With the introduction of the RX 6000 and the “spontaneous combustion” on the part of all those attending the online event, it was clear that AMD is going to fight with NVIDIA for the throne of performance. One of the features that was presented as a novelty was AMD Smart Access Memory or SAM, a technology that promises to increase the performance of the GPU by an interesting percentage, but what exactly is it and how does it work?
With Smart Access Memory AMD has a trump card to displace Intel from the high performance CPU market and at the same time could make a difference with NVIDIA on GPU. The problem is that this technology is quite hidden in terms of how it works technically, so there is no document or White Paper to extract official information as such, so we will have to play with speculation a little and update this article when we know for sure how it works, so that you have first-hand information.
AMD Smart Access Memory, up to 11% more performance
As expected, a series of parameters and requirements are needed for SAM to work and to increase the performance of the GPU and therefore to offer more FPS. What do we need specifically? According to AMD:
- A 500 series motherboard .
- A Ryzen 5000 processor.
- One RX 6000 series GPU.
- Radeon driver 20.11.2 or later.
- Have the board’s BIOS updated to one based on AGESA 1.1.0.0 or higher.
With this in mind we will proceed to explain how this technology works and what Smart Access Memory can do for us. First of all, we have to bear in mind that in current gaming systems the CPU only has access to the VRAM of the GPU in a very curious way: in 256 MB blocks , where you can only see the first 256 MB of it. .
It doesn’t take a scholar to understand that this forces programmers to adapt their game engines to this system, thereby creating a very high bottleneck at higher resolution and Hertz.
AMD eliminates bottleneck, but only on its X570 platform
The 256 MB blocks granted for the CPU are first a waste of the speed of the PCIe bus, and secondly a fractionation of the textures that does not make sense and therefore forces the vast majority to be stored in RAM, requiring a lot of quantity of the latter and very fast to alleviate the opposite effect.
This type of access slows down performance greatly, both in timing itself and in total bandwidth, so to solve it, specific hardware is needed that connects directly and without restrictions between them, hence the limitations in the series of processors, boards and GPU.
How much return can we earn? According to AMD up to 11% on average , but it specifies it better in certain games:
What is it we don’t know for sure? Well, many questions remain on the table, such as:
- What improvement will we have at 1080p and high FPS?
- Will we lose performance with riser cables?
- Is PCIe 4.0 essential?
- Could it be used in other chipsets as long as we have a Ryzen 5000 CPU and a RX 6000 GPU at the very least?
- Are we facing an improved Remote Direct Memory Access or is there something else?
Well, we will have to find out later, but for now it requires hardware and software for it to work and it is an interesting and totally free improvement.
