If you're upset about the dire lack of consumer RAM knocking around, this 480 GB Intel GPU isn't going to cheer you up
by https://www.pcgamer.com/uk/author/jacob-fox/ · PC Gamer
Share this article
0
Join the conversation
Follow us
Add us as a preferred source on Google
Newsletter
Subscribe to our newsletter
With companies, tech execs, and outlets such as yours truly telling everyone over and over again—sometimes very literally—that it's difficult to get your hands on memory right now and will be for some time, it can be a bit jarring to think about what the AI bigwigs are getting up to. I picture that scene where The Joker slides down a pile of cash, except it's DRAM. Anyway, case-in-point is a new AI-focused GPU from Intel that's rocking more than just a little memory.
Intel's new GPU D-Series, previously codenamed 'Crescent Island', crams up to 480 GB of LPDDR5X memory alongside its processor:
"Built on Intel's latest Xe3P architecture, the GPU combines up to 480 GB of LPDDR5X memory, 1.5 TB/s of memory bandwidth and 1.3 petaflops of FP8 compute within a 350-watt power envelope. This balance of memory capacity, compute performance and power efficiency reflects Intel's approach to optimizing infrastructure for real-world inference workloads."
Latest Videos FromPC Gamer
Watch full video here:
The GPU, Intel says, is "designed to simplify deployment within existing data center environments while supporting application portability across Intel GPU platforms."
Intel's previous attempts at AI GPUs, codenamed Gaudi, failed to compete with Nvidia to any significant degree, which led to the cancellation of its follow-up 'Falcon Shores' GPUs. This time around, though, Intel is more directly targeting AI inference workloads rather than training, too.
Intel's VP of AI Products and GTM Anil Nanduri explains: "Inference is becoming the defining AI workload, and enterprises need infrastructure that delivers the right performance at the right cost. With Intel GPU D-series, we're focused on making high-capacity inference practical to deploy and economical to scale."
For context on the memory front, Intel's previous Gaudi 3 chip features 128 GB of unified HBM. Obviously this isn't apples-to-apples, as there's also a switch in the kind of memory it's using, from HBM to slower but more efficient LPDDR5X memory. "Target applications," Intel says, "range from multi-tenant AI services to robotics and other latency-sensitive workloads."
The biggest gaming news, reviews and hardware deals
Keep up to date with the most important stories and the best deals, as picked by the PC Gamer team.
Contact me with news and offers from other Future brandsReceive email from us on behalf of our trusted partners or sponsors
Whatever the case, it's a lot more than we see in our gaming GPUs, and perhaps worryingly is closer to the kind of memory we see in our consumer devices—worrying, that is, because it signals yet more production lines that could have been tailored to us being tailored to AI use cases.
At least we might be getting more memory in the RTX 5080 next year. Except oh, wait, if that happens it'll be tailored for AI, too, and almost certainly priced accordingly. I'll just wait for all this to blow over.
Best graphics card 2026
1. Best overall: AMD Radeon RX 9070
2. Best value: AMD Radeon RX 9060 XT 16 GB
3. Best budget: Nvidia RTX 5050
4. Best mid-range: AMD Radeon RX 9070 XT
5. Best high-end: Nvidia GeForce RTX 5090