Share Share by:
- Copy link
- X
Share this article Join the conversation Follow us Add us as a preferred source on Google
Our friend Michael over at Phoronix spotted today that Intel’s XPU Manager 1.3.5 now includes explicit support for Intel’s “BMG-G31,” marking another software-level confirmation that Intel’s larger “Big Battlemage” die is in active development and driver testing. This sort of entry in an Intel-maintained tooling package is low-noise evidence, because vendors generally don’t add device IDs for hardware that doesn’t exist.
This update joins a steady trickle of corroborating traces in the open-source world. Over recent months Mesa has merged patches to recognize distinct Battlemage IDs and specifically classify a G31/BMG variant, and multiple new PCI device IDs associated with Battlemage have turned up in Linux driver trees. Together these driver-level sightings make a compelling case that Intel is prepping at least one larger Battlemage SKU for desktop or workstation use.
Public leaks and manifests that have circulated point to a BMG-G31 configuration far larger than the G21 used in current Arc B5x0 parts like the Arc B580. The recurring rumor set includes: roughly 32 Xe2/Xe cores, a 256-bit memory bus paired with at least 16 GB of GDDR6 memory, and board-level power envelopes that could reach as high as ~300 W for higher-binned desktop SKUs. Those numbers, if true, would position BMG-G31 as a bona fide midrange, mainstream part rather than another entry-level Arc.
You may like
-
Intel might be working on its own multi-frame generation tech — “XeSS MFG” appears in Arc driver files -
Intel Core Ultra 5 338H appears in new Geekbench listing, confirming new Xe3 iGPU and naming scheme -
New Panther Lake Core Ultra X7 358H leaks in new benchmark
A bigger die is a double-edged sword. On the one hand, the spec sheet that’s been condensing from the mists of leaks and rumors would lift Intel into performance brackets it hasn’t consistently hit with prior Arc generations. It would mean stronger 1440p performance potential and more credible competition against Nvidia and AMD in that tier. On the other hand, larger dies cost more per good chip (lower wafer yields) and can be harder to price aggressively without eating margins. It’s the classic manufacturing/competitiveness trade-off.