more into advertising, maybe they'll find a competent marketing scheme, and utilize being first to market, at good prices to upgrade to, and gain actual market share traction.

Look, Nvidia had ok GPUs, but Maxwell's BIGGEST selling point, was it was an upgrade option, with no alternative for a LONG time. Being first to market is hard to ignore. Many people couldn't even wait from the 980Ti release to the Fury X release to buy the 980Ti.

If Polaris holds the GPU crown for a couple months, while we wait for Nvidia, I mean it won't even take me more than 1 month. I'll buy it if it's the fastest thing out and an upgrade option and it's available first. I'm not waiting around. If there is something better, I'll sell it, or figure it out later, but I'm enjoying this gpu gen asap and if the gains of node shrink GPUs/new architecture GPUs are significant over the 980Ti/Fury X, I doubt people will care to wait much longer either if AMD prices aggressively.

To suggest though that if AMD released first, their only option would be to use high prices is really just a low level look at the situation.....

You're literally suggesting that this GPU release is the same as the last 2 GPU releases....l which I mean, that's just a low intelligence look at the situation there is no other way to put it.

Exactly.

Any top management must know that for the share price to substantially improve is to have a large increase in corporate profits which will need to include a significant sales improvement. Notice I said include, not exclusively have. I can imagine a lot of work in arriving at the perceived optimal pricing points.

 

[Headfoot]

I forget, is this a consumer technology enthusiast forum or a stockholders annual meeting? I can never tell

 

[railven]

I forget, is this a consumer technology enthusiast forum or a stockholders annual meeting? I can never tell

And this is why AMD is going to fail. When their own customers aren't willing to shell out a little extra.

 

[Headfoot]

I shell out extra for returns in performance. I couldn't care less who makes the thing that performs faster. If a company wants my money they should make a faster card or have the same speed card cost less. AMD got my business this round because 2x290 was roughly around 1x980 Ti but a lot cheaper. A little thing people on this forum used to care about called "bang for the buck"

 

[tential]

And this is why AMD is going to fail. When their own customers aren't willing to shell out a little extra.

Your whole goal is maximizing today's profits.

No one is saying the GPUs will be dirt cheap.
But if at $350 AMD gains 10% additional marketshare for releasing a GPU first, or if they release at $270 AMD gains 20% additional marketshare, than clearly the $270 is better. It's more money, and more marketshare.

Even still, you have 0 idea what AMD's goal is. No company sits there and worries just about today. You worry about tomorrow, you worry about the competition, etc.

AMD has been in the red for awhile now. A couple more months of being in the red won't hurt AMD if it sets them up for future gains in the future. Seriously I can't believe you are suggesting that the only way AMD can release Polaris is at the highest possible price the consumer will charge.

That was smart with Fiji, when it was the end of the 28nm process, and it wasn't going to gain ANY market share. Fiji sucked, and had lots of pitfalls.

Polaris has no competition, is released first to market. Just like how Nvidia aggressively priced Maxwell when AMD had no answer to it with the GTX 970, I expect AMD to have an aggressively priced mid tier chip.

It doesn't mean all chips have to be cheap but don't expect AMD to come out of the gate first to market FINALLY, and not attempt to eat into Nvidia's midrange marketshare with whatever tactics possible. Marketshare is important, just ask Nvidia.... they mindshare purchase a ton of people.

So please, lets try and act like we can actually use more than just a blanket statement "OMG, AMD is going to release these super cheap GPUs and they're going down the same path again."

It's just literally the most basic of statements you can make with just no analytical value there what so ever.
------
On a side note, it's not hard to raise money from investors, while consequently losing money, while consequently focusing on volume and pushing brand awareness. I mean, I literally just spent my time doing that for a company. The short term gains of revenue were NOT worth it. The CEO and investors were not investing for that. They were investing for the future gains of 5-10 to 20 years down the line.....

So I mean, short term thinking is great, but that's just not everything.

And this is not another 28nm to 28nm GPU transition so comparing it to other ones is futile. This is a double node shrink... so use a comparison to something like that.

 

[railven]

I shell out extra for returns in performance. I couldn't care less who makes the thing that performs faster. If a company wants my money they should make a faster card or have the same speed card cost less. AMD got my business this round because 2x290 was roughly around 1x980 Ti but a lot cheaper. A little thing people on this forum used to care about called "bang for the buck"

People love bang for buck. They love it so much most people probably jumped on AMD cards AFTER they collapsed in value. I stuck a 290X into my basement computer to replace an aging HD 5870. It cost me $165.

Unfortunately, bang for buck is what is slowly killing AMD. If AMD is in a position to actually rattle Intel/NV, I hope they don't go down the "bang for buck" hole again.

 

[maddie]

They went that route. They've been trying to sell a faster/cheaper card since Nov 2013. How did that work? Pretty bad.

When did you see AMD gain back market share? You think it was Fiji selling strong? It was Grenada. A higher priced Hawaii with more RAM.

It was received better, thus it sold better. To us "intelligent forum hopper" we knew it was a dolled up bargain bin <$200 290X. Joe Schmoe didn't. But AMD saw GPU growth. Meanwhile Fiji got it's teeth kicked in by GM200 on a pegged leg.



If the lead is that great, more of a reason for AMD to milk this for all they can. AMD needs money, perhaps if they had more money they could actually coax more devs on the PC side to give them a shot. They controll 100% of the console market. What good does that do them if they can't even keep a previous customer (Tomb Raider/Square-Enix) on the catalog list?


AMD needs to make some quick cash. So they can fund other aspects of their corporation. If you're solution is to sell superior GPUs at discount prices for probably a year or longer, that's just losing money.


Again, why didn't AMD milk the bitmining scene? Radeon Bitming Edition $1000 MSRP, you don't think it would have sold? Meanwhile NV is selling $1000 Titans. Then down the road (ala Nvidia) they release a proper 290X.

Rinse repeat (NV sure did, and they're laughing dragging bags of money to the bank, meanwhile AMD still couldn't sell <$200 290X).

Build your brand then marketshare will come. CPU wise they have the best bargain CPUs. Isn't doing jack to save them.

How do you explain this 2 day old statement to gamecrate by Roy Taylor? He's talking about the entry requirement for VR. I think you're being way too myopic. Notice he says "much lower price"

My belief is that we will get R9 390 performance at the 380X price point.

Previously Koduri said that there are 2 versions of Polaris, 10 and 11, not two GPUs in total. One interpretation is a GPU design with a GDDR5 memory controller [Polaris 10] and the other with a HBM controller [Polaris 11]. this will explain the statement that Polaris can use Gddr5 and HBM, just not on the same version. Within each version, we might see more than one GPU.

GC: You've mentioned the existence of quality content as one of the drivers of VR adoption. What are the other obstacles limiting widespread adoption of VR?

RT: They are two-fold. First of all, it's expensive for a lot of people. Now the people who really want it, just don't care. Quite frankly Oculus could have sold their headsets for a thousand dollars and I'm pretty convinced they still would have sold all of them. Because this first wave of adopters just aren't going to care. But when you get to the second wave of adopters, to people who are not technically savvy, then the price becomes tentative. And right now it's cost-prohibitive for a lot of people.

If you look at the minimum spec for either the Oculus or the HTC, and then you look at how many units of the minimum spec have been sold since their launch, so I'm talking about the Radeon 290 or GeForce GTX 970, according to Jon Peddie Research, the total install base of those parts or better is 7.5 million units. So we're going to have to make it possible to run good quality VR at a much lower price. And I'm confident with Polaris we're going to have a big impact to help that.

The second barrier is that, right now, you try VR out and your reaction is, invariably, "That's really really cool. That's super cool." But then you take the headset off. I would defy anybody, I haven't met a single person yet who could say to me: "I tried something and it was so great, I would have spent 10 hours in there if I could have done." That doesn't yet exist. So I think this content thing is absolutely critical, and it's why I talk about it so much, and why we now have a worldwide team, frankly in search of that content.

 

[railven]

How do you explain this 2 day old statement to gamecrate by Roy Taylor? He's talking about the entry requirement for VR. I think you're being way too myopic. Notice he says "much lower price"

My belief is that we will get R9 390 performance at the 380X price point.

Previously Koduri said that there are 2 versions of Polaris, 10 and 11, not two GPUs in total. One interpretation is a GPU design with a GDDR5 memory controller [Polaris 10] and the other with a HBM controller [Polaris 11]. this will explain the statement that Polaris can use Gddr5 and HBM, just not on the same version. Within each version, we might see more than one GPU.

We've had that. We've had better.

But now they're going to use a most likely costlier die to repeat it? The only perks would be improved efficiency.

 

[maddie]

We've had that. We've had better.

But now they're going to use a most likely costlier die to repeat it? The only perks would be improved efficiency.

I feel as if I'm talking to an alien with no common reference points.

I guess we will have to disagree and see what happens. Time will certainly tell.

edit: I think Nvidia would want AMD to price as high as possible.

Also, why won't you comment on what Roy Taylor said?

 

[JDG1980]

We've had that. We've had better.

But now they're going to use a most likely costlier die to repeat it? The only perks would be improved efficiency.

The runaway success of Maxwell demonstrates that efficiency is actually quite important to GPU buyers. If people only cared about performance per dollar, the deeply discounted R9 290 and 290X cards would have flown off the shelves. Instead, it was the GTX 970 that was 2015's big hit. If people only cared about performance per dollar, no one would have bought any GM206 card except for people who needed its HEVC decoding and/or HDMI 2.0 output capabilities.

Moreover, there is really no good reason to believe that a FinFET chip around 200mm^2 is going to be more expensive than an equivalent 28nm planar chip at 438mm^2. The "doomsday charts" that ShintaiDK loves to post don't hold water. Not only will a refined, die-shrunk performance equivalent to Hawaii probably be cheaper to make, it will also be able to use a narrower memory bus due to the introduction of memory compression and possibly GDDR5X. That drives down the total bill of materials.