The AMD Vega Memory Architecture Q&A With Jeffrey Cheng

Eager for AMD to launch the AMD Vega GPUs to take on the NVIDIA Pascal GPUs? Well, they are not quite here yet. But if you read our AMD Vega GPU Architecture Tech Report or saw our first look of DOOM running on AMD Vega, you know they are very close to the final silicon.

In the meantime, we have more details on AMD Vega from our coverage of the AMD Tech Summit. In this article, we will share with you our Q&A session with AMD Senior Fellow Jeffrey Cheng on his area of expertise – the AMD Vega memory architecture. Check it out!

AMD Vega Memory Architecture Q&A With Jeffrey Cheng


The AMD Vega Memory Architecture

Jeffrey Cheng is a AMD Senior Fellow whose expertise is in memory architectures. Now, the AMD Vega memory architecture here refers to how the AMD Vega GPU manages memory utilisation and handles large datasets. It does not deal with the memory hardware design, including the High Bandwidth Cache and HBM2 technology.

For those who prefer a summary of what Jeffrey talked about, here are the key takeaways :

The AMD Vega GPU Architecture Tech Report

  • Large amounts of DRAM can be used to handle big datasets, but this not the best solution.
  • Using large amounts of DRAM greatly increases cost and power consumption.
  • AMD chose to design a heterogenous memory architecture to support various memory technologies like HBM2 and even non-volatile memory (e.g. Radeon Solid State Graphics).
  • At any given moment, the amount of data processed by the GPU is limited, so it doesn’t make sense to store a large dataset in DRAM. It would be better to cache the data required by the GPU on very fast memory (e.g. HBM2), and intelligently move them according to the GPU’s requirements.
  • The AMD Vega’s heterogenous memory architecture allows for easy integration of future memory technologies like storage-class memory (flash memory that can be accessed in bytes, instead of blocks).
  • The AMD Vega has a 64-bit flat address space, giving it potentially up to 512 TB of addressable memory.

The AMD Vega GPU Architecture Tech Report

  • Game developers currently try to manage data and memory usage, often extremely conservatively to support graphics cards with limited amounts of graphics memory.
  • With the introduction of AMD Vega, AMD wants game developers to leave data and memory management to the GPU. Its High Bandwidth Cache Controller and heterogenous memory system will automatically handle it for them.

MORE : You can read more about AMD Vega in our articles – the AMD Vega GPU Architecture Tech Report, and Watch AMD Vega Run DOOM On Vulkan!


Support Tech ARP!

If you like our work, you can help support our work by visiting our sponsors, participating in the Tech ARP Forums, or even donating to our fund. Any help you can render is greatly appreciated!



About The Author

Related posts

Leave a Reply

%d bloggers like this: