Would it be better for them to have released something and 6 months later have a far superior product available?
Imho, yes. Around summer of 2014, if you told me AMD won't have anything to launch from July 1, 2014 to July 1, 2015, I would have said OK then do a re-spin of R9 290, add 10-15% higher clocks, do a transistor level improvement (ala GTX480 --> GTX580), maybe add an AIO CLC and relaunch a 10-15% faster, cooler and quieter reference R9 295XT PE with performance near GTX980 for $429, and add an after-maret cooler/AIO on the 290 with 5% higher clocks and sell it for $329. It would have been a lot easier to sell a card that uses more energy but it's as fast as a 970/980 than the position R9 290/290X are in now where out of the box they are slower and use more energy than the 970/980. Secondly, it would have allowed AMD to turn the image of the R9 290 series around because all new launch reviews would have painted the R9 295X and 295XT PE as cool and quiet! This is crucial because today even a $250 R9 290 with 50-60% more performance than a 960 can't sell. That's shocking and never in the 15 years of GPU history have I seen something like this.
Also, with a new card launch in the fall of 2014, AMD could have added HDMI 2.0, new H.264 hardware decoder, etc. These basic check-mark features would have at least put them on equal footing with Maxwell. Either R9 300 series will completely pummel R9 200 series into the ground, or AMD has overstretched its resources too much and couldn't develop a brand new series in time.
And the question of a looming Pascal in 2016 still rings true with R300's late release (we don't know the exact date of its release, so how are you coming up with 1 yr after R300?).
NV releases brand new architectures roughly every 2-2.5 years. This has been consistent since GTX200 series. You can look it up yourself. In fact, it took NV 3 years to go from GTX580 to 780Ti and 2.5 years to go from GTX680 to its 980 successor. Therefore, I would say we won't see Pascal's mid-range GP204 until Q3-Q4 2016. We might see some lower end card like 750Ti successor by Q2 2016 though. NV likes to maintain a 2 year+ cadence between its flagship cards (GK100/110 --> GM200). They are almost doing a tick-tock of sorts where they bifurcate a generation into 2 halves. That have followed this strategy with Kepler and Maxwell and there is little reason to indicate Pascal won't be the same execution.
Nobody wants an old 28nm product when the new hotness from NVIDIA is around the corner.
Is this a serious statement? NV just launched the Titan X, and no other GM200 series product. What exactly will NV have "around the corner" on 14nm/16nm by August 2015? You honestly think NV will overcome the laws of physics and its historical trends and succeed GM200 in 12 months with a Big Daddy Pascal? It took them 3 years to do that with Fermi to Kepler and it's already been more than 2 years since 780 came out and NV still hasn't even released consumer GM200 cards. Therefore, it's only logical to assume that GP200 or w/e they call the flagship isn't going to be out until Q1-Q2 2017. That gives R9 300 series plenty of time to make a dent.
As for the bold part, nobody is talking about the notebook market because AMD has been absent for years and that isn't the reason for recent shifts in market share.
You are not following the discussion closely enough. If 100 AMD cards are sold in Q3 2014, and NV sells 200 cards, the market share for
that quarter is 66% NV, 33% AMD. Then in Q4 2014, NV sells 250 cards, AMD sells 90 cards, the market share for
that quarter becomes 26% AMD, 74% NV. You are now going to say that AMD only has 26% overall market share in the industry - this is incorrect. The overall market share for that 2H of the year is actually (100+90) / (200+250+100+90) = 30%. Therefore, by you only looking at quarterly market share numbers, you are ignoring millions of AMD (and NV) customers who haven't upgraded. That means if AMD were to get R9 300 series notebook design wins, its market share will go up
no matter what NV does.
If you ship 0 product and next quarter you ship 100 units, your market share goes up automatically. Since the GPU market fluctuates between 14-17 million units a quarter, if it was NV that gained market share at AMD's expense, the overall market would have remained at 14-17 million -- it isn't. It has fallen sharply to just 11.5-12.5 million while the number of units NV sells is basically flat. This is not the same if NV took over AMD's OEM design wins because if it did, the overall GPU market would continue to be 14-17 million units sold a quarter. You keep ignoring this.
Your argument that Windows 10 + Skylake will somehow move R300 top end GPUs doesn't make sense as you pointed out, the notebook market has been growing quite big and if anything, that will benefit NVIDIA more than AMD.
I specifically outlined all 4 market segments, not just the top end. If R9 300 series is competitive and can get notebook design wins, it'll gain market share against NV. If what you said is true, AMD would never be able to recover market share EVER but it has done so many times before. Most people in this thread and online are ready to burry AMD but until we see how good (or bad) R9 300 series is, this type of thinking is premature and has no evidence to support it. Last time AMD was bombing with 2900 and 3800 series and 4800 series made a huge mark.
What you think is not relevant. You're speculating, you don't know why it's late. More importantly, it doesn't matter why it's late. Whatever the reason, it's not a good thing. Having a crap product that's on time isn't a good thing either.
When the consumers in the market keep blaming AMD for being late, having inferior products but they have bought NV for 5, 10, 15 years in a row, where is the firm supposed to get free cash flows to invest into future products? I love it how consumers who keep talking smack about AMD the most are NV owners who haven't purchased an AMD card since the ATI buyout. Profits and R&D investments don't just come from thin air. The PC gaming community proved 100% their bias because they never supported AMD during perf/watt days, during price/perf days nor during AMD's leadership days (5870/ 7970Ghz). In fact, even during the mining craze when AMD cards made $
which made them free (which means one could have built multiple rigs with HD4870->R9 290 and used the money earned to
buy countless NV cards), 95% of the PC market ignored it too. So really, it's ultimately the loyal NV consumer that put AMD in the position the are in today. That core 50% market share group that always buys NV has remained. What happened are the brand agnostic PC gamers who stopped waiting for AMD and bought NV. But it doesn't change the fact that even during the generations when AMD was the clear choice based on all the metrics that are used to hype up Maxwell today, AMD never even managed to get > 50% market share.
As an example, 970 is hyped as some spectacular card but when HD5850 smashed everything NV had to offer in that price range, and still pummeled GTX460 in perf/watt 9 months later, what happened? NV gained market share with Fermi. Oops. So really the key themes that have emerged since 2008 when HD4000 series launched is that the NV's loyal customer base will use any metric possible to justify sticking to NV and over the years that metric has changed countless times. During the Titan days, DP was key, now it's irrelevant. During GTX200/Fermi days, perf/watt was irrelevant, now it's THE key metric. Also, even on this forum people make uninformed statements like Maxwell is the greatest overclocking GPU generation of all time but HD7950 overclocked on air from 800mhz to 1.25Ghz-- not a single Maxwell card can accomplish this. Every generation all key metrics where AMD leads have been downplayed and there is no doubt in my mind that the average PC gamer buys NV first because of his experience with NV for years/decades and because of the brand name/popularity.
I am not trying to pick on you but just focusing on your cards, HD7970Ghz CF smashes 680 2GB SLI into the ground in modern games. Like it's not even close, but did you buy 7970Ghz CF that paid for itself? No, you didn't. So how is AMD supposed to get money to develop future generations when people keep buying inferior products?
In modern games today, HD7970Ghz CF are consistently 20-30% faster than a GTX690, while a single 7970Ghz beats 680 by 15-20% constantly. All of this has been ignored by NV users who keep focusing on Maxwell but they never want to talk about Kepler flopping, or GTX570/580 running out of VRAM, etc.
Looking at the performance in GTA V, ARES II 7970Ghz CF is getting 57 fps but a 690 only manages 38 fps. That's mind-blowing, like a generation apart and yet most people on this forum chose 680 SLI/690 over 7970Ghz CF and now they will never admit that HD7000 series was way better than GK204. This type of brand loyal following is killing AMD's chances long-term even when they deliver superior products. Once again, we see R9 295X2 smashing 980 into the ground and today these 2 cards cost nearly the same......