Tobias Mann

Tested: AMD's Strix Halo vs Nvidia's DGX Spark Hands On: Two tiny boxes, 128 GB apiece – but very different strengths

#ICYMI I tested Nvidia's GB10-based DGX Spark against the AMD Ryzen AI Max+ Pro 395-based HP Z2 Mini G1a.

For AI development, the GB10 is hard to beat, but for an multi-purpose PC, AMD's Strix Halo is more flexible.

Full story: @theregister.com

www.theregister.com/2025/12/25/a...

#AI #PC

Tested: AMD's Strix Halo vs Nvidia's DGX Spark Hands On: Two tiny boxes, 128 GB apiece – but very different strengths

#AMD's Strix Halo (Ryzen AI Max+ 395) and #Nvidia's DGX Spark (#GB10) are an interesting class of hardware for local AI #development.

Which is right for you depends on whether you want a #PC that runs AI or a PC for #AI.

My latest for @theregister.com
www.theregister.com/2025/12/25/a...

The one exception, of course, is Google's TPU pods which are still using optically switched torus topologies.

If you hadn't noticed, AI infra all pretty much looks like #Nvidia's NVL72 now.

#AWS' Trainium3 UltraServers are the latest example

For more on the convergent evolution of #AI infra and the future of the #datacenter, find my latest on @theregister.com

forums.theregister.com/forum/all/20...

Amazon keeps pressure on Intel, AMD with 192-core Graviton5 re:invent: The homegrown chips now account for half of all new CPUs added to AWS over the past three years

Here’s what you need to know about Graviton5
- 192 cores
- 2MB L2 per core
- 192MB of L3
- DDR5 7200 support - 8800MT/s planned
- 25% higher perf than AWS dual G4 M8g instances (192 cores vs 2x96 cores)

My latest for @theregister.com #reinvent #reinvent25

www.theregister.com/2025/12/04/a...

And that’s a wrap on Re:Invent

#reinvent #reinvent25

Funny how Amazon's Trn3 UltraServer looks a lot like Nvidia's NVL72...

#reinvent #reInvent25

Amazon to fuse Nvidia's NVLink into Trainium4 accelerators Re:Invent: Meanwhile, Trainium3 makes its debut promising million-chip training clusters

To compete with Nvidia, Amazon is embracing Nvidia, fusing the GPU giant's NVLink interconnects into its next-gen Trainium4 UltraServers.

My latest for @theregister.com

theregister.com/2025/12/02/a...

#reinvent #reInvent25 #aws #ai #cloud

Blackhole QuietBox, Tenstorrent's AI workstation reviewed hands on: $12K machine promises performance that can scale to 32 chip servers and beyond but immature stack makes harnessing compute challenging

The Tenstorrent's Blackhole QuietBox is an impressive bit of #AI kit.
- 128GB of GDDR6
- 2TB/s of MEM BW
- 3 petaFLOPS of FP8
- 12.8 Tbps of interconnect BW
- and an ambitious software strategy that... needs some polish

Find my full review on @theregister.com

www.theregister.com/2025/11/27/t...

Europe joins the US as an exascale superpower SC25: EuroHPC's biggest iron still has more to give with Universal Cluster expansion expected to come online next year

Europe has joined the US as an exascale super power. EuroHPC's biggest iron has crested 1 EF on the Top500's HPL benchmark.

Is HPL still the bench to watch when Blackwell offers more than 200x the FP8 perf as FP64? #sc25

My latest for @theregister.com

www.theregister.com/2025/11/17/e...

I'll be at SC25 next week repping @theregister.com for the fourth year running. Give me a shout if you're going to be in St Louis for the event.

TPU v7, Google's answer to Nvidia's Blackwell is nearly here : Chocolate Factory's homegrown silicon boasts Blackwell-level perf at massive scale

Nvidia's biggest scale up domain is 72 GPUs. Google's is 9,216 TPUs.

Historically TPUs have trailed on FLOPS, memory, & bandwidth. That's no longer the case with Ironwood.

Google has a Blackwell-class TPU with absurd scale. More on @theregister.com ⬇️

www.theregister.com/2025/11/06/g...

I might be crucified for saying this, but OEM GPU servers are boring now.

Everything is just a rebadged HGX box or NVL rack now.

The only point of differentiation becomes whose lights out management interface does it have.

Side view of a Tenstorrent QuietBox (Blackhole) with the side panel removed.

Say what you will about the hardware or the software, @tenstorrent.bsky.social's Blackhole QuietBox is a gorgeous bit of kit.

Full review is already in the works #AI #Workstation #watercooling #Tech

DGX Spark Nvidia's desktop supercomputer: first look hands on: This relatively-affordable AI workstation isn’t about going fast; it’s about doing everything well enough

This was a fun review. I don't think folks realize how big a deal the DGX Spark is. A year ago an #Nvidia #workstation with 128GB+ of VRAM cost tens of thousands of dollars. Is it better than AMD's Strix Halo? Now that's the real question.

#DGXSpark #AI
www.theregister.com/2025/10/14/d...

On the road again.

If you can't use AI then it's bye bye, Accenture tells staff ai-pocalypse: Consultancy says machine learning advice is making bank

If you can't use AI then it's bye bye, Accenture tells staff.

TT-QuietBox™ The TT-QuietBox™ Liquid-Cooled Desktop Workstation offers superior performance per dollar for developers looking to run, test, and develop AI models, or port and develop libraries for HPC.

tenstorrent.com/hardware/tt-...

It’s the Blackhole variant

I’m embarrassed to admit that I’ve never held a QSFP DD cable before. It’s enormous compared to the SFP+ DACs I’m used to.

I love my dog. He's usually a very good boy. But having adopted him as a 10 week-old puppy less than a year before COVID hit, his anxiety can be overwhelming at times.

He had a vet visit today. He got his shots, but wouldn't let the Dr. do a proper exam. We get to come back in 4 weeks and try again

TT-QuietBox (Blackhole)

Look what just landed in the lab

Terminal output showing Neofetch: tobiasmann@uranus ----------------- OS: Ubuntu 24.04.3 LTS x86_64 Host: TRX50 AERO D -CF Kernel: 6.14.0-29-generic Uptime: 1 hour, 50 mins Packages: 795 (dpkg) Shell: bash 5.2.21 Resolution: 1920x1080 Terminal: /dev/pts/0 CPU: AMD Ryzen Threadripper 7960X s (48) @ 5.364GHz GPU: NVIDIA GeForce RTX 3090 Ti GPU: NVIDIA GeForce RTX 3090 Ti Memory: 1068MiB / 128295MiB

Something curious I’ve noticed is I’m using virtualization less in my homelab preferring instead to run bare metal with containers.

I still keep a PVE box for VMs when they’re warranted but a lot of the stuff I’m doing can be achieved using containers.

#Linux #VM #Homelab #tech

Nvidia's context-optimized Rubin CPX GPUs were inevitable Analysis: Why strap pricey, power-hungry HBM to a job that doesn't benefit from the bandwidth?

Ever since Nvidia started talking about disaggregated inference architectures at GTC this spring, I had a feeling a HBM-less prefill accelerator was only a matter of time.

My latest for @theregister.com

www.theregister.com/2025/09/10/n...

#Nvidia #AI #Datacenter #Servers #HPC

R9 9900X on a MSI X870E-P MB.

Matrix.org homeserver grinds to a halt after RAID meltdown : Engineers wrangle 55 TB restore and traffic replay as millions of messages queue up

Oof. I can relate. Last night my email was filled with TrueNAS warnings. A drive reported Smart Errors.

Logs: extended test failed.
short test fails. Yep dead drive. 😟

Backup my core files and drop in the cold spare. 6 hours of resilvering left to go.

www.theregister.com/2025/09/03/m...

Yes indeed. Stout little card. Though I’m rather eager to put the Radeon Pro AI r9700 to the test.

Here’s batch 1 perf figs if you’re curious.

My Bugatti and Ferrari are on loan.

The 6000 Ada is in the other system.