www / QHT , re: QCOM accelerators / NVLink licence?,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
Hence, the NVLink licence?
QHT>>>
QC has patented every single breakthrough with that "innovative memory architecture". Either QC license it for tons of money to AMD and Nvidia or we keep everything close to the chest in hopes of creating a moat for inference servers. We'll see if they can bring this innovative memory to edge devices as well.
>>>>>>>>>>>>>>>>>>>>>>
Copilot <<<<<<<<<<<<<<<<<
Q1 - NVLink license- from who the whom
Nvidia licensed its NVLink technology to Qualcomm and Fujitsu, among others, as part of its NVLink Fusion initiative announced in May 2025.
?? NVLink Licensing Details
- Licensor: Nvidia
- Licensees: Qualcomm, Fujitsu, MediaTek, Marvell, Alchip Technologies, Astera Labs, Synopsys, Cadence
- Technology Licensed: NVLink Fusion — includes NVLink port designs and memory atomic protocol
- Purpose: To enable third-party CPUs and accelerators to connect directly to Nvidia GPUs using NVLink’s high-speed interconnect
- Strategic Goal: Expand Nvidia’s ecosystem by allowing semi-custom AI infrastructure with tight CPU-GPU integration
Qualcomm plans to use NVLink Fusion in its upcoming data center CPUs, allowing them to pair directly with Nvidia GPUs for rack-scale AI workloads.
Q2 why no INTC / AMD yet
Intel and AMD are not part of Nvidia’s NVLink Fusion licensing because they are pursuing their own competing interconnect standards — and strategically avoiding dependency on Nvidia’s proprietary ecosystem.
?? Why Intel and AMD Aren’t in NVLink Fusion
- Competing Standards:
- AMD is leading the UALink Consortium, which includes Intel, Arm, Broadcom, Google, and others. UALink aims to create an open, vendor-neutral memory-coherent interconnect for AI accelerators and CPUs.
- This directly challenges Nvidia’s NVLink, which remains proprietary and requires tight coupling with Nvidia GPUs.
- Strategic Independence:
- Intel and AMD both build their own AI accelerators (e.g., Gaudi, MI300X) and want to avoid architectural lock-in to Nvidia’s stack.
- NVLink Fusion still requires Nvidia silicon in the system, limiting true mix-and-match flexibility.
- Control and IP Restrictions:
- Nvidia retains control over critical software and hardware layers in NVLink Fusion, which may deter companies like Intel and AMD that prioritize full-stack autonomy.
- Market Positioning:
- Intel and AMD are positioning themselves as alternatives to Nvidia, not collaborators. Joining NVLink Fusion could dilute their differentiation in the AI infrastructure market.
|
