NVIDIA prepares a graphics card with 7,424 shaders, it could be the RTX 3070 Ti

It seems that NVIDIA is preparing to respond to the launch of the new Radeon RX 6000 graphics cards from AMD, and that the RTX 3070 Ti could be one of the most important novelties of the American company.

In a recent leak, collected by VideoCardz, we have seen that NVIDIA has in hand a version of the GA102 graphics core, the same one used by the RTX 3080 and RTX 3090, which would have a total of 7,424 active shaders.

This GPU would be used to shape the RTX 3070 Ti, a graphics card that would be much more powerful than the RTX 3070, since the latter uses the GA104 core and has 5,888 active shaders.

In addition to the difference in the number of shaders, which as we see is quite large, it is said that the RTX 3070 Ti would have a 320-bit bus like the RTX 3080, which implies an increase in bandwidth compared to the RTX 3070.

The RTX 3070 Ti would have higher bandwidth and more power, but it would also have more graphics memory. According to the source, this graphics card would have a total of 10 GB of GDDR6X memory at an unspecified frequency, but we imagine that it should be around 18 or 19 GHz.

This graphics card should be NVIDIA’s answer to the Radeon RX 6800 XT, a model that has already leaked in some benchmarks, and which has offered results that are quite difficult to understand, as it has significant irregularities.

READ ALSO  The Nvidia GeForce RTX 3090 is 49.4% faster mining Ethereum vs the RTX 3080

Right now, it is best to wait to see the actual performance tests in AMD’s next-generation games before drawing conclusions. The only reliable thing we have is the data that AMD offered during the Zen 3 presentation. At that event, the Big Navi performance was seen in three games, and it was confirmed that it was slightly below the RTX 3080.

Specifications of the supposed GeForce RTX 3070 Ti
ModelRTX 3080RTX 3070 Ti?RTX 3070
Base platePG132 SKU 10N / APG142 SKU 10
GPU8nm GA102-2008nm GA102-1508nm GA104-300
Die Size 628 mm2 628 mm2 392 mm2
Transistors 28.3 B 28.3 B 17.4 B
CUDA Cores 8704 7424 5888
Silicon Occupation 81% 70% 96%
Tensor Cores 272 232 184
RT Cores 68 58 46
Freq. Base 1440 MHzN / A 1500 MHz
Freq. Turbo 1710 MHzN / A 1725 MHz
Shader Perf. 29.8 TFLOPSN / A 20.3 TFLOPS
Memory 10GB G6X 10GB G6X 8GB G6
Freq. Memory 19 GbpsN / A 14 Gbps
Memory interface 320 bits 320 bits 256 bits
Bandwidth 760 GB / sN / A 448 GB / s
TDP 320WN / A 220W