Intel has issued a fresh driver update, version 32.0.101.8804, targeting its Arc Pro discrete graphics cards and integrated GPUs. The release concentrates on refining dynamic memory allocation for the Arc Pro B370 and Arc Pro B390 models.
Expanded Memory Allocation Window
Under the new driver, a system equipped with 64 GB of RAM can now allocate up to 59.5 GB—equivalent to 93 percent of total system memory—directly to the integrated GPU. This marks an increase from the previous ceiling of 87 percent. The allocation process remains entirely dynamic and automatic, continuing Intel’s approach of letting the driver manage the split without manual intervention, in contrast to rival solutions such as AMD’s Ryzen Strix Halo platform that require users to partition memory into fixed pools.
Measurable Workload Gains
Intel states that the updated memory behavior, combined with other driver-level optimizations, translates into tangible performance improvements in certain professional applications. The company points specifically to a 15 percent speed increase in Blender on the Arc Pro B390, and a 5 percent uplift for older integrated GPUs, signaling benefits across both current and prior-generation hardware.
Bandwidth Considerations and Practical Limits
While the flexibility of borrowing such a large share of system memory is a clear advantage, the underlying memory technology imposes its own constraints. Panther Lake systems tap LPDDR5X-8533 memory, which on paper delivers a memory bandwidth of roughly 68.26 GB/s. By comparison, a discrete Nvidia GeForce RTX 5060 laptop GPU offers up to 384 GB/s, while Apple’s M5 Max pushes bandwidth to 614 GB/s. For workloads that depend more heavily on raw memory bandwidth than on capacity—common in high-resolution 3D rendering or AI inferencing—Intel’s integrated graphics remain at a measurable disadvantage despite their expanded memory allocation range.
Sources: www.intel.com, videocardz.com