Think of a processor, it contains a graphics unit with RTX 5070 power and offers high-level performance in both games and artificial intelligence. Nvidia’s new processor, RTX Spark, with an internal graphics processor powered by RTX 5070, was introduced. We also interpreted it in the image!

Nvidia’s new processor, expected as N1X, was introduced under the name RTX Spark. Nvidia is now in direct competition with Intel, AMD, Apple and Qualcomm in the PC market. Nvidia CEO Jensen emphasized that this platform combines CUDA, RTX and artificial intelligence technologies for the first time in a consumer product of this scale. Now let’s look at the details of the RTX Spark processor.

Nvidia RTX Spark features

The processor, which will be produced by TSMC with the 3nm production process, consists of 70 billion transistors. RTX Spark, which is ARM-based like Apple and Qualcomm processors, was developed in cooperation with Nvidia and MediaTek and is designed to have 20 cores at the highest level. On the GPU side, the graphics unit, which has 6,144 CUDA cores with Blackwell architecture and is expected to perform at the RTX 5070 level, serves. This graphics unit can show artificial intelligence performance at the level of 1 petaflop.

The processor, which comes with up to 128 GB LPDDR5X combined memory support, offers 600 GB / s bandwidth. Thus, both the processor and the graphics unit can benefit from the common 128 GB memory pool. In particular, the high memory requirement required for artificial intelligence is met. Naturally, not every product that uses this processor will come with 128 GB of RAM; Other affordable models with much lower RAM may also come.

What can you do with RTX Spark?

Can handle 90 GB 3D scenes,

Can process images with 12K resolution,

Can play high-end gaming at 100 FPS at 1440p with Ray Tracing and DLSS turned on

1 million context support for large language models  and you can run AI models with 120 billion parameters locally.

RTX Spark, which consumes 80w of power under load, brings together the processor, graphics card and artificial intelligence unit under one roof in a high-performance structure.

So how do you like this new ARM-based processor platform? Will RTX Spark change your decision to buy a computer? Let’s meet in the comments.

hardware architecture

RTX Spark is actually last year’s DGXSpark personal artificial intelligence used in the wonder computer Consumer version of the GB10 chip .

• CPU: Based on MediaTek Grace CPU Scalable up to 20 cores
• GPU: Blackwell architecture RTX GPU – It is expected to offer performance at the level of RTX 5070.
  • 6,144 CUDA cores
  • AI performance up to 1 petaFLOP
• Memory: up to 128GB LPDDR5X unified memory (CPU + GPU shared pool)
  • 600 GB/s bandwidth
• Connection: NVLink C2C (Something like Apple UltraFusion, low latency CPU–GPU communication)
• Production: TSMC 3nm manufacturing process
  • ~70 billion transistors
• PCIe Gen 5 support (5 lanes)

Nvidia also offers smaller configurations Models with up to 16 GB RAM He says it will happen.

Performance and usage goals

No official benchmark was given, but Nvidia shares the following goals:

• Ability to process 90 GB 3D scenes
• 12K image editing
• ~100 FPS AAA game at 1440p with ray tracing + DLSS turned on (example: Indiana Jones and the Great Circle )
• Running AI models with 120 billion parameters locally
• Support for 1 million token context length

GPU power in some scenarios It could be close to the laptop-class RTX 5070 level. It is predicted but not confirmed.

AI and software side

RTX Spark’s biggest focus is artificial intelligence:

• Local AI model running (up to 120B parameters)
• 1M context support for large language models
• Advanced AI software stack
• Special optimization preparations for Adobe (Premiere Pro, Photoshop)

Because it is ARM based x86 applications emulation (Windows on ARM / Prism) It can work with . This may cause performance loss in some software. However, many professional applications come native to ARM:

• blender
• DaVinci Resolve
• Cinema4D
• Redshift
• Topaz Photo
• CapCut
• Cubase / Bitwig Studio
• affinity

On the gaming side, Riot Games ( League of Legends , Valorant ) and Krafton ( PUBG Companies such as ) are working on ARM reinforcement. Expansion in anti-cheat systems (Easy Anti-Cheat, BattlEye, Denuvo) is also planned.

Manufacturers and devices

The first manufacturers to use the RTX Spark platform are:

• Asus ProArt P14/P16
• Dell XPS 16
• HP OmniBook X14 / Ultra 16
• Lenovo Yoga Pro 9N
• MSI Prestige N16 Flip AI
• Microsoft Surface Laptop Ultra

Also including Acer, Asus, Dell, Gigabyte, HP, Lenovo and MSI 30+ laptop and 10+ desktop systems are being developed .

Power consumption and limitations

• Single digit wattage levels at low load
• Up to ~80W output under load
• No external graphics card (eGPU) support
• There is no clear information about Linux support and handheld/console usage