AMD Radeon VII 1440p, 4K & Ultrawide Gaming Performance

In the lead-up to last month’s CES, the rumor mill seemed to be on fire with AMD speculation. One rumor was of an RX Vega follow-up, and lo and behold, Lisa Su announced the Vega-based Radeon VII during AMD’s first-ever CES keynote. The card becomes the world’s first gaming GPU built on a 7nm process, and with 16GB of memory under its hood, it’s built for heavy workloads.

The “VII” in Radeon VII represents the fact that the card is built using 7nm, and also that it’s a follow-up to the original Vega, which spawned RX Vega 56 and 64. The new card has a cool name, but it doesn’t exactly open the doors for bigger or smaller models, and that’s likely on purpose.

During a briefing with AMD a few weeks ago to discuss the Radeon VII, we were surprised by the fact that the company jumped into a look at creative workloads before gaming ones, despite the card being touted as the “world’s first 7nm gaming” GPU. Highlighting VII’s gaming focus is likely for marketing more than anything else, because “world’s first 7nm jack-of-all-trades card” doesn’t have the same ring to it. With its “do everything” ambition, the Radeon VII is similar to NVIDIA’s TITAN series. It just doesn’t have the Tensor or RT cores that the TITAN RTX does (but its saving grace is that it costs 1/4th as much).

With 16GB of HBM2, Radeon VII feels like a professional-grade card out-of-the-gate, and at launch, we questioned whether or not it should have been called the Radeon Pro VII (or even Frontier Edition 2). We’re talking 1TB/s of available memory bandwidth, after all. Creative users will lust for that long before gamers will.

The move to 7nm helps AMD develop a GPU that’s more power efficient, but what fun would there be in releasing a product that only delivered the same performance, but at less power? It’s much better when we get far greater performance for the same power – or perhaps 5W more than RX Vega 64, as AMD’s spec sheet shows.

As our power testing will highlight later, AMD’s TDP spec doesn’t align with our real-world testing, but that’s a good thing, since VII draws a good deal less power than the Vega 64, despite its big performance boost, and having twice the amount of HBM2.

On the topic of VRAM, AMD’s marketing push with this card hovers a lot around the importance of future-proofing ourselves with lots of memory. As games with bigger textures get released, and we play our games at higher resolutions, the need for more GPU memory is undoubtedly going to increase.

I admit I do find some of AMD’s assumptions to be a little slanted, though. I don’t personally think 16GB of VRAM is going to do many people much good with gaming now, or in the next few years – and by the time a few years pass, we’ll have new GPUs to use.

In the press deck (slide seen below), AMD highlights 11GB as being “required” for gaming in 2019, which isn’t even close to being the case. Also strange is that the marker is notches higher than the 11GB it references, exaggerating the differences a bit.

The reviewer’s guide breaks down a few modern games that can use more than 8GB of VRAM, but I consider this flawed based on the fact that current memory tools only detect whether or not memory is allocated, not if all of that allocated data is required. Do you really believe Black Ops 4 needs 12GB of VRAM? Or that Star Control: Origins needs 9GB? I’m not convinced.

With VII being derived from Instinct, it feels to me like AMD was forced to a 16GB VRAM config, and now has to justify that reality through what I’d consider to be misleading marketing. Ultimately, it feels like AMD didn’t even know where to target this card, so it just decided to target it everywhere. That even includes 1080p resolution. That’s a resolution we haven’t used in our high-end GPU launch reviews since 2014. How many people need 300+ FPS in Siege, really?

To add to this, the night prior to this article’s embargo lift, AMD told us that it decided to increase the double-precision (FP64) performance on this card to twice what was previously announced – which was already twice what we were told after CES. AMD’s RX Vega 64 offered about 0.8 TFLOPS of FP64, whereas Radeon VII, after AMD’s whim, becomes 3.52 TFLOPS. That means the card doesn’t just target gaming and creative, but now also science and finance. For comparison’s sake, the 2080 Ti offers 0.42 TFLOPS of FP64 performance.

The upside to any confusion surrounding Radeon VII is that the card is a good jack-of-all-trades. It’s going to be a solid option for most workloads, and now with increased FP64, it becomes a relative steal at $700 to those who need it. Two of these GPUs would give us the same spec’d FP64 performance as one TITAN V.

With gaming, which happens to be the focus of this article (I know, it’s hard to tell), we have many benchmarks that will help us turn our gaming assumptions into answers. Another article takes a look at the card’s workstation performance, so feel free to open that one in a new tab!

A Look At Testing Methodology

Games Tested & Vendor Neutrality

A total of eight games are included in our current test suite. Recent additions include Battlefield V, Forza Horizon 4, and Shadow of the Tomb Raider. Beyond these eight titles, UL’s 3DMark and VRMark, as well as Unigine’s Superposition, are used for some quick and dirty tests that you may be able to run at home.

Here’s the full list of tested synthetic benchmarks, games, and developer allegiances:

• Battlefield V
• Deus Ex: Mankind Divided – AMD partner
• F1 2018
• Far Cry 5 – AMD partner
• Forza Horizon 4
• Monster Hunter World
• Shadow of the Tomb Raider – NVIDIA partner
• Tom Clancy’s Ghost Recon Wildlands – NVIDIA partner
• UL 3DMark & VRMark
• Unigine Superposition

Battlefield 1 - Techgage Tested Settings (1)

Battlefield V as-tested settings

Battlefield 1 - Techgage Tested Settings (2)

Battlefield V as-tested settings

Deus Ex Mankind Divided - Techgage Tested Settings (1)

Deus Ex: Mankind Divided as-tested settings

Deus Ex Mankind Divided - Techgage Tested Settings (2)

Deus Ex: Mankind Divided as-tested settings

Deus Ex Mankind Divided - Techgage Tested Settings (3)

Deus Ex: Mankind Divided as-tested settings

F1 2018 - Techgage Tested Settings (1)

F1 2018 as-tested settings

F1 2018 - Techgage Tested Settings (2)

F1 2018 as-tested settings

Far Cry 5 - Techgage Tested Settings (1)

Far Cry 5 as-tested settings

Far Cry 5 - Techgage Tested Settings (2)

Far Cry 5 as-tested settings

Far Cry 5 - Techgage Tested Settings (3)

Far Cry 5 as-tested settings

Forza Horizon 4 - Techgage Tested Settings

Forza Horizon 4 as-tested settings

Forza Horizon 4 - Techgage Tested Settings

Forza Horizon 4 as-tested settings

Forza Horizon 4 - Techgage Tested Settings

Forza Horizon 4 as-tested settings

Monster Hunter World - Techgage Tested Settings (1)

Monster Hunter World as-tested settings

Monster Hunter World - Techgage Tested Settings (2)

Monster Hunter World as-tested settings

Rise of the Tomb Raider - Techgage Tested Settings (1)

Shadow of the Tomb Raider as-tested settings

Rise of the Tomb Raider - Techgage Tested Settings (2)

Shadow of the Tomb Raider as-tested settings

Tom Clancy's Ghost Recon Wildlands - Techgage Tested Settings (1)

Tom Clancy's Ghost Recon Wildlands as-tested settings

Tom Clancy's Ghost Recon Wildlands - Techgage Tested Settings (2)

Tom Clancy's Ghost Recon Wildlands as-tested settings

For our apples-to-apples testing, the graphics settings seen above apply to every one of our tested resolutions so as to deliver easily comparable results. In most cases, each configuration is tested twice, with more runs added if the initial results make the extra testing necessary (which isn’t required too often).