Which nvidia gpu generation began to loose in performance to amd gpus in cpu limited…
- User
Which nvidia gpu generation began to loose in performance to amd gpus in cpu limited scenarios?
Assistant
NVIDIA GPUs started showing relative performance losses to AMD GPUs in CPU-limited scenarios beginning with the RTX 40 series (Ada Lovelace), particularly when compared to AMD's RDNA 3 GPUs like the RX 7900 XTX and 7900 XT.
This happens because NVIDIA's recent driver architecture has higher CPU overhead compared to AMD's implementation. In scenarios where the CPU becomes the bottleneck (typically at lower resolutions like 1080p with high frame rates, or in CPU-intensive games), AMD GPUs can sometimes deliver better performance despite having lower raw GPU compute power.
The issue became more noticeable with:
- RTX 4090 and 4080 showing smaller performance advantages over AMD competitors at 1080p than at 4K
- Some esports titles and CPU-heavy games where AMD GPUs could match or exceed NVIDIA performance at lower resolutions
- Situations with older or lower-end CPUs where the driver overhead difference becomes more pronounced
However, this is highly dependent on the specific game, API used (DX11 tends to show bigger differences than DX12/Vulkan), and the rest of the system configuration. NVIDIA has been working on reducing driver overhead, but the architectural differences mean AMD still tends to have an advantage in CPU-limited scenarios.
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.