AMD Vega 10 Coming End Of 2016 With Powerful V9 Architecture, Vega 11 Due 1H 2017
AMD’s most powerful GPU yet code-named Vega 10 is set to debut at the end of the year with Vega 11 following early next year. The company also has a new board code-named “Magnum” that will be showcased at SC 2016 this upcoming November. It’s not clear yet whether Vega 10 will make its debut alongside “Magnum” next month but several of our sources have confirmed that we’ll definitely see Vega 10 before year’s end.
Magnum is a unique chip, it features a matrix of logic blocks that can be configured and programmed individually for any desired application or program. In other words, it’s the company’s first ever FPGA and its greatest attempt yet to expand its penetration into the high performance embedded market.
AMD’s Vega 10 Debuting Before Year’s End – Features New V9 GCN Architecture
Moving on to Vega 10, the GPU that most of you reading this are probably interested in. Vega 10 will be the very first GPU based on the company’s brand new GCN graphics architecture and 9th generation collection of visual IP. Vega 10 has been long in the making and we’ve heard so much about it over the past several months.
The specifications for AMD’s upcoming flagship Radeon graphics card, Fury X successor, based on Vega 10 have been leaked on one of the company’s leading engineers’ LinkedIn page. Packing an impressive 4096 next generation V9 cores and an unbelievable 12 teraflops of compute performance. Putting it ahead of Nvidia’s top dog the $1200 GTX TItan X.
Since then we have actually spotted the card on AMD’s & Sapphire’s official websites. The card was touted as a 4K gaming capable high-end offering with a 2016 release date. Not long after that Raja Koduri announced that the Radeon Technologies Group was celebrating a major milestone with regards to Vega 10’s development. Subsequently, the head of PR at the Radeon Technologies Group teased the launch venue for Vega 10 on his facebook page.
Completely Redesigned V9 GCN Architecture With Significant Advances
The V9 nomenclature refers to AMD’s 9th graphics architecture. Which represents a significant departure from the previous 8th generation GCN architecture of Polaris ,RX 400 series, and AMD’s “Islands” family of R9 200 & HD 7000 graphics cards.
The structure and configuration of SIMD units is entirely new. Each Vega NCU is now capable of simultaneously processing variable length wavefronts. To understand why this is such a big deal we have to look at AMD’s current GCN implementation.
In AMD’s current GCN implementation, each compute unit has four 16-wide vector SIMD units, capable of executing four 16-wide wavefronts ( a group of threads ) over four cycles. In addition to one scalar unit, capable of executing one instruction per cycle. This unit is delegated time-critical tasks, where the four-cycle turnaround of the SIMD units isn’t sufficient.
Unfortunately, these 16-wide SIMD units work exactly the same no matter how small of a wavefront they’re fed. Executing a 16-wide wavefront would take just as long as executing a 4-wide wavefront, rendering the other 12 ALUs inside the SIMD completely useless. And as graphics workloads are inherently non uniform it’s effectively impossible to find any scenario where all 16-wide SIMD units are fully occupied at any given time.
Variable Width SIMDs, Getting More Performance Out Of Fewer Cycles
This is no longer the case in AMD’s new GCN implementation inside Vega. The V9 architecture includes new incredibly clever schedulers and coherency subsystems that allow several smaller wavefronts to be executed simultaneously inside any SIMD that’s able to accommodate the workload. This in effect allows each NCU to finish considerably more work in the same amount of time compared to its predecessor. In addition to freeing up valuable cache and memory resources for other compute units.
It’s very hard to predict how much of a difference this big of an improvement in resource utilization and CU occupancy will yield given how unpredictable and inherently fluctuant graphics workloads are. Which brings us neatly to Vega’s rumored specs.
Vega 10 will come in several configurations for gaming & professional computing applications. All configurations will feature second generation high bandwidth memory but in various capacities. Radeon Pro Vega 10 will feature the highest configuration of 4096 stream processors and 16GB of memory. The board has a TDP of ~230W and is powered by eight and six pin PCIe power connectors. Cooling wise the Radeon Pro will come in active and passively cooled configurations for servers.
AMD Vega 10 & Vega 11 GPUs
|Graphics Card||R9 Fury X||RX 480||TBA||TBA|
|GPU||Fiji XT||Polaris 10||Vega 11||Vega 10|
8.6 (FP16) TFLOPS
5.8 (FP16) TFLOPS
25 (FP16) TFLOPS
|Memory||4GB HBM||8GB GDDR5||TBA||16GB/8GB HBM2|
|Bandwidth||512 GB/s||256 GB/s||TBA||512 GB/s|
AMD’s Vega 11 Coming Early 2017, Faster & More Power Efficient Than Polaris 10
Vega 11 will feature 8GB of HBM2 and is aimed at delivering the highest possible performance in notebook devices as well as replacing Polaris 10 in the midrange on the desktop. The GPU will be faster than Polaris 10 and use less power. Vega 11 will come in mini-ITX SFF and standard PCIe form factors. Most of the GPU’s extra oomph comes from the updated graphics architecture in addition to higher clock speeds compared to the RX 480 and RX 470. Vega 11 based cards will target the same ~$250 mid-range target that’s currently occupied by Polaris 10 and Nvidia’s GTX 1060.
Radeon Pro Vega 10 will be announced later this year with availability expected immediately after. Gaming Radeon graphics cards based on Vega 10 are expected to be available on shelves in the first quarter of next year. Vega 11 Radeon graphics cards are expected to be soon after that, but more towards the summer.
AMD Graphics Architectures
|WCCFTech||Year||Product||GPU||Process||Transistors In Billions||Performance||Memory||Bandwidth|
|Southern Islands||2012||HD 7970 Ghz||Tahiti||28nm||4.3||4.1 TFLOPS||3GB GDDR5||264GB/s|
|Volcanic Islands||2013||R9 290X||Hawaii||28nm||6.2||5.6 TFLOPS||4GB GDDR5||320GB/s|
|Caribbean Islands||2015||R9 Fury X||Fiji||28nm||8.9||8.6 TFLOPS||4GB HBM1||512GB/s|
|POLARIS||2016||RX 480||Polaris 10||14nm||5.7||5.2 TFLOPS||8GB GDDR5||256 GB/s|
|VEGA||2017||TBA||Vega 10||14nm||Up To 18||12 TFLOPS||16GB HBM2||512 GB/s|
|VEGA||2018||TBA||Vega 20||7nm||-||-||32GB HBM2||1 TB/s|
|NAVI||2019||TBA||Navi 10||7nm||-||-||Nextgen Memory||-|