Exclusive | NVIDIA planning a Turing-based MX350 successor to take on Intel Tiger Lake's Xe DG1 iGPU, will be a GTX 1650-class chip with PCIe Gen4 support

A short while ago, we reported exclusive details about NVIDIA's upcoming GeForce MX330 and MX350 entry-level dGPUs. These GPUs are still based on the Pascal architecture despite Turing being a couple of years old by now. However, we are getting to hear from our sources that NVIDIA is indeed working on a Turing-based MX GPU, and that it will take on Xe DG1 in Tiger Lake later this year.

Our sources tell us that a successor to the MX350 (N17S-G5), which is internally being referred to as the N18S-G5, is in the works for a 2H 2020 launch to coincide with the launch of Intel Tiger Lake. So, while the current MX330 and MX350 will cater to laptops based on Intel Ice Lake and Comet Lake, the N18S-G5 is intended to counter the Xe DG1 iGPU that will be part of Tiger Lake processors. Laptops powered by Intel Tiger Lake will, therefore, offer both MX350 and N18S-G5 dGPU options. 

The N18S-G5 will come in two variants, A and B, neither of which currently have a marketing name. Both the N18S-G5 variants will be based on the Turing TU117 GPU, which essentially makes them a gimped GTX 1650. They will support PCI Gen4 speeds, 64-bit GDDR6 VRAM, and have a 25 W TDP envelope. The difference, however, lies in the package size with the N18S-G5-B slated to come in a 29 x 29 mm package compared to the N18S-G5-A1's 23 x 23 mm.

The N18S-G5-A can be configured to have either GDDR5 VRAM at 3.5 GHz or GDDR6 at 5 GHz memory clocks while the N18S-G5-B features GDDR6 VRAM at 5 GHz memory frequency. GPU core clocks and CUDA Core counts are still unknown, but expect something on the lines of a GTX 1650 Max-Q. 

The table below illustrates what we know of these Turing-based MX GPUs so far:

Based on what we are hearing from our sources, the Turing-based N18S-G5 MX GPUs and the MX350 (N17S-G5) will be the upsell options for Tiger Lake laptops. For Comet Lake and Ice Lake, the MX350 and MX330 will remain. This is certainly a welcome news as we can now expect GTX 1650 Max-Q level performance in the entry-level.