The AMD Radeon VII Evaluation: An Surprising Shot At The Excessive-Finish

For AMD's Radeon Applied sciences group, 2018 was a tad annulment 12 months. After launching the Polaris Structure in 2016 and the Vega Structure in 2017, AMD started its first full 12 months of Vega. As a substitute of getting to introduce a 3rd structure in three years, the corporate is specializing in additional increasing the household by bringing Vega's laptop computer and server variants to market. And whereas AMD's laptop computer efforts in 19459002 have gone in an odd course their Radeon Intuition server efforts have put some pep again into their figurative stride, giving the corporate a declare to the primary 7nm GPU.

After releasing a last-generation product replace within the type of the Radeon RX 590 in November, we had anticipated the AMD client aspect to be prepared for some time. As a substitute, AMD introduced a slightly sudden announcement at CES 2019 final month: The corporate would difficulty a brand new high-end client card, the Radeon VII (Seven). Radeon VII is predicated on the aforementioned server GPU and is positioned as the newest flagship card for players and content material creators alike. AMD is once more focused on AMD's gamers. We're beginning immediately – on Day 7, in a well timed method – to check out AMD's newest map to see how the Radeon VII is as much as the problem.

At first look, the Radeon VII appears uncomplicated. The cardboard's silicon base is AMD's Vega 20 GPU, a by-product of the unique Vega 10, which has been enhanced for scientific computing and machine studying, and is predicated on TSMC's state-of-the-art 7nm course of to enhance efficiency. An necessary milestone for AMD's server-GPU efforts – it’s primarily the server-class first high-end GPU since Hawaii in 2013 – AMD was anxious to current the Vega 20 within the later levels of its launch on the coronary heart of AMD's comparatively new Radeon Intuition MI50 and MI60 server accelerators.

gif;base64,R0lGODlhAQABAAAAACH5BAEKAAEALAAAAAABAAEAAAICTAEAOw== - The AMD Radeon VII Evaluation: An Surprising Shot At The Excessive-Finish

Designed primarily for servers, Vega 20 shouldn’t be the GPU class that would give shoppers an affordable means. Or not less than it appeared to be AMD's authentic thought. Then again, one thing sudden occurred: NVIDIA didn’t transfer the counter very a lot when it comes to efficiency per greenback. The brand new Turing-based GeForce RTX playing cards are features as a substitute. They need to introduce a brand new paradigm of rendering video games with real-time ray tracing results, allocating giant elements of the already giant Turing GPUs to this objective. The top end result was comparatively excessive costs for the GeForce RTX 20 Collection playing cards, whereas their efficiency will increase in standard gaming are far decrease than the standard era improve.

Going through a much less hostile value setting than many anticipated, AMD has determined to convey Vega 20 to shoppers as NVIDIA duels with a kind of larger value factors. The Radeon VII is on the highway with $ 699 and comes up with the GeForce GTX 2080 as the brand new Radeon flagship.

Comparability of the specs of the AMD Radeon collection

AMD Radeon VII
AMD Radeon RX Vega 64
AMD Radeon RX 590
AMD Radeon R9 Fury X

Energy processors
(60 CUs)
(64 CUs)
(36 CUs)
(64 CUs)


Primary clock
1400 MHz
1247 MHz
1469 MHz
N / A

Increase Clock
1750 MHz
1546 MHz
1545 MHz
1050 MHz

Reminiscence Clock Gbps HBM2
1.89 Gbps HBM2
eight Gbps GDDR5
1 Gbit / s HBM

Reminiscence Bus Width
2048 bit
256 bits

eight GB
eight GB

single precision
13.eight TFLOPS
eight.6 TFLOPS

Double Precision
three.5 TFLOPS
(1/four fee)
(1/16 fee)
(1/16 fee)
(1/16 fee)

On-board pressure

Reference cooling
Out of doors triple fan
N / A

Manufacturing Course of
TSMC 7nm
GloFo 14nm
GloFo / Samsung 12nm
TSMC 28nm

Vega 20
(331 mm²)
Vega 10
(495 mm²)
Polaris 30
(232 mm2)
(596 mm2)

(GCN 5)
(GCN 5)
GCN four
GCN three

Transistor Rely

Date of introduction

introductory value
$ 699
499 US
$ 279
649 US

In our specification desk, Radeon VII comes with a "peak clock" of 1800 MHz, whereas the official increase clock is 1750 MHz. In comparison with RX Vega 64s peak clock, which was solely 1630 MHz, AMD is one other peak of about 10%. Due to an open-air cooler and a redesigned SMU, Radeon VII ought to be capable of improve and keep its larger clock frequencies somewhat extra typically. Whereas AMD's newest card doesn’t add another ROPs or CUs (it's really a small decline from the RX Vega 64), it constantly good points throughput.

The most important change from the RX Vega 64, nonetheless, is that AMD has greater than doubled reminiscence measurement and reminiscence bandwidth. That is because of the 7nm die shrink know-how, which delivers the newest AMD GPU with a comparatively modest die measurement of 331mm2. The additional area has created AMD area on the Interposer for 2 extra HBM2 stacks, permitting for extra VRAM and a wider reminiscence bus. AMD additionally barely elevated the clock pace, from 1.89 Gbps for the RX Vega 64 to a flat 2 Gbps for the Radeon VII.

gif;base64,R0lGODlhAQABAAAAACH5BAEKAAEALAAAAAABAAEAAAICTAEAOw== - The AMD Radeon VII Evaluation: An Surprising Shot At The Excessive-Finish

Apparently sufficient, the Radeon VII is actually a Radeon Intuition MI50. So for AMD, cannibalizing Intuition gross sales is feasible if the efficiency of the Radeon VII is simply too good for skilled pc customers. Consequently, AMD has decreased some options of the chip solely barely to assist distinguish merchandise. We'll speak extra about that later, however the primary cause is that the cardboard works at a non-native FP64 fee, loses its full-chip ECC help, and naturally, for a client product, the Radeon software program makes use of drivers as a substitute of the skilled intuition -Treiberstapels.

In fact, each time you speak about integrating a server GPU right into a client or prosumer card, you're speaking in regards to the potential of a strong card, and this definitely holds true for the Radeon VII. Finally, the angle which AMD is aiming for with its newest flagship card, relies on its competitiveness and can also be mixed with its world-class 16GB HBM2 reminiscence. As one among AMD's few distinctive specs in comparison with NVIDIA, VRAM capability is an enormous a part of the advertising and marketing facet of AMD. You’ll focus closely on content material creation and VRAM-intensive gaming. Additionally new to this card and one thing that can delight AMD is the triple fan cooler that replaces the warmly obtained fan of the Radeon RX Vega 64/56 playing cards.

As well as, AMD, as a board vendor, makes a pleasant change to retail and sells the brand new card instantly for a similar value of $ 699. As AIBs immediately additionally convey their model reference playing cards to the market, that is an choice to keep away from extreme beginning costs.

gif;base64,R0lGODlhAQABAAAAACH5BAEKAAEALAAAAAABAAEAAAICTAEAOw== - The AMD Radeon VII Evaluation: An Surprising Shot At The Excessive-Finish

For those who take a look at the aggressive panorama, there are some things to sort out immediately. A giant a part of the combination is ​​(as normal currently) a recreation bundle. The continuing Elevate the Recreation Absolutely Loaded Package deal sees Satan Might's Cry 5, The Division 2 and Resident Evil 2 free on the Radeon VII, RX Vega and RX 590 playing cards. In the meantime, the RX 580 and RX 570 playing cards qualify for 2 out of three playing cards. Usually a bundle would add worth on to a direct competitor – on this case the RTX 2080 – however NVIDIA has its personal duel game-on bundle with Anthem and Battlefield V. In a state of affairs the place the Radeon VII is predicted to be punished With the RTX 2080, as a substitute of successful instantly, these worth creations have gotten more and more necessary.

gif;base64,R0lGODlhAQABAAAAACH5BAEKAAEALAAAAAABAAEAAAICTAEAOw== - The AMD Radeon VII Evaluation: An Surprising Shot At The Excessive-Finish

The launch of the Radeon VII can also be the primary product launch because the current shift within the aggressive panorama for variable refresh replace applied sciences. Variable refresh screens have change into essential for players. Because the creation of variable picture replace know-how earlier this decade, there was a transparent separation between AMD and NVIDIA playing cards. AMD playing cards help VESA Adaptive Sync – higher often known as AMD's FreeSync branding – whereas NVIDIA desktop playing cards solely help their proprietary G-Sync characteristic. However final month NVIDIA surprisingly introduced that their playing cards would help VESA Adaptive Sync on the desktop underneath the label "G-Sync Compatibility". Particulars on the construction of this program are sparse, however on the finish of the day adaptive synchronization in NVIDIA drivers is usable, even when a FreeSync panel shouldn’t be licensed as G-Sync compliant.

The online result’s that though NVIDIA AMD's announcement doesn’t hinder performance, it undermines AMD's FreeSync benefit. All low cost VESA Adaptive Sync Screens beforehand solely out there for AMD playing cards might now be helpful on NVIDIA playing cards as nicely. In fact, AMD favored to emphasise the "free" a part of FreeSync, so it was clearly dulled as a weapon in opposition to NVIDIA. The official line of AMD believes that it is a win for FreeSync and client alternative, although the truth is commonly a bit extra unpredictable.

The introduction of the Radeon VII and its aggressive place over the GeForce RTX 2080 signifies that AMD should additionally give attention to the present useful hole between its playing cards and NVIDIA's newest Turing machines. For this objective, AMD's place in DirectX Raytracing (DXR) and AI-based picture high quality / efficiency strategies equivalent to DLSS has remained the identical. In brief, AMD's argument is that the efficiency hit and markup for these options will not be definitely worth the total distinction in picture high quality. Within the meantime, AMD shouldn’t be standing nonetheless and along with the DXR fallback drivers, they’re working to help WinML and DirectML for his or her playing cards. The danger for AMD, after all, is that the thicknesses of the hole between DXR or NVIDIA (DLSS) will improve quickly. Then the useful hole turns into greater than a theoretical annoyance.

All in all, pushing out a 7nm GPU for finish customers is now a really aggressive step, so early within the lifecycle of this course of, particularly as Q1 is usually at a low cycle and Q2 is down. Nonetheless, AMD doesn’t have a lot time to attend. The one main impediment can be pricing, which is appropriate to shoppers.

That brings us to immediately's begin. For $ 699, NVIDIA has already upgraded the value vary when it comes to devoted for accelerating raytracing and machine studying. For the Radeon VII, the phrases are 16GB HBM2 and the prosumer / content material creator worth. All that is still is their gaming efficiency.

2/2019 GPU value comparability


$ 1299
GeForce RTX 2080 Ti
(play on bundle)

Radeon VII
(Elevate the bundle of toys)
$ 699 / $ 719
GeForce RTX 2080
(play on bundle)

$ 599

Radeon RX Vega 64
(Enhance the Bundle of the Recreation)
499 US
GeForce RTX 2070
(Recreation On Bundle, 1 Recreation)

Radeon RX Vega 56
(Pack the bundle of toys)
$ 429

$ 349
GeForce RTX 2060
(Recreation On Bundle, 1 Recreation)

Leave A Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.