Intel's mythical Big Battlemage GPU pops up in the MESA Linux driver making a launch later this year that little bit more likely

We've covered Intel's purported plans to launch a bigger, badder version of its latest Battlemage GPU tech on several occasions. Mostly, those centred on a GPU codenamed G31 being spotted in Intel shipping manifests. Now, the very same G31 has popped up in the official MESA OpenGL drivers for Linux, and there's even more reason to believe Intel might actually launch a high-performance gaming GPU later this year.

Right now, Intel has the Arc B570 and B580 graphics cards, based on the G21 Battlemage chip and positioned at the very low end with MSRPs of $190 and $250, respectively. The G21 has 20 EUs or graphics execution units, and the widespread consensus is that the G31 sports 32 EUs. So, it's a much more powerful GPU.

Depending on clock speeds and performance scaling, not to mention driver quality, 32 EUs could put Big Battlemage, which could possibly be called the Intel Arc B770, up with the likes of Nvidia's RTX 4070 or RTX 5070 GPUs.

A more pessimistic outcome would be RTX 5060 Ti performance. However, it's hard to imagine Intel launching a G31-based GPU to compete with the mere RTX 5060 Ti. That's because the G21 chip in the B570 and B580 is already substantially larger and more expensive to manufacture than the GB206 GPU in the RTX 5060 Ti.

Based on the 32 EU assumption, the G31 will be a very large GPU with a die size comparable to the RTX 4080 and RTX 5080 chips, AD103 and GB103, respectively. While all the evidence is that Intel currently isn't attempting to make huge profits out of Arc gaming GPUs, even in that context, selling GPUs with comparable costs to a $1,000 RTX 5080 for around $400 in order to compete with the 5060 Ti seems unlikely.

The G31 is also expected to be a 256-bit GPU in terms of memory access. So, it would likely ship with 16 GB. That would be a handy advantage over the 192-bit buses and 12 GB of VRAM of the RTX 4070 and 5070. As a gaming GPU, the G31 probably only makes sense if it can at least trade blows with those mid-range GPUs priced at $500 or so.

Of course, the other scenario is that the G31 may indeed be slated for retail availability, but not as a gaming GPU. Instead, Intel could be lining up the G31 as a professional AI solution to go up against the likes of Nvidia's RTX Pro workstation AI cards.

RTX Pro cards are based on the same silicon as Nvidia's gaming graphics cards, as opposed to its dedicated Blackwell AI hardware, but sell at much higher price points than gaming GPUs. Intel could be planning something similar for G31. It could still massively undercut Nvidia on price while entirely sidestepping unflattering comparisons with both Nvidia and AMD's gaming GPUs, which put a much lower realistic price cap on any gaming graphics card.

The idea then would be a pro card with a load of VRAM priced much higher than would make any sense at all as a gaming solution, but still very compelling for workstation AI applications. In that scenario, the G31 would indeed come to market later this year, but make zero impact on gaming.

For now, it's all speculation. The entry of the G31 doesn't prove for sure that Intel is intending to launch the thing as a commercial product, let alone guarantee it will be a compelling gaming GPU. But it does show that Intel is still working on a Big Battlemage chip, and that alone is pretty exciting, given how desperate we are for more competition in the gaming graphics card market.