AMD officially reveal the Radeon Vega’s specifications

AMD officially reveal the Radeon Vega's specifications

AMD officially reveal the Radeon Vega’s specifications

 
We have all been waiting a long time for this, AMD’s official Radeon Vega specifications, which will give is a fairly clear picture of how well AMD’s new flagship will perform in a variety of applications. 
 
To start off, we now know that Vega has a total of 64 compute units and 4096 Stream Processors and peak FP32 performance of around 13TFLOPs, which is over 50% higher than AMD’s Fiji-based R9 Fury X. 

  

AMD officially reveal the Radeon Vega's specifications

 

In terms of raw processor count, the Radeon Vega is a lot like the R9 Fury X, offering a stream processor count of 4096, though Vega thanks to architectural changes and process node changes/optimisation will be able to offer increased clock speeds, giving AMD’s Radeon Vega around 50% more FP32 compute performance. 

AMD’s Radeon Vega also supports Rapid Packed Math and FP16 instructions, allowing this GPU to offer even more compute performance when less precision is required. 

Given the RX Vega’s stated 13 TFLOPs of compute performance and 4096 stream processors, AMD’s Radeon Vega Frontier Edition will offer boost clock speeds that are around 1587MHz, which is very similar to recent leaks which have suggested boost clock speeds of 1600MHz. 

One interesting thing is that the AMD Radeon Vega will offer slightly lower levels of memory bandwidth than the R9 Fury X though, in reality, Vega will offer a lot more memory performance thanks to several optimisations in the memory pipeline and the addition of AMD’s new Draw Stream Binning Rasterizer (DSBR) to Vega. These changes will allow AMD to do more work with the memory bandwidth that they have available, allowing them to have more effective memory bandwidth.  

 

  RX 570 RX 580 R9 Fury X Radeon Vega
Frontier Edition
GPU Polaris 10 Pro Polaris 10 XT Fiji XT Vega 10
Process node 14nm 14nm 28nm 14nm
Shade Engines 4 4   4
CUs per shader Engine 8 9   16
Shaders Per CU 64 64 64 64
Stream Processors 2048 2304 4096 4096
Performance (FP32) 5.095 TFLOPs 6.175 TFLOPs 8.6TFLOPs 13 TFLOPs
Performance (FP16) 5.095 TFLOPs 6.175 TFLOPs 8.6TFLOPs 25 TFLOPs
Render Output Units 32 32 64 64
Texture Mapping units 126 144 256 256
Hardware Threads 4 4 8
Memory Interface 256-bit 256-bit 4096-bit 2048-bit
Memory 4/8GB GDDR5 4/8GB GDDR5 4GB HBM 16GB HBM2
Memory Bandwidth 224GB/s 256GB/s 512GB/s 480GB/s

  
AMD officially reveal the Radeon Vega's specifications

 

AMD didn’t show many benchmarks for their Radeon Vega GPU, though what they did show was highly impressive, with the results below showing the Vega easily beat Nvidia’s Titan Xp in a range of 3D modelling applications. 

One of the big causes of this high level of performance here is due to Vega’s new Geometry Engine and Pixel Engine designs, which allows AMD’s new architecture to offer 2x the peak geometry performance per clock of their previous architectures. This significantly speeds up 3D modelling and design workflows, allowing more complex designs to be rendered in real-time. 

  
AMD officially reveal the Radeon Vega's specifications

 

AMD has shown that their Radeon Vega Frontier Edition GPUs will be available in both Air-cooled and liquid cooled varieties, both of which offer the same internal specifications.   

This GPU is not intended to be used by gamers, with drivers that are optimised for professional use and pricing to match. AMD states that gamers will have to wait for their RX Vega GPU to be released at a later date, as it will come at a lower price with optimisation for gaming workloads.   

AMD’s Radeon Technologies Group will be revealing more about Vega in Mid-June, so expect to hear more about Vega soon. 

 

You can join the discussion on AMD’s Radeon Vega GPU on the OC3D Forums. 

 

AMD officially reveal the Radeon Vega's specifications

AMD officially reveal the Radeon Vega’s specifications

 
We have all been waiting a long time for this, AMD’s official Radeon Vega specifications, which will give is a fairly clear picture of how well AMD’s new flagship will perform in a variety of applications. 
 
To start off, we now know that Vega has a total of 64 compute units and 4096 Stream Processors and peak FP32 performance of around 13TFLOPs, which is over 50% higher than AMD’s Fiji-based R9 Fury X. 

  

AMD officially reveal the Radeon Vega's specifications

 

In terms of raw processor count, the Radeon Vega is a lot like the R9 Fury X, offering a stream processor count of 4096, though Vega thanks to architectural changes and process node changes/optimisation will be able to offer increased clock speeds, giving AMD’s Radeon Vega around 50% more FP32 compute performance. 

AMD’s Radeon Vega also supports Rapid Packed Math and FP16 instructions, allowing this GPU to offer even more compute performance when less precision is required. 

Given the RX Vega’s stated 13 TFLOPs of compute performance and 4096 stream processors, AMD’s Radeon Vega Frontier Edition will offer boost clock speeds that are around 1587MHz, which is very similar to recent leaks which have suggested boost clock speeds of 1600MHz. 

One interesting thing is that the AMD Radeon Vega will offer slightly lower levels of memory bandwidth than the R9 Fury X though, in reality, Vega will offer a lot more memory performance thanks to several optimisations in the memory pipeline and the addition of AMD’s new Draw Stream Binning Rasterizer (DSBR) to Vega. These changes will allow AMD to do more work with the memory bandwidth that they have available, allowing them to have more effective memory bandwidth.  

 

  RX 570 RX 580 R9 Fury X Radeon Vega
Frontier Edition
GPU Polaris 10 Pro Polaris 10 XT Fiji XT Vega 10
Process node 14nm 14nm 28nm 14nm
Shade Engines 4 4   4
CUs per shader Engine 8 9   16
Shaders Per CU 64 64 64 64
Stream Processors 2048 2304 4096 4096
Performance (FP32) 5.095 TFLOPs 6.175 TFLOPs 8.6TFLOPs 13 TFLOPs
Performance (FP16) 5.095 TFLOPs 6.175 TFLOPs 8.6TFLOPs 25 TFLOPs
Render Output Units 32 32 64 64
Texture Mapping units 126 144 256 256
Hardware Threads 4 4 8
Memory Interface 256-bit 256-bit 4096-bit 2048-bit
Memory 4/8GB GDDR5 4/8GB GDDR5 4GB HBM 16GB HBM2
Memory Bandwidth 224GB/s 256GB/s 512GB/s 480GB/s

  
AMD officially reveal the Radeon Vega's specifications

 

AMD didn’t show many benchmarks for their Radeon Vega GPU, though what they did show was highly impressive, with the results below showing the Vega easily beat Nvidia’s Titan Xp in a range of 3D modelling applications. 

One of the big causes of this high level of performance here is due to Vega’s new Geometry Engine and Pixel Engine designs, which allows AMD’s new architecture to offer 2x the peak geometry performance per clock of their previous architectures. This significantly speeds up 3D modelling and design workflows, allowing more complex designs to be rendered in real-time. 

  
AMD officially reveal the Radeon Vega's specifications

 

AMD has shown that their Radeon Vega Frontier Edition GPUs will be available in both Air-cooled and liquid cooled varieties, both of which offer the same internal specifications.   

This GPU is not intended to be used by gamers, with drivers that are optimised for professional use and pricing to match. AMD states that gamers will have to wait for their RX Vega GPU to be released at a later date, as it will come at a lower price with optimisation for gaming workloads.   

AMD’s Radeon Technologies Group will be revealing more about Vega in Mid-June, so expect to hear more about Vega soon. 

 

You can join the discussion on AMD’s Radeon Vega GPU on the OC3D Forums.