Trio of new AMD GPU references for possible Mac refresh found in latest macOS Sierra beta
Users delving into the latest macOS 10.12.2 beta have discovered strings aligning with potential support for a trio of new GPU families in future hardware -- a refreshed Polaris 10, the unknown Polaris 12, and the Vega 10 series.
The code and new KEXT files associated with drivers were found in the latest macOS beta by the Hackintosh community, on the tonymacx86 forums. The last time something like this was discovered, it was the Polaris 11 chipset, which turned out to be the debut of the Radeon Pro 450, 455, and 460 in the Retina MacBook Pro with Touch Bar.
Speculation is running wild, based on little else than the codenames. The Polaris 10 is referred to as the 10XT2, and is probably a process improvement over the existing Polaris 10 series found in the Retina iMac given previous naming conventions.
Polaris 12 is completely unknown at this time.
Some details about the Vega 10 are known. The Vega 10 will be based on AMD's GCN architecture, and the sole offering currently utilizing the nomenclature is more commonly known as the "Fury Pro." At present, the Fury Pro is intended for desktop configurations, like the Mac Pro, and pushes 12 teraflops versus 9 teraflops with the Nvidia 1080 GPU, and has with 16GB of memory, and a 230W thermal design profile.
The code and new KEXT files associated with drivers were found in the latest macOS beta by the Hackintosh community, on the tonymacx86 forums. The last time something like this was discovered, it was the Polaris 11 chipset, which turned out to be the debut of the Radeon Pro 450, 455, and 460 in the Retina MacBook Pro with Touch Bar.
Speculation is running wild, based on little else than the codenames. The Polaris 10 is referred to as the 10XT2, and is probably a process improvement over the existing Polaris 10 series found in the Retina iMac given previous naming conventions.
Polaris 12 is completely unknown at this time.
Some details about the Vega 10 are known. The Vega 10 will be based on AMD's GCN architecture, and the sole offering currently utilizing the nomenclature is more commonly known as the "Fury Pro." At present, the Fury Pro is intended for desktop configurations, like the Mac Pro, and pushes 12 teraflops versus 9 teraflops with the Nvidia 1080 GPU, and has with 16GB of memory, and a 230W thermal design profile.
Comments
I don't understand this statement, the current iMacs do not use the Polaris 10 GPU. The Polaris 10 is the new generation which in fact replaces the M300 line found in the current iMacs.
Maybe this just means that Apple tests a wider range of hardware options than they actually release as products.
Luckily we have the "can't innovate my ass" 3+ year old MacPro available to show us how its done. Make sure you configure it with the $1000US (upgrade price) D700 that gets you a whopping 3.5 teraflops (you get two of them but you can count on 1 hand the number of software that supports dual cards). Meanwhile you can buy the 9 teraflop Nvidia 1080 for under $600 (full price). Oh yah... that's right, Apple does not allow us to upgrade graphics cards in our workstations. "can't innovate my ass"
Actually that is one of the reasons we have seen such a long period between updates. GPUs on 14nm processes just arrived late this summer and that was only a small selection of cards. Frankly the only way you can get decent performance per watt is to go with 14nm class chips as the GPU workd has been stagnet otherwise.
Actually not true. This is Metal that supports dual cards on behalf of the applications. Thanks to Metal, all the applications on macOS support dual GPUs, provided they code to Metal. The function of the Metal platform is to allow developers to "write to the metal", i.e. get most granular control on implementing the parallelism between GPU's and CPU. If your CPU waits until the GPU completes some shader then this is not very intelligent way to spend your money. In contrast if you distribute the task in a so granular fashion that all CPU, GPU1 and GPU2 are busy all the time then this is true parallel computing. In order to get that granularity you must "write to the metal" of the GPU. With Nvidia you cannot get that. Only AMD and the GPU in Apple's A series chips can be used that way.
If they put two of these in a Mac Pro, could that be cooled in the current trash can? Does this signify a change in form factor?