Nvidia appears to be feeding OEMs two different versions of its MX150 mobile GPU, and it has been less than forthcoming about the performance differences.
As first reported by Notebookcheck, there are apparently two different variants of the GeForce MX150 in the wild. These differences are easily identifiable in a pair of GPU-Z screenshots taken from an Asus Zenbook laptop and a Lenovo IdeaPad 320S Ultrabook. The Zenbook sports a MX150 GPU with the device ID of 10DE 1D10, whereas the IdeaPad is labeled as 10DE 1D12. The publication also claimed to have only found the 1D12 variant in 13” Ultrabooks so far, including the HP Envy 13, Asus Zenbook 13, and Xiaomi Mi Notebook.
Although the memory capacity and CUDA core count of the MX150 GPUs are identical, the clockrates (and therefore, performance) are vastly different, with the 1D12 GPU downclocked (in comparison to the 1D10) by 523MHz, 494MHz, and 249MHz on the base, boost, and memory clocks, respectively. The report also asserted the MX150 variants have different TDPs, with the 1D10 sitting at a 25W TDP and the 1D12 drawing less than half that at 10W.
The difference in clock rate and power consumption of the 1D12 version are similar to the differences found in Nvidia’s Max-Q graphics, which allows powerful GTX-branded GPUs (such as the GTX 1070 and 1080) to fit into smaller devices by lowering the power ceiling and GPU clocks via firmware. However, Nvidia plainly advertises these differences with the higher-end GPUs, and the company doesn’t appear to have any differentiated marketing materials for the two MX150 GPUs.
To be fair, Nvidia’s GeForce MX150 specification page clearly notes that GPU implementation will vary by OEM, and that consumers should refer to respective vendors’ websites for actual shipping specifications. We're accustomed to seeing differences in memory capacity in these situations, and the statement suggests that OEMs are ultimately in control of how to utilize the GP108 GPU that Nvidia ships them for use in their products. This could mean that Nvidia isn't to blame for the lack of transparency, but the differentiated clock rate and device ID indicate that this is implemented at a firmware level, which more or less shifts responsibility back to Nvidia (which provides the firmware).
We’ve reached out to Nvidia for comment and will update this article if and when any new information is available.