Md Ray
Member
I know, a bit late with the PC performance-related thread for RDR2 now, but since DLSS is right around the corner for this game and a lot of you might jump back into it to test it so I figured this might be helpful. Anyway, I'll cut to the chase.
Specs:
Ryzen 7 3700X
RTX 3070
16GB DDR4-3200
(This is not a graphics tweaking guide)
Graphics quality used: Custom (Ultra-High)
Resolution: 1440p
DX12: ReBAR: off, transfer queues: off, async compute: off - FPS: 78.13 - BASELINE
If you are playing the game using DX12 API on an Ampere GPU for some reason, then consider enabling resizable bar, if your system (motherboard+CPU) happens to support it. Guide here.
Let's enable ReBAR only.
DX12: ReBAR: on, transfer queues: off, async compute: off - FPS: 83.66 +7% BOOST!
Enabling transfer queues (more on this below) may add another 0.5 or 1fps further.
Let's switch to Vulkan API now with ReBAR off for a baseline:
VULKAN: ReBAR: off, transfer queues: off, async compute: off - FPS: 87.82 +5% BOOST!
You may think enabling ReBAR w/ Vulkan will carry over the perf improvement you saw with DX12, but that's not the case. I found no improvement w/ ReBAR on + Vulkan. In fact, there was a slight regression. Margin of error stuff, really. So it doesn't make much of a difference whether ReBAR is on or off when using Vulkan.
VULKAN: ReBAR: on, transfer queues: off, async compute: off - FPS: 87.46 -0.4% REGRESSION!
Now, let's enable transfer queues and keep ReBAR on as is!
VULKAN: ReBAR: on, transfer queues: on, async compute: off - FPS: 92.49 +6% BOOST!
An overall boost from DX12 baseline (78fps) to Vulkan w/ transfer queues enabled (92.5fps) is a nice, near +20% uplift with no impact to visual quality on Ampere architecture!
That's like a difference between a juddery, unstable ~50fps to a locked 60fps, all without reducing any graphics settings.
Here's how you enable transfer queues:
Documents > Rockstar Games > Red Dead Redemption 2 > Settings > system.xml
"true" = on
false = off
-Transfer queues is likely a Vulkan-specific feature as enabling/disabling this using DX12 made no difference in perf whatsoever, at least on an Ampere GPU, however, there was a noticeable uplift with VK API.
-Enabling ReBAR benefits DX12, but it made no difference with Vulkan (so I recommend keeping ReBAR on, anyway).
-Finally, async compute. Enabling this decreased perf by a small percentage using Vulkan. With DX12 - there were straight-up crashes every time on RTX 3070. So it's best to keep it off on NVIDIA hardware (Ampere specifically). I suspect it's meant for AMD GPUs. Those GCN and RDNA architectures will likely benefit from it. I feel even NVIDIA's Turing architecture may benefit from it for reasons. I don't know, play around with async compute if you have a non-Ampere card.
Async on vs off stats:
Enabled: 85.88.5fps
Disabled: 87.5fps
As usual, your mileage may vary, these are my findings using an RTX 3070, so these tweaks may not apply to, say, RDNA 2, and the performance characteristics will likely be completely different for you. I encourage you to experiment with different combinations of APIs, async compute on/off along with transfer queues, etc.
Specs:
Ryzen 7 3700X
RTX 3070
16GB DDR4-3200
(This is not a graphics tweaking guide)
Graphics quality used: Custom (Ultra-High)
Resolution: 1440p
DX12: ReBAR: off, transfer queues: off, async compute: off - FPS: 78.13 - BASELINE
If you are playing the game using DX12 API on an Ampere GPU for some reason, then consider enabling resizable bar, if your system (motherboard+CPU) happens to support it. Guide here.
Let's enable ReBAR only.
DX12: ReBAR: on, transfer queues: off, async compute: off - FPS: 83.66 +7% BOOST!
Enabling transfer queues (more on this below) may add another 0.5 or 1fps further.
Let's switch to Vulkan API now with ReBAR off for a baseline:
VULKAN: ReBAR: off, transfer queues: off, async compute: off - FPS: 87.82 +5% BOOST!
You may think enabling ReBAR w/ Vulkan will carry over the perf improvement you saw with DX12, but that's not the case. I found no improvement w/ ReBAR on + Vulkan. In fact, there was a slight regression. Margin of error stuff, really. So it doesn't make much of a difference whether ReBAR is on or off when using Vulkan.
VULKAN: ReBAR: on, transfer queues: off, async compute: off - FPS: 87.46 -0.4% REGRESSION!
Now, let's enable transfer queues and keep ReBAR on as is!
VULKAN: ReBAR: on, transfer queues: on, async compute: off - FPS: 92.49 +6% BOOST!
An overall boost from DX12 baseline (78fps) to Vulkan w/ transfer queues enabled (92.5fps) is a nice, near +20% uplift with no impact to visual quality on Ampere architecture!
That's like a difference between a juddery, unstable ~50fps to a locked 60fps, all without reducing any graphics settings.
Here's how you enable transfer queues:
Documents > Rockstar Games > Red Dead Redemption 2 > Settings > system.xml
"true" = on
false = off
-Transfer queues is likely a Vulkan-specific feature as enabling/disabling this using DX12 made no difference in perf whatsoever, at least on an Ampere GPU, however, there was a noticeable uplift with VK API.
-Enabling ReBAR benefits DX12, but it made no difference with Vulkan (so I recommend keeping ReBAR on, anyway).
-Finally, async compute. Enabling this decreased perf by a small percentage using Vulkan. With DX12 - there were straight-up crashes every time on RTX 3070. So it's best to keep it off on NVIDIA hardware (Ampere specifically). I suspect it's meant for AMD GPUs. Those GCN and RDNA architectures will likely benefit from it. I feel even NVIDIA's Turing architecture may benefit from it for reasons. I don't know, play around with async compute if you have a non-Ampere card.
Async on vs off stats:
Enabled: 85.88.5fps
Disabled: 87.5fps
As usual, your mileage may vary, these are my findings using an RTX 3070, so these tweaks may not apply to, say, RDNA 2, and the performance characteristics will likely be completely different for you. I encourage you to experiment with different combinations of APIs, async compute on/off along with transfer queues, etc.
Last edited: