After much waiting and a ton of speculation, Intel has finally announced its first discrete GPU in the form of the Iris Xe Max (or the DG1) for ultra-portable laptops and notebooks. Technically speaking, you’re basically looking at the Gen 12 Xe LP iGPU found on Tiger Lake with its own bus and L3 cache. Sure, the clocks have been ramped up and the GPU gets its own memory in the form of 4GB LPDD4x (not GDDR6), but ultimately it’ll mostly bounce around the iGPU space.
|Iris Xe MAX||Gen12 Xe||Gen 11|
|FP32 Rating (Peak)||2.46 TFLOPs||2.1 TFLOPs||1.13 TFLOPs|
|Memory Bus Width||128-bit||128-bit|
The Xe Max’s primary competitors will be the integrated Radeon Vega GPUs featured on the Renoir and upcoming Lucienne and Cezanne APUs. The GeForce MX350 and MX450 also play in the same territory, but it’ll mostly be the former that the DG1 will trade blows with. Considering Gen12’s poor performance in DX12 titles, it’ll be a surprise if it manages to beat even the top-end Ryzen 5000 APUs which are slated to arrive sometime next year.
At present, there are three laptops shipping with the Xe Max, one from Acer, Dell, and ASUS, each. The sales for these notebooks starts from the beginning of November or the midnight of the 31st. Intel is primarily aiming the Xe Max towards the Asia-Pacific region which is where most entry-level GPUs such as the RX 580, 570, and the GTX 1650/1660 sell the most.
The GPU core architecture is the exact same as the featured in the Tiger Lake-U CPUs. You can read more about that here. There is one, big Xe Slice with six subslices, with pairs sharing the pixel backend. Here, unlike the Gen12 Xe iGPU, DG1 gets the full 16MB of L3 cache along with a dedicated 128-bit bus paired with 4GB of LPDDR4x-4266 VRAM. It has a peak memory bandwidth of 68GB/s, similar to the MX350/450.
The other primary difference compared to the Gen12 iGPU is with respect to the core clocks. While the Tiger Lake-U graphics can boost to a maximum of 1.35, the Iris Max manages to reach an impressive 1.65GHz. However, the effective clock will vary depending on the workload, thermals and TDP. On paper, the Iris Max or the DG1 is around 20% faster than the vanilla Gen12 integrated graphics.
Speaking of the TDP, the Iris Max has a default TDP of 25W, roughly the same as the entire Tiger Lake-U SoCs. Once again though, this will be regulated by the OEMs and may go up or down as per the chassis design.
Overall, don’t expect the Iris Max to change Intel’s position in the GPU market. It’s simply an integrated GPU with its own dedicated memory, bandwidth and power budget. It won’t be until the HPG parts land next year that we’ll get an idea of how effective Intel’s offering will be in dGPU market.