Radeon Instinct MI6 

Released December, 2016
  • 1.1 GHz
  • 8 GB GDDR5

While CUDA 4 already had unified virtual addressing support which allowed x86 and GPU memory to be addressed together in a single space, it only simplified memory management and did not solve some issues that such a solution had and still required memory copying that will allow GPU to work with it.
by Slobodan-Simic (Nov, 2013)
The stated official performance figures are on par with the original Mantle figures, with Star Swarm showing a major performance increase and Battlefield 4 showing a small but significant performance increase.
by Syed-Muhammad-Usman-Pirzada (May, 2014)

Features Key features of the Radeon Instinct MI6

memory bandwidth Rate at which data can be read from or stored in onboard memory

Radeon Instinct MI6
224 GB/s
GeForce GTX 1080 Ti
484.4 GB/s
Radeon RX 480
256 GB/s

pixel rate Number of pixels a graphics card can render to the screen every second

Radeon Instinct MI6
39.5 GPixel/s
GeForce GTX 1080 Ti
139.2 GPixel/s
Radeon RX 480
40.5 GPixel/s

texture rate Speed at which a graphics card can perform texture mapping

Radeon Instinct MI6
177.6 GTexel/s
GeForce GTX 1080 Ti
354.4 GTexel/s
Radeon RX 480
182.3 GTexel/s

floating point performance How fast the gpu can crunch numbers

Radeon Instinct MI6
5,682 GFLOPS
GeForce GTX 1080 Ti
11,340 GFLOPS
Radeon RX 480
5,834 GFLOPS

shading units Subcomponents of the gpu, these run in parallel to enable fast pixel shading

Radeon Instinct MI6
2,304

texture mapping units Built into each gpu, these resize and rotate bitmaps for texturing scenes

Radeon Instinct MI6
144

In The News From around the web

The stated official performance figures are on par with the original Mantle figures, with Star Swarm showing a major performance increase and Battlefield 4 showing a small but significant performance increase.
Now this particular update, along with the various performance updates and bug fixes will include 2 major updates.
14
Nov

Nvidia announces new CUDA 6

by Slobodan Simic |
www.fudzilla.com
Nvidia also revealed that while we are currently seeing software implementation of unified memory with CUDA 6, Maxwell will most likely have some sort of hardware functionality for implementing unified memory thus removing any performance penalties and most likely doing it much better than software solution with CUDA 6.
While it does not resolve technical limitations that require memory copying, like limited bandwidth and latency of the PCIe, it moves memory management to CUDA rather than other toolkit, thus simply making the life of the programmer much easier.
14
Nov

NVIDIA Announces CUDA 6: Unified Memory for CUDA

by Ryan Smith |
www.anandtech.com
Kicking off next week will be the annual International Conference for High Performance Computing, Networking, Storage, and Analysis, better known as SC.
Maxwell will have some kind of hardware functionality for implementing unified memory (and presumably better performance for it), though it’s not something NVIDIA is talking about until Maxwell is ready for its full unveiling.
18
Apr

Intel Reports First-Quarter Revenue of $12.6 Billion

by Tarun Iyer |
www.tomshardware.com
We are working with our customers to introduce innovative new products across multiple operating systems.

Specifications Full list of technical specs

gpu

GPU brand AMD
GPU name Ellesmere
Clock speed 1,120 MHz
Turbo clock speed 1,233 MHz
Is dual GPU No
Reference card None

raw performance

Shading units 2,304
Texture mapping units 144
Render output processors 32
Compute units 36
Pixel rate 39.5 GPixel/s
Texture rate 177.6 GTexel/s
Floating-point performance 5,682 GFLOPS

memory

Memory clock speed 1,750 MHz
Effective memory clock speed 7,000 MHz
Memory bus 256 bit
Memory 8,192 MB
Memory type GDDR5
Memory bandwidth 224 GB/s

noise and power

TDP 150W
Report a correction

Comments

comments powered by Disqus