RTX 5070 Ti Specs May Have Leaked, GPU Up to 300W

Nvidia has yet to officially announce plans for its next-generation high-end RTX GPU, which is expected to be called the RTX 5070 Ti, but leaks continue to pour in.

The latest batch of reports from major tech leakers contain quite a bit of interesting information about what appears to be an RTX 5070 Ti being actively developed by Nvidia. Based on the new specifications leaked from reputable sources, NVIDIA's GeForce RTX 5070 Ti will be a great high-end GPU choice for gamers.

A few weeks ago, Videocardz reported on the new specs of the NVIDIA GeForce RTX 5070 Ti, which talked about its GPU core count and TGP. Now, we have more information about this particular variant.

Starting with previously shared details, the NVIDIA GeForce RTX 5070 Ti is rumored to pack 8960 cores into one of the GB203 SKUs it's expected to use. The reference board is said to be labeled "PG147 SKU 60," and will be a 300W TGP design. It appears that NVIDIA has changed the specs a bit from previous rumors.

The NVIDIA GeForce RTX 5070 Ti will most likely use the GB203-300 GPU SKU, and the specific board will actually be the "PG147 SKU 60." The GB203 GPU is also the chip that will power the higher-end RTX 5080 (GB203-400). The RTX 5080 will use the full configuration, while the RTX 5070 Ti will run a cut-down variant.

Other details we can now confirm are a TBP increase to 350W, up 50W from the previously rumored specs. This would mark a 23% higher TBP than the RTX 4070 Ti and RTX 4070 Ti SUPER, both of which are rated at 285W. It should be noted, however, that this is just the TBP figure and the average power consumption will be lower depending on the application.

Next up is the memory design. Since the NVIDIA GeForce RTX 5070 Ti will be using the GB203 GPU, it will come with a 256-bit memory bus resulting in 16GB of GDDR7. The memory will be configured at 28Gbps, resulting in a total bandwidth of 896GB/s. This is a 33% improvement in bandwidth over the RTX 4070 Ti SUPER.

Back to the core count issue. The RTX 5070 Ti is said to have 16% more cores than the RTX 4070 Ti, which is a significant increase, but only 6% more than the 4070 Ti Super. However, when compared to the 4080, we see that the 5070 Ti may fall behind current-gen GPUs, as it is said to have nearly 8% fewer cores.

With so many architectural improvements in Blackwell, the CUDA core difference may not matter. There's also the performance-per-dollar metric to consider here, which could give the 5070 Ti the edge.

As for the launch timing, it looks like the NVIDIA GeForce RTX 5070 Ti will be the third product to hit the retail market this year instead of the RTX 5070. NVIDIA's internal plan is to launch all six RTX 50 "Blackwell" models in the first half of 2025, with the majority of the products launching in Q1. So January through June next year will be a very busy first half of the year for the market, with a slew of AMD & Intel GPUs also expected to launch in the same timeframe.

Nvidia is expected to launch the RTX 50 series at CES 2025 in January next year, so it won't be long before we get some confirmation on the full specifications of each GPU.