https://videocardz.com/newz/geforce-rtx-4090-limited-to-120hz-on-samsung-odyssey-neo-g9-dual-4k-240hz-monitor The Odyssey G9 Neo, a flagship offering from Samsung, boasts a 32:9 aspect ratio, DisplayHDR 1000 certification, and a 1000R curvature. Notably, this monitor delivers dual 4K resolution (7680×2160) with a 240Hz refresh rate, provided the GPU can support it. Regrettably, only some GPUs meet this requirement. The Odyssey G9 Neo was unveiled just prior to the release of the Radeon RX 7800 series, announced as the first monitor in the market to adopt the DisplayPort 2.1 standard. However, the adoption of DisplayPort 2.0, later upgraded to 2.1, has been slow, to the extent that NVIDIA opted not to support this standard with their RTX 40 series nearly a year ago.
QuasarZone’s testing has revealed that neither NVIDIA nor Intel graphics cards can fully support the 7680×2160 resolution at a smooth 240Hz refresh rate through their HDMI 2.1 implementation, despite the use of the full 48 Gbps bandwidth. The exact reason for this limitation remains unclear, but it is speculated that NVIDIA GPUs may lack sufficient Display Stream Compression (DSC) pipelines to handle it.
It should be noted that 8K/120p stream with DSC requires 32 Gbps. The Odyssey monitor only needs half of the pixels (it’s dual 4K, while 8K is quad 4K) but at twice the refresh rate. Exactly why RTX 4090 is not able to output at 240Hz is not clear, but
Redditor rushes with a possible explanation:
DSC uses display pipelines within the GPU silicon itself to compress the image down. Ever notice how one or more display output ports will be disabled when using DSC at X resolution and Y frequency? That is because the GPU steals those display lanes to process and compress the image.
So what does this mean? It means if the configuration, in silicon, does not allow for enough display output pipelines to be used by a single output port, THAT is where the bottleneck occurs.
But there are deeper things with DSC than bandwidth. There is also how the compression is done, both ratio wise and slice wise. DSC will happily allow a 3.75:1 ratio for 10 bit inputs as long as the driver/firmware of the GPU allows for it (as it is part of DSC spec). Nvidia’s VR API tools for developers only allow for a max of 3:1 it should be noted.
The allowable slice dimensions and count (how the screen is divided for compression) also determines how much throughput can be achieved (by way of increasing parallelism during compression). This is a silicon/hardware limitation, although again, could be limited by firmware.
The Samsung G9 monitor serves as an example of why NVIDIA’s decision to forego DP 2.1 plans was not the right decision, yet also why it may not be a significant concern. Firstly, the Odyssey G9 Neo commands a high price of $2,499 in the US, surpassing even the cost of an RTX 4090 GPU. Furthermore, the NVIDIA RTX 4090 card faces no obstacles in supporting high refresh rates at 4K resolution or ultra-wide DQHD (dual 1440p), which has been gaining more attention from gamers.
I think next gen NVIDIA GPUs will have DP 2.0 or higher.
