AMD Radeon Pro Vega 64X videocard review

AMD Radeon Pro Vega 64X

Radeon Pro Vega 64X videocard released by AMD; release date: 23 June 2017. The videocard is based on GCN 5.0 microarchitecture codenamed Vega 10.

Core clock speed - 1250 MHz. Boost clock speed - 1475 MHz. Texture fill rate - 377.6 GTexel/s. Pipelines - 4096. Manufacturing process technology - 14 nm. Transistors count - 12500 million. Power consumption (TDP) - 250 Watt.

Memory type: HBM2. Maximum RAM amount - 16 GB. Memory bus width - 2048 bit. Memory clock speed - 786 MHz (1572 MHz effective). Memory bandwidth - 402.4 GB/s.

Benchmarks

GFXBench 4.0
Car Chase Offscreen
Top 1 GPU
This GPU
34770 Frames
13123 Frames
GFXBench 4.0
Car Chase Offscreen
Top 1 GPU
This GPU
34770.000 Fps
13123.000 Fps
GFXBench 4.0
Manhattan
Top 1 GPU
This GPU
27823 Frames
3715 Frames
GFXBench 4.0
Manhattan
Top 1 GPU
This GPU
27823.000 Fps
3715.000 Fps
GFXBench 4.0
T-Rex
Top 1 GPU
This GPU
69225 Frames
3357 Frames
GFXBench 4.0
T-Rex
Top 1 GPU
This GPU
69225.000 Fps
3357.000 Fps
Geekbench
OpenCL
Top 1 GPU
This GPU
352097
75438
PassMark
G2D Mark
Top 1 GPU
This GPU
1306
948
PassMark
G3D Mark
Top 1 GPU
This GPU
38733
14631
Name Value
GFXBench 4.0 - Car Chase Offscreen 13123 Frames
GFXBench 4.0 - Car Chase Offscreen 13123.000 Fps
GFXBench 4.0 - Manhattan 3715 Frames
GFXBench 4.0 - Manhattan 3715.000 Fps
GFXBench 4.0 - T-Rex 3357 Frames
GFXBench 4.0 - T-Rex 3357.000 Fps
Geekbench - OpenCL 75438
PassMark - G2D Mark 948
PassMark - G3D Mark 14631

Specifications (specs)

Essentials

Architecture GCN 5.0
Code name Vega 10
Launch date 23 June 2017
Place in performance rating 172

Technical info

Boost clock speed 1475 MHz
Compute units 64
Core clock speed 1250 MHz
Manufacturing process technology 14 nm
Peak Double Precision (FP64) Performance 755.2 GFLOPS
Peak Half Precision (FP16) Performance 24.17 TFLOPS
Peak Single Precision (FP32) Performance 12.08 TFLOPS
Pipelines 4096
Pixel fill rate 94.40 GPixel/s
Texture fill rate 377.6 GTexel/s
Thermal Design Power (TDP) 250 Watt
Transistor count 12500 million

Video outputs and ports

Display Connectors No outputs

Compatibility, dimensions and requirements

Interface PCIe 3.0 x16
Supplementary power connectors None

API support

DirectX 12.1
OpenCL 2.0
OpenGL 4.6
Shader Model 6.4
Vulkan

Memory

High bandwidth memory (HBM)
Maximum RAM amount 16 GB
Memory bandwidth 402.4 GB/s
Memory bus width 2048 bit
Memory clock speed 786 MHz (1572 MHz effective)
Memory type HBM2

Technologies

Unified Video Decoder (UVD)
Video Code Engine (VCE)