vacancies advertise contact news tip The Vault
facebook rss twitter

AMD reveals Radeon Instinct MI25, a Vega GPU accelerator

by Mark Tyson on 22 June 2017, 10:02

Tags: AMD (NYSE:AMD)

Quick Link: HEXUS.net/qadiws

Add to My Vault: x

At the end of the AMD Epyc 'New Era in the Datacenter' presentation, CEO Dr Lisa Su came on stage to talk about GPUs. Of course the talk was about GPUs being used for machine intelligence and other IT data centre processing growth areas. With this established Dr Su launched the Radeon Instinct line of accelerators, starting with the MI25. Following the launch event AMD has uploaded its Radeon Instinct MI25 product pages providing an overview of performance, features, and specifications.

This accelerator is claimed by AMD to be the "world’s fastest training accelerator for machine intelligence and deep learning". Key performance specifications of the Radeon Instinct MI25 can be seen below in a screenshot from the press deck.

  • GPU Architecture: 14nm FinFET AMD Vega10
  • Stream Processors: 4,096
  • GPU Memory: 16GB HBM2
  • Memory Bandwidth: Up to 484 GB/s
  • Performance: Half-Precision (FP16) 24.6 TFLOPS, Single-Precision (FP32) 12.3 TFLOPS, Double-Precision (FP64) 768 GFLOPS
  • ECC: Yes
  • MxGPU Capability: Yes
  • Board Form Factor: Full-Height, Dual-Slot
  • Length: 10.5”
  • Thermal Solution: Passively Cooled
  • Standard Max Power: 300W TDP
  • OS Support: Linux 64-bit
  • ROCm Software Platform: Yes
  • Programming Environment: ISO C++, OpenCL™, CUDA (via AMD’s HIP conversion tool) and Python5 (via Anaconda’s NUMBA)

The Radeon Instinct MI25 wasn't the only member of the family unveiled. AMD also launched the Radeon Instinct MI8 and Radeon Instinct MI6 based upon its older GPU architectures. For example the MI8 is based upon a Fiji GPU and made for SFF HPC providing 8.2 TFLOPS of peak FP16|FP32 performance at less than 175W board power and 4GB of High-Bandwidth Memory (HBM) on a 512-bit memory interface. The MI6 is based upon a Polaris GPU with 5.7 TFLOPS of peak FP16|FP32 performance at 150W peak board power and 16GB of ultra-fast GDDR5 GPU memory on a 256-bit memory interface.



HEXUS Forums :: 5 Comments

Login with Forum Account

Don't have an account? Register today!
Seems like the name is causing some issues here: First it's “SU” and then “Du” instead of “Su”
;-)
HollyDOL
Seems like the name is causing some issues here: First it's “SU” and then “Du” instead of “Su”
;-)
Machine Intelligence at work… :P
AMD is comparing with Nvidia P instead of V, on the CPU side you rock but GPUS? Sorry folks
lumireleon
AMD is comparing with Nvidia P instead of V, on the CPU side you rock but GPUS? Sorry folks

Well, that's a bit unfair. You can't compare against something that doesn't exist. Volta may have been announced, has many “here's what it can do” pages but there is no physical product yet.

As an aside, we don't really know how powerful this GPU is but it's supposedly the world's fastest accelerator, which suggests it's faster than P100 in some tasks, and more importantly they believe it is noticeably more power efficient. When was the last time Radeon was more power efficient? Plus, the gaming versions are hinted to be faster than Frontier and will likely have higher clockrates than MI25, too.

Now that I've defended this, it is time to confess I am also disappointed that there aren't more examples of MI25 outperforming Pascal somehow.
Ozaron
Well, that's a bit unfair. You can't compare against something that doesn't exist. Volta may have been announced, has many “here's what it can do” pages but there is no physical product yet.
Agreed. Only way AMD could have legitimate comparisons against Volta would be if they had insider, pre-release access to NVidia's future hardware. So err, yeah.


Ozaron
As an aside, we don't really know how powerful this GPU is but it's supposedly the world's fastest accelerator, which suggests it's faster than P100 in some tasks, and more importantly they believe it is noticeably more power efficient. When was the last time Radeon was more power efficient? Plus, the gaming versions are hinted to be faster than Frontier and will likely have higher clockrates than MI25, too.

Now that I've defended this, it is time to confess I am also disappointed that there aren't more examples of MI25 outperforming Pascal somehow.
AMD have been noticeably picky and cagey about pre-release benchmark publications lately.

That said, we know the TDP and it's equal to that of the Tesla P100, and we know they're claiming it's ~10% more powerful than P100, so there's fairly decent ballpark figures. As far as compute goes, it's easier to predict relative performance anyway; it'll certainly be competitive with, if not outright outperforming Pascal.