CONFIRMED: Marvell announced the Structera X 3406 and 3506 in a press release on 12 October 2026. “Industry-first” claims and all performance figures are Marvell’s own. Marvell expects samples in Q1 2027.
Marvell has announced the second-generation Structera X 3406 and Structera X 3506, which it calls “CXL memory expanders” for cloud, enterprise and AI systems [1]. CXL (Compute Express Link) is a standard that lets servers attach extra memory over a fast connection, beyond the memory slots next to the processor. Marvell says AI inference, the work of running trained models, is held back by memory.
What Marvell announced
According to Marvell [1]:
- Six memory channels: “Six DDR4/DDR5 memory channels deliver 50% more channels than the previous generation”. Marvell calls this “industry-first”.
- Capacity: “up to 18 DDR4 RDIMM slots supporting 9TB for Structera X 3406 and 12 DDR5 RDIMM slots supporting 6TB for Structera X 3506”. The chips use PCIe Gen 6 connections.
- Sharing: memory can be shared “with up to four hosts and up to 256GB/s PCIe6.0 x16 of host bandwidth”, which Marvell also calls “industry-first”.
- Compression: “Up to 250GB/s compression/decompression throughput unlocks up to 2–3x more effective memory capacity, depending on workloads”, using LZ4 compression.
- Security: post-quantum cryptography, Caliptra 2.0, CXL IDE and AES-256 memory encryption.
Marvell’s performance claims
For AI chatbots and other large language model services, Marvell claims the chips deliver “up to 40–50% faster TTFT at equivalent context length, or up to 2x longer supportable context windows at equivalent latency compared to the previous generation” [1]. TTFT, or time to first token, is how long a model takes to start answering. For recommendation systems, Marvell claims “up to 2x higher queries per second” [1]. These figures have not been independently tested.
Availability
Marvell said the Structera X 3406 and 3506 “are expected to sample in Q1 2027” and are being demonstrated at the OCP Global Summit [1]. Marvell said the first-generation Structera X 2404 and 2504 are “already shipping to hyperscalers” [1]. No pricing was given.
Comment
The release quotes Jim Handy, general director of Objective Analysis: “today’s memory shortage and subsequent high prices are driving IT managers to use CXL to re-deploy used DDR4 DIMMs into DDR5 systems for immediate cost savings” [1].
“We are delivering industry-leading CXL controllers, co-developed with leading hyperscalers to address the most demanding AI infrastructure requirements,” said Will Chu, executive vice president and general manager, Custom Cloud Solutions at Marvell [1]. The release also carries statements from AMD, Arm, Alibaba, Samsung Electronics and SK hynix [1].
Sources
- Marvell Technology, “Marvell Unveils Second-generation Structera X, Industry-first Six-channel CXL 3.2 Memory Expanders for AI Inference”, press release (Business Wire), 12 October 2026. https://investor.marvell.com/news-events/press-releases/detail/1037/marvell-unveils-second-generation-structera-x-industry-first-six-channel-cxl-3-2-memory-expanders-for-ai-inference
