… however, the GPU curiously lacks hardware support for DirectX 9.
I find it surprising that this misconception still prevails and that journalists don't see through Intel's BS excuses for not supporting DirectX 9.
If they can support DirectX 9 through a translation to DirectX 12,
then they do have the underlying hardware features they need.
It should be basic knowledge that none of the current GPUs support either DirectX 12, OpenGL or Vulkan directly, it is the responsibility of the GPU driver to manage the API's state and translate operations into the GPU's native API. A GPU can support any API which can be translated into its hardware capabilities, which is why GPUs have no problem supporting API features released long after the GPU (e.g. Vulkan on Kepler).
While Arc GPUs lack DirectX 9 support, foolproof API translation technologies exist, which convert DirectX 9 API instructions into DirectX 12. This is not fundamentally unlike how 32-bit applications work on 64-bit Windows (using WOW64 machine-architecture translation).
This is not comparable at all. x86-64 is basically a superset of x86-32, and it doesn't involve emulating a complex state machine on top of another.
What hw support? That shtick came from MooreLawIsDead/wccftech,
Anyone surprised?
TPU should really check their sources better, even indirect sources. Sometimes people repeat false claims until others just assume them to be true, that may be the case here.
While my BS alarm goes off immediately when I hear people making such claims, others shouldn't have to know graphics programming to see through this. Anyone deep into tech show know that the driver's responsibility is to translate graphics APIs into the GPU's native API and manage the API's internal state.
I was also disappointed back when Intel had their "ARC apology tour", visiting major reviewers/content creators like LTT, GN and PCWorld, and they all failed to reveal Intel's marketing BS and excuses, not just for DirectX 9, but the obvious bogus claims like haven't had time to mature, and the nonsense that newer DirectX 12 games performed better on ARC (when the data actually showed otherwise). While these guys obviously spend thousands of hours benchmarking and overclocking, they could benefit a lot from learning more about how GPUs actually work and how games work. This would not only help them see through marketing BS and excuses from the makers, but also help determine what kind of issues they are observing, and obviously make better content for their viewers.
I'm not really into the hole architecture and instruction sets of GPUs but I'm pretty sure fixed hw blocks have nothing to do with DirectX9 implementation.
It's nearly two decades since GPUs were mostly fixed function. Programmable GPUs are backwards compatible in the sense that it's easy for the driver to implement the "fixed function features" through programmable shaders etc. And just simple logical deduction would make this obvious; OpenGL is still supported and have way more legacy features than DirectX 9.
Their drivers suck and don't have years of experience implementing high performance Dx9 acceleration so they went the "throw the bucket at microsoft" route with translation - this and it's dis/advantages were already discussed pretty much ad nauseum.
They actually had support in their old driver, but apparently found it too much effort to rewrite it for the new GPU architecture.
But otherwise you pretty much nailed it; they just didn't want to spend the resources, and blame someone else when it inevitably fails.
No matter how much manure they spread on top of it, the underlying facts will remain; they are emulating one complex API on top of another very different and complex API, there will be a lot of glitches and slowdowns, and it will never be close to 100%. The best case to hope for would be something comparable to Wine in Linux.
