AMD Helios rackscale solution for AI offers 15% more AI Compute than NVIDIA Vera Rubin NVL72

Q: How does Helios achieve the claimed 15% more peak FP4 compute than NVIDIA's Vera Rubin NVL72?

AMD Helios achieves the claimed 15% more peak FP4 compute by packing 72 AMD Instinct MI455X GPUs together with 6th Gen AMD EPYC Venice CPUs, AMD ROCm software, and AMD Pensando networking to deliver a combined 2.9 ExaFlops of FP4 performance. AMD highlights Helios's higher subsystem specs versus Vera Rubin, including 31 TB of HBM4 memory (50% more), 43 TB/s scale-out bandwidth, and 1.7 PB/s of memory bandwidth, which together support the higher peak FP4 figure.

Q: What are the GPU and CPU components used in the AMD Helios rack-scale system?

The AMD Helios rack-scale system uses 72 AMD Instinct MI455X GPUs paired with the latest 6th Gen AMD EPYC "Venice" server CPUs. It also integrates AMD ROCm software and AMD Pensando networking as part of the solution.

Q: What networking hardware does Helios use for rack-to-rack and cluster connectivity?

AMD Helios uses AMD Pensando networking, specifically the new AMD Pensando Vulcano 800 AI NICs, for rack-to-rack and cluster connectivity. These Vulcano 800 AI NICs provide the high-bandwidth, low-latency connectivity needed for scaling to gigawatt-scale AI clusters.

Q: What total FP4 compute performance (in ExaFlops) does a single Helios rack deliver?

A single AMD Helios rack delivers 2.9 ExaFlops of FP4 compute.

Built for AI infrastructure and large-scale foundational-model training and inference, AMD Helios is an impressive new rack-scale solution.

VIEW GALLERY - 5

Kosta Andreadis

Senior Editor

Published Jul 23, 2026 1:59 PM CDT

1-minute read time

TL;DR: AMD announced Helios, a rack-scale AI supercomputer combining 72 Instinct MI455X GPUs, 6th Gen EPYC Venice CPUs, ROCm, and Pensando networking to deliver 2.9 ExaFlops FP4, 15% higher peak FP4 compute than NVIDIA Vera Rubin NVL72, 31TB HBM4, and claimed 30% more tokens per dollar for scalable AI clusters.

Voice: Kosta AndreadisSpeed

0:00 / 2:37

AMD has officially launched its AMD Helios rackscale solution for AI infrastructure and large-scale foundational-model training and inference. As a supercomputer, AMD Helios combines the new AMD Instinct MI455X GPU, 72 of them, with the latest 6th Gen AMD EPYC 'Venice' server CPUs, AMD ROCm software, and AMD Pensando networking. It's a lot of cutting-edge tech coming together, delivering 2.9 ExaFlops of FP4 Compute.

AMD Helios rackscale solution for AI offers 15% more AI Compute than NVIDIA Vera Rubin NVL72 5

VIEW GALLERY - 5 IMAGES

According to AMD, it's enough to deliver 15% more compute (peak FP4 performance) than NVIDIA's latest Vera Rubin NVL72 rack-scale solution. And when it comes to raw specs, there is a notable difference in favor of Helios. For example, the 31TB of HBM4 memory capacity is 50% more than Vera Rubin. And with 43 TB/s scale-out bandwidth and 1.7 PB/s of memory bandwidth, AMD notes that Helios can offer AI companies 30% more tokens per dollar.

"As demand grows for AI factories and frontier AI, customers need infrastructure that scales efficiently from a single rack to large gigawatt-scale AI clusters while maximizing performance, performance per watt and cost per token," AMD explains. "AMD Helios delivers an integrated rack-scale building block that simplifies deployment while providing the compute, networking and open software foundation needed for the next generation of AI infrastructure."

Read more: AMD shows off next-gen Zen 6-based EPYC 'Venice CPU, Instinct MI455X GPU for Helios AI racks

AMD Helios rackscale solution for AI offers 15% more AI Compute than NVIDIA Vera Rubin NVL72 4

Frequently Asked Questions

TweakBot answers common questions about this news using TweakTown's own coverage from this page and related content from our archive. Tap a question to reveal the answer, or type your own below.

Question #1

How does Helios achieve the claimed 15% more peak FP4 compute than NVIDIA's Vera Rubin NVL72?

Question #2

What are the GPU and CPU components used in the AMD Helios rack-scale system?

Question #3

What networking hardware does Helios use for rack-to-rack and cluster connectivity?

Question #4

What total FP4 compute performance (in ExaFlops) does a single Helios rack deliver?

Have a question not listed here? Ask below and TweakBot will answer it.

Of course, AMD Helios isn't confined to a single rack, as it's designed for gigawatt-scale AI clusters thanks to the new AMD Pensando Vulcano 800 AI NICs providing high-bandwidth, low-latency connectivity for growth. And with that, AMD Helios is already being deployed to a wide range of leading AI companies like Anthropic, Meta, Microsoft, Oracle, OpenAI, as well as supporting the infrastructure of companies like Dell Technologies, Lenovo, IBM, and others.