NVIDIA RTX Spark systems consume significantly less energy than traditional PCs with a discrete graphics card. But still, the requirements for the power supply units remain important. RTX Spark brings together a CPU, GPU, and AI accelerator on a single chip, so its power consumption is determined by the entire platform.
Unlike traditional PCs, the RTX Spark uses a single System-on-Chip based on the Grace Blackwell architecture. This means that all the CPUs, GPUs, neural processors, and shared memory operate within one power budget.
According to official specifications, NVIDIA RTX Spark consumes 240 W for the entire system. At the same time, the GB10 Grace Blackwell chip itself requires a TDP of 140 watts. The 100-watt difference is used for SSDs, networks, USB-C, and other components powered as required by the platform design. Therefore, NVIDIA recommends using the included power adapter.
How to calculate the required PSU wattage and power headroom
To get the accurate figure on how many watts the chosen PSU must deliver to the system, start with defining the maximum power consumption of every PC component. Include everything it contains – CPU, GPU, motherboard, fans, peripherals, and others. Then, sum them up and add an extra 20-30% power headroom to easily upgrade your build in the future or replace some components without losing the PSU power. Keep in mind that the noted 20-30% is a guideline, not a strict rule, as every configuration may require specific headroom volume.
Also remember that headroom doesn’t mean that your PC will take these extra watts all the time. This figure simply provides the power supply with a reserve for peak loads and also allows it to operate within the range of highest efficiency.
Supporting peak loads and rapid power consumption changes
Since modern GPUs can consume significantly more watts in just seconds, the PSU should keep up with not only the average loads but also peak system loads. The ATX 3.1 standard includes requirements for handling short-term power spikes for exactly this reason.
Transient power spikes last for a very short time, but even a few seconds is enough to exceed the typical GPU power consumption. If the power supply doesn’t have sufficient headroom or doesn’t meet current standards, this may cause various system issues, reboots, or trigger a protection mechanism. That’s why Seasonic engineers advise looking for those PSU that has enough power headroom specifically for such power-jump cases, and ATX 3.1 support for new NVIDIA RTX Spark systems.
Power efficiency, voltage stability, and low-ripple performance
The power supply unit for the NVIDIA RTX Spark system should deliver a stable 240 W, according to the official technical specification. For this type of power source, a more efficient PSU wastes less power while converting AC to DC. This is especially important for a compact system.
RTX Spark can dynamically adjust its power consumption based on CPU, GPU, and AI computing loads. A well-designed PSU must keep the output voltage within the required limits during load changes.
Low ripple is also important for the RTX Spark as part of its overall power quality. That’s why: a stable DC voltage without significant ripple gives the sensitive GB10 electronics a more stable power environment. In ATX 3.1, the maximum ripple for +12 V is 120 mV peak-to-peak, and for +5 V and +3.3 V, it is 50 mV.
Required connectors, cable capacity, and power delivery compatibility
The PSU specs for NVIDIA RTX Spark builds should align with concrete output power and connection specifications:
– Output voltage of 48 V DC is a key requirement for powering the DGX Spark. A standard ATX +12 V output cannot be connected directly to the system.
– Maximum current is 5 A. Combined with 48 V, this results in a rated power of 240 W, which corresponds to the DGX Spark’s stock power adapter.
– The cable must be rated to handle the required 48 V / 5 A and physically match the system’s connector. There’s a warning from NVIDIA about using adapters or cables not approved for the DGX Spark.
– The RTX Spark has four USB-C ports, but only one of them supports Power Delivery. The total power available via USB-C is limited to 30 W.
Electrical protection and thermal management for continuous AI workloads
The power supply unit should have a built-in set of protection mechanisms – against overloads, overvoltage, overcurrent, short circuits, and overheating. This helps prevent damage to core system components and unexpected PSU shutdowns.
During intensive use when working with AI models, the PSU may operate near its rated load for extended periods. Therefore, it should deliver 240 W of power output and provide thermal headroom. NVIDIA recommends using the RTX Spark and its adapter in a well-ventilated environment with a temperature of 5–30°C.
How to select a reliable power supply for an RTX Spark desktop in 2026
Follow this step-by-step guide:
– Match the exact power requirements: a stable 240 W the PSU should deliver.
– Use an NVIDIA-approved adapter for optimal performance.
– Check connector and cable compatibility.
– Consider thermal conditions, as NVIDIA specifies an operating temperature range of 5–30°C for RTX Spark.
– For GB10 partner systems, check the OEM specification.
Conclusion
Choosing a PSU for the NVIDIA RTX Spark should be considered from all sides. Don’t look only at the rated PSU power, but also at its stability, efficiency, compatibility, and the ability to operate reliably under sustained load. A properly selected PSU provides a stable foundation for the productive and long-lasting operation of the entire system.