Intel Battlemage display drivers coming soon for Linux — functional drivers to focus on power efficiency first

Lunar Lake Architecture
(Image credit: Intel)

Intel is hard at work building Linux-ready graphics drivers for its next-generation Battlemange GPUs, which will soon arrive with Intel's Lunar Lake CPUs. Phoronix reports that Intel will have a functional display driver for Battlemage once the upcoming Linux 6.11 kernel cycle rolls in. This new display driver is by no means complete, but it provides display functionality for Battlemage GPUs for the first time in Linux, which is a big deal regardless.

Right now, Intel is actively working on Direct Rendering Manager driver changes, HPD display enabling, and a number of power-efficiency-focused features for the upcoming Linux driver. These abilities include panel replay functionality, which enables a driver to just refresh portions of the display showing new information, rather than refreshing the entire frame buffer. It also has Auxless advanced link power management, link-off-between-frames support, and content match refresh rate, most of which are more power-saving techniques. Content match refresh rate is a variant of VRR, which purportedly prevents frame drops in certain types of content.

It makes a lot of sense for Intel to be prioritizing power-saving techniques in its Linux-based drivers first. There's a good chance Battlemage will be released as an integrated graphics solution first in Lunar Lake processors before discrete GPU variants arrive. This means Battlemage's primary competition will be against integrated graphics solutions in Qualcomm's Snapdragon X Plus and X Elite processors, as well as AMD's all-new Zen 5-based Ryzen AI 300 processors. Power efficiency is one of the biggest strengths of these chips, especially the Qualcomm Snapdragon counterparts, so it's reasonable for Intel to prioritize power efficiency techniques first in its driver as it tries to close the gap on Qualcomm with its next-gen Lunar Lake architecture.

Even if mobile Battlemage doesn't arrive first, the architecture will be used to a much greater extent in Intel's integrated graphics solutions. Intel produces significantly more CPUs than GPUs, so inevitably, Battlemage will have significantly larger adoption as an integrated graphics solution rather than a dedicated gaming GPU or enterprise GPU.

In integrated graphics form, Intel is touting a 1.5x performance upgrade over its previous-gen Arc Alchemist-based integrated graphics chips. The architecture has been optimized from the ground up, boasting larger XMX engines (for AI workloads), enhanced vector engines, larger RT units, and larger caches. The XVE vector engines are now capable of doing eight 512-bit multiplies per clock and eight 2048-bit vectors per clock via the XMX engines. Intel is also increasing the SIMD engine to 16 lanes.

The vertex fetch throughput and mesh shading performance have been increased by 3x, as well, which are important attributes that increase 3D graphics performance.

Intel's integrated graphics projections for Battlemage won't be the same as the discrete GPU versions, because Intel is using low-powered transistors for its integrated graphics versions. But there's a good chance Battlemage in discrete GPU form will be a big upgrade regardless, due to all the changes Intel has made to its upcoming Battlemage GPU architecture.

Aaron Klotz
Contributing Writer

Aaron Klotz is a contributing writer for Tom’s Hardware, covering news related to computer hardware such as CPUs, and graphics cards.

  • "This new display driver is by no means complete, but it provides display functionality for Battlemage GPUs for the first time in Linux, which is a big deal regardless."

    First time ?

    Intel's Lunar Lake iGPU Battlemage "Xe2" graphics architecture already got support for the display, & also audio via open-source Linux support patches, and Direct Rendering Manager (DRM) kernel graphics/display driver updates.

    "hda: Add Intel BMG PCI ID and HDMI codec vid." Date: Wed, 15 May 2024
    https://lore.kernel.org/lkml/87zfsr5v2b.wl-tiwai@suse.de/

    I will post the display support kernel link as well. There are even more patch entries !

    But, Linux 6.10 also had support for actually driving a display attached to the Xe2 integrated graphics with Lunar Lake. This was the main pull request for the drm subsystems for 6.10.

    https://lore.kernel.org/dri-devel/Zh_Q72gYKMMbge9A@intel.com/
    i915/xe display:
    - Lunar Lake display enabling, including cdclk and other refactors
    - BIOS/VBT/opregion related refactor
    - Digital port related refactor/clean-up
    - Fix 2s boot time regression on DP panel replay init
    - Remove duplication on audio enable/disable on SDVO and g4x+ DP
    - Disable AuxCCS framebuffers if built for Xe
    xe:
    - improved error capture
    - clean up some uAPI leftovers
    - devcoredump update
    - Add BMG mocs table
    - Handle GSCCS ER interrupt
    - Implement xe2- and GuC workarounds
    Reply
  • You forgot to mention these patch entries, since they are related to the above patches for Battlemage iGPU.

    Because, Intel's Lunar Lake CPUs also got support, via a new patch that brings in support for DLVR (Digital Linear Voltage Regulator), which reduces the power input and in turns add to the overall efficiency of these ultra-low power SOCs. It will be merged with the upcoming Linux 6.11 cycle.

    https://lore.kernel.org/linux-pm/20240619124600.491168-4-srinivas.pandruvada@linux.intel.com/
    Digital Linear Voltage Regulator can help in reducing the power consumption of the processor. The patch for the Linux kernel extends the int340x_thermal driver for dealing with Lunar Lake's DLVR.

    Lunar Lake's DLVR behaves the same as with Meteor Lake and Arow Lake. The MMIO offset and bit positions have been changed though, when compared to the Meteor Lake and Arrow Lake lineup.


    https://cdn.wccftech.com/wp-content/uploads/2024/06/Intel-Lunar-Lake-CPU-DLVR-Support.png
    Reply
  • TechyIT223
    Linux always stays ahead of windows in terms of early patches and support of any hardware. I wonder why 🤔

    Also why does the author think battlemage GPUs will find more use in integrated graphics rather than desktop variants? Low powered transistors doesn't make sense as outlined in the last paragraph of this article.

    Im certainly fine with battlemage GPUs targeting the desktop market as previou rumors have been hinting. Of course iGPU is also xe2's domain, but desktop variants will share same architecture as integrated graphics
    Reply
  • Eximo
    They'll certainly sell more Battlemage iGPUs than discrete unless some miracle happens with gaming performance.

    The node they are using for the integrated tile isn't the same as the one they will use on the desktop/mobile discrete GPUs. One is low power (N3B), the other is going to be the higher power version of the node (N3P, presumably)
    Reply
  • Metal Messiah. said:
    First time ? There are even more patch entries !

    April support for the Xe2 GPUs in the display drivers, for enabling display output handling specifically for Battlemage. BMG graphics cards will only support a maximum DisplayPort bandwidth of UHBR13.5.

    https://lists.freedesktop.org/archives/intel-gfx/2024-April/348477.html


    https://i.imgur.com/4WgvUtP.png


    This was the December 2023 pull request. "Xe" graphics driver replaced the traditional i915 DRM kernel driver at Linux. Following are all the changes added for Xe2 (Battlemage) GPUs:

    https://lore.kernel.org/dri-devel/ZXzTA75G5VhCrDis@intel.com/T/#udrm/xe: Add performance tuning settings for MTL and Xe2drm/xe/xe2: Add MOCS tabledrm/xe/xe2: Respond to TRTT faults as unsuccessful page faultdrm/xe/xe2: Add initial workaroundsdrm/xe/xe2: Set tile y type in XY_FAST_COPY_BLT to Tile4drm/xe/xe2: Update MOCS fields in blitter instructionsdrm/xe/xe2: Determine bios enablement for flat ccs on igfxdrm/xe/xe2: Modify main memory to ccs memory ratio.drm/xe/xe2: Allocate extra pages for ccs during bo createdrm/xe/xe2: Updates on XY_CTRL_SURF_COPY_BLTdrm/xe/xe2: Update chunk size for each iteration of ccs copydrm/xe/xe2: Update emit_pte to use compression enabled PAT indexdrm/xe/xe2: Update context image layoutsdrm/xe/xe2: Handle fused-off CCS enginesdrm/xe/xe2: AuxCCS is no longer useddrm/xe/xe2: Define Xe2_LPG IP featuresdrm/xe/xe2: Define Xe2_LPM IP featuresdrm/xe/xe2: Track VA bits independently of max page table leveldrm/xe/xe2: Program GuC's MOCS on Xe2 and beyonddrm/xe/xe2: Program correct MOCS registersdrm/xe/xe2: Update SVG state handling
    And the list goes on. There have been more entries/patches for Battlemage support.
    Reply
  • TechyIT223
    Eximo said:
    They'll certainly sell more Battlemage iGPUs than discrete unless some miracle happens with gaming performance.

    The node they are using for the integrated tile isn't the same as the one they will use on the desktop/mobile discrete GPUs. One is low power (N3B), the other is going to be the higher power version of the node (N3P, presumably)

    Are you sure about the nodes used? Because N3P it seems only has higher performance characteristics. Not sure about the power though, but it could also be the same.

    But anyway Intel should be better off if they use the same nodes on both iGPU and discrete gaming cards.
    Reply
  • Eximo
    Only rumor/information I can find is that N4P is likely to continue to be used for Nvidia Blackwell desktop cards and was supposed to be for Battlemage, with plans to use N3P for Celestial. Likely for RDNA4 as well. Seems the smaller nodes are going to be reserved for CPU logic for now. Since Intel put the iGPU on the logic tile they are kind of stuck with having two nodes for the same generation.

    They didn't do too badly, looks like clock speeds are still well over 2GHz on the iGPU. Arc only maxes out at 2400Mhz on N6. Next 'shrink' might gain them a few hundred Mhz. They are talking 3Ghz from Blackwell.

    But that is all rumor.
    Reply
  • N3P is actually an optical shrink of N3E, and it is compatible with its predecessor in terms of IP blocks, process/design rules, electronic design automation (EDA) tools etc.

    N3P = at least 5% higher performance than N3E, 2% optical shrink, and 1.04x transistor density.
    Reply
  • TechyIT223
    Thanks for the clarification and input guys
    Reply