[bryanlarsen]

My theory on why Apple prefers AMD over nVidia:

AMD hardware at a specific price point is more powerful than nVidia hardware at the same price point. However, nVidia has superior drivers that eliminates the difference.

And since Apple prefers to use their own drivers, nVidia loses their main point of differentiation.

But of course the "Apple" drivers for video cards are basically vendor drivers with Apple doing QA & release management. But the driver is nVidia's secret sauce, they're not going to show it to Apple, since Apple is now a very competitive GPU manufacturer. At some point Apple will probably put their own GPU's inside Macs, so nVidia doesn't want to give them a head start.

AMD cares less about giving Apple a head start because they care more about the short term than the long and competing against nVidia.

[dogma1138]

AMD forked over the driver code, NVIDIA doesn’t let Apple handle the drivers and pushed CUDA which apple didn’t want anything to do with.

As for the performance per buck thing it’s not accurate. For gaming and 3D applications AMD has a huge bottleneck in thier geometry pipeline which causes the to underperform in games especially pre VEGA. NVIDIA also used tile rasterization and shader output caching on die to increase performance this is something that AMD adopted only with VEGA.

Pure flops don’t mean much even for compute.

[walshemj]

The new Vega boards lag nvidias 1080ti apart for some edge cases of interest to cryptocurrencies

[EdgarVerona]

Though to be fair, they also are significantly cheaper than the 1080ti. Well, if you can find one at MSRP at least.

[hotmail]

Coding with Vulkan directly seems to eliminate the nvidia perf advantage in many cases, so...is it still the hardware or the software lagging?

[joombaga]

Vs. coding for nvidia or vs. coding for DirectX independent of card-maker?

[lostboys67]

Quite they are all being snapped up by miners

[hotmail]

Looking at AMD’s server strategy, hey are going directly at NVidia, nvidia just doesn’t know it yet. They Epyc line is built for heavy PCIe comma and low comms overhead from NIC to GPGPU. Of course, guess which gpgpu works best with them?