A visual depiction of where things are rumored to be headed:

Cats outta the bag finally:
http://www.extremetech.com/computing/227...-chinese-market

AMD Might be catching up sooner than people think with Zen Family,
Intel has announced they wil "Scale Back Desktop Processor Development Significantly", And aim more for mobile market.

AMD also has new Deals with Nintendo, Sony, and Microsoft ( For Polaris Products, separate contracts from the XBO/PS4/NX agreements ).

Worth the Watch, Definitely

Just another statement confirming that AMD Polaris 10 based cards are the "flagship" products, this year. Presumably, they're out in June/July. Nvidia seems to be targeting the same time-frame for their new flagship, from what I've read (but, I don't follow Nvidia closely).

Polaris Replaces R9-390X and Below, Fury will remain in production.
All the Tahiti Rebrands and Hawaii Rebrands will discontinue production.

Navi is where it's at though.

NGM, Scalable GPU, Instead of one BIG EXPENSIVE CHIP that will have low Yields per wafer,
They will put 4 smaller Chips on a Single Interposer w/ NGM. As far as the PC is Concerned it's one GPU, when it's really 4.

So Top End w/ Be 4, Mid will be 3, Mainstream 2, Entry 1.

The videos above describe perfectly why this will save Money left and right.

nVidia's New Monster Chip, is selling for $5000, because they only get so many good ones per/wafer.

Finally had the time to watch the videos. Thanks for posting them.

The presenter's "speculation" regarding the future of AMD, Nvidia, and Intel sounds quite thoughtful and believable. As a fan, it was fun to watch/listen to a positive evaluation of AMD's possible near-future.

Conclusion: Bye bye, Nvidia (well, maybe)

Anyhow, at first blush, it sounds like I should plan on a Polaris 10 sooner than later. Then, I can anxiously look forward to the Navi as the card to follow my Polaris 10 -- and hope Navi becomes what the presenter "speculated" it would be.

I might have to bite the bullet and upgrade my GPU First,

None of my Lightnings work w/ VR Headsets (no HDMI, all MiniDP and Single Link DVI).

Tried 4 Different MiniDP-> HDMI Adapters, 1.4 Active, 1.4 Passive, 2.0 Active, etc,

None worked 100%,

So I've resorted to downgrading my GPU's to the 7870XT Backup (7930), as I gave my 7950 OC to my brother.

Seems I can Get a 4 GB R9-380X for lil over $200, and I expect that to drop a bit when 400 Series Launches,

So I might see the 400 Series Prices as well.

But I need more than 1536 Sh and 2GB of Ram to Run VR. lol.

Hopefully, a single highest end Polaris 10 will be nearly as fast as my 2x HD7970 Crossfire. That would be enough incentive to buy a new card at around $350 (the rumored price range). But, rumors are rumors. Maybe we'll see some facts at the end of this month.

Ditto,

Replace the Xfired Lightnings, Keep them preserved, When I build a new system, put lightnings back in the old one and use that one for studio rendering, and the new rig for gaming

GTX1080 Priced at $599, Titan Crushing Performance, half the TDP.

Big Beefy Chip, they gotta be selling them at a loss to try and maintain market share.

1080 it is for me........just waiting for the £££ conversion but no doubt it will be £699 as we usually get ripped off.

From there press release:

At a special event today, Nvidia announced their new flagship graphics card: the GeForce GTX 1080. The card uses a new Pascal GPU built on a 16nm FinFET process, as well as GDDR5X memory from Micron. Nvidia says they spent several billions of dollars developing Pascal, and the result is a massive GPU capable of powering today and tomorrow's games at ultra quality settings. It's also energy efficient thanks to improvements in the way Nvidia delivers power to the GPU, making Pascal the company's most efficient architecture yet.

According to Nvidia, the GTX 1080 is "a whole lot" faster than the Titan X, and faster than two GTX 980s in SLI, while consuming a lot less power. The chart below sees the GTX 1080 consume around 180W of power, compared to the 165W TDP of the GTX 980, with performance around 30% faster than the Titan X. An on-stage demo of the GTX 1080 had the card's core running at 2.1 GHz air cooled, with a temperature of less than 70°C. While these aren't the stock clock speeds for the 1080, it shows just how overclocking friendly this card will be. The stock clock speeds for the GTX 1080 are 1607 MHz with a boost of 1733 MHz, on 2,560 CUDA cores. This suggests the card is using a GP104 core rather than the full GP100 GPU used on the Tesla P100, which comes partially-disabled with 3,584 CUDA cores.

The GTX 1080 will retail for $599 in its basic edition, and $699 for "Founder's Edition", although it's not completely clear what benefits the Founder's Edition will bring. The card will be available worldwide from May 27th. There will also be a GTX 1070 available for $379 ($449 for the Founder's Edition), which Nvidia says is faster than a GTX Titan X, packing 6.5 TFLOPs of performance and 8 GB of GDDR5 memory. The GTX 1070 will be available from June 10th.
GTX1080

Over to you AMD,intersting the stats on Nvida's website are against the 980 not the 980ti or titan and the recomended power supply is only 500 watts!!?

Competition is good -- increases options and decreases prices

By the way, it will be interesting to compare AMDs 14nm manufacturing process "performance" to Nvidia's 16nm manufacturing process.

Until the actual comparison-test reports come in, we can't say which is better. There are rumors -- but they are rumors.

Nonetheless, "on paper", AMD 14nm should have a "measurable" performance advantage -- in FPS and/or power use. They should also cost less per chip at a given performance level.

Thing is, AMD is said to be aiming at "very low power requirements" and "low price" with the first Polaris cards (versus highest possible FPS). Nvidia may have emphasized something else. So, we won't know anything for sure until both AMD 14nm and Nvidia 16nm cards are on the market and tested on a spectrum of different games.

A few interesting features on the Pascal cards for VR too

http://www.geforce.com/hardware/10series/geforce-gtx-1080

1 Power Connector :-O

The Nvidia "1080" is anticipated to be a $600 card and the 1070 closer to $500. The May 27 launch is for a $700 card. Previous rumors indicated Nvidia production was behind AMD. So, the first cards out may be in "short supply" -- almost a "paper launch" (Nvidia has done this before -- we'll see how it turns out this time).

Recent AMD Polaris 10 rumors imply AMD will first release a $299 to $350 product. AMD's rationale is rumored to be that the vast majority of Gamers won't spend much more than $300 for a card (although, SimHQ members might be exceptions to that rule ).

Anyhow, the new AMD marketing plan is rumored to be "high sales volume" (with excellent performance) versus Nvidia's "fastest card on the market" (that few will actually buy because of the high price). Nontheless, the Nvidia strategy does keep Nvidia in the "public eye" -- which has great impact on sales volume overall.

Hopefully, AMD will keep the "late May to early July" time frame that they are rumored to be targeting.

AMD is seeing the bigger picture,

nVidia is still living in the 2002 market.
PC Sales are Struggling w/ Mobile Devices taking over, Gaming PC's of the $2000+ Variety are losing out to Consoles, etc etc.

The only way to bring customers back to PC Gaming is to give them cheaper hardware.

nVidia's Full PASCAL GPU is selling for $5000,

1080 is about 1/2 of the Full GPU at $500
1070 slightly less

nVidia's GPU's are Huge, they get less per wafer and even less yields due to defects,

Almost guaranteed they are selling the 1000 series at a loss, they dont want to lose their 80% Discrete Marketshare.

AMD is Fabricating smaller GPU's, while they wont have a Fastest in the world this round, they will have more GPU's per wafer and more complete GPU's yielded.

Next Gen (Navi), will be scalable via interposer (2/3/4 GPU's on the Interposer w/ HBM2. As Far as System and DX12 is concerned it's one GPU, when it's really in fact multiple.

Crossfire and SLi will likely be dead in 2 years, there's no point when DX12+ Supports Explicit Multi GPU, and Vulkan will have it's own version of the same tech.


The Bigger Picture?
Discrete GPU's are 15% of the Total GPU Market Share right now, w/ Consoles Taking up 75% (65% Current 10% past)

AMD owns about 78% of the Total Market share via owning 100% of the Consoles, part of past consoles, and 100% of the 3 next gen consoles.

Developers, Who have relied on nVidia Pro Cards to develop on, will start shifting their Studios to AMD Pro/Sumer Cards,
NO Point developing on nVidia Architecture when 80+% of the Market will be running on GCN1.1, 1.2 and 1.3.

No there's the "Port to PC" Market, The Games will be coded to be efficient on GCN as the consoles run them, no one is gonna fork out $600+ FOR A gpu when a Entry/Mid Level $260 and Below GPU can Run the Games at 1080p/60fps which is still the standard.

Their 80% PC Marketshare is about to take a nose dive.
They cant afford to turn people away w/ another $1499 Card that will be replaced by a faster cheaper one a few months layer (Titan/x80Ti's).

Now PC GPU Market Share is about 15%,
Multiple GPU's is about 2% of that, if not Less.
4K and VR are still a small % of that 15% as well.
XFire / SLi is a hassle to code for, developers would rather not deal with it at all. the amount of XFire/SLi Supported titles being released drops like a rock every quarter.

Further proof SLi is dead, a majority of the new "GameWorks ready Effects/Engines" in nVidia's GameWorks SDK do not support SLi, but will work fine w/ DX12 EMA.

Even when Navi/Scalable GPU's Launch, the GPU Die's themselves will be cheaper to manufacture and they will get more per wafer, so a 2 GPU Die on a Single Interposer Card would cost significantly less than a card w/ a GPU that has 1.5-2.5x the die size.

I'll just say this--market share is irrelevant when a majority of your products are sold at low margins.

My company, which is a semiconductor company that sells chips to many of the big companies out there, including Intel, GM, and others I'm not allowed to name, recently dumped manufacturing many high-profile yet low-margin chips for them. I believe one was for the Xbox, actually.
If you need to make 500,000 chips that cost you 75 cents, that you only sell for $1, but have yield issues such that you're only making $50,000 profit on that order...why bother? Using that manufacturing capacity (which is finite) to make smaller numbers of higher margin chips just makes sense.

Having large market share for name recognition is not more meaningful than having the fastest card for name recognition. Larger profits are.

It's a simple investor fact--if you're selling a small number of chips at a large profit margin, well you can sell more to increase profits. That looks good. If you're selling a large number at a small margin, well you're not going to be able to increase the price...so you have to sell even MORE chips which becomes harder and harder to do as numbers increase. That looks iffy.

It looks to me like you're trying to spin things as positive or negative because you want them to be so, not because there's any real truth to it. Intel and nvidia aren't worried, and AMD isn't feeling like they're about to crush the competition.

Simply put AMD is trying to fight a two front war, when none of its competitors are, whilst having the least amount of resources. nvidia went whole-hog into discrete and mobile, never cared about CPUs. Intel went with CPUs and mobile, never really cared about discrete although they've done some integrated. AMD's mobile side isn't as strong as either of the others as they tried to do the CPU/GPU/APU thing. Intel and nvidia really only compete with AMD, not each other.

AMD's purchase of ATI was a gamble that might have paid off if they'd come out with great APUs within a year or two of the acquisition. They took too long, WAY too long, and the market has changed away from PCs and consoles to the point where the area they're strongest in is now the weakest part of the market-. It's like declaring in 2016 you have the best mounted cavalry division in the world. Great, if it was 150 years ago, but how does that help against mechanized infantry and air support?

IMO AMD needs to bite the bullet and redivide into the old AMD/ATI again so they can focus their efforts correctly. ATI might be able to best nvidia, but AMD/ATI never will because they're worried about Intel. AMD might beat Intel, but AMD/ATI never will because they're worried about nvidia. They're unwilling or unable to ignore one and focus on the other.
If AMD's lackluster showings weren't obvious, all you need to do is look at the evolutionary, not revolutionary, products nvidia and Intel have been shipping the past few years. I went from an i7-2600k to a Skylake and saw modest improvements only...because Intel doesn't need to push it. Likewise my 770 to 970 jump helped things but not in any jaw-dropping way...because nvidia doesn't need to push it.

It's been years since I bought an AMD CPU or GPU because they've not had the best performance in the price range I look at of roughly $350 each for my lone PC. There was a time when I had an FX and a Radeon, but W was president.
Like HP's purchase of Compaq, or 3dFX of STB, the ATI buy was a strategic error that has had lasting effects.



The Jedi Master

Good analysis of "price, volume, profit".

Its cool to have some positive things to say about AMD -- to an AMD fan

And, we fans would sure like things to get better businesswise for AMD because:

Competition is Good and AMD is the only realistic competition for Intel and Nvidia (I'd hate to think how slow CPU and GPU progress would be and how much things could cost without competition).

However, we're not trying to "spin". Spin is a form of lying.

Rather, we're simply regurgitating rumors regarding AMD -- and hoping they're true "this time". They've been in error in the past. So, as usual, we hope and wait for the test reports and the financial statements.