Puget Systems

10/12/2026

When we tested Noctua’s first 240mm liquid cooler against our previous Asetek OEM unit, the initial results were substantial enough that we reran the tests.

On our Ryzen 9 9950X test system, the hottest CPU core reached 96°C with the Asetek cooler and 86°C with the Noctua during our combined CPU and GPU stress test.

The repeat testing added important context: the Noctua briefly peaked at 91°C before settling back to 86°C. Across the evaluation, cooling improvements ranged from 3–10°C.

Then we moved Noctua’s upgraded fans onto the Asetek radiator. Temperatures still peaked at 96°C, suggesting the fans alone weren’t responsible for the bulk of the improvement.

A good first result is interesting. Checking it is the part that matters.

Patrick Eaton walks through the testing, pump-noise observations, and qualification process in our latest article.

Read the full evaluation:
https://www.pugetsystems.com/labs/articles/noctuas-debut-240mm-aio-cooler-impresses/

10/06/2026

Wow, what a fantastic night! 🙌

Thank you to everyone who came out to the Future Experience Showcase! Seeing everyone all immersed in the LED vignettes and experiencing real-time tracking and integration in person was exactly what we had in mind when we built this event.

And the PLAY vignette? Let’s just say the Greedo shot first 👀

If you missed this one, make sure you’re following us, because there will be a next one.

Huge thank you to our co-host and title sponsor for making this all possible, and to our incredible supporting partners:

, , , , , and

We couldn’t do it without you. 🙏

09/24/2026

What happens when a customer needs software Puget Systems hasn’t tested before?

Recently, a reseller came to us with a customer who needed a 1U rackmount server for StorageDNA’s DNAfabric platform. It was a new use case for us: our Labs team had not benchmarked DNAfabric, and our consultants had not previously configured a server for that workflow.

So we started where we always do: with the actual environment.

We worked with the reseller to understand the customer’s current setup, requirements, expected growth, and budget. Then we went directly to StorageDNA’s technical team to understand how their software would perform best on the hardware we carry.

That collaboration produced two viable paths:
🔹 A lower-cost configuration that met the immediate need.
🔹 A higher-performance option designed to grow with the customer’s workload.

The customer chose the scalable option: a Puget Server E100-1U with an AMD 8-core processor, 96GB of DDR5 memory, mirrored M.2 boot drives, mirrored SATA SSDs for local storage, and an ATTO 8-port external SAS/SATA controller.

When a workflow is new, the right response is to listen closely, validate the requirements, and collaborate until the hardware fits the work.

09/21/2026

Topaz Video's Starlight models are some of the most demanding local AI workloads we've tested and until now, there was no public data on how well consumer GPUs actually handle them.

We ran our own benchmarks across all four Starlight variants (Mini, Sharp, Fast 2, and Precise 2.5) on NVIDIA and AMD's current lineup. A few things stood out:
- The RTX 5090 leads every single variant, and on Starlight Fast 2 it's over 10x faster than the 5080, 5070 Ti, and 5060 Ti which all land within a couple seconds of each other, likely due to a VRAM bottleneck
- Starlight Sharp and Fast 2 both require at least 16GB of VRAM, which knocks the RTX 5070 out of contention entirely
- AMD's Radeon RX 9070 XT is only compatible with one variant (Precise 2.5), and its processing time was roughly 3x slower than the weakest NVIDIA card we tested

If your workflow leans on Starlight, VRAM capacity matters just as much as raw compute.
Full breakdown in the comments.

09/14/2026

Preparations are in full swing for The Future Experience Showcase on Sept 24th-26th!

If you’re in the LA area come check us out! We’ll be opening our doors and showcasing LED applications for film/Tv, live events, brand activations, and more! See it, Touch it, Experience it.

📅 Sept 24th-25th Private Demo - Reach out directly to book!

📅 Sept 26th: Open House - limited tickets available on Eventbrite, Link in our Bio.

Huge thank you to our partner, and our main sponsor for making this possible.

09/09/2026

We've been sitting on this for a few weeks: we've launched a NEW PODCAST!!!

Built Different goes inside the studios, labs, and workspaces of people doing the wildest, most demanding creative and scientific work out there and the hardware built to keep up with them.

Episode 1 is personal: our own founder, Jon Bach, on 25 years of building this company, why we publish our benchmark data for the whole industry to see, and the values behind it all.

We make the machine. You make the magic.

Link in bio to watch or listen. New episodes monthly!

08/31/2026

We recently worked with a client who builds their own systems regularly, but needed something outside their usual range: a rock-solid testbed that could image multiple USB drives simultaneously as part of their deployment pipeline. Rather than chase down workstation-grade parts on an unfamiliar platform during a supply crunch, they came to us.

We built them a Threadripper platform with 14 USB ports, letting them image the drives in under 30 minutes; a job that previously took 2-3 separate test benches to accomplish.

This is the value of a supply chain team that maintains real OEM and distributor relationships. Even when the broader market is tight, we can still build the right system configurations for the workflows that need them.

Read the full story: https://hubs.ly/Q04v-N0m0

08/25/2026

Liquid-cooled GPU systems are now closer to home.

The partnership between Puget Systems and Comino Liquid Technology is evolving to bring Comino’s unique liquid-cooled GPU platforms to North American customers with final assembly, integration, testing, and support handled by Puget Systems.

What this means for you:
• Final integration, system validation, testing, and burn-in in the US
• A more direct path from configuration to deployment
• 3-year coverage backed by Puget Systems Support

The lineup is designed for demanding deep learning & AI inference, HPC & scientific research, and other GPU-intensive workflows where system density, deployment timelines, and support ownership matter.

Explore Puget Systems, Powered by Comino: https://hubs.ly/Q04v8sZM0

08/18/2026

The most useful thing we found testing two AMD Radeon AI PRO R9700s wasn't a benchmark number. It was what didn't work.

Stock vLLM multi-GPU does not run on RDNA 4. Not tensor parallelism, not pipeline parallelism. Both abort during RCCL collective initialization before a model ever loads.

If you're speccing a dual-R9700 workstation on the assumption that the standard vLLM recipe applies, you'd find out after the PO cleared.

The working path is llama.cpp's ROCm backend, which distributes over direct HIP transfers with no collectives. It runs Qwen3.6-27B at Q4 across both cards at 23.4 tok/s single-user, 56 tok/s aggregate at eight concurrent users. That's genuinely usable for the 27B tier.

Full testing and setup guide:

Can two Radeon AI PRO R9700 cards deliver production-quality local LLM inference and image generation? How do they compare against Intel's Arc Pro B70 GPUs?

08/13/2026

SIGGRAPH 2026 was a blast and we had more going on than we could fit into one booth!

We teamed up with Form Studios and AMD to bring the Robo Dojo to life a real-time motion capture experience where attendees fought AI-powered robots and walked away with a branded clip generated on the spot.

Down the hall, District Cinema's "Roll for Reality" turned an actual play D&D session into a full-on virtual production experience.

And our president, Jon Bach, led a Birds of a Feather discussion on where workstations are headed as AI workloads and memory constraints reshape infrastructure decisions.

Watch the recap video below!

SIGGRAPH 2026 is a wrap, and what a week it was!From AI-powered m...