PCIe® 7.0 Technology – Too Soon? Not Fast Enough?

Many technologists seem to agree on one thing: innovation has never been faster. Why? Isn’t it obvious? AI. The race is on to offer differentiated hardware solutions for optimized AI performance at the edge and in the data center.

Data scientists, AI system architects, IC designers, optical engineers, interconnect providers like Samtec and other solutions providers are rethinking system topologies. GPU clustering, disaggreated computing, cache-coherent memory, optical interconnects, and 200 Gbps channels are among the solutions defining AI system architectures.

Given those realities, a number of system architects ponder if PCIe 7.0 technology 128 GT/s raw bit rate is fast enough. That seems fair given GPU-to-GPU and GPU-to-CPU bridging protocols reach beyond 200 Gbps already. Many wonder if PCIe 7.0 can keep up as an alternative to GPU-to-XPU interconnects.

But is that a fair a question to ask?

PCIe 7.0 vs GPU-to-XPU Use Cases

Frankly, equating PCIe 7.0 technology with GPU-to-XPU interconnects is not an apples to apples comparison. Fairly, the comparison is more apples to oranges. Think about the use cases.

PCI Express® (and it’s predecessor PCI®) have historically linked x86 CPUs with a multitude of I/Os in general compute applications – desktops, laptops, workstation, servers and more. Examples of general compute I/Os may include graphics cards, storage devices, network adaptors or even AI accelerators.

PCI-SIG®, the consortium that owns and manages PCI specifications as open industry standards, consistently works with member companies to double PCIe specification performance every three years. With the planned delivery of the PCIe 7.0 spec in 2025, they are right on schedule.

Let’s look at the oranges. GPU-to-XPU interconnects were designed to enable high-speed, point-to-point GPU-to-GPU communications. In the earliest versions of accelerated computing, system architectures may have included one GPU for every CPU. Oversimplified, this is a 1 CPU : 1 GPU architecture.

As AI models have grown, a 1 CPU : n GPU architectures have quickly emerged. With LLMs and other models having TRILLIONS of parameters, thousands of GPUs need “clustered” to handle the necessary parallel computing. Faster and faster GPU-to-XPU interconnects have been the result. 200 Gbps GPU-to-XPU interconnects are the standard as of this writing.

So what’s the end result? Is PCIe 7.0 technology too fast or is it DOA? It depends who you ask. In this writer’s opinion, PCIe 7.0 technology will complement GPU-to-XPU interconnects for the I/O uses cases transferable between general compute, acclerated compute, and ever evolving AI system topologies.

The question comes up, though, how real is PCIe 7.0 technologies?

Samtec/Alphawave Semi 128 Gbps PAM4 Interoperability

The short answer is PCIe 7.0 capable technology is real and ready for protoype implementation. Recently at ECOC and again at OCP, Alphawave Semi and Samtec demostrated 128 Gbps PAM4 system interoperability. The set-up combined Alphawave Semi’s IP with Samtec’s high-performance interconnects.

The Alphawave Semi PipeCORE™ PCI-Express PHY transmits 128 GT/s data to a 2.5+ meter transmission line consisting of various Samtec high-performance interconnect systems. Despite the 2.5+ meter overall length and eight connection points, the system still achieves an excellent pre-FEC BER of e-10 or better.

The NovaRay® I/O Panel Mount Cable System is a high-speed, high-density cable system that provides ~4 Tbps aggregate data rate. It is designed for cable-to-cable bulkhead panel connections. The cable assembly in this demo is 2.0 meters long.

So, is PCIe® 7.0 technology coming too soon? Is it fast enough? Just like Goldilocks and her porridge, PCIe 7.0 technology is the right solution for the right time.

Please visit www.samtec.com/pcie for more information.