AMD announced the Radeon VII graphics card at its first ever CES keynote, calling it the “world’s first 7nm gaming GPU.” It’s the die shrink for the second generation of the AMD Vega graphics architecture and the company’s next-generation of high-performance GPUs.
The shrink from 14nm to 7nm alone will be responsible for a saving extra efficiency and the ability to run the new Radeon VII at a higher clock speed for greater gaming performance. But from the early announcement it’s not necessarily clear what else has changed with the second-gen GPU design down at the consumer level.
But we do know when it’s being released next month, and what sort of level of gaming performance AMD is aiming at with its new top-end graphics card. It’s interesting that team Radeon had previously stated its reticence to bring the pricey 7nm Vega core down to the consumer level, given the lukewarm reception that the original 14nm RX Vega cards received, and yet now gamers are being treated to the most expensive single-core gaming GPU AMD has ever created. Maybe someone proved people were willing to spend $699 on a graphics card…
Read more: the best gaming GPUs to buy today
If it really is able to offer that level of gaming performance across more than just the AMD-friendly Far Cry 5 and Battlefield V titles, then it’s got a good chance of being a successful high-end Radeon GPU.
It’s not long until the new Radeon VII card is on the shelves, with AMD giving a February 7, 2019 release date for the world’s first 7nm gaming GPU. Soon, my precious…
The 60 CU 7nm Vega 20 GPU used in the Radeon VII comes with 3,840 GCN cores, and 16GB HBM2 memory. That’s fewer cores than the Vega 64, but twice the memory, and around 300MHz higher boost clock speeds.
Given the on-stage performance comparison with the RTX 2080 it’s not a surprise to see AMD price the Radeon VII at $699 MSRP. Fingers crossed it doesn’t suffer the Vega 64 and 56 cards’ fates where launch prices were quickly replaced by far higher actual retail prices.
As we mentioned early, AMD is targeting RTX 2080-level performance, indeed the only gaming benchmarks it actually showed during its CES keynote had the Radeon VII performing a little higher than the RTX 2080 in Far Cry 5 and Battlefield V, and much quicker in the Vulkan-based Strange Brigade.
The official Radeon VII release date is February 7, 2019. It was first announced to the world at AMD’s CES 2019 keynote, where CEO, Dr. Lisa Su, took to the stage professing AMD’s love for gamers before introducing the “world’s first 7nm gaming GPU.”
At the moment we don’t know if there will just be an AMD reference design being sold to end users at launch, with the triple-fan shroud shown on stage, but it would seem a good bet that it is. Though AMD has also said that “several leading add-in board partners plan to offer the cards.”
Whether those third-party cards, or any factory-overclocked versions, will all be launched on the February 7 release date too is still not entirely clear.
The 7nm AMD Vega 20 GPU is the chip at the heart of the new Radeon VII, the same graphics silicon that is currently doing the rounds as part of the latest Radeon Instinct. It’s not just a ‘simple’ 7nm die-shrink, though, aside from the extra accelerators in the GPU for the pro crowd AMD has also doubled the ROP-count from 64 in the Vega 64 to 128 in the Radeon VII.
That extra silicon in the rendering backend, along with the double-width 4,096-bit memory bus, should enable the card to really take advantage of the massive 16GB of second-gen high-bandwidth memory (HBM2) that AMD has packed the card out with.
|Radeon VII||RX Vega 64||RTX 2080|
|GPU||Vega 20||Vega 10||Turing TU104|
|Memory||16GB HBM2||8GB HBM2||8GB GDDR6|
Though, if the pixel-pushing performance increases are verified on launch, the Radeon VII will then highlight just where the original Vega GPUs were too unbalanced in the backend to offer competitive gaming performance.
In terms of the actual core-count of the card’s Vega 20 GPU, there are actually fewer compute units (CUs) inside the Radeon VII. The RX Vega 64 had 64 CUs and 4,096 GCN cores, while the new GPU has just 60 CUs and therefore 3,840 GCN cores. But that will be offset by other features of the new silicon.
As well as the TSMC 7nm chip (not GlobalFoundries this time…) offering nominally lower power and more die area, it also allows AMD to hike up the clock speeds. The Vega 20 GPU has been pushed up to 1,800MHz for the boost clock, with the base clock starting at 1,450MHz – just below the top speed of the RX Vega 64.
It doesn’t look like there’ll be an RX Vega 56-esque version
We don’t yet know what the TDP of the new card is. You might have hoped that because of the 7nm shrink you’d be looking at a TDP lower than the 295W of the Vega 64, but given the frequency of the GPU we’d be surprised if it didn’t match the 300W TDP of the M150 Radeon Instinct card bearing the same Vega 20 silicon.
The solitary nature of the Radeon VII naming scheme does seem to preclude any other, lower-spec versions being announced. So, it doesn’t look like there’ll be an RX Vega 56-esque version to compete with the Nvidia RTX 2070. That could mean this is the only high-end option team Radeon has in store for gamers, with the upcoming AMD Navi GPUs taking care of more mainstream matters.
The $699 MSRP for the Radeon VII has been set by AMD, but with little control over what the add-in board (AIB) partners or retailers actually charge for the cards, we’d be surprised if you saw many versions available for $699 anytime around launch.
We’re hopeful this card doesn’t fall victim to the same ‘launch price’ shenanigans AMD pulled with the RX Vega 64 and Vega 56 GPUs, where the initial price given to reviewers was only a short-term price used around their release. After launch the previous Vega cards became unfeasibly expensive, and not just because of the mining boom. There were precious few cards in retail and therefore prices also skyrocketed.
Hopefully AMD has arranged a high volume of GPU supply so the cards don’t suffer from any sort of supply and demand pricing issues. Reportedly the biggest issue in terms of manufacturing, and pricing, was the HBM 2 last time out. Fingers crossed the nightmare AMD had with packaging the GPU and memory together has been ironed out, and that won’t be such a problem this time around.
This is where the “world’s first 7nm gaming GPU” is going to sink or swim; in just how competitive it is in terms of real-world gaming performance. The AMD slides show the Radeon VII beating the Nvidia RTX 2080 in some key modern titles. On-stage Dr. Lisa Su showed the Vega 20 chip outpacing the Turing GPU in Far Cry 5, Battlefield V, and Strange Brigade at 4K – showing gaming leadership in DirectX 11, DirectX 12, and the Vulkan APIs respectively.
In the press deck for the card it shows the Radeon VII absolutely smashing the RTX 2080 in Battlefield V at 1440p, and level-pegging with Forza Horizon 4 at 4K.
In terms of how it performs compared with the previous version of the Vega graphics architecture, AMD’s own performance benchmarks have the card between 25% and 42% faster than the RX Vega 64. The overall average of all the tests it has carried out, across a host of different games, comes out at 29% faster than the previous generation.
With AMD promising the new 7nm GPU being able to offer 25% extra performance at the same power that would maybe seem to suggest the Radeon VII is going to need a little extra juice to get there.
And that shows a classic AMD situation of throwing everything it can at a graphics card in order to hit the necessary performance level. Here, it needs to at least match the RTX 2080, but to do that AMD has thrown caution to the wind with regards to the clock speeds and the power draw of the card itself. It’s going to be mighty interesting to see just how hot the 7nm GPU gets running at that sort of level – Vega ran mighty hot running much slower, but hopefully the die-shrink may alleviate some of those thermal issues.
There will be times where the Radeon VII does outperform the RTX 2080
One other thing to mention with regards its touted relative performance to Nvidia’s Turing GPU is that, as is traditional with vendor performance benchmarks, AMD has chosen to highlight the games where its architectures generally do well. Battlefield, and its Frostbite engine, were AMD-buddies long before Nvidia got DICE ray tracing, and Far Cry 5 loves itself some Vega architecture too.
So while there are going to be times where the Radeon VII does outperform the RTX 2080, there will equally be occasions where the reverse is true and the GeForce GPU takes performance precedence.
At the risk of garnering some AMD hate, it’s also worth pointing out that this performance level essentially puts the Radeon VII on par with the GTX 1080 Ti. And if you look at it purely on those terms, as two straight rasterising GPUs, it’s taken AMD two years to be able to create a GPU that’s capable of matching the old Pascal card’s princely performance. And despite them being 7nm versus 14nm GPUs, the older Nvidia card is still going to have the edge in terms of efficiency too.
So while it’s great that AMD is releasing a high-end gaming card that can compete with Nvidia’s Turing GPUs, that excitement has to be tempered by the fact that realistically it will still be more expensive than the RTX 2080 around launch, slower in some games, and possibly demand a whole lot more power to get there too.