For this testing, only official PC games with in-game benchmarks will be used that I own; only official Game Developers FPS counters will be used.
The In-game benchmarks will show less difference between GPUs because Game Developers include a representative cross-sections of their PC game. Most PC games have entire segments of maps that are less demanding for GPUs, which wipe out a portion of the FPS differences between GPUs around similarly performance points compared to the most demanding parts of PC games.
For example, Hitman 2 has 7 maps and only 2 maps are demanding of GPUs: a) Miami; b) Mumbai. 71% of this game is played on maps that will show low differences between GPUs around a similar performance metric.
AMD Labs' released FPS result for 25 PC games at 3840x2160p. The average difference was 28.59% between the RX Vega 64 LC versus Radeon VII. The switch to in-game benchmarks will show around a 4% lower difference between the two GPUs.
New target Metric=24.59%
Next, Fallout 76 (68.35%), Doom 2016 (33.18%) Battlefield One (35.98%), Battlefield V (33.19%), The Witcher 3 (33.82%) cannot be benchmarked as they do not have the Official Benchmarks Tool provided by their Game Developers, which will reduce FPS difference by 3.3%.
New Target Metric=21.3%
Finally, my Powercolor Red Devil Vega 64 averages out to be 2.7% slower than the RX Vega 64 LC model on the stock Power Plan used by AMD Labs.
New Target Metric=24%
Therefore, using Official Benchmarks Tools, we should expect to see around a 24% performance increase between these two GPUs:
1) Powercolor Red Devil RX Vega 64 100%
2) Radeon VII 124%
Test SystemRyzen 2700X PBO - Powerdraw allowed up to 150watts.
DDR4-3421-CL14 Low Latency Subtimings at 1.47volts.
7 PC case fans and one 240mm AIO for the CPU.
Gigabyte Auros X470 7 WiFi (BLCK 100.7) with onboard Audio used.
500GB SSD.
Corsair 850Watt Platinum PSU 94% Efficiency.
GPUs
Powercolor Red Devil Vega 64
Adrenalin 19.2.1 335watts Averaged Out -/+ 10watts depending on the PC game Engine.
Performance 106.3% over Reference Blower Vega 64
Sapphire Radeon VII (AMD manufactured)
Adrenalin 19.2.1 255watts Averaged Out -/+ 10watts depending on the PC game Engine.
Testing with a Watt-Meter from the Plug Socket
Default Power Plans for both GPUs.
Strange Brigade Benchmark - every option set to Ultra
Red Devil Vega 64 Total System Power Draw =501watts
Radeon VII Total System Power Draw =390watts
110watt power draw reduction during the Strange Brigade Benchmark.
Shadow of the Tomb Raider - Highest Preset.
Red Devil Vega 64 Total System Power Draw =484watts
Radeon VII Total System Power Draw =404watts
80watt power draw reduction during the Shadow of Tomb Raider Benchmark.
A total of 10 PC games were measured at the wall and Power Draw reductions where observed in every single PC game and average out around 80watts amount.
FPS Result Performance Increase at 3840x2160p 13 GamesAshes of the Singularity Vulkan Crazy Preset
Red Devil Vega 64 =49FPS (100%)
Radeon VII =58.3FPS (119%)
AC Odyssey highest Preset (Cloudless Day)
Red Devil Vega 64 =34FPS (100%)
Radeon VII =41FPS (120.5%)
Deus EX Mankind DX12 Divided Ultra
Red Devil Vega 64 =31.3FPS (100%)
Radeon VII =39.6FPS (126.5%)
Far Cry 5 Ultra TAA
Red Devil Vega 64 =47FPS (100%)
Radeon VII =60FPS (127.5%)
Forzia Horizon 4 Ultra
Red Devil Vega 64 =66.7FPS (100%)
Radeon VII =75.9FPS (114%)
Hitman DX12 Ultra
Red Devil Vega 64 =64.30FPS (100%)
Radeon VII =78.15FPS (121.5%)
Hitman 2 Ultra Mumbia Benchmark
Red Devil Vega 64 =50.5FPS (100%)
Radeon VII =62.47FPS (124%)
Middle-Earth Shadow War HD Textures 8.8GBs
Red Devil Vega 64 =43FPS (100%)
Radeon VII =53FPS (123%)
Rise of the Tomb Raider DX12 Highest Preset
Red Devil Vega 64 =49.95FPS (100%)
Radeon VII =58.88FPS (118%)
Rainbow Six Seige Ultra 100% TAA and 100% Render Scaling
Red Devil Vega 64 =51FPS (100%)
Radeon VII =65FPS (127.5%)
Shadow of Tomb Raider DX12 Highest Preset
Red Devil Vega 64 =36FPS (100%)
Radeon VII =45FPS (125%)
Strange Brigade DX12 Ultra
Red Devil Vega 64 =64FPS (100%)
Radeon VII =87FPS (136%)
The Division DX12 Highest Preset
Red Devil Vega 64 =48.1FPS (100%)
Radeon VII =59.9FPS (124.5%)
The Performance Increase over the 13 games is 23.6%. Therefore, no real surprise in the performance delta for the Radeon VII, but a big surprise in power consumption.
Returning to power consumption, Radeon VII draws identical power to a GTX 1080TI Founders Edition for PC gaming and about 30watts more than the RTX 2080 Founders Edition.
The 300watt TDP rating appears to be Compute or OpenCL productivity watt limit for things like Blender, Luxmark, etc.
Ryzen 2700X plus Radeon VII Comparative analysisFar Cry Ultra TAA (RTX 2080 FE with I9 7920 X @4.4 GHZ All Core)
RTX 2080 FE =56FPS
https://www.youtube.com/watch?v=b03PqtYUp-0Far Cry Ultra TAA (Ryzen 2700X PBO)
Radeon VII =60FPS
The performance increase is 7.1% in the Official Benchmark for Far Cry 5.
Forza Horizon 4 Ultra (Intel I7 6700K at 4.7GHz with RTX 2080 OC)
RTX 2080 OC =69.9FPS
https://www.youtube.com/watch?v=CDoepsPwFlAForza Horizon 4 Ultra (Ryzen 2700X PBO)
Radeon VII =75.9FPS
The performance increase is 8.6% in the Official Benchmark for Forza Horizon 4.
Strange Brigade all settings at Ultra (Intel Core i7 8700K at 5GHz)
Asus Strix RTX 2080 O8G Gaming =67FPS
https://www.youtube.com/watch?v=EIeFoG4k_IYStrange Brigade all settings at Ultra (Ryzen 2700X PBO)
Radeon VII =87FPS
The performance increase is 29.8% n the Official Benchmark for Strange Brigade.
As can be seen, for owners of Ryzen CPUs, buying the Radeon VII is a bit of a no-brainer decision, as expected the Radeon Drivers perform better with a Ryzen CPUs installed than there Intel CPU counterparts at 3840x2160p.
Notes on testing:Hitman 2 in-game benchmark throws away the first 5 seconds of the Simulation; there is an FPS cap of 45FPS at the beginning of the Simulation as the graphics engine initializes assets for Vega GPUs. However, the graphics engine can take up to 8 seconds to initialize the assets with the 45FPS framecap in place for Vega GPUs. Therefore, the Game Developers should increase the starting dustbin on FPS results to 10 seconds in the Simulation to give the final pop up result reading more consistency between runs of the Simulation.
