AMD Radeon RX Vega Benchmark Review: Vega 64 and Vega 56 Tested

Jump To:

Test System

Before we look at the numbers, let’s take a brief look at the test system that was used as we switched over to a new system in April 2017 when Windows 10 Creators Update was released. All testing was done using a fresh install of Windows 10 Pro 64-bit version 1703 and benchmarks were completed on the desktop with no other software programs running.  There has been some concern of people testing a cold card versus a hot card, but we’ve always done out testing ‘hot’ since the site started back more than a decade ago.

Video Cards & Drivers used for testing:

  • AMD Radeon Software Crimson Edition 17.7.02 for Radeon Cards
  • NVIDIA  GeForce 385.01 for GeForce Cards

Intel X99 Platform

The Intel X99 platform that we used to test the all of the video cards was running the ASUS X99-E-10G WS motherboard with BIOS 0603  that came out on 03/15/2017. We went with the Intel Core i7-6950X Broadwell-E processor to power this platform and overclocked it up to 4.0GHz on all cores. The Corsair Vengeance LPX DDR4 memory kit we used was a 64GB kit (4x16GB) and while it is rated at 3600MHz we actually ran it at 3333MHz at 1.30V with 16-16-16-30 1T memory timings. The Samsung SSD 960 EVO 1TB M.2 PCIe NVMe SSD was run with latest firmware available. A Corsair RM1000x power supply provides clean power to the system and is also silent as the fan hardly ever spins up. This is critical to our testing as it lowers the ambient noise level of the room and gives us more accurate sound measurements.

Here are the exact hardware components that we are using on our test system:

The Intel X99 Test Platform

Component

Brand/Model

Live Pricing

Processor Intel Core i7-6950X
Motherboard
ASUS X99-E-10G WS
Memory
64GB Corsair Vengeance LPX 3600MHz DDR4
Video Card Various
Solid-State Drive Samsung SSD 960 EVO 1TB
Cooling Corsair Hydro H115i
Power Supply Corsair RM1000x
Case HighSpeed PC Top Deck Tech Station
Operating System Windows 10 64-bit
Monitor ASUS PB287Q 28″ 4K

 

Let’s move on to the Battlefield 1 benchmark results!

Print
Jump To:
  • chrisday85

    Ok, here’s some things ticking me off: The Vega 64 is literally twice the gflops, and twice the memory bandwidth of the RX 580. The gap from the Vega 64 to RX 580 should be huge. The GTX 1060 to 1080 may be basically as large Gflops wise, but is minimal memory bandwidth wise. And what do you notice in these benchmarks? What the bloody hell is going on? The Vega is a new architecture, it’s supposed to be better in all sorts of ways. Yet here we see the same architecture generation from Nvidia performing much better in line with the increases. Allow me to demonstrate:

    Cross multiply Nvidia GTX 1060 Gears 4 1080, using the RX 580 and you find the AMD if scaling as well, should be average 157 in the frame rate, before accounting for any leap from the new architecture. What does it turn in? 128. Nvidia is scaling better in the same generation than AMD is in the new one?

    Fallout IV: It should be 125 if it scaled as well as the NVIDIA card. It is instead 112.

    Ghost Recon:

    95 instead of 106

    Battlefield:

    131 instead of 153

    Deus Ex:

    112 instead of 125.

    Grand Theft Auto is the only one it scales a tiny bit better.

    It’s scaling WORSE in DX 12 than it is in DX 11 I might add, do the math.

    I bought the Vega 64 and I have to say I’m disappointed. I’m hoping these beta drivers are the reason, because this is absurd, and I might add, the card gets to 86c unless I turn the fans to “wind turbine” as well. I also seem to be having an issue with Batman Arkham Knight in frame rate, which I’m just going to chalk up to some weird sort of glitch, because it’s nowhere near where benchmarks peg it. This has got to be one of AMD’s most disappointing launches of a new architecture I have seen.

  • Stephen

    test system, 6950x ivy bridge-e? error

    • Nathan Kirsch

      haha, winner! Fixed and thank you.

  • Sean Kumar Sinha

    Was hoping for a card that’d make NVidia push Volta and create a realistic GPU war. Unfortunately, we got cards that compete with stuff that NVidia has had out for over a year and that draw too much power, in comparison. These are good cards, for sure, but they are a year too late. Vega 56 seems like a good option for some 1080P Freesync gaming, though.

  • James Yarno

    Ethereum hash rate?