Наконец, третьим фактором успеха следующих микроархитектур NVIDIA станет технология виртуальной памяти. Причём виртуальная память будет реализована уже в следующем году в микроархитектуре Maxwell. Заметим, в компании NVIDIA считают ошибочным путь AMD, которая добивается унификации доступа к памяти с помощью когерентности кэша. Такой путь, заявляет Далли, ведёт к генерированию дополнительного трафика, что может привести к эффекту бутылочного горлышка. В NVIDIA будут реализовать другой подход. Для предоставления параллельного доступа к памяти центральному процессору и графическому процессору будет создан единый пул с использованием указателей и исключений.
In CUDA 6, Unified Memory is supported starting with the Kepler GPU architecture (Compute Capability 3.0 or higher), on 64-bit Windows 7, 8, and Linux operating systems (Kernel 2.6.18+)
kl
- Very nice from CUDA 6. Really eager to get started with this. Just one question; does unified memory reads both PC and GPU as one combined memory or is still seen as seperate and CUDA automatically does all the copying instead of the programmer?

Mark Harris
- On current hardware, the latter -- in a PC today the GPU and CPU memories are physically discrete. But the same programming model could be used for physically unified memory systems, such as Tegra SOCs.

Vincent Roberge
- Will CUDA Unified Memory be supported on GTX cards in Windows 7 and 8 or will it be limited to Tesla cards (due to requirement for TCC driver)? I am really looking forward to using Unified Memory in my CUDA applications, but do not want to limit my clients to using Tesla cards only.

Mark Harris
- Unified Memory will be supported on Compute Capability 3.0 and later (sm_30 - so K10 or GTX 680 or later), on 64-bit Linux, Windows 7 and Windows 8 in CUDA 6.0. Support for other operating systems will come at a later date

Unified Memory creates a pool of managed memory that is shared between the CPU and GPU, bridging the CPU-GPU divide. Managed memory is accessible to both the CPU and GPU using a single pointer. The key is that the system automatically migrates data allocated in Unified Memory between host and device so that it looks like CPU memory to code running on the CPU, and like GPU memory to code running on the GPU.