Our Jeremy said it best when he reported on the surprise announcement of an Intel-Nvidia collaboration yesterday: it’s cats living with dogs. Naturally, such a strange occurrence has led people to question just how it will affect current product roadmaps—especially Intel Arc. On that front, thankfully, Intel says things are business as usual.
An Intel spokesperson has supplied a statement to the media saying that: “collaboration is complementary to Intel’s roadmap and Intel will continue to have GPU product offerings” (via PC World and Hot Hardware). That’s pretty unambiguous, at least for the short term.
Intel’s roadmap, to be clear, currently includes at least two new generations of processor containing its Arc Xe3 iGPUs: Panther Lake and Nova Lake. The former should launch relatively soon, will be mobile (ie, laptop)-only, and will at least in part be made using the company’s new 18A node—although early yields apparently aren’t looking too promising. The latter will arrive later and should have both desktop and mobile options, just like current-gen Arrow Lake, and looks to be made on TSMC’s N2 node.
One key bit in the midst of this is that Arc Xe architecture and outside of that its discrete graphics cards. Intel first launched an Arc GPU in 2022, and since then it’s pushed out two generations of discrete Arc graphics cards. It also features in laptops as integrated graphics, but Intel already had the less powerful Iris Xe for that.
We’re currently on the Arc Battlemage generation (B-series) but Celestial looks to be just around the corner. Presumably that’s part of the roadmap that Intel will be continuing to travel.
Nvidia CEO Jensen Huang (Image credit: Bridget Bennett/Bloomberg via Getty Images)
The newly announced Intel-Nvidia partnership, however, will cover both datacentres and presumably laptop/mobile chips. Intel server silicon will be bundled alongside Nvidia GPU accelerators with the help of the NVLink interconnect, and for the consumer market, system-on-chips (SoCs) will be made that combine Intel CPUs with Nvidia RTX GPU chiplets.
This, apparently, will be achieved by combining Intel Foveros (3D packaging) tech with TSMC-made Nvidia chiplets in “a multi-technology packaging capability and multi-process packaging technology.”
It’s clear why people might worry this could threaten the future of Arc, because if Nvidia starts covering Intel’s mobile GPU bases then there might be less incentive for Intel to put so much time and effort into furthering its own GPU architectures. That’s especially the case given Arc has seemed on a knife’s edge for a while now regardless, at least to those of us on the outside looking in.
There is at least some talk that Nvidia GPUs might entirely replace Intel’s own ones in its future products, though this is very speculative.
Intel CEO Lip-Bu Tan (Image credit: Intel)
Tech YouTuber Moore’s Law is Dead, for instance, claims that one Nvidia source told him: “Honestly, it’s the datacentre side that’s the bigger part of this story when it comes to the deal, but yes, we [Nvidia] are making Intel’s future iGPU tiles. However, I must caution that this is years out, and I wouldn’t expect these ‘Nvidia x Intel’ products until Titan Lake at the earliest.”
Titan Lake is expected to launch around 2028. This quote, to my eyes at least, is ambiguous as to whether the supposed Nvidia source is essentially just reiterating the news that Intel and Nvidia will be working together on SoCs and whether “future iGPU tiles” means it will be making some iGPU tiles or all future iGPU tiles.
In other words, the question is: Where does Intel’s current roadmap end? When Intel says any new Nvidia-Intel products will be complementary to the planned line-up, is that only the case for the next few years? Will 2028 and beyond see a shift towards Nvidia GPUs in place of Intel ones? Will the Druid architecture ever see the light of day?
No answers here. But for now, at least, we know that Arc should be around for a while longer. In some guise.
From PCGamer latest via this RSS feed