RTX 4060 Ti ‘AD106’ GPU Benchmarked Extensively, As Fast as the RTX 3070 and RTX 2080 at Just 180W
Over at chiphell, user panzerlied has posted a thread showcasing the performance of the AD106 GPU. This GPU is set to be used in NVIDIA’s upcoming RTX 4060 Ti and the RTX 4060. We are not sure about the actual GPU used. However, we can suppose that the user got their hands on an RTX 4060 Ti, as the specs resemble a full-fat AD106.
RTX 4060 Ti ‘AD106’ GPU Benchmarks
As a reminder, the user mentions that these benchmarks are not a representation of the actual performance. In fact, the performance will be better on launch. For the sake of comparison, the user also benchmarked the RTX 3070 (GA104) and the RTX 2080 (TU104). Spoiler alert, the RTX 4060 Ti is extremely promising, being as fast as both the aforementioned GPUs.
Starting off with the specifications, the AD106 GPU will feature 4608 Cuda Cores amounting to 36 SMs. We currently do not have any GPU that fits this criterion, so the GPU used may be an early engineering sample. In addition, the full-fat AD106 chip can boost to a whopping 2.535GHz. Additionally, the memory used is supposedly GDDR6 running at 18Gbps over a narrow 128-bit memory bus. The total effective bandwidth accumulates to 228GB/s, much slower than the other GPUs.
Harukaze compiled this data into a single data graph to show us the relative performance of each GPU. Surprisingly, AD106 is quite competitive and trades blows with high-end SKUs of the last 2 generations. In Time Spy Extreme, the GPU manages to amass 6651 points, which despite not being record-breaking is impressive. It loses to the RTX 3070 by a very narrow margin but is significantly faster than the RTX 2080.
What we can conclude from this is that the RTX 4060 Ti (full-fat AD106) is faster than the RTX 2080 and on par with the RTX 3070. Better yet, it consumes just 180W of power in contrast to the ~240W consumed by both the RTX 2080 and the RTX 3070. In mathematical terms, the RTX 4060 Ti is equal to the RTX 2080/3070 whilst consuming 27% less power.
Conclusion
The results are in front of us and the conclusion is rather simple. The RTX 4060 Ti will be the go-to choice for budget gamers, provided NVIDIA prices it accordingly. It is faster than the RTX 2080 and equal to an RTX 3070 at least in raw raster performance. This coupled with DLSS 3.0 will make it extremely intriguing for the budget market.
AMD also has various options in the making, coming from their Monolithic Navi33 GPU. Since AMD has a cheaper design, we expect them to be competitive, at least with Navi32 and Navi33. The ripple effect will cause NVIDIA to lower prices, though that entirely depends on how Navi32 and Navi33 perform in the first place.
SKU | Chip | FP32/CUDA | SMs | Max Clock | Cache | Memory Bus | VRAM | Memory Spec | Speed (Gbps) | TDP | |||
RTX 40 Titan | AD102-450 | 18432 | 144 | 2.4GHz? | 96MB | 384 | 48GB | GDDR6 ECC | – | 800W | |||
RTX 4090 Ti | AD102-400-A1 | 18176 | 142 | 2.6GHz? | 96MB | 382-bit | 24GB | GDDR6X | 24 | 600W? | |||
RTX 4090 | AD102-300-A1 | 16384 | 128 | 2.52GHz | 96MB | 384-bit | 24GB | GDDR6X | 21 | 450W (TGP) | |||
RTX 4080 Ti | AD102 | 14848 | 116 | 2.7 GHz? | 80MB? | 320-bit | 20GB | GDDR6X | 23 | 420W | |||
RTX 4080 | AD103-300-A1 | 9728 | 76 | 2.505GHz | 64MB | 256-bit | 16GB | GDDR6X | 22.5 | 320W(TGP) | |||
RTX 4070 Ti | AD104-400-A1 | 7680 | 60 | 2.61GHz | 48MB | 192-bit | 12GB | GDDR6X | 21 | 285W (TGP) | |||
RTX 4070 | AD104-251-A1 | 5888 | 46 | 2.475GHz | 36MB | 192-bit | 12GB | GDDR6X | 21 | 250W | |||
RTX 4060 Ti | AD106-350-A1 | 4352 | 34 | 2.535GHz | 32MB | 128-bit | 8GB | GDDR6 | 18 | 160W | |||
RTX 4060 | AD106 | 4096 | 32 | 2.6GHz? | 32MB | 128-bit | 8GB | GDDR6X | 18 | 120W |