Back in 2016 AMD introduced their new lineup of Radeon GPU accelerator known as Radeon Instinct. These new Radeon Instinct GPU accelerators are targeted towards handling deep learning and machine intelligence tasks. Where AMD will be targeting directly with what NVIDIA have to offer with their Tesla family. Which is currently being used in the world most powerful supercomputers.
Now AMD Radeon Instinct lineup will features three types of GPU’s. Which includes the lower end Radeon Instinct M16 based on Polaris architecture, the mid end Radeon Instinct M18 based on Fiji. And finally the top of the line Radeon Instinct M125, which features 24.6 TFlops of computing power.
According to AMD their Radeon Instinct GPU accelerators will offer performance in both single and double precision scenarios. Now with the launch of Radeon Instinct AMD also revealed their EYPC 7000 series server grade chips. This is what AMD have been working hard for as their EYPC 7000 series chips have support for up to 128 PCIe Express lanes. Aside from that both the platforms from AMD can also link up with each other, without the need for a PICe Express switch. A big plus point that AMD has over Nvidia right now.
Though what is the most intersecting aspect of these two platforms is about its performance. Where according to AMD four of their top end Radeon Instinct M125 GPU’s, coupled with AMD EYPC server. Can perform as well as the human brain in computing power. This is not where AMD stops as in the future AMD will also be providing a rack. That can support up to 125 Radeon Instinct M125 GPU’s. Offering a mind boggling 3 PetaFlops of computing power.
Though for companies who can’t wait, AMD will be providing their Falconwitch server. That can support up to 16 Radeon Instinct GPU accelerators. Which can provide its users with 400 TFlops of computing power.