For AMD's Radeon Technologies group, 2018 was a small year of congestion. After launching the Polaris Architecture in 2016 and the Vega Architecture in 2017, AMD began its first full year of Vega. Instead of having to introduce a third architecture in three years, the company is focusing on further expanding the family by bringing Vega's laptop and server variants to market. And while AMD's laptop efforts have taken a strange turn, the efforts of the Radeon Instinct server have attracted some attention and the company has claimed the first 7nm GPU.
Following the release of a last-generation product update in November, the Radeon RX 590 was expected to take AMD's consumer side a while longer. Instead, AMD announced a rather unexpected announcement at CES 201
At first glance, the Radeon VII seems uncomplicated. The card's silicon base is AMD's Vega 20 GPU, a derivative of the original Vega 10, which has been enhanced for scientific computing and machine learning, and is based on TSMC's state-of-the-art 7nm process to improve performance. An important milestone for AMD's server GPU efforts – it is essentially the server's first high-end GPU since Hawaii in 2013 – AMD was anxious to present the Vega 20 in the later stages of its launch as the GPU , the centerpiece of AMD's relatively new Radeon Instinct MI50 and MI60 server accelerators.
First of all designed for servers, Vega 20 is not the GPU class that could give consumers a cheap way. Or at least it seemed to be AMD's original idea. On the other hand, something unexpected happened: NVIDIA did not move the counter very much in terms of performance per dollar. The new Turing-based GeForce RTX cards are functions instead. They want to introduce a new paradigm of rendering games with real-time ray tracing effects, allocating large parts of the already large Turing GPUs to this purpose. The end result was a relatively high price for the GeForce RTX 20-Series cards, while their performance increases in traditional games are much lower than in the usual generations.
Facing a less hostile price environment than many had expected, AMD has finally decided to bring Vega 20 to consumers, with NVIDIA dueling for one of those higher price points. The Radeon VII has taken the road with $ 699 and is on the field with the GeForce GTX 2080 as the new Radeon card for the flagship.
|AMD Radeon Series|
|AMD Radeon VII||AMD Radeon RX Vega 64  AMD Radeon RX 590||AMD Radeon R9 Fury X|
|Power Processors|| 3840
( 60 CUs)
|ROPs||64||64  32||64|
|Basic clock||1400 MHz 1494 MHz 1494 MHz||N / A|
|Boost Clock||1750MHz||1546MHz||1545MHz||1050rpm up to 1050 MHz and 1050 MHz; 2450 MHz; 2450 MHz; Memory clock||2.0 Gbps HBM2||. 19659014] 1 Gbps HBM|
|memory bus width||4096-bit||2048-bit||256-bit||4096-bit|
|Single Precision||13.8 TF LOPS||12.7 TFLOPS||7.1 TFLOPS||8.6 TFLOPS|
|Double Precision|| 3.5 TFLOPS
| 794 GFLOPS
(1/16 Rate) 19659014] 445 GFLOPS
| 538 GFLOPS
|Board Power||300W||295W||225W||225W||275W||] Reference Cooling||Outdoor Triple Fan||Blower||N / A||AIO CLC|
|Manufacturing process||TSMC 7nm||GloFo 14nm||19659014] TSMC 28nm|
|GPU|| Vega 20
| Vega 10
| Polaris 30
|Fiji ] (59 6 mm2)|
|GCN 4||GCN 3||13.2 B||12.5B  5.7B||8.9B|
|Date of introduction||07.02.2014||14.08.2017||11/15/2018||06/24/2015|
|Introductory price||$ 699||$ 499||$ 279||$ 649|
In our specification table, Radeon VII is shipped with a "Peak Engine Clock" of 1800 MHz is 1750 MHz. Compared to RX Vega 64s peak clock, which was only 1630 MHz, AMD is another peak of about 10%. Thanks to an open-air cooler and a redesigned SMU, Radeon VII should be able to increase and maintain its higher clock frequencies a little more often. While AMD's newest card does not add any other ROPs or CUs (it's actually a small decline from the RX Vega 64), it does gain in throughput across the board.
If anything, the biggest change from the RX Vega 64 is that AMD has doubled its memory size and more than doubled its memory bandwidth. This is due to the 7nm die shrink technology, which delivers the latest AMD GPU with a relatively modest die size of 331mm2. The extra space has created AMD space on the Interposer for two more HBM2 stacks, allowing for more VRAM and a wider memory bus. AMD was also able to increase storage speed slightly, from 1.89 Gbps for the RX Vega 64 to a flat 2 Gbps for the Radeon VII.
Interestingly, the Radeon VII, based on its basic specifications, is essentially a Radeon Instinct MI50. So for AMD, cannibalizing Instinct sales is possible if the performance of the Radeon VII is too good for professional computer users. As a result, AMD has reduced some features of the chip only slightly to help distinguish products. We'll talk more about it later, but the main reason is that the card works at a non-native FP64 rate, loses its full-chip ECC support, and of course uses the Radeon software for a consumer product driver rather than the professional instinct -Treiberstacks.
Of course, you're always talking about using a server GPU in a consumer or prosumer card, you're talking about the potential for a powerful card, and this certainly applies to the Radeon VII. Ultimately, the perspective is Aiming for competitiveness with its flagship AMD, AMD is also combined with its world-class 16GB HBM2 memory. As one of AMD's few unique specifications compared to NVIDIA, VRAM capacity is a big part of the marketing aspect of AMD. You will focus heavily on content creation and VRAM-intensive gaming. Also new to this card and something that will thrill AMD is its triple-vented radiator, which replaces the warmly received fan of the Radeon RX Vega 64/56 cards.
In addition, AMD throws a proper change He came as a board vendor in the retail and can sell the new card directly for the same price of 699 USD. As AIBs today also bring their brand reference cards to the market, this is an option to avoid excessive starting prices.
If you look at the competitive landscape, there are a few things to tackle today. A big part of the mix is (as usual lately) a game package. Devil May Cry 5, The Division 2, and Resident Evil 2 are included with the Radeon VII, RX Vega, and RX 590 cards for free in the ongoing "Fully Loaded" package. Meanwhile, the RX 580 and RX 570 cards qualify for two out of three cards. Normally a bundle would add value directly to a direct competitor – in this case the RTX 2080 – but NVIDIA has its own duel game-on bundle with Anthem and Battlefield V. In a scenario where the Radeon VII is expected to be punished With the RTX 2080, instead of winning directly, these value creations are becoming increasingly important.
The launch of the Radeon VII is also the first product launch since the recent shift in the competitive landscape for variable refresh monitoring technologies. Variable refresh monitors have become a must-have for gamers. Since the advent of variable image update technology earlier this decade, there has been a clear separation between AMD and NVIDIA cards. AMD cards support VESA Adaptive Sync – better known as AMD's FreeSync branding – while NVIDIA desktop cards only support their proprietary G-Sync feature. However, last month NVIDIA announced the surprise that their cards would support VESA Adaptive Sync on the desktop under the "G-Sync Compatibility" label. Details on the structure of this program are sparse, but at the end of the day adaptive sync is usable in NVIDIA drivers even if a FreeSync panel is not & gt; G-Sync Compatible & # 39; is certified.
The end result is that The NVIDIA announcement does not hamper AMD's features, it undermines AMD's FreeSync advantage – all cheap VESA Adaptive Sync monitors previously only available for AMD cards are now available for NVIDIA Cards of use. Of course, AMD liked to emphasize the "free" part of FreeSync, so it was clearly dulled as a weapon against NVIDIA. The official line from AMD is that this is a win for FreeSync and consumer choice, though the reality is often a bit more unpredictable.
The launch of the Radeon VII and its competitive position over the GeForce RTX 2080 means that AMD also needs to learn about the current gap between its maps and NVIDIA's latest Turing machines. For this purpose, AMD's position in DirectX Raytracing (DXR) and AI-based image quality / performance techniques such as DLSS has remained the same. In short, AMD's argument is that the performance hit and markup for these features are not worth the overall difference in image quality. In the meantime, AMD is not standing still and together with the DXR fallback drivers, they are working to support WinML and DirectML for their cards. The risk to AMD, of course, is that the functional gap becomes more than a theoretical hassle as DXR or NVIDIA's DLSS efforts quickly gain in importance.
All in all, a 7nm GPU is being pushed out of the game for consumers. This is a very aggressive step, and so early in the lifecycle of this process, especially as Q1 generally falls flat on a cyclical basis and Q2 fails , However, AMD does not have much time to wait. The only major obstacle would be pricing, which is acceptable to consumers.
This brings us to today's start. For $ 699, NVIDIA has already upgraded the price range in terms of dedicated hardware for accelerating raytracing and machine learning. For the Radeon VII, the terms are 16GB HBM2 and the prosumer / content creator value. All that remains is their gaming performance.
|2/2019 Comparison of GPU Prizes|
|1299|| GeForce RTX 2080 Ti
(Game On Bundle))
|Radeon VII [19459021[RaisetheGameBundle]||$ 699 / $ 719|| GeForce RTX 2080
(Game On Bundle)
| Vega 64
(Raise the Game Bundle )
|$ 499|| GeForce RTX 2070
(Game on Bundle, 1 game)
|Radeon RX Vega 56   $ 429|
|$ 349|| GeForce RTX 2060
(Game On Bundle, 1 game)