Intel Arc B580 release date, price, and specs

Intel just revealed the full specs of its new gaming GPUs, and is making some bold claims about them. Specifically, the company says its new $249 Intel Arc B580 Battlemage graphics cards are on average 10% faster than the Nvidia GeForce RTX 4060 if you’re playing at 2,560 x 1,440. Not only that, but Intel says the B580 even outperforms the RTX 4060 when ray tracing is enabled.

If that does indeed turn out to be true, then Intel may well have the best graphics card for budget buyers on its hands. Nvidia has left the sub-$250 graphics card market alone on the desktop recently, with the $299 GeForce RTX 4060 being its cheapest offering in its current-gen lineup. Meanwhile, our AMD Radeon RX 7600 review showed this $269 GPU outperforming the RTX 4060 in most standard gaming tests, but falling behind when it comes to ray tracing. If Intel’s claims are to be believed then it has made a sub-$250 GPU that can do it all.

Intel Arc B580 specs

Ray tracing is one of the areas where Intel has focused its efforts for Battlemage (that’s the codename for the architecture used by these GPUs), with the new ray tracing units in the GPUs featuring three traversal pipelines rather than just two, as in the previous Alchemist units, as well as two triangle intersection units, rather than just one in the previous GPU architecture.

Meanwhile, the rest of the BMG-G21 GPU on which the B580 and B570 are based is divided into five render slices, which is Intel’s GPU building block. Each of these render slices contains four Xe2 cores and four ray tracing units, and each Xe2 core is further divided into eight XMX cores for AI (a bit like Nvidia’s Tensor cores), and eight XVE (vector engine) cores for shader processing.

Intel says the new Xe2 render slices offer a “significant increase in performance per core” compared to its previous GPU architecture, and the company also claims this new GPU is targeted at 1440p gaming, rather than the 1080p resolution usually targeted by budget GPUs.

To this end, it also comes with 12GB of GDDR6 VRAM, rather than the 8GB found on the RTX 4060 and Radeon RX 7600, and it’s attached to a 192-bit memory interface, giving it a total memory bandwidth of 456GB/s. That’s significantly wider than the 272GB/s of bandwidth offered by the RTX 4060 with its 128-bit interface.

Unlike Nvidia, Intel is also choosing to stick with the classic 8-pin PCIe power connector (just one of them) for the B580, quoting a total board power of 190W for stock speed cards. In addition, Intel has also announced the Arc B570, which has 10GB of VRAM and a 160-bit memory interface, along with 18 RT cores, 144 XMX cores, and a 2,500MHz GPU clock.

Intel XeSS 2 frame generation

Those aforementioned XMX AI cores are going to get some use in some games too, thanks to the introduction of Intel’s new XeSS 2 tech. Much like Nvidia DLSS 3, XeSS 2 sees Intel introducing its own frame generation system, called XeSS-FG, which will use AI to insert new frames between the ones rendered by the GPU, smoothing out the frame rate.

A handful of games will support the new XeSS 2 frame generation tech at launch, including F1 24, Assassin’s Creed Shadows, Quidditch Champions, and Robocop Rogue City. However, Intel also points to its XeSS upscaling tech now being supported in over 150 games, and it’s possible that XeSS 2 frame generation could end up being supported by many more games in the future.

Intel Arc B580 performance

The big question, of course, is how the Arc B580 will perform, and Intel is making some big claims. For a start, the company says the new GPU is on average 24% faster than Intel’s previous Arc A750 Limited Edition, a card that launched at a price of $289.

Perhaps more importantly, though, according to Intel’s figures, the B580 beats both the RTX 4060 and Radeon RX 7600 in games using rasterization without the addition of ray tracing. In fact, Intel says that, across 47 games at 2,560 x 1,440 with Ultra settings, the Arc B580 is on average 10% quicker than the RTX 4060.

Not only that, but Intel also says the B580 can beat the RTX 4060 at ray tracing, which is the one area where the Radeon RX 7600 struggles. Intel gives an example of Forza Motorsport, where at the high ray tracing preset, the B580 averages 64fps compared to 57fps on the RTX 4060, while at the ultra ray tracing preset, the B580’s 51fps average towers over the RTX 4060’s 31fps, thanks to its extra memory. Intel has also produced a graph showing the performance relative to price.

If the $299 RTX 4060 is the 100% baseline for ray tracing performance, then according to Intel’s figures from nine ray-traced games, the $249 Arc B580 figure is 125%, while the $269 Radeon RX 7600 is 88%. This is a bit of a tortuous graph to understand, though, and at first glance it makes it look as though the Arc B580 is 25% quicker than the RTX 4060.

It’s also worth bearing in mind that Intel’s figures come from benchmarks run at 2,560 x 1,440, where the 8GB RTX 4060 doesn’t perform at its best, and the extra VRAM gives the new Intel GPU an advantage. We’d be interested to see some of these benchmark figures at 1,920 x 1,080.

Intel Arc B580 price

The Intel Arc B580 price is $249, undercutting both the Nvidia GeForce RTX 4060 and AMD Radeon RX 7600, which looks like good value if the performance delivers on Intel’s promises.

Intel Arc B580 release date

The Intel Arc B580 release date is Friday, December 13, 2024, just in time for the holiday season. Comparatively, the new Nvidia RTX 5000 lineup isn’t expected to be unveiled until CES at the start of January 2025.

Of course, the biggest issue for Intel will be driver support and compatibility issues, which plagued its first desktop Arc GPUs at launch. However, the company points to the huge amount of work it’s done with drivers over the last few years, including over 50 new driver releases, and over 120 new games supported as soon as they came out. If Intel can really nail driver support, and deliver all this for $249, then it could be onto a winner.

In the meantime, check out our Intel Arc A770 review to see how we get on with benchmarking one of the company’s previous graphics cards.