SparkPilot turns iPhone, iPad and Mac into the control plane for your DGX Spark(s). Onboard Sparks and clusters, see live telemetry, manage models, power App Intents (Apple Foundation) with local AI, and launch your favourite harness at the click of a button. Local when you want it, connected when you need it. Beta out next week. @NVIDIAAI fancy testing it?

FB marketplace find £20 2009 MacBook in immaculate condition but the HDD was dead. Chucked a 256GB SSD in there along with Ubuntu 24. Thought I would use it as an additional red team box for experiments, siloed, running @pidotdev and being driven by Deepseek V4 0731 Flash from the dual spark cluster. I should install Kali really. Let the fun begin 😎

2 billion tokens served from my dual Spark setup running DeepSeek V4 0731 Flash. From reverse engineering the RDMA protocol to building some of the Spark mobile app. You have been a faithful servant. Qwen 3.8 Flash is now loading.

Ha. Ask fable anything about firmware, it shits the bed and cries cyber security. I tried on my RDMA project and it was like nar fam. I’ll just stick to using it for finding coupons and energy deals.

I got sick of managing my DGX Sparks from the terminal, so I started building a native app for iPhone, iPad and Mac. You will be able to set up a fresh Spark with a tap of a few buttons, adding vLLM, anti-OOM, and your favourite models that fit. Additionally, you will be able to set up clusters easily by connecting the CX7 cable and clicking Pair. I’m also building a custom Apple Foundation Models provider backed by Spark, with Siri and Shortcuts exposed through App Intents. Very early, but the first setup flow is working. This can be connected via USBC MCDMA, Wi-Fi, or remotely via Tailscale/ZeroTrust.

I wanted to see if there was any real difference between the DGX Spark and its siblings. Seven GB10 Sparks, same silicon, different chassis, airflow, fan design, power tuning and sustained behaviour. I compared the HP ZGX, Acer GN100, MSI EdgeXpert, Lenovo PGX, Gigabyte AI TOP ATOM, ASUS GX10 and Dell Pro Max on thermals, sustained performance and noise. • HP ZGX Nano G1n Best overall balance. Peak CPU 77.3°C, GPU 69°C. HP rates it at 27.6 dBA under intensive workloads. My ZGX also runs about 10°C cooler than my PGX under sustained loads. • Acer Veriton GN100 Best thermals. Peak CPU 74.7°C, GPU 69°C, GPU power 69.2W. • ASUS Ascent GX10 Strong thermal design. Peak CPU 87.3°C, GPU 82°C, GPU power 69.8W. • MSI EdgeXpert MS-C931 Performance focused. MSI claims 1,729 tok/s vs ~1,600 tok/s for the reference design, with the rear chassis 15°C cooler, top 9.1°C cooler and SSD 9°C cooler in its own testing over a DGX Spark. • Lenovo ThinkStation PGX Performance focused. Sustained system draw has been measured at roughly 104 to 160W. In longer heat-soaked runs, it can edge out DGX Spark by a few percent. • Gigabyte AI TOP ATOM Aggressive power tuning. Peak CPU 90°C, GPU 81°C, GPU power 75.5W. • Dell Pro Max with GB10 Runs warmer. Peak CPU ~87.7°C, GPU ~80°C, GPU power ~70W. Sustained decode settles around 80 to 83°C CPU and 68 to 71°C GPU. The performance gap between these machines is usually small; the thermal gap is a different story.

MCDMA | Release Update 🚀 After a couple of conversations with @alexocheema, it became clear we’re very aligned on what MCDMA could unlock for local AI and what I’ve been trying to achieve with it. Alex originally reached out because he’d been following the work and saw a lot of overlap with what @exolabs has been building. Once we started digging in, I got pretty excited about what becomes possible when you combine the two. So I’ve decided to collaborate with EXO on the next stage of MCDMA. The goal stays exactly the same. If anything, this just lets us be more ambitious and explore some things that weren’t possible before. I want to spend a little more time exploring that before sharing the next step. Genuine thanks to Alex for reaching out and offering support before any collaboration was even on the table. Really looking forward to what we can build together. More soon.

MCDMA | Metal CUDA Direct Memory Access 🚀 If you have a Spark and an Apple Silicon Mac, MCDMA gives you a direct RDMA path between CUDA memory and Metal-side unified memory over USB-C. Registered memory, rkeys, one-sided READ/WRITE, two-sided SEND/RECV with credit flow control. Same verbs both ways, no master/slave. The Mac writes straight into CUDA-mapped memory on the Spark, and the Spark writes straight back into Mac memory. My setup takes it a little further: Spark 1 ⇄ CX7 ⇄ Spark 2 (prompt processing) Spark 1 ⇄ USB-C ⇄ Mac Studio (Decode) Spark 2 ⇄ USB-C ⇄ Mac Studio (Decode) Two independent MCDMA USBC links, so the Studio isn't stuck behind one cable; both Sparks move data concurrently, and it writes results back into either. Measured, every byte delivery verified: • 939 MB/s single link • 1.80 GB/s Mac → both Sparks, concurrent • 1.25 GB/s both Sparks → Mac, concurrent • 24 µs round-trip, 41k msg/s small-message One Spark + One Mac works. Two is just how I'm using it: DeepSeek prompt processing across the Sparks, decode on the Studio. Benchmarks, tests, Open Source, and write-up this week. @NVIDIARTXSpark @NVIDIAAI @NaderLikeLadder @msharmavikram There’s still a lot of performance headroom here. If the currently locked USB4 controller can be allowed to train at full capability, I’d love to test how far we can push this. Please check your DMs.