This will be the lowest overall performance chip in the Ada Lovelace family and, as with laptops, there will be at least two more versions according to rumours. And it is that this AD107 has been seen in its full version, which will give life to the RTX 4060 desktop and two more GPUs that are going to hit the market as entry-level. For this reason, and confirming its technical specifications from NVIDIA itself, we now know that Chip in its full version It is much smaller than its predecessors. This is the AD107-400 of the RTX 4060.
It’s really small and compact, which is usually not a good sign because it has less heat dissipation surface, but NVIDIA has given it a really low consumption, which will help a lot so that the temperature and noise generated by each model are really controlled. Yes for quite an interesting power.
NVIDIA AD107-400, the GPU that comes from laptop to PC
And it is that this chip, although first conceived for PCs, arrived earlier in the mobility sector to be included in entry-level and low-end gaming laptops, which gave life to the RTX 4060 Mobile and RTX 4050 Mobile. What has been leaked is a confirmation that NVIDIA designed the AD107-400, which in turn is a full version of it, ie not a more powerful AD107.
This is interesting, because if we compare it with their predecessors, NVIDIA and TSMC have achieved quite an interesting feat and it is nothing more than reducing the field by 27%. In other words, within the mid-range, this is a really compact chip that will only measure 146mm2.
For comparison’s sake and being manufactured in the TSMC 4N, a node based on the company’s 5 nm but improved somewhat, they are 146 mm2.
They stand up against the 200 mm2 of the GA107 and TU117, the first manufactured in 8 nm and the second in 12 nm. We don’t know if it has millions of transistors, but with the GA107 having no less than 8.7 billion, that number should increase substantially.
Full specifications of this chip
NVIDIA not only offered up its theoretical performance under its comparison last week, but also revealed a bulk of its final specifications. And here comes the really cool part, because from the performance of the shaders we can mathematically calculate their final frequency for versions that follow the NVIDIA specifications.
As you can see, the RTX 4060 will be powered by 15 TFLOPS at FP32, so knowing that it will have 3,072 confirmed shaders, we can safely say that the frequency will be 2,441 MHz, which is 41 MHz higher than we expected. Is. Estimated at that time.
The other confirmed key aspect is its 8GB of VRAM, which will come with 24MB of L2 for gaming’s way. These are less gains than expected, as it was believed to come with 32MB. Where it improves is in TGP, as it was supposed to come with 150W, and instead, it will eventually be 115W, a significant reduction.
This RTX 4060 with the AD107-400-A1 will arrive next month for $299 as an MSRP, and from the looks of it, it won’t have an FE version like its big sister, the RTX 4060 Ti.