AMD Addresses Potential Fiji 4GB HBM Capacity Concern - Investing In More Efficient Memory Utilization

AMD's CTO of Calculating and Graphics Joe Macri addressed concerns around potential memory chapters limitations with HBM, especially on loftier end GPUs. As AMD's upcoming ultra enthusiast GPU code named FIji is rumored to characteristic 4GB of memory. The same capacity as Nvidia's electric current high end GTX 900 series offerings and AMD'southward last generation Hawaii ( R9 290 series ) graphics cards.
AMD HBM 12
The potential business organisation is that on such a powerful GPU as Fiji is rumored to be, 4GB of memory may pose a limitation. Fiji is AMD's upcoming flagship ultra enthusiast class GPU. Information technology 'due south also the globe'southward first GPU to leverage stacked High Bandwidth Memory.
The logic is that the faster the GPU, the more than likely that users will push the graphical settings of a game. And thus more memory chapters will be utilized to return the effects of these higher in-game graphical settings. This however isn't the issue hither because 4GB of memory is really quite plenty. But a question effectually the possibility of 4GB posing an effect at very high resolutions such as 4K and high resolution multi-monitor setups can arise.

AMD's Joe Macri Addresses Potential 4GB HBM Capacity Business With High End GPUs Such As Republic of the fiji islands

Arstechnica managed to get Joe Macri on the telephone to talk more well-nigh HBM. And in his comments Macri addressed multiple aspects of the concern mentioned in a higher place. Is HBM express to 4GB on loftier end GPUs ? and if so will 4GB be enough ? The short answers are no and aye. Just let'due south hear what Joe Macri had to say for a more comprehensive view.

"Y'all're not limited in this world [die stacking globe] to any number of stacks, but from a chapters point of view, this generation-one HBM, each DRAM is a two-gigabit DRAM, so yep, if yous take four stacks you lot're limited to iv gigabytes. Yous could build things with more stacks, you could build things with less stacks. Capacity of the frame buffer is just one of our concerns. There are many things y'all tin can do to utilise that capacity better. So if you have iv stacks y'all're limited to four [gigabytes], but nosotros don't really view that as a performance limitation from an AMD perspective."

For obvious reasons Macri did not reveal the actual memory capacity of Fiji. Specifications will come once the GPU is launched which AMD says you'll be able to buy this summer. Macri pointed out that in that location is no 4GB limitation. Products can be designed with as many memory stacks as is desired. Note, with "stacks" Macri is referring to HBM cubes so to speak. Because each cube is still express to 4 DRAM dies, each 256MB large, layered on-top of ane another. What is not limited all the same, is the number of cubes / stacks yous want to add to the pattern.

With 4, 256MB dies, you get a total of 1GB of retentiveness with each HBM1 stack. Four of these would internet 4GB of retention. For college capacities more HBM stacks can be arranged around the host processor, in the case it'southward the Fiji GPU.

Macri also addressed the second question, which is if 4GB is all nosotros had ,is it going to be enough ? Macri says yes.

"If you lot really await at frame buffers and how efficient they are and how efficient the drivers are at managing capacities beyond the resolutions, y'all'll find that there's a lot that can exist done. We do not run across 4GB as a limitation that would cause performance bottlenecks. We simply need to exercise a better task managing the capacities. We were getting free capacity, because with [GDDR5] in social club to go more bandwidth we needed to brand the memory arrangement wider, and so the capacities were increasing. Every bit engineers, we always focus on where the bottleneck is. If y'all're getting capacity, you don't put as much endeavor into improve utilising that capacity. 4GB is more than than sufficient. We've had to get do a little flake of investment in order to amend utilise the frame buffer, but we're not really seeing a frame buffer capacity [problem]. You'll be blown away by how much [capacity] is wasted."

According to Macri, GDDR5 fed GPUs actually take likewise much unused memory today. Because to increase GPU retention bandwidth, wider memory interfaces are used. And considering wider memory interfaces crave a larger amount of GDDR5 memory chips, GPUs ended up with more than memory chapters than is actually needed.
Macri also stated that AMD invested a lot into improving utilization of the frame buffer. This could include on-die memory compression techniques which are integrated into the GPU hardware itself. Or more clever algorithms on the driver level.

We have no doubt that one time the GPU is released, we'll see very extensive tests of the memory arrangement. Both to analyze the issue of the huge increase in memory bandwidth that is enabled by HBM has on operation. And to discover if the retentiveness capacity will be as sufficient as Joe Macri believes. The truth is that nosotros haven't seen a unmarried GPU however that can actually make use of 4GB all on its own all the while maintaining playable framerates. You lot'll simply run out of performance mode before you run out of memory.

Multi-GPU setups of course are the exception. You'll actually accept enough functioning to turn the settings up and fill up the memory buffer while maintaining playable framerates. And that'due south when it may very well become a limitation. Fortunately however that's where low level APIs such equally DX12 and Vulkan may prove to be the solution. With advanced multi-GPU rendering techniques, the retentivity buffers of multiple GPUs tin can be combined to form a larger memory pool that can exist utilized and that would solve the outcome.

There are and then many moving parts in the GPU industry today that are sure to exist important in the future. A new memory standard, multiple new low level APIs and diverse rendering innovations practice bring back much needed excitement to this infinite.