⋮    ⋮  

AMD’s Raja Koduri Confirms Vega 10 Die Size At 484mm² – RX Vega To Be Faster In Gaming Than Frontier Edition


AMD has finally confirmed the die size of its new flagship graphics chip powering its Vega based graphics products. The first of the new Vega lineup, the Radeon Pro Vega Frontier Edition and Radeon instinct MI25 launched late last month. The rest of the Vega family, namely the gaming oriented RX Vega graphics products and professional Radeon Pro Vega parts are set to release later this month at SIGGRAPH.

AMD's Vega 10 GPU - 484mm² Of Magic Sand

Vega 10 is generally believed to be one of two Vega graphics chips that AMD has developed based on the Vega architecture. The second being Vega 11, in addition to the fact that very little is known about this chip AMD has yet to even officially acknowledge its existence.  We've only seen it mentioned in an official capacity once, by Anandtech's Ryan Smith back in March of last year.

NVIDIA & AMD Graphics Card Prices Drop by up to 18% Across the Board

With that being said, Vega 10 is going to be our primary focus in this piece. We finally have official confirmation regarding the actual die area of this chip and it comes directly from the man himself, Raja Koduri.

Vega 10 measures 484mm², 252mm² larger than AMD's previously largest 14nm based graphics chip, Polaris 10, which powers the company's Radeon RX 540 and 470 graphics cards. This makes it by far the largest chip the company has designed on 14nm FinFET.

AMD Confirms July Launch At SIGGRAPH For RX Vega Gaming Optimized Graphics Cards

Some rumors have been floating around over the past couple of days that AMD may be delaying its RX Vega launch. The company dispelled these rumors earlier today, confirming that RX Vega will in fact debut at SIGGRAPH later this month as previously announced.

AMD’s Open Source Vulkan Ray Tracing Engine Debuting In Games This Year – Radeon Rays 2.0

AMD's Jason Evangelho also went on record to state that gamers should not draw conclusions about Vega's gaming performance based on Vega Frontier Edition benchmarks. Vega Frontier Edition was designed and optimized for a vastly different set of tasks and use cases aimed at the professional market. On the other hand RX Vega graphics cards coming at the end of the month are specifically designed and optimized for games.

We've also seen similar comments from an MSI representative, the very same one that hinted at Vega's power requirements a week earlier.

"please dont compare this card too much with RX Vega" via Tweakers.com

These comments echo Raja Koduri's previous statements about RX Vega graphics cards having significantly better optimized drivers for games and "some flavors" of RX Vega actually outperforming Vega FE.

Raja Koduri - Chief Architect Radeon Technologies Group - Reddit AMA

RX will be fully optimized gaming drivers, as well as a few other goodies that I can't tell you about just yet....But you will like FE too, if you can't wait:)

Raja Koduri - Chief Architect Radeon Technologies Group - Reddit AMA

Consumer RX will be much better optimized for all the top gaming titles and flavors of RX Vega will actually be faster than Frontier version!

Q : 6) Many argue that vega is just a refined polaris gpu, how would you respond to this ?

Raja Koduri - Chief Architect Radeon Technologies Group - Reddit AMA

A: My software team wishes this was true:)

Vega is both a new GPU architecture and also completely new SOC architecture. It's our first Infinity Fabric GPU as well

It appears that Vega, being a new architecture, poses significant challenges for AMD's driver team. There's very little doubt that AMD's driver team is still optimizing for Vega and hasn't rolled out specific game by game optimizations in the drivers just yet. AMD will likely not be doing so until closer to RX Vega's official release. To what extent this will affect Vega's current gaming performance? We'll have to wait until those drivers are out and we have RX Vega in the labs to say for certain. Don't expect any miracles though.

What's Inside The Magic Sand?

Vega 10 Block Diagram

Vega 10 features 256 texture mapping units and 64 next generation Vega compute units arranged in two islets, each housing two compute engines. Every compute engine includes two distinct compute clusters. Each of those clusters features 512 stream processors and 32 texture mapping units. The chip in its entirety has a total of 4096 stream processors and 256 texture mapping units.

On the front-end side of things there are 64 render output units that make up 16 distinct render back-ends that connect to the 2048-bit HBM2 memory interface. The whole Vega 10 die sits on an interposer and is 2.5D stacked with two HBM2 stacks. Every stack can be configured with up to 8 GB of memory for a total of 16 gigabytes as can be found in the Radeon Pro Vega Frontier Edition.

For a more detailed look at the Vega 10 GPU specs make sure to check out our in-depth Vega 10 spec break down here.

AMD Vega 10 GPU Specifications

GPUPolaris 10 XTVega 10 XT
Process Node14nm14nm
Shader Engines44
Stream Processors23044096
Performance5.8 TFLOPS
5.8 (FP16) TFLOPS
25 (FP16) TFLOPS
Render Output Units3264
Texture Mapping Units144256
Hardware Threads48
Memory Interface256-bit2048-bit
Memory8GB GDDR5Up To 16GB HBM2

AMD Radeon Vega Lineup:

Graphics CardRadeon R9 Fury XRadeon RX 480Radeon RX Vega Frontier EditionRadeon RX Vega 64Radeon RX Vega 56(Radeon Pro Vega 64Radeon Pro Vega 56
GPUFiji XTPolaris 10Vega 10Vega 10 XTX/XTVega 10 XLVega 10Vega 10
Process Node28nm14nm FinFETFinFETFinFETFinFETFinFETFinFET
Compute Units64366464566456
Stream Processors4096230440964096358440963584
Performance8.6 TFLOPS
8.6 (FP16) TFLOPS
5.8 (FP16) TFLOPS
26 (FP16) TFLOPS
Up to 13+ TFLOPS
26+ (FP16) TFLOPS
~25 (FP16) TFLOPS
22 (FP16) TFLOPS
Texture Mapping Units256144256256TBA256224
Render Output Units64326464TBA6464
Memory Bus4096-bit256-bit2048-bit2048-bit2048-bit2048-bit2048-bit
Launch20152016June 2017July 2017July 2017December 2017December 2017
Price$649 US$199 (4 GB)
$229 (8 GB)
$999 (Reference)
$1499 (Liquid)
$499 (Reference)
$549 (Limited Air)
$599 (Liquid)
$649 (Liquid LE)

The Vega Architecture

High Bandwidth Cache And Memory Sub-System

With the Vega architecture AMD is introducing several new cutting edge technologies, chief among which is a brand new unique memory engine. In Vega 10 the HBM2 storage acts as a superfast cache thanks to a specialized processor that AMD dubs the High Bandwidth Cache Controller. The HBCC works to seamlessly stream data in and out of the memory, allowing Vega GPUs to have an insanely large address space of up to 512TB. This address space is only limited by the system's overall storage space.

Vega Next Generation Compute Engine

The next generation compute unit the company is debuting with Vega can execute half precision 16-bit floating point ops at twice the rate of FP32, which software can opportunistically take advantage of to increase throughput and reduce the thermal and power footprints of the GPU.

Next-Gen Compute Units (NCUs) provide super-charged pathways for doubling processing throughput when using 16-bit data types.1 In cases where a full 32 bits of precision is not necessary to obtain the desired result, they can pack twice as much data into each register and use it to execute two parallel operations. This is ideal for a wide range of computationally intensive applications including image/video processing, ray tracing, artificial intelligence, and game rendering.

Geometry Engine

Vega also features a new programmable geometry engine that delivers twice the performance per clock. In combination with the engine’s new primitive shader discard capability Vega is now significantly faster at tessellation and rendering of complex geometry and detail rich scenes.

The most challenging workloads for a GPU can present it with millions of geometry primitives per frame, all of which must be evaluated to determine their contribution to the final image.  New primitive shader technology allows Radeon Pro Vega graphics to perform geometry culling at an accelerated rate, eliminating unnecessary work for the rest of the GPU.  An advanced workload distribution mechanism then assigns processing tasks to the available pipelines in a way that maximizes their utilization and avoids idle time.  The result is Radeon Pro Vega graphics is capable of rendering extremely complex 3D models and scenes smoothly in real time.

The gemoetry pipeline also includes a new Primitive Discard Accelerator that detects parts of the gemoetry that are obscured by other objects or sit outside the scene and discards them, saving power and performance. The PDA ensure only visible parts of the scene are rendered and no energy is wasted on rendering invisible geometry. The issue of wasting cycles on rendering the invisible has led to unnecessarily slow performance in numerous games including Crysis 2, where it would make GPUs wastefully tessellate entire oceans of invisible water hidden below the surface.

Pixel Engine

Another key part of the Vega architecture is AMD's brand new pixel engine which is able to break work down into batches that then can enter the cache directly rather than reside in memory. This saves power, cycles, increases overall bandwidth and renders the scene faster.

Another clever technology that will be debuting with the Vega architecture is the shade-once technology which works just like the Primitive Discard Accelerator but on the pixel scale. It analyses pixels early in the graphics pipe and discards any that are hidden behind other objects in the scene. Again saving power, cycles and rendering the scene faster.

Another key advantage with the new Pixel engine is the fact that AMD has now linked it directly to the on-chip cache rather than the off-chip memory. This approach allows for some key optimization opportunities that developers are already familiar with on the gaming consoles.


Vega Architecture Key Features

– 4x Power Efficiency
– 2x Peak Throughput/Performance Per Clock
– High Bandwidth Cache
– 2x Bandwidth per pin
– 8x Capacity Per stack (2nd Generation High Bandwidth Memory)
– 512TB Virtual Address Space
– Next Generation Compute Engine
– Next Generation Pixel Engine
– Next Generation Compute Unit optimized for higher clock speeds
– Rapid Packed Math
– Draw Stream Binning Rasterizer
– Primitive Shaders

You can read about the Vega architecture in full detail here.