Steady Progress for Intel’s Next-Gen GPUs
Intel has quietly added four new device IDs (e220-e223) for its unreleased Battlemage G31 GPU in the latest Compute Runtime update. This open-source project (supporting OneAPI/OpenCL) confirms active development of high-end Battlemage graphics cards, though Intel remains officially silent. The update follows recent Mesa driver additions, reinforcing that BMG-G31 is advancing toward release.
What BMG-G31 Reveals About Intel’s Plans
While not all device IDs become retail products, this aligns with Computex leaks about the Arc B770—Battlemage’s flagship successor to the A770. Expected specifications include:
- 32 Xe2-Cores (vs. 20 in current B580)
- 256-bit memory bus with 16GB GDDR6
- PCIe 5.0 x16 support (first for Intel GPUs)
- Bandwidth targeting ~456 GB/s
For context, the mid-range BMG-G21 (B570/B580) uses PCIe 4.0 x8 and tops out at 18 Xe2-Cores.
Market Positioning & Challenges
The Battlemage G31 targets a late 2025 launch, potentially competing with NVIDIA’s RTX 5060 Ti and AMD’s RX 9600 XT. Key hurdles remain:
- Driver maturity: Intel’s software still lags behind rivals
- Pricing: Must undercut competitors at ~$299-$349
- Board partner commitment: Only 4 BMG-G21 models launched despite 9 device IDs
Why Gamers Should Watch Closely
If specs hold, the Arc B770 could offer:
- 2x ray tracing performance over Alchemist
- AV1 encoding parity with NVIDIA/AMD
- DX12 Ultimate feature support
Leaks suggest a 190W TDP—requiring robust cooling but staying below NVIDIA’s power-hungry designs.
Also, Read
- Intel Arrow Lake Refresh – Only Minor Clock Boosts, No NPU Upgrade
- AMD Launches Ultra-Budget Ryzen AI 5 330 APU for Entry-Level Laptops
- Intel Nova Lake-AX Leaked – High-Power APU to Challenge AMD Strix Halo
Key Takeaways
- Open-source trails don’t lie: Consistent ID sightings signal Battlemage G31’s reality.
- Performance leap likely: 256-bit bus and PCIe 5.0 could finally make Intel competitive.
- Patience required: Late 2025 launch leaves room for spec changes.
Intel’s silence continues, but the code speaks volumes.
Source: Wccftech, intel/compute-runtime