Great Reading If You’re Having Trouble Sleeping
NVIDIA hasn’t been pushing the envelope with Blackwell, and that hasn’t changed for the RTX 5060 Ti, RTX 5060 or RTX 5050 if the leaks are to be believed. The RTX 5060 Ti will continue the tradition of releasing in both 8GB and 16GB flavours, with both offering a 128-bit memory interface, 4,608 CUDA cores and a 180W TDP which suggests it will have slightly higher frequencies than the previous generation. The RTX 5060 and RTX 5050 will both only have 8GB, the 5060 with GDDR7 but the RTX 5050 will continue to use GDDR6 but both will keep a 128-bit bus. You will see 3,840 CUDA cores on the RTX 5060 which is a nice jump from the 3,072 of the RTX 4060 and means a large bump in TDP to 150W. The RTX 5050 on the other hand scrapes by with 2,560 CUDA cores though that does keep the TDP to a reasonable 130W.
The reasons for these rather milquetoast specifications, assuming the leaks are accurate of course, are varied but can likely be guessed. The biggest is HPC and AI, for NVIDIA makes far more money on those parts compared to their gaming cards and it only makes sense to focus on those products. The problem is that this means less resources to design cutting edge consumer GPUs. The second likely reason is pricing, with the current state of the GPU market you have to be very careful about how much a mid-level GPU should cost. The word “should” is very important, for it is almost impossible to find an NVIDIA card, let alone one close to what the announced MSRP was. In theory a lower MSRP will also mean the ridiculous markups we have to put up with will be a little less painful. It is also important to keep the cards affordable so that system builders can sell their custom builds at a price that consumers will find acceptable.
The final likely reason might well be the saddest one, why would they bother pushing the envelope. The RX 9070 and 9070 XT are great cards, but their performance isn’t up to the RTX 5080 and nowhere near the RTX 5090. Intel’s Arc cards are great for their theoretical prices, but again they don’t hold a candle to NVIDIA’s cards. Without competition driving NVIDIA to push the envelope on their mid-range lines, why would they expend their limited resources to do so?
One thing to hope for is that these cards can be made in vast quantities, and will actually be available once released.
