AMD's New DX11 Enhancing Preview Driver Underwhelms in 50 Game Test Suite

News
By
published

Only 3% faster across 50 titles

AMD
(Image credit: AMD)

AMD's new Software Preview Driver for May 2022, has brought about some astonishing performance gains for several DX11 titles, with some titles gaining a 30% boost in minimum fps. However, such gains may not be very common, based on the results of a 50-game benchmark suite tested by Techspot. AMD's miracle-working driver helped in a few cases, but the 'magic' was clearly lacking elsewhere.

AMD's claim of 10% higher graphics performance in DX11 titles only applies to a few select games, less than 10 to be more precise. Techspot's 50-game average revealed that the preview driver provided only minor performance benefits over the current Adrenalin 22.5.1 driver in general.

It's worth noting that Techspot tested both DX11 titles and DX12 titles in its suite of 50 games. AMD specifically stated that it's May 2022 preview driver only boosted games that use the older DX 11 API. Even there, Techspot found several DX11 games that didn't improve much if at all, though elsewhere a could of DX12 titles showed some decent improvements.

Techspot used a Ryzen 9 5950X testbed with 32GB of dual-rank memory powering an AMD Radeon RX 6700 XT. Due to the sheer amount of games being tested, only one card was tested rather than checking multiple GPUs. The card was tested with the current 22.5.1 drivers along with the preview driver, at 1080p, 1440p, and 4K.

Out of the 50 games tested, only about one third showed more than a minor performance benefit with the preview driver, and two-thirds showed a 1% or higher improvement. The rest had no benefit or incurred a slight performance penalty. That's for the 1080p testing, and the 1440p and 4K results showed even lower gains.

At 1440p, 20 of the 50 games saw zero performance improvement from the preview driver, with four games suffering a performance deficit. At 4K, this pattern extends to 24 titles, or basically half of the test suite.

Games that showed notable performance increases include Watch Dogs: Legion (it's not clear if that was in DX12 or DX11 mode), Death Stranding (a DX12-only game), The Witcher 3, Apex Legends, and Forza Horizon 5 (another DX12-only game). All of those saw at least a 5% or higher bump in performance at all three resolutions. Watch Dogs: Legion and Death Stranding were the only games to get above the 10% mark with the preview driver at 1080p and 1440p, while the largest improvement at 4K was 9%.

Interestingly, Godfall consistently showed worse performance. That's another DX12-only game, and performance was 6-7% lower at 1440p and 4K. That's an AMD-promoted game, but it's always been a bit inconsistent in our own testing. Plus, this is likely a "preview" driver for good reason.

Overall, the preview driver averaged a 3% performance benefit over the current Adrenalin 22.5.1 driver at 1080p, and just 2% at 1440p and 4K. If you happen to play one of the handful of games that showed larger gains, you might appreciate the improvements, but don't expect miracles across a broader selection of titles.

Aaron Klotz
Aaron Klotz
Contributing Writer

Aaron Klotz is a contributing writer for Tom’s Hardware, covering news related to computer hardware such as CPUs, and graphics cards.

15 Comments Comment from the forums
  • Makaveli
    Is tomshardware not going to do their own review on this driver?
    Reply
  • Alvar "Miles" Udell
    Is anyone really surprised a driver doesn't give a massive performance boost across the board for a card that's been on the market for over a year, especially when those cards are using an architecture that's just a refresh of a five year old architecture?

    Reply
  • Foxlum
    It's really a good thing we even get improvements as cards age. 3-10% performance boost in most titles is impressive in a preview driver, especially because it's a wide range of affected DX11 and DX12 games.
    Reply
  • lothear
    This lazy review is underwhelming. DX11 should be tested exclusively, by including DX12/vulkan you muddle the results. Every other tech site review separate that data. Very lazy.
    Reply
  • -Fran-
    lothear said:
    This lazy review is underwhelming. DX11 should be tested exclusively, by including DX12/vulkan you muddle the results. Every other tech site review separate that data. Very lazy.
    This is a very good point, actually. Didn't see it that way before.

    AMD never said DX12 or Vulkan would be improved, so the conclusion by including them is misleading. Looking at only DX11, the driver does give a good impression to me. Free performance, even when minimal, shouldn't be scoffed at.

    Regards.
    Reply
  • hannibal
    Well, in anyway not bad to get improvements :)
    But also we know that the test was not pure dx11 test,, but gamers play games, not DX or Vulcan versions, so this give a picture what overall increase is. But it would be nice to see pure DX11 increase also.
    Reply
  • waltc3
    AMD never said it would improve all of your D3d11 games, and AMD listed the "select" games they said showed improvement. It's hard to see how it "underwhelms" when you investigate a claim AMD never made...;) Indeed, you said a 30% improvement in a game you tested--way above what AMD claimed in its game list. So, it really doesn't matter if all D3d11 games improve, does it? And yet, other people tested and found results that differ from yours, apparently. I'm quite sure that improvements exist in more games than are on the list AMD supplied with the driver, however.

    I'm using it now--I prefer the May Performance driver to 22.5.1 WHQL.
    Reply
  • hotaru.hino
    Alvar Miles Udell said:
    Is anyone really surprised a driver doesn't give a massive performance boost across the board for a card that's been on the market for over a year, especially when those cards are using an architecture that's just a refresh of a five year old architecture?
    But AMD cards supposedly "age like fine wine." 🙃
    Reply
  • KananX
    lothear said:
    This lazy review is underwhelming. DX11 should be tested exclusively, by including DX12/vulkan you muddle the results. Every other tech site review separate that data. Very lazy.
    On top they used a weaker GPU in 6700 XT which is of course way less bottlenecked than a 6900 XT. I think hardware unboxed made a better video about this same topic.
    Reply
  • -Fran-
    KananX said:
    On top they used a weaker GPU in 6700 XT which is of course way less bottlenecked than a 6900 XT. I think hardware unboxed made a better video about this same topic.
    TechSpot = HardwareUnboxed.

    Regards.
    Reply