AMD Radeon VII in the second test

AMD Radeon VII: Benchmarks mit aktuellen Spielen und (Async) Compute

The AMD Radeon VII is undoubtedly interesting as a 7 nm product with a 16 GB memory – despite its weaknesses. This second test on the topic contains new game benchmarks and measurements for DirectX 12, Async and Compute. The new games paint a somewhat better picture, but cannot change the conclusion.

Further tests with the Radeon VII

Before the fall of the press embargo, many players had high hopes for the new AMD Radeon VII (test). AMD's new flagship promised with the 7 nm process GPU and the 16 GB HBM2 memory technical highlights that no other graphics card can offer – and possibly RTX-2080 performance without memory limit at a lower market price.

In practice, however, the 3D accelerator has to struggle with problems: performance, volume, price and now also availability (for RRP). Based on the comprehensive first benchmarks and measurements, the GeForce RTX 2080 remained the more recommendable product in a direct comparison.

However, the editorial team was not yet able to take a look at some facets in the past week, and many top current games were also left out. BitcoinMinersHashrate is making up for it today.

More games, more Async Compute and more (or first time) Compute

The focus is on several new games such as Assassin's Creed Odyssey, Battlefield V, Call of Duty: Black Ops 4 as well as the topics DirectX 12, Vulkan and the associated Async Compute. The same applies to performance away from games: The speed with GPU compute.

The tests show that the Radeon VII has an easier time in the newest games than in the traditional test course. A tie is still not enough. This does not change the general assessment for the time being, because prices and availability in retail have developed negatively for the last few days.

Test system and test results

This test of the Radeon VII also uses the usual graphics card test system. An Intel Core i7-8700K overclocked to 4.7 GHz serves as the processor. Two 16-gigabyte DDR4 modules (dual-rank) are operated with DDR4-3200 at the timings 16-16-16-38-1T. Windows 10 with the October update (version 1809) and all patches were used for the benchmarks.

All three test series are completed by three graphics cards. In addition to the Radeon VII, the Radeon RX Vega 64 in reference design and the GeForce RTX 2080 Founders Edition take part in the comparison. The two AMD graphics cards are fired with the launch driver of the Radeon VII, the Nvidia accelerator with the GeForce 418.81.

Details of the game tests

The game benchmarks are all performed in 3,840 × 2,160. The level of detail is maximized, but performance is slightly reduced in Assassin's Creed Odyssey, Resident Evil 2, Hitman 2 and Just Cause 4.

New game benchmarks with Battlefield V, Resident Evil 2 and co.

It looks a little better for the Radeon VII in newer games. In 3840 × 2.160 the GeForce RTX 2080 FE is on average four percent faster, in the traditional graphics card test course it is nine percent. There is an even bigger increase in frame times. Instead of twelve percent, the gap in the new games is only three percent. Here, the Radeon VII can more than halve the gap to the Nvidia graphics card.

Read This Now:   Chinese factories close due to an increase in Covid-19 cases, with this production is paralyzed

Depending on how much a title can benefit from the high memory bandwidth, the performance increase compared to the Radeon RX Vega 64 is more or less clear. The plus of the Radeon VII to the Radeon RX Vega 64 climbs on average in the new course from 25 to 28 percent. However, what is more important is the bonus that Vega generally draws from the new title.

In the newer game selection, a new king has also been found for the Radeon VII: Call of Duty: Black Ops 4 runs with 20 percent more frames per second than with the GeForce RTX 2080 FE. In this case, however, this is not due to the generally good performance of Vega technology, because the Radeon VII is also significantly faster than the RX Vega 64 with an increase of 44 percent. The latest part of the action series can obviously benefit significantly from the doubled memory bandwidth , And also in Battlefield V the AMD graphics card performs five percent better, in Resident Evil it is still three percent.

In contrast, the GeForce RTX 2080 FE delivers significantly more FPS in Hitman 2, the plus is 17 percent. And in the new Apex Legends Free 2 Play shooter, the Nvidia accelerator is 15 percent faster. The remaining games (Assassin's Creed Odyssey, Fallout 76 and Just Cause 2), on the other hand, show the familiar picture, where the GeForce RTX 2080 is twelve to three percent faster.

DirectX 12, Vulkan and Async Compute

With DirectX 12 and Vulkan and with the Async Compute switched on or off, the Radeon VII behaves very similar to the Radeon RX Vega 64, which is not surprising due to the identical architecture.

AMD advertises improvements to Async Compute on the Vega 20 GPU, but this is hardly or not at all apparent in the tests. In fairness, however, it must be noted that Async Compute can only be switched on or off in the rarest of cases. Most of the time, the low-level feature is automatically switched on when DirectX 12 or Vulkan is activated, which blurs the image.

Read This Now:   Intel Vice President shows us the first official photograph of Alder Lake-S

In the best case, the Radeon VII with Async Compute performs one percent better than the Radeon RX Vega 64, which is within the range of the measurement inaccuracy.

It also shows that if the Radeon RX Vega 64 benefits from DirectX 12 or Vulkan, so does the Radeon VII. The leap in the not yet final beta of F1 2018, for example, is enormous, increasing by 24 percent on the Radeon RX Vega 64 through DirectX 12 and even by 29 percent on the Radeon VII. With the low-level API, the AMD flagship then achieves the same performance as the GeForce RTX 2080, which can only increase by eleven percent.

However, there are also games that are slower with DirectX 12 on all three graphics cards than DirectX 11. This is the case in Battlefield V, Resident Evil 2, Star Wars: Battlefront 2 and World of Warcraft. It is striking that games that were exclusively developed with a low-level API from the start work absolutely perfectly and also very quickly. This applies to Strange Brigade and Wolfenstein 2.

GPU compute when rendering with ray tracing

With Compute, the Radeon VII can often clearly differentiate itself from the Radeon RX Vega 64. AMD's new flagship in the Baikal benchmark is finished in just 69 percent of the time of its “predecessor”. Blender saves 40 percent of the time, while Luxmark scores 61 percent and 28 percent more points, respectively. Only in the VRay benchmark is the time saved only 15 percent. But that also shows that the Radeon VII can often benefit significantly more from the higher memory bandwidth at Compute than in games.

The GeForce RTX 2080 FE is nevertheless the faster graphics card at Compute, even if the backlog turns out to be significantly smaller than with the Radeon RX Vega 64. The Nvidia graphics card in the Baikal test is finished seven percent earlier, with Blender is still a clear 40 percent and 34 percent of Vray – Nvidia has made massive gains against Pascal with Turing at Compute. Only in Luxmark is the Radeon VII, with an increase of 67 percent compared to Vega 64, 12 percent ahead of the Turing GPU.

Read This Now:   Asus and Noctua announce the RTX 3070 Noctua Edition


With the additional benchmarks, the performance of the Radeon VII can now be classified even better than in the test a week ago. Among other things, they show that the graphics card is more dependent on the game selection than the Radeon RX Vega 64 and apparently the well-known graphics card test course tastes less good. In the course now used, the lead of the Radeon VII to the Radeon RX Vega 64 increases from 25 to 28 percent, while the gap to the GeForce RTX 2080 FE is reduced from nine percent to four percent. The frame times improve significantly in the same constellation, from a minus of twelve percent to a minus of three percent.

AMD's flagship still does not achieve the performance of a GeForce RTX 2080 FE and thus a slightly overclocked GeForce RTX 2080, but in the latest titles the performance level is at least at the level of a slow GeForce RTX 2080 and slightly above that of a GeForce GTX 1080 Ti ,

Game performance gets better, other problems remain

However, this does not change the recommendation for a GeForce RTX 2080. Yes, the Radeon VII is almost equally fast in new games – and even faster in Call of Duty: Black Ops 4. On average, the GeForce RTX 2080 is and remains ahead and the problems of volume and price (and meanwhile also availability) remain. If AMD still loosened one of the two screws, the Radeon VII would be worth considering in connection with positive features such as memory expansion and undervolting potential, but currently the GeForce RTX 2080 is still the graphics card of choice.

There are no special features in the benchmarks with DirectX 12, Vulkan and Async Compute. The Radeon VII behaves exactly like the Radeon RX Vega 64, which is not surprising due to the same basic architecture. There are no improvements in Async Compute in the benchmarks.

Compute is growing significantly in some cases

Compute, on the other hand, has this all the more. Since compute applications such as the tested rendering of a ray tracing image can benefit significantly more from a high memory bandwidth than games, the Radeon VII can outperform the Radeon RX Vega 64 in this area. A plus of around 50 percent or more is not uncommon, which can save a lot of time. However, Nvidia Compute has also improved significantly on Turing. There are applications where the Radeon VII can beat the GeForce RTX 2080. In general, the GeForce RTX 2080 offers slightly higher performance when rendering an image.

This article was interesting, helpful, or both? The editors appreciate any support in the form of deactivated ad blockers or a subscription to BitcoinMinersHashrate. More on the topic of ads on BitcoinMinersHashrate.

Notice: ob_end_flush(): failed to send buffer of zlib output compression (1) in /home/gamefeve/ on line 5373

Notice: ob_end_flush(): failed to send buffer of zlib output compression (1) in /home/gamefeve/ on line 5373