The NVIDIA DGX Spark is a personal AI supercomputer desktop built on the GB10 Grace Blackwell Superchip with 128GB unified memory, launched October 2025 for developers who need to run large AI models locally. It is the entry point to the DGX family for individuals.
The NVIDIA DGX Spark shrinks a data-centre idea onto a desk. It is a 150mm-square box weighing about 1.2kg, built on the GB10 Superchip that marries a 20-core Grace Arm CPU to a Blackwell GPU on a single package, with 128GB of unified memory and a claimed petaFLOP of FP4 AI compute. NVIDIA pitches it as a personal AI supercomputer for developers who want to prototype, fine-tune and run large models locally rather than renting cloud time, and it went on sale on 15 October 2025.
The DGX Spark is a mini desktop AI computer with the GB10 Superchip, 128GB of unified LPDDR5x memory and a 4TB NVMe SSD, starting around 3,999 US dollars before tax and import costs. It is built for running and fine-tuning large local models, not for gaming, and only makes sense if your work genuinely needs that much on-device AI memory.
The headline part is the GB10 Superchip, which pairs a Blackwell-generation GPU with a 20-core Arm CPU on one package. The CPU carries 10 high-performance Cortex-X925 cores paired with 10 power-efficient Cortex-A725 cores. NVIDIA's NVLink-C2C interconnect ties the CPU and GPU together so both see the full memory pool at far higher bandwidth than a normal PCIe link.
That memory pool is the real story. 128GB of unified LPDDR5x is coherent across CPU and GPU, which means the full model loads into one contiguous pool without being fragmented or shuttled between separate memory domains. For AI work, capacity at this scale is what lets a single desk-side box hold models that would otherwise need multiple high-end graphics cards.
It helps to be blunt about what this is not. The DGX Spark is not a gaming machine and it is not a general workstation you would buy for video editing on price-performance grounds. Its value is the 128GB unified memory feeding an AI-tuned chip, and FP4 compute measured for AI inference rather than rasterised frames. Judge it on the size of model you can load, not on benchmarks meant for a desktop GPU.
The standard configuration ships with a 4TB NVMe M.2 SSD, which matters because model checkpoints and datasets are large and you do not want to be constantly clearing space. Networking includes 10 Gigabit Ethernet, Wi-Fi 7 and Bluetooth 5.3. The 10GbE port is there for a reason: NVIDIA designed two Spark units to be linked via their ConnectX-7 ports so they can run a model too large for a single 128GB box. That direct peer connection creates a 256GB unified memory pool and enables inference on models that would otherwise need rack-mounted hardware, which is unusual at this size and tier.
The DGX Spark ships with NVIDIA AI Enterprise software, giving access to the same model development frameworks -- NIM microservices, RAPIDS, and the full CUDA toolkit -- that run on large data-centre systems. That software continuity is a genuine advantage for teams: a developer prototypes on the Spark, then deploys to cloud or on-prem infrastructure without rewriting their pipeline. It runs Ubuntu Linux and can also run Windows 11, so integration into an existing developer environment is straightforward.
This is a developer and researcher tool. If you are building AI applications, fine-tuning language models that run into the hundreds of billions of parameters, or you need a private on-device environment for sensitive data, the DGX Spark removes the cloud bill and the latency. If you mainly want to run a model or two for personal use, a high-memory desktop with a strong consumer GPU is the more sensible spend -- the AI PC range at Evetech shows what those configurations look like at local Rand pricing. For the graphics-card route to local AI, the GPU best sellers list shows which high-memory cards SA buyers are choosing for the job.
The DGX Spark draws up to 170W sustained and ships with an external 240W power brick, so it plugs into a standard wall socket without specialist electrical work. The fan noise is measured at around 38 dB at idle, comparable to a quiet desktop rather than a server. That matters for a desk environment: this is not a rack unit that needs a separate server room. Connectivity on the rear includes multiple USB-C ports, HDMI, the 10GbE port, and the dual QSFP ports for the ConnectX-7 networking. The footprint is smaller than most external hard drives stood upright, so desk real estate is not a constraint.
International pricing starts near 3,999 US dollars before tax, with retail listings elsewhere ranging higher depending on bundled software and storage. For SA buyers the landed cost includes import duties, VAT and shipping, so treat the dollar figure as a floor rather than a shelf price. Because it is a specialist product rather than mass-market, confirm current availability and lead time before committing, and be honest about whether 128GB of unified memory is something your workload will actually use.
The combination of the GB10 Superchip with 128GB of unified memory and FP4 AI compute, all in a tiny chassis. It is tuned for running and fine-tuning large AI models on-device, which a standard PC cannot do without multiple high-end cards.
It is not designed for gaming. The GPU is tuned for AI compute and the software stack targets developers, so a gaming desktop with a consumer GPU is a far better and cheaper choice if games are your goal.
128GB of unified LPDDR5x memory, shared coherently between the 20-core Arm CPU and the Blackwell GPU. That shared pool is what lets it hold large models a typical graphics card cannot fit.
One handles models that fit in 128GB. NVIDIA allows two units to be linked directly via ConnectX-7 to run a single larger model on a combined 256GB pool, which is only worth it for serious research workloads.
If you regularly fine-tune large models or need a private on-device setup, the Spark is purpose-built. For occasional local model use, a high-memory desktop with a strong consumer GPU usually gives better value once import costs are added.
The DGX Spark is a specialist tool for serious local AI work. If you are weighing it against a GPU-based build, compare your options in the AI PC range at Evetech and match the hardware to the model sizes you actually run.
The combination of the GB10 Superchip with 128GB of unified memory and FP4 AI compute, all in a tiny chassis. It is tuned for running and fine-tuning large AI models on-device, which a standard PC cannot do without multiple high-end cards.
It is not designed for gaming. The GPU is tuned for AI compute and the software stack targets developers, so a gaming desktop with a consumer GPU is a far better and cheaper choice if games are your goal.
128GB of unified LPDDR5x memory, shared coherently between the 20-core Arm CPU and the Blackwell GPU. That shared pool is what lets it hold large models a typical graphics card cannot fit.
One handles models that fit in 128GB. NVIDIA allows two units to be linked directly via ConnectX-7 to run a single larger model on a combined 256GB pool, which is only worth it for serious research workloads.
If you regularly fine-tune large models or need a private on-device setup, the Spark is purpose-built. For occasional local model use, a high-memory desktop with a strong consumer GPU usually gives better value once import costs are added.
Wi-Fi 6E and Wi-Fi 6 affect SA PC builders through motherboard support, adapter choice, and router placement. This difference between Wi-Fi 6E and Wi-Fi 6 explained guide helps plan a stable desktop network before parts are chosen.
Read more
LGA 1851 vs LGA 1700 changes AAA gaming builds through platform features and upgrade path. For SA buyers, the real-world difference is usually CPU pairing, motherboard cost, memory support, and whether a high-end GPU is already the main limit.
Read more
B850 vs B650 depends on Strategy and sim games needs, platform limits, and budget fit. Compare compatibility, latency, upgrade path, and real setup constraints before choosing for an SA build.
Read more
IPS and VA for strategy and sim games differ in text clarity, contrast, viewing angles, and long-session comfort. Use the real world difference between IPS and VA to plan an SA monitor choice for maps, menus, and detailed dashboards.
Read more
USB4 vs Thunderbolt 4 depends on Competitive esports needs, platform limits, and budget fit. Compare compatibility, latency, upgrade path, and real setup constraints before choosing for an SA build.
Read more
FSR 4 vs FSR 3.1 for cinematic story games depends on game support, image quality preferences, latency goals, and GPU fit. SA buyers should check the game and hardware path first.
Read more
Pre-Black Friday wired gaming mouse buying planning should focus on shape, specs, warranty, and usual price bands. Use this SA guide to shortlist sensible options before deal periods without assuming live discounts.
ATX 3.1 vs ATX 3.0 for Competitive Esports comes down to GPU power connectors, PSU quality, cable routing, headroom, and upgrade plans. SA buyers should match the choice to their setup instead of assuming one option always wins.
Wi-Fi 6E vs Wi-Fi 6 depends on Couch co-op gaming needs, platform limits, and budget fit. Compare compatibility, latency, upgrade path, and real setup constraints before choosing for an SA build.
Best gaming setup for SA university residence is a cautious guide for buyer planning. Compare spec fit, budget, compatibility, warranty, and SA buyer context for SA buyers without assuming live prices, availability, or exact benchmark results.
Remote camera viewing fails when the manufacturer's P2P relay cannot punch through a strict NAT or carrier-grade NAT, which blocks the outbound connection the app needs to reach the camera from outside the home network.
To fix a microphone that is too quiet on an audio interface, add an inline preamp like a Cloudlifter for up to +25dB of clean gain before the signal hits the preamp. Checking cable condition and the gain knob position rules out simpler causes first.
CAMM2 uses a dense flat connector with 644 or 666 contacts near the CPU, a completely different form from the 262-pin DDR5 SO-DIMM stick. Knowing how to read a spec sheet means you won't discover upgradeability only after opening the case.
The Razer Core X V2 launches at $349.99 with Thunderbolt 5 and 140W power delivery to the host but ships with no GPU and no PSU, requiring a separately purchased ATX PSU in its dedicated bay. It accepts cards up to 4 slots wide and includes a TB5 cable.
RX 9070 XT build frame pacing issues needs a step-by-step check before replacing parts. Review drivers, settings, thermals, power, background apps, display limits, cables, and game options before changing hardware.
Wi-Fi 6E and Wi-Fi 6 affect SA PC builders through motherboard support, adapter choice, and router placement. This difference between Wi-Fi 6E and Wi-Fi 6 explained guide helps plan a stable desktop network before parts are chosen.
Low fps on rtx 5080 build in south africa usually needs a settings, driver, thermal, or power check first. Work through safe SA troubleshooting steps before replacing parts or assuming a full upgrade.
LGA 1851 vs LGA 1700 changes AAA gaming builds through platform features and upgrade path. For SA buyers, the real-world difference is usually CPU pairing, motherboard cost, memory support, and whether a high-end GPU is already the main limit.
