We're less than 10 hours away from AMD's preview of Vega, its cutting-edge GPU architecture for high-end Radeon RX 500 series graphics cards. The event is set to take place tomorrow morning at 9:00AM ET in Las Vegas, Nevada. The preview will include 4K gaming demos featuring AMD's exciting new high-end Ryzen processors and unreleased ultra high-end enthusiast class Radeon graphics cards powered by next generation Vega GPUs.
AMD's Biggest CPU & GPU Architectural Leaps In Half A Decade, Ryzen & Vega
Zen, now known as "Ryzen", represents AMD's first major architectural iteration in the CPU arena since the introduction of Bulldozer in 2011. Vega is a similarly big deal. It's the biggest architectural overhaul in GPUs for the company since the introduction of GCN -- Graphics Core Next -- for the very first time in 2011 with the HD 7000 series family of graphics cards.
Unlike Zen however, which is an entirely new CPU core design, Vega maintains the core DNA of GCN. However, with Vega the basic GCN building block known as the compute unit -- CU for short -- is entirely new. We've detailed Vega's Next Compute Unit -- NCU -- architecture in detail in our exclusive last month. We'd highly encourage you go check out that article if you're interested in reading about why exactly Vega's NCU is such a big deal, what's new about it and what exactly it can do that previous GCN CUs cannot.
What To Expect At The Vega Preview
We know that Vega will be demoed running various triple A games at 4K to showcase the architecture's impressive gaming performance and power efficiency. Furthermore, thanks to one particularly crafty AMD fan, we know that the company will reveal and discuss Vega's biggest architectural features and improvements over Polaris and previous generation GCN graphics cards.
We've compiled a list of the features that were spotted in the code-base of the ve.ga teaser website.
We don't yet know whether these performance and power efficiency claims were made in relation to Polaris or Radeon R9 Fury series graphics cards based on the Fiji GPU. Regardless of whichever of these two is the reference point, the performance and power efficiency improvement figures remain quite extraordinary.
Vega Architecture
- 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 compute unit architecture
- Rapid packed math
- Draw stream binning rasterizer
- Primitive shaders
Vega's Confirmed Specs
Members of the press were able to spot some key specifications pertaining to Vega at last month's Vega demo event. 8GB of HBM2 memory for the consumer version of Vega and 16GB for the professional deep-learning MI25 variant were confirmed.
Additionally, an employee slipped a key specification that wasn't supposed to be made public yet and it's that Vega 10 features 512GB/s of memory bandwidth. The memory capacity and bandwidth are clear indications that Vega 10 has a 2048-bit wide memory interface. Half that of its older sibling, Fiji. However, because HBM2 is rated at twice the speed of HBM1, Vega 10 is able to achieve the same 512GB/s of memory bandwidth.
In terms of graphics horsepower, the Vega 10 powered MI25 accelerator is rated at a staggering 12.5 TERAFLOPS of single precision floating point compute and double that in half precision FP16 compute. That's 1.5 TERAFLOPS more than NVIDIA's Tesla P100 accelerator, powered by the monstrous 610mm² GP100 GPU and 2.5 TERAFLOPS more than the GTX 1080.
The MI25 is a professional, passively cooled product. The gaming oriented variant of Vega, equipped with more aggressive cooling solutions and running at higher clock speeds, would naturally be expected to achieve an even higher figure.
Vega's Rumored Specs
One of the few things that AMD has not talked about regarding Vega's specifications to date are the number of GCN stream processors it actually has. Vega 10 is believed to have 4096 GCN stream processors, according to the LinkedIn page of a leading engineer which leaked earlier this year.
Assuming that this figure is accurate, Vega 10 would have to operate at a frequency 20% higher than Polaris 10 to achieve the 12.5 TFLOPS of the Radeon Instinct MI25. We're talking 1520Mhz+, on a passively cooled enterprise GPU. A clock speed that few, mostly liquid cooled, overclocked RX 480 cards can achieve. None of AMD's current or past professional grade graphics cards and/or accelerators come close to that. We've also never seen such a large hike in clock speeds from one graphics generation to another at the same process node generation.
AMD Vega Lineup
| Graphics Card | Radeon R9 Fury X | Radeon RX 480 | Radeon RX Vega Frontier Edition | Radeon Vega Pro | Radeon RX Vega (Gaming) | Radeon RX Vega Pro Duo |
|---|---|---|---|---|---|---|
| GPU | Fiji XT | Polaris 10 | Vega 10 | Vega 10 | Vega 10 | 2x Vega 10 |
| Process Node | 28nm | 14nm FinFET | FinFET | FinFET | FinFET | FinFET |
| Stream Processors | 4096 | 2304 | 4096 | 3584 | 4096 (?) | Up to 8192 |
| Performance | 8.6 TFLOPS 8.6 (FP16) TFLOPS | 5.8 TFLOPS 5.8 (FP16) TFLOPS | ~13 TFLOLPS ~25 (FP16) TFLOPS | 11 TFLOLPS 22 (FP16) TFLOPS | >13 TFLOLPS >25 (FP16) TFLOPS | TBA TBA |
| Memory | 4GB HBM | 8GB GDDR5 | 16GB HBM2 | TBA | TBA | TBA |
| Memory Bus | 4096-bit | 256-bit | 2048-bit | 2048-bit | 2048-bit | 4096-bit |
| Bandwidth | 512GB/s | 256GB/S | 480GB/S | 400GB/S | TBA | TBA |
| TDP | 275W | 150W | TBA | TBA | TBA | TBA |
| Launch | 2015 | 2016 | June 2017 | June 2017 | July 2017 | TBA |
