NVIDIA’s upcoming RTX Spark platform may arrive in two hardware configurations, according to details found inside an early Windows 11 graphics driver.
The driver files reference two versions of the RTX Spark N1X. One configuration appears to include a Blackwell GPU with 6,144 CUDA cores, while the second is listed with 5,120 CUDA cores.
NVIDIA has already confirmed that RTX Spark will combine a Blackwell based GPU with a processor developed alongside MediaTek. However, the company has not yet provided a complete specification list or explained how many versions of the platform will reach consumer devices.
The driver information suggests NVIDIA may offer a higher performance model alongside a reduced configuration for less expensive systems. Both could support as much as 128GB of LPDDR5X memory, allowing manufacturers to provide large memory capacities even on the lower tier variant.
Windows 11 driver files point to two RTX Spark configurations
The information was found in NVIDIA’s first graphics driver associated with RTX Spark. Version 616.00 is described as a developer preview rather than a finished consumer release.
A file inside the package lists two devices under the RTX Spark N1X name. The higher configuration contains 6,144 CUDA cores, which would equal 48 Streaming Multiprocessors if NVIDIA retains the usual relationship of 128 CUDA cores per SM.
The second configuration has 5,120 CUDA cores, corresponding to 40 Streaming Multiprocessors.
| RTX Spark configuration | CUDA cores | Expected SM count | Possible CPU pairing |
|---|---|---|---|
| Higher tier model | 6,144 | 48 | 20 core MediaTek CPU |
| Lower tier model | 5,120 | 40 | 18 core MediaTek CPU |
| Maximum memory support | Up to 128GB LPDDR5X | Not applicable | Expected on both versions |
The possible CPU pairings have not been confirmed. The more powerful GPU may be matched with the previously announced 20 core processor, while the reduced model could use an 18 core version.
This would allow NVIDIA and its partners to create several types of products from the same platform. Premium laptops and compact workstations could use the full configuration, while lower priced systems could use the reduced version.
CUDA core counts do not reveal final gaming performance
The 6,144 core figure matches the CUDA core count of some existing NVIDIA graphics products, including the desktop GeForce RTX 5070. The 5,120 core version has the same count as the laptop GeForce RTX 3070.
These comparisons do not mean RTX Spark will provide identical performance.
Graphics performance depends on more than CUDA core count. Clock speed, power limits, memory bandwidth, cache, architecture, cooling, and software support all affect real results.
RTX Spark is also expected to use shared LPDDR5X memory rather than the dedicated GDDR memory found on desktop graphics cards. Shared memory can provide large capacity and improve efficiency in compact systems, but it may offer lower bandwidth than a dedicated desktop GPU.
The platform’s power target will be particularly important. A laptop or small form factor PC cannot provide the same cooling and electrical limits as a full desktop graphics card.
Early demonstrations suggest strong graphics capability
Although full benchmarks are not available, RTX Spark has reportedly demonstrated demanding games with advanced visual settings.
Alan Wake 2 and Pragmata have both appeared to run smoothly on the platform with path tracing enabled. These demonstrations suggest that the Blackwell GPU could support modern ray tracing features despite operating inside a more efficient integrated design.
Upscaling and frame generation are likely to play an important role. NVIDIA may use its latest DLSS technologies to improve frame rates while reducing the amount of native rendering work required.
That approach would be similar to current GeForce laptops, where performance depends on a combination of GPU power, neural rendering, and carefully selected settings.
However, demonstration footage cannot replace independent testing. Resolution, image quality mode, frame generation settings, and power consumption must be known before meaningful comparisons can be made.
Large memory support could benefit AI and professional workloads
The reported support for up to 128GB of LPDDR5X memory may be one of RTX Spark’s strongest features.
Because the CPU and GPU are expected to share the same memory pool, the graphics hardware could access far more memory than most conventional laptop GPUs. This could help with local AI models, 3D rendering, video production, engineering applications, and other workloads that require large datasets.
A traditional laptop GPU may include between 8GB and 16GB of dedicated video memory. RTX Spark could potentially access a much larger shared pool, depending on how NVIDIA and system manufacturers configure memory allocation.
Shared memory does not guarantee better performance, but it may allow workloads to run that would otherwise exceed the capacity of a standard mobile GPU.
Key specifications remain unknown
NVIDIA has not confirmed the final clock speeds, memory bandwidth, power limits, device names, or launch pricing for either configuration.
It is also unclear whether both versions will appear in consumer laptops or whether one will be reserved for developer systems, compact desktops, or professional devices.
The driver is an early developer release, so the listed configurations could change before launch. Hardware identifiers inside preview drivers sometimes refer to test designs that do not become commercial products.
Still, the discovery provides the clearest indication yet that RTX Spark will not be limited to one specification. A two tier lineup would give manufacturers more flexibility and help NVIDIA compete across several price and performance categories.
The 6,144 core model appears positioned as the flagship, while the 5,120 core version could reduce cost and power consumption without sacrificing maximum memory capacity. Independent benchmarks will ultimately determine whether either version can deliver desktop class gaming and AI performance inside a compact Windows PC.
Discussion (0)
Be the first to comment.