By Jeff Sigua, Product Manager
As the world is on the brink of a tech refresh, scale-out solutions will get a significant boost in both TDP and PCIe transfer speeds. Our CPUs jump from 250w up to 400w in Gen5. Our PCIe lane architecture increases from ~32 GB/s to ~64 GB/s - in each direction. The extra watts will push the boundary of I/O solutions and speeds. Scale out I/O is expanding capacity of new hardware resources instead of increased capacity of old hardware.
Over the past two decades, One Stop Systems has been specializing in cable PCIe expansion using passive PCB routing expertise, or switch architecture, to drive the tech roadmaps into the future. Recently, we have developed PCIe5 Host Interface Boards (HIB) to work up to 2 meters over copper cables and connecting to the latest Gen5 PCIe expansion backplanes or rack mount enclosures.
As next Gen motherboards become available, it seems the common theme among them is fewer native x16 PCIe5 slots coming from the CPU chips. Those extra PCIe5 lanes from the CPU have been routed to internal network or storage options built into the PCIe5 motherboards and connectors. This highlights the importance of using PCIe5 over cable, to allow for more scaled-out array of I/O cards, such as NVIDIA’s H100 GPU and other high data capture card options in development from multiple third-party companies. At the hub of the scale-out will be One Stop Systems' new Gen5 Short Depth Server (SDS). This server appliance can host up to 16x NVMe U.2 storage with removeable drive canister feature + GPU node + PCIe5 cable expansion to our sister product, the 4U Pro or our Centauri storage expansion solution.
We’re seeing full scale-out I/O rack deployments running off One Stop Systems' SDS or EOS server. The SDS or EOS can handle up to 128 end point devices, with the help of custom motherboard BIOS optimized for PCIe enumeration. As we move into PCIe Gen5, the rate of the sensor and data ingestion doubles in speeds talking back to the host server to the scaled out I/O devices. These speeds become a critical path with AI on the Edge and/or machine learning software packet sizes. NVIDIA talks about the increased tensor cores AI benefits here, as it also gets a power bump to 350w TDP per PCIe5 GPU.
Scale-out I/O with Gen5 opens the tech boundaries for our future as a society by doubling its performance. We should see a rapid growth in autonomous cars/trucks/machines and in our infrastructure around use with smart city. Increasing the edge technologies of radar, lidar, 4k/8k camera capture systems with enough processing speeds to get near real time results using machine learning.
Click the buttons below to share this blog post!
By: Jaan Mannik – Director of Commercial Sales
The term AI, or Artificial Intelligence, is everywhere nowadays and has quietly woven itself into the fabric of our daily lives. It powers the recommendations we see on streaming platforms, the navigation apps that guide us through traffic, and even the virtual assistants that answer our questions in seconds. From optimizing energy use in smart homes to predicting market shifts in finance, AI has become the invisible engine driving convenience, efficiency, and insight across industries.
In manufacturing, AI-driven robots collaborate with humans to streamline production. In agriculture, machine learning models monitor crops, forecast yields, and conserve resources. Retailers use predictive analytics to anticipate consumer needs before customers even express them. The reach of AI is no longer confined to futuristic labs, it’s in our phones, vehicles, and cities, constantly learning and adapting to serve us better.
OSS PCIe-based products deliver critical advantages for modern military sensor systems by enabling real-time data acquisition, processing, and transmission in rugged, mission-critical environments. These benefits stem from their ability to support high-bandwidth, low-latency interconnects, modular scalability, and environmental resilience, all of which are essential for today’s advanced military platforms.
Companies today are being asked to do more with data than ever before. Bigger AI models, faster insights, and workloads that don’t stay in one place, it’s a lot to keep up with. Traditional infrastructure just isn’t built for this kind of speed and flexibility.
The answer isn’t about throwing more hardware at the problem. It’s about building smarter, more agile infrastructure that adapts as demands change. And that’s where scale-out and increasingly, a blend of scale-out and scale-up come into play.
Get the latest news, product releases, upcoming events, blog posts and more!
Chana Small
March 03, 2023
Dear onestopsystems.com admin, Your posts are always well-supported by facts and figures.