Tuesday, September 1st 2026
NVIDIA's "Rubin CPX" GPU Reborn with HBM4 Memory, No More GDDR7
NVIDIA has reportedly revived its "Rubin CPX" AI accelerator project, which had previously been put on hold. Today, well-known supply chain analyst Ming Chi Kuo reported that NVIDIA is reintroducing the "Rubin CPX," now featuring HBM4 memory instead of the previously planned GDDR7. Late last year, NVIDIA announced its plans to develop a specialized accelerator derived from the "Rubin" GPU family, designed for large-scale agentic AI workloads. After the announcement, the project was reportedly paused, but NVIDIA has now redesigned the memory architecture. Previously, the "Rubin CPX" was slated to have 128 GB of GDDR7 memory, but it will now include 168 GB of HBM4 memory. This GPU SKU is housed in a separate rack with configurations ranging from 64 to 256 standalone CPX GPUs per rack.
This GPU is dedicated to prefill workloads, powering racks that manage input context for LLMs and KV cache processing. Essentially, NVIDIA is separating prefill from decode tasks, with CPX GPUs handling prefill and regular "Rubin" GPUs managing decode. A rack tray with eight CPX GPUs can handle 1.34 TB of long-context prefill and the associated KV cache, all utilizing HBM4 memory for faster processing and increased bandwidth. Since the previous design did not require HBM4 integration, NVIDIA has redesigned the package, likely using TSMC's CoWoS-S or CoWoS-L for packaging.
The chip delivers 30 PetaFLOPS of NVFP4 compute performance on a monolithic die. It includes four integrated NVENC and four NVDEC video encoders directly on-chip, enabling streamlined multimedia workflows without the need for external processing. For LLMs now containing trillions of parameters, having dedicated racks for prefill while others handle decode allows for much faster token delivery. NVIDIA also recommends maintaining a 1:1 ratio of CPX GPUs to regular GPUs in the decode rack.
Source:
Ming Chi Kuo
This GPU is dedicated to prefill workloads, powering racks that manage input context for LLMs and KV cache processing. Essentially, NVIDIA is separating prefill from decode tasks, with CPX GPUs handling prefill and regular "Rubin" GPUs managing decode. A rack tray with eight CPX GPUs can handle 1.34 TB of long-context prefill and the associated KV cache, all utilizing HBM4 memory for faster processing and increased bandwidth. Since the previous design did not require HBM4 integration, NVIDIA has redesigned the package, likely using TSMC's CoWoS-S or CoWoS-L for packaging.
9 Comments on NVIDIA's "Rubin CPX" GPU Reborn with HBM4 Memory, No More GDDR7
Not quite. Enterprise already uses HBM for AI accelerators because of its 2048-bit bus and massive bandwidth. it isn't hogging GDDR7. GDDR7 is designed specifically for consumer GPUs because HBM is way too expensive and complex (requiring silicon interposers) to put on budget cards like an RTX 6060. Plus, GDDR7 actually has lower raw latency (~12-15 ns vs HBM's 15-20 ns), which gaming workloads care about far more than ultra-wide AI memory buses
You seem to have some gaps in your knowledge that you should fill before correcting others...
GDDR7 is being used in lots of enterprise-focused products, not only consumer ones.