A significant part of the issue (other than the AD107 really being a xx50 class GPU) is that the 3070 has a substantial fill-rate uplift over any of the 4060 variants and a decent amount of memory bandwidth to go with it. The 3070 and 4060 Ti are pretty much matched in terms of compute and texturing, but there isn't quite enough L2 cache in the latter to make up for the fact that its fill-rate is pretty poor.This is the first gen in a long time since the 60 series is not faster then older 70 series.
I was expecting something but not that bad.
Thanks for the review! As expected Steve mop the floor with the 4000 series.
"The RTX 4060 isn't as egregiously poor as the 4060 Ti, but it's still disappointing, and it's certainly not worth buying at $300. The limited 8GB VRAM buffer is more forgivable at this price, but it's a bitter pill to swallow given that the RTX 3060 offered 12 GB."
Again this makes me feel good for going used 3070 at $300.
This image also highlights another potential drawback of the 4060 -- it only uses 8 PCI Express lanes (just follow the traces from the bottom of the chip package). Yes it's version 4, so offers the same bandwidth as x16 3.0, but stick this into a PC that only has PCIe 3.0 and you'll take a minor ding in performance, in some games. It's not really a big deal, but it's a bit of a penny-pinching design choice by Nvidia (and AMD for some of its cards).
The "mighty" MSI Ventus X2 board looks so empty and that GPU there... so little.
4 phase VRM for GPU and 1 phase VRM for GPU.
This could make a 1 slot design posible again and maybe a low profile dual slot.
That's because it really is a laptop GPU -- it's used for 4060 Mobile, 4050 Mobile, and the RTX 2000 Ada Generation products. The only discrete card it's currently in is the 4060 (though it will appear in the 4050 eventually).Wow. That looks like a laptop GPU to me. It's tiny!