I'm getting the fear about the next generation of mid-range graphics cards. The latest rumours surrounding the specs of the upcoming RTX 4060 GPU are pretty meagre, to say the least. Regularly reliable leaker, kopite7kimi (opens in new tab), has listed the expected specifications of the AD107 chip at the heart of the RTX 4060 and they seem to almost be a like-for-like match with the mobile version of the same GPU.
Except a tiny bit worse.
The leaker has listed the core configuration as just 3,072 CUDA cores, while the upcoming RTX 4070 is expected to come with near double that figure (opens in new tab). The current rumoured config has the RTX 4070 sporting the same AD104 GPU as the RTX 4070 Ti (opens in new tab) model, but with 5,888 CUDA cores.
Personally, I'd have hoped Nvidia would drop the RTX 4060 onto the AD106 GPU, and get close to the core count of the mobile RTX 4070 at 4,608. And yes, I agree, it's getting confusing talking across mobile and desktop chips with the same damned name but completely different specs.
But that doesn't mean I'm down with Nvidia now reportedly sticking to the same specs for its mobile and desktop versions of the RTX 4060. That actually makes me a bit sad for how this mainstream card is actually going to perform in the real world. On the plus side, there have been leaked performance numbers for a laptop running the full RTX 4060 at 140W (opens in new tab), and the good news there is that the mobile version seems to be running beyond the gaming levels of the 12GB RTX 3060 (opens in new tab) on the desktop.
The delta is 17% in some games, but just 5% in others. So still not a large increase for the mobile chip.
RTX 4060 still uses PG190. AD107-400-A13072FP328G GDDR6 18Gbps115W24M L2I will try to remain neutral about any leaks in the future. 😁😁😁February 13, 2023
But, as I say, the laptop AD107 GPU is actually little better than the rumoured desktop spec of the RTX 4060 GPU, because the leaker is suggesting that it only has 24MB of L2 cache as opposed to the 32MB of the laptop chip. And runs with a lower TGP than the mobile version.
With the Ada architecture (opens in new tab) Nvidia has managed to increase gaming performance on otherwise like-for-like GPU specs by increasing both the clock speed and the L2 cache component of the chips.
Best gaming PC (opens in new tab): The top pre-built machines from the pros
Best gaming laptop (opens in new tab): Perfect notebooks for mobile gaming
For reference, the RTX 3060 had 3,584 CUDA cores, with just 3MB of L2 cache, and an average clock speed of 1,837MHz on our tested Zotac model. The likelihood is the RTX 4060 will run around 1GHz higher than that and, combined with the 24MB L2 cache, it will push the 3,072 cores of the chip beyond its Ampere ancestor.
But how far beyond is now the concern after seeing the performance of the laptop RTX 4060, even more so if Nvidia continues the existing precedent of increasing the price of its new graphics cards. Still, it does mean we're almost guaranteed now to get an RTX 4060 Ti using the AD106 GPU. Probably priced at $500.
Aww, now I've made myself sad again.