⋮    ⋮  

AMD Vega 10 & Vega 11 GPUs Detailed – Radeon 500 Series Showcase Coming Feb 28

Author Photo
Feb 22, 2017
427Shares
Submit

Today AMD confirmed via a press release that it will be holding an exclusive showcase event for its Radeon 500 series Vega GPUs on Feb 28. The press event which is going to be part of GDC 2017 is called Capsaicin And Cream. We already reported on what the company plans to do in this event last week, so make sure to check out that article if you want to know what AMD has in store for this upcoming exclusive Radeon 500 series Vega preview.

Press release excerpt:

intel-8th-generation-core-i7-8000-seriesRelatedIntel 6 Core Coffee Lake i7 8700K & i5 8600K Specs Leaked – Launching Later This Year

Hosted by Radeon Technologies Group’s Senior Vice President and Chief Architect, Raja Koduri, the show will be a celebration of PC gaming, the technology steering its future, and the developers who work tirelessly to transform their imaginations into tomorrow’s blockbuster games.

The one-hour live event will offer PC gaming enthusiasts and developers around the world a preview of AMD’s latest graphics and processor technologies, reveal exciting new details surrounding Vega, and showcase the summer’s most anticipated PC and VR games from visionary game developers.

Last week we also detailed AMD’s plan to roll out its next generation Radeon 500 series Vega family of graphics cards in May. Today, our friends over at VCZ have corroborated that report, stating that the new Radeon 500 series should indeed land in May.

AMD Vega 10 GPU die/chip + 2 HBM2 stacks

So, where exactly will Vega GPUs compete in the market and what kind of performance are they going to offer?

Vega 10 Is AMD’s Next Fury, Will Compete With The GTX 1080 & Upcoming GTX 1080 Ti

Let’s start off with the basics. There are two new Vega GPUs called Vega 10 and Vega 11. Vega 10 is the bigger of the two. This chip will be the flagship of the new Radeon 500 series stack and is set to compete with the GTX 1080 and the upcoming GTX 1080 TI. An early engineering sample of this GPU has already been demonstrated in action, outperforming the GTX 1080 by approximately 10% whilst running on 300 series Fury drivers.

amd-radeon-rx-vega-air-cooled-edition-dark-silver-2-wccftechRelatedAMD RX Vega Leaked Benchmark Shows It Ahead Of GTX 1080 – Specs Confirmed, 1630MHz Clock Speed, 8GB HBM2 & 484GB/s Of Bandwidth

The performance is only bound to get better with more post silicon optimization and more fine-tune drivers. We’ve already heard reports that 80% of AMD’s current driver team is working on optimizing and fine tuning the drivers for Vega. So that’s a good sign.

AMD Vega 10 prototype board via VCZ

Vega 11 To Compete At The Higher End Of The Mid-Range, Will Be AMD’s 1070

The smaller chip of the two is said to be AMD’s new high performance mid-range weapon, just like the GTX 1070 is to Nvidia’s arsenal. Whether this GPU will actually be on par, outperform or underperform the GTX 1070 is still unknown. This GPU hasn’t been part of any public showcase or demonstration to date. Making it a a prime target for AMD’s marketing time come February 28th where we might see the first ever preview of Vega 11.  You can find the schedule for the AMD’s GDC 2017 activities here. Which will include a special session about optimizing games for the Vega architecture.

AMD Vega Lineup

Graphics CardRadeon R9 Fury XRadeon RX 480Radeon RX Vega Frontier EditionRadeon Vega ProRadeon RX Vega (Gaming)Radeon RX Vega Pro Duo
GPUFiji XTPolaris 10Vega 10Vega 10Vega 102x Vega 10
Process Node28nm14nm FinFETFinFETFinFETFinFETFinFET
Stream Processors4096230440963584TBAUp to 8192
Performance8.6 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
Memory4GB HBM8GB GDDR516GB HBM2TBATBATBA
Memory Bus4096-bit256-bit2048-bit2048-bit2048-bit4096-bit
Bandwidth512GB/s256GB/S480GB/S400GB/STBATBA
TDP275W150WTBATBATBATBA
Launch20152016June 2017June 2017July 2017TBA

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 find our full architectural deep-dive article here.

 

Submit