Nvidia has silently disabled the Peer-to-Peer functionality in its regular high-end graphics cards – indirectly forcing users to upgrade to more expensive ones for this support.
This was revealed in a test by Puget Systems, where RTX 4090 GPU is found to have a faulty P2P system – and later confirmed by an Nvidia representative in a forum! Businesses relying on this feature for their overhead works are now forced to buy a much more expensive card – like the RTX A6000 GPU.
P2P Functionality in Nvidia GPUs
Puget Systems, a leading custom PC builder, has recently tested the AMD and Intel PC server CPU-based systems with NVIDIA GeForce GPUs from the last generation and the current generation to determine how good they’re in overall features.
During the test, Puget found out that the Peer-to-Peer (P2P) functionality in Nvidia GeForce GPUs was ripped off in the latest generations while dedicating this support only to many experience cards. The test included comparing RTX 6000 Ada GPU vs. GeForce RTX 4090 GPU, where the latter has its P2P support ripped off!
And this is official, as noted by an Nvidia representative in the community forums. This sparked a debate on why Nvidia did this without any wide announcement to the public, as many businesses have relied on this for their overhead works.
In the past few years, Nvidia has seen most of its GeForce RTX GPUs being used for other purposes (content creation and cryptocurrency mining) than the intended purpose (gaming). And one of the major factors that enabled them to do it is the P2P support – which Nvidia wanted to cash on – or fairly, wanted to stop from exploiting.
This led the company to rip this feature off from the GeForce RTX GPUs – used by general users- and allocate it only for the high-end resourceful GPUs like the RTX 6000 Ada cards used by servers. Thus, it’s an indirect push by Nvidia to move users to buy more expensive cards for popular features.