No. We are talking about CPUs. Not GPUs. No clue what you are talking about with CPUs and Drivers.. GPUs don't use the multi thread or single threaded stuff. They just get the info from the CPUs. The multi/single threaded stuff is what causes the limitations seen with DX11. DX12 is almost entirely focused on the CPU not the GPU so what I was saying earlier was I wonder how much it would help amd's current FX lineup of CPUs and by how much, since DX12 is much much better with multithreading and better allocation of CPU time than DX11 and AMD's FX are very good with optimized multithreaded applications. Just remember the benefits you gain with FPS is because the CPU is being used more efficiently and can therefore better handle tasks and better at handing them off to the GPU to do more work quicker. It all stems from the CPU
Draw calls are all about CPU performance.
One of the best examples of where draw calls matter is the view distance setting in games.
It doesn't matter if you have a Titan X or a Fury X, if you pair it with an inadequate CPU, your performance will plummet in many titles when you crank up the view distance because you're hitting the limit of how many draw calls your CPU can handle.
If you monitor CPU usage, you will usually see that you are probably at or near 100% CPU usage on a single core, while your GPU is well below 100% usage.
Reducing the draw distance would lower CPU usage, lifting that bottleneck, allowing your card to hit 100% GPU usage again and giving you higher framerates.
Or if you upgraded to a CPU with faster single-thread performance, you would also see the GPU usage and game performance increase.
NVIDIA's DX11 drivers are multi-threaded and see an increase of more than 2x the number of draw calls they can handle on a multi-core system. (1.25M to 2.75M draw calls)
AMD's DX11 drivers do not support multi-threading and see no performance benefit whatsoever.
Both the single-threaded and multi-threaded DX11 tests have basically identical performance. (1M draw calls)
However if you look at the DX12/Mantle test results, the 290X is actually outperforming the 980 by 30% or thereabouts. (20M vs 15M draw calls)
So it's not that the card is slower, it's that their DX11 drivers are holding it back by being both less optimized, and by not supporting multi-threaded DX11 instructions.
If we assume that DX12 is hitting the limit of what both cards can do, single-threaded DX11 performance on the 290X would be around 1.6M draw calls if it was as optimized as NVIDIA's drivers, and adding multi-threaded DX11 support on top of that could have pushed that number above 3.5M draw calls in the multi-threaded test.
But with the current state of AMD's DX11 drivers being single-threaded, the only way to get good performance with AMD cards running DX11 titles is to brute-force it with strong single-threaded CPU performance, and that's where Intel is currently leading.
AMD's CPUs are more focused on having lots of weaker cores, than fewer strong ones.
DX12/Mantle/Vulkan should bring huge performance improvements to AMD cards when paired with AMD CPUs, since that will be the first time their drivers take advantage of all those cores.
But many game operations can't really be multi-threaded at all, so single-thread performance is always going to remain very important - even with DX12/Mantle/Vulkan.
Zen might hit the right balance between the number of cores and per-core performance for gaming in DX12/Vulkan at the typical enthusiast price-point (around £200) because I suspect you'll be getting a 6/8 core Zen CPU for the cost of a quad-core CPU from Intel, and DX12/Vulkan definitely benefits from having more than 4 cores.
But single-threaded performance is always going to matter for anything prior to DX12--
especially with an AMD GPU--and I suspect that you'll still want to go with an Intel chip for that.
It's the heading
