AMD has picked the crossroad where it's picking compute over gaming because they don't have the resources nor the architecture to tackle both gaming and compute both architectures.
AMD made a gamble where they hoped they could reduce investment, software development on R and D for the architecture because of the console market.
They hoped by coding for their architecture in consoles, they would have a natural advantage that would naturally carry over to the discrete gaming market.
What they underestimated was the amount of work to code for directx 12 from consoles and the bugginess of directx 12 made it unattractive to developers. There is no cheap port of directx 12 from consoles and most ports perform worse on directx 12 on both sides.
Generally the best directx 12 performing titles on AMD, run poorly on Nvidia which shows the weakness and closed nature of directx 12. Directx 12 was not designed to be a mutually beneficial open system. It was a trojan horse by AMD which was meant for GCN to florish while kepler would flounder if Nvidia stayed on the same architecture. Luckily for Nvidia, Directx 12 was problematic to code for an open environment with a wide variety of hardware configurations and was limited to Windows 10 systems. This made directx 11 the preferable path for developers where driver strength and savings for developers from reduced code allowed Nvidia to flourish because of a much stronger GPU software driver team which is reflected by their strong day 1 performance.
At this point AMD is in an ackward place. Directx 12 won't take off until it benefits Nvidia as much as AMD because they represent the majority of the market. And they don't have the man power to code the drivers for direct x 11 which is why day 1 performance tends to suffer on top of development teams coding for Nvidia since of their marketshare.
AMD can shine with directx 11 as games like call of duty shows but they don't have the resources or marketshare using directx 11 to go head to head with Nvidia in this space.
AMD has to make a decision where they invest more on software and marketing and essentially bribe software developers to make GCN the lead platform in the PC market. It's too late for this and AMD does not have the resources.
Less likely but if AMD wants to save money on software development is collaborate fully with Nvidia on making directx 13, where both companies mutually benefit, rather than just AMD mostly. This way, developers have incentives to use the platform because it benefits the whole market. Not just AMD.
Unfortunately for us, I think AMD is picking an exit strategy and intends to go full compute and abandon the PC discrete market to prevent getting squashed by Intel and Nvidia. Right now AMD appears to be following this road because they haven't given a damn about the gaming market with their hardware release. This is reflected by the lack of gaming focused cards and the bandaid solution at the moment of bundling games rather than releasing new hardware.
I don't blame them. 7nm is such an expensive node which doesn't leave room for 3 players to make money because there is not enough revenue to go around. Without mining, the discrete videocard market might be worth 6 billion dollars for gaming annually. With the cost of designs, wafers and the fierce competition from Nvidia, continuing to compete in the discrete market where an increasing more expensive market where the rewards are not worth the risk. A 6 billion dollar pie where you need 40% marketshare to make money vs a 50 billion dollar pie(the CPU market) where you just need 15-20% to make money. Intel represents a dark horse at the moment, but I have a feeling they will land microsofts next console. I know there are some ryzen/navi rumors out there but Intel unlike Nvidia can swallow their pride and give their products away to get a contract. With Apple moving away from Intel, Intel will have extra capacity at their fabs which they need to feed. In addition, with AMD closing in on Intel's CPU IPC advantage, Intel will need need need their graphics, particularly laptops to keep their marketshare. Getting a console is too important strategically and if I was Intel, I would be willing to spend billions to get it. Because it not only raises the chances of a successful Discrete launch, it weakens AMD and Nvidia who are biting into their markets. In addition, a console trojan horse from Intel has a strong chance to succeed since the incentive is there for developers.
Unlike AMD, Intel has the marketshare and money to get developers to code for their system on top of provide stronger software support through funds. If Intel is able to get something like a 4tflop into their laptops/discrete integrated by 2020-2021, with Intel strong chip sales, developers will have a strong reason to code for Intel as the lead platform since they represent 70% of the market overall.
https://www.extremetech.com/gaming/...pu-market-shifts-between-intel-amd-and-nvidia
That means, if they can land a console, it means very big things which puts uncomfortable pressure on Nvidia and squeezes out AMD altogether. I also feel the integrated graphic market more closely mirrors to closed ecosystem of consoles in terms of hardware meaning a low level API is more likely to work in this market.
Facebook
Twitter