The first rumors arrive about the specifications and performance of the graphics series NVIDIA GeForce…
The NVIDIA GeForce RTX 4060 It is the great hidden of the family of GeForce RTX 40 Series GPUs. Until today, it is the model that we know least about, and today the information is extended thanks to the chip that will give it life, the AD106.
This latest information lowers the performance expectation that we can expect from this GPU, and it is that as revealed by the leaker @kopite7kimithe graphics chip AD106 will give a score less than 7,000 points in the 3DMark Time Spy Extreme benchmarking software.
First hint of the final performance of the NVIDIA GeForce RTX 4060
A score of 6,500 to 7,000 points in 3DMark implies being up to 18% faster than the GeForce RTX 3060 Tiand even a 69% faster than its predecessor, the GeForce RTX 3060, so it’s a noticeable performance jump. Depending on the score it finally reaches, it would have a performance very much on par. to that of a GeForce RTX 3070since this GPU moves over 6,700 points.
With this performance improvement, together with the use of NVIDIA DLSS technology, it is already a perfect option to enjoy games at a resolution 4K at 60FPS, or resort to 1440p monitors with a high refresh rate. Now if you think about 1080p, you better also think about renewing the monitor for a GPU of this range, or resorting to a more basic model, which will also be more affordable, and should have no problem offering hundreds of FPS at said resolution, thinking of the eSports.
Additionally, the leaker indicates that both the AD106 and AD107 chips will use one PCI-Express 5.0 x8 interfacewhich could mean potential performance losses on older platforms such as PCIe 3.0.
Specifications we could expect from this mid-range GPU
Rumors say that this GPU could arrive with a configuration of 4,608 CUDA Cores under the Ada Lovelace architecture, manufactured by TSMC using its most advanced manufacturing process, the 4nmthe same with which the Apple A16 Bionic debuted a few days ago.
This chip could work at a maximum Turbo frequency that is around 2.60GHzto which could be added a memory configuration of 8GB GDDR6 with a 128-bit interface, and offering bandwidth that could move over 320GB/swith an energy consumption that would be around 220Wwhich is 50W more compared to its predecessor (+30%), or the same consumption than the GeForce RTX 3070.
GeForce RTX 4060
|Lithography||TSCM 4N||Samsung 8nm||Samsung 8nm|
|TMUs / ROPs||N/A / 48||152 / 80||112 / 64|
|Tensor / RT Cores||N/A||152 / 38||112 / 28|
|Freq. Turbo||2.20 to 2.60GHz||1,665MHz||1,780MHz|
|FP32 Performance||~21 TFLOPs||16 TFLOPs||13 TFLOPs|
|RT TFLOPs||N/A||32 TFLOPs||25 TFLOPs|
|Tensor-TOPs||N/A||192 tops||top 101|
|Memory||8GB GDDR6||8GB GDDR6||12GB GDDR6|
|memory interface||128 bit||256 bit||192 bit|
|Launch||Q1 2023||December (2020)||February (2021)|