AMD Instinct

Short description: Brand name by AMD; professional GPUs for high-performance-computing, machine learning

AMD Instinct

Release date	June 20, 2017 (2017-06-20)
Architecture	GCN 3rd gen GCN 4th gen GCN 5th gen CDNA CDNA 2
Models	MI Series
Transistors	5.7B (Polaris10) 14 nm 8.9B (Fiji) 28 nm 12.5B (Vega10) 14 nm 13.2B (Vega20) 7 nm 25.6B (Arcturus) 7 nm 58.2B (Aldebaran) 6 nm
Fabrication process	TSMC 28 nm GloFo 14 nm TSMC 7 nm TSMC 6 nm
History
Predecessor	AMD FirePro Radeon Sky series

AMD Instinct is AMD's brand of professional GPUs. It replaced AMD's FirePro S brand in 2016. Compared to the Radeon brand of mainstream consumer/gamer products, the Instinct product line is intended to accelerate deep learning, artificial neural network, and high-performance computing/GPGPU applications.

The Radeon Instinct product line directly competes with Nvidia's Ampere and Intel Xeon Phi and incoming Intel Xe lines of machine learning and GPGPU cards.

Before MI100 introduction in November 2020, the Instinct family was known as AMD Radeon Instinct, AMD dropped the Radeon brand from its name.

Supercomputers based on (AMD CPUs and) AMD Instinct GPUs now take the lead on the Green500 supercomputer list with over 50% lead over any other, and top the first 4 spots, including the second, which is the current fastest in the world on the TOP500 list, Frontier.

Products

The three initial Radeon Instinct products were announced on December 12, 2016, and released on June 20, 2017, with each based on a different architecture.

MI6

The MI6 is a passively cooled, Polaris 10 based card with 16 GB of GDDR5 memory and with a <150 W TDP. At 5.7 TFLOPS (FP16 and FP32), the MI6 is expected to be used primarily for inference, rather than neural network training. The MI6 has a peak double precision (FP64) compute performance of 358 GFLOPS.

MI8

The MI8 is a Fiji based card, analogous to the R9 Nano, and expected to have a <175W TDP. The MI8 has 4 GB of High Bandwidth Memory. At 8.2 TFLOPS (FP16 and FP32), the MI8 is marked toward inference. The MI8 has a peak (FP64) double precision compute performance 512 GFLOPS.

MI25

The MI25 is a Vega based card, utilizing HBM2 memory. The MI25 performance is expected to be 12.3 TFLOPS using FP32 numbers. In contrast to the MI6 and MI8, the MI25 is able to increase performance when using lower precision numbers, and accordingly is expected to reach 24.6 TFLOPS when using FP16 numbers. The MI25 is rated at <300W TDP with passive cooling. The MI25 also provides 768 GFLOPS peak double precision (FP64) at 1/16th rate.

MI300 Series

The MI300A and MI300X are data center accelerators that use the CDNA 3 architecture, which is optimized for high-performance computing (HPC) and generative artificial intelligence (AI) workloads. The CDNA 3 architecture features a scalable multi-chip module (MCM) design that leverages TSMC’s advanced packaging technologies, such as CoWoS (chip-on-wafer-on-substrate) and InFO (integrated fan-out), to combine multiple chiplets on a single interposer. The chiplets are interconnected by AMD’s Infinity Fabric, which enables high-speed and low-latency data transfer between the chiplets and the host system.

The MI300A is an accelerated processing unit (APU) that integrates 24 Zen 4 CPU cores with four CDNA 3 GPU cores, resulting in a total of 228 CUs in the GPU section and 128 GB of HBM3 memory. The Zen 4 CPU cores are based on the 5 nm process node and support the x86-64 instruction set, as well as AVX-512 and BFloat16 extensions. The Zen 4 CPU cores can run general-purpose applications and provide host-side computation for the GPU cores. The MI300A has a peak performance of 61.3 TFLOPS of FP64 (122.6 TFLOPS FP64 matrix) and 980.6 TFLOPS of FP16 (1961.2 TFLOPS with sparsity), as well as 5.3 TB/s of memory bandwidth. The MI300A supports PCIe 5.0 and CXL 2.0 interfaces, which allow it to communicate with other devices and accelerators in a heterogeneous system.

The MI300X is a dedicated generative AI accelerator that replaces the CPU cores with additional GPU cores and HBM memory, resulting in a total of 304 CUs and 192 GB of HBM3 memory. The MI300X is designed to accelerate generative AI applications, such as natural language processing, computer vision, and deep learning. The MI300X has a peak performance of 653.7 TFLOPS of TP32 (1307.4 TFLOPS with sparsity) and 1307.4 TFLOPS of FP16 (2614.9 TFLOPS with sparsity), as well as 5.3 TB/s of memory bandwidth. The MI300X also supports PCIe 5.0 and CXL 2.0 interfaces, as well as AMD’s ROCm software stack, which provides a unified programming model and tools for developing and deploying generative AI applications on AMD hardware.

Accelerator	Architecture	Lithography	Compute Units	Memory	Memory Type	PCIe Support	Form Factor	FP16 Performance	BF16 Performance	FP32 Performance	FP32 Matrix Performance	FP64 Performance	FP64 Matrix Performance	INT8 Performance	INT4 Performance	TBP Peak
MI6	GCN 4	14 nm	36	16 GB	GDDR5	3.0	PCIe	5.7 TFLOPS	N/A	5.7 TFLOPS	N/A	358 GFLOPS	N/A	N/A	N/A	150 W
MI8	GCN 3	28 nm	64	4 GB	HBM			8.2 TFLOPS		8.2 TFLOPS		512 GFLOPS				175 W
MI25	GCN 5	14 nm	64	16 GB	HBM2			26.4 TFLOPS		12.3 TFLOPS		768 GFLOPS				300 W
MI50	GCN 5	7 nm	60			4.0		26.5 TFLOPS		13.3 TFLOPS		6.6 TFLOPS		53 TOPS		300 W
MI60	GCN 5		64	32 GB				29.5 TFLOPS		14.7 TFLOPS		7.4 TFLOPS		59 TOPS		300 W
MI100	CDNA		120					184.6 TFLOPS	92.3 TFLOPS	23.1 TFLOPS	46.1 TFLOPS	11.5 TFLOPS		184.6 TOPS		300 W
MI210	CDNA 2	6 nm	104	64 GB	HBM2e			181 TFLOPS		22.6 TFLOPS	45.3 TFLOPS	22.6 TFLOPS	45.3 TFLOPS	181 TOPS		300 W
MI250			208	128 GB			OAM	362.1 TFLOPS		45.3 TFLOPS	90.5 TFLOPS	45.3 TFLOPS	90.5 TFLOPS	362.1 TOPS		560 W
MI250X			220					383 TFLOPS		47.92 TFLOPS	95.7 TFLOPS	47.9 TFLOPS	95.7 TFLOPS	383 TOPS		560 W
MI300A	CDNA 3	6 & 5 nm	228	128 GB	HBM3	5.0	APU SH5 socket	980.6 TFLOPS 1961.2 TFLOPS (With Sparsity)		122.6 TFLOPS		61.3 TFLOPS	122.6 TFLOPS	1961.2 TOPS 3922.3 TOPS (With Sparsity)	N/A	550 W 760 W (Liquid Cooling)
MI300X			304	192 GB			OAM	1307.4 TFLOPS 2614.9 TFLOPS (With Sparsity)		163.4 TFLOPS		81.7 TFLOPS	163.4 TFLOPS	2614.9 TOPS 5229.8 TOPS (With Sparsity)	N/A	750 W

Software

ROCm

Following software is, as of 2022, regrouped under the Radeon Open Compute meta-project.

MxGPU

The MI6, MI8, and MI25 products all support AMD's MxGPU virtualization technology, enabling sharing of GPU resources across multiple users.

MIOpen

MIOpen is AMD's deep learning library to enable GPU acceleration of deep learning. Much of this extends the GPUOpen's Boltzmann Initiative software. This is intended to compete with the deep learning portions of Nvidia's CUDA library. It supports the deep learning frameworks: Theano, Caffe, TensorFlow, MXNet, Microsoft Cognitive Toolkit, Torch, and Chainer. Programming is supported in OpenCL and Python, in addition to supporting the compilation of CUDA through AMD's Heterogeneous-compute Interface for Portability and Heterogeneous Compute Compiler.

Chipset table

Model (codename)	Release Date	Architecture & Fab	Transistors & Die Size	Core		Fillrate		Processing power (GFLOPS)			Memory				TBP	Bus interface
				Config	Clock (MHz)	Texture (GT/s)	Pixel (GP/s)	Half	Single	Double	Bus type & width	Size (GiB)	Clock (MT/s)	Bandwidth (GB/s)
Radeon Instinct MI6 (Polaris 10)	2016	GCN 4th gen 14 nm	5.7×109 232 mm2	2304:144:32 36 CU	1120 1233	177.6	39.46	5800	5800	358	GDDR5 256-bit	16	7000	224	150 W	PCIe 3.0 x16
Radeon Instinct MI8 (Fiji XT)		GCN 3rd gen 28 nm	8.9×109 596 mm2	4096:256:64 64 CU	1000	256.0	64.0	8200	8200	512	HBM 4096-bit	4	1000	512	175 W
Radeon Instinct MI25 (Vega 10 XT)		GCN 5th gen 14 nm	12.5×109 510 mm2	4096:256:64 64 CU	1400 1500	384	96.0	24600	12300	768	HBM2 2048-bit	16	1704	436.2	300 W
Radeon Instinct MI25 mxgpu (Prototype, 2017)	Unreleased		2× 12.5×109 510 mm2	2× 4096:256:64 64 CU	1400 1500	2× 384	2× 96.0	2× 24600	2× 12300	2× 768		2× 16		2× 436.2	300 W
Radeon Instinct MI50 (Vega 20 GL)	2018	GCN 5th gen 7 nm	13.2×109 331 mm2	3840:240:- 60 CU	1450 1746	419.04	-	26800	13400	6700	HBM2 4096-bit	16	2000	1024	300 W	PCIe 4.0 x16
Radeon Instinct MI60 (Vega 20 GL)				4096:256:- 64 CU	1500 1800	460.8	-	29450	14725	7362.5		32		1024	300 W

See also

• ROCm - AMD's open compute software stack
• AMD FirePro - AMD's predecessor to Radeon Instinct
• AMD Radeon Pro - AMD's workstation graphics and GPGPU solution
• Nvidia Quadro - Nvidia's competing workstation graphics solution
• Nvidia Tesla - Nvidia's competing GPGPU solution
• Xeon Phi - Intel's competing massively-parallel multicore processor line
• List of AMD graphics processing units

External links

• AMD Instinct accelerators web page