# RTX Spark vs M5 Max (November 2026): specs, memory and prices

> Both chips reach 128GB of unified memory, but Apple lists 614GB/s of bandwidth while NVIDIA quotes 1 FP4 petaflop and runs CUDA on Windows on Arm.

URL: https://claudegpt.si/news/rtx-spark-vs-m5-max  
Published: 2026-11-18T15:03:09.731Z  
Author: ClaudeGPT Si Newsroom  
Section: Reviews  
Topics: Nvidia, Apple, AI chips, Comparison, Hardware  
Publisher: ClaudeGPT Si

## Key takeaways

- Both NVIDIA RTX Spark and Apple M5 Max top out at 128GB of unified memory.
- Apple lists 614GB/s of bandwidth for the 40-core M5 Max; NVIDIA publishes no figure, and ASUS lists up to 300GB/s for its RTX Spark desktop.
- As of November 18, 2026, a 128GB Dell XPS 16 with RTX Spark costs $6,999.99, and M5 Max MacBook Pros start at $4,099 with 36GB.

RTX Spark vs M5 Max comes down to this: both chips top out at 128GB of unified memory, but on the published figures the M5 Max reads that memory about twice as fast, while RTX Spark brings CUDA and Windows. NVIDIA says RTX Spark compact desktops from ASUS, Dell and MSI [went on sale on November 18, 2026](https://blogs.nvidia.com/blog/rtx-spark-desktop-local-ai-available-now/), joining the RTX Spark laptops that shipped in October. Apple's M5 Max ships in the 14-inch and 16-inch MacBook Pro.

If you run AI models locally, the choice is mostly about software. RTX Spark runs NVIDIA's CUDA stack on Windows. The M5 Max runs macOS, with Apple's MLX framework and Metal. Price is the other big difference, and it depends on which machine you buy, not just the chip.

This comparison is based on the makers' published specs and results, not hands-on testing.

## RTX Spark vs M5 Max specs side by side

Every figure below comes from NVIDIA's and Apple's own pages, plus ASUS for the one RTX Spark bandwidth number published so far. "Not published" means the maker hasn't said.

| | NVIDIA RTX Spark (N1X) | Apple M5 Max |
| --- | --- | --- |
| CPU | 20-core NVIDIA Grace (18-core in the lower configuration) | 18 cores: 6 super cores and 12 performance cores |
| CPU architecture | Arm (runs Windows on Arm) | Apple silicon |
| GPU | Blackwell RTX GPU, 6,144 CUDA cores (5,120 in the lower configuration) | 32-core or 40-core GPU, with a Neural Accelerator in each core |
| Tensor or AI hardware | Fifth-generation Tensor Cores with FP4 | Neural Accelerators in the GPU, plus a 16-core Neural Engine |
| Max unified memory | Up to 128GB LPDDR5X (64GB on the lower configuration) | Up to 128GB (40-core GPU only; the 32-core version has 36GB) |
| Memory bandwidth | Not published by NVIDIA. ASUS lists up to 300GB/s for its ProArt GR1X | 460GB/s (32-core GPU) or 614GB/s (40-core GPU) |
| AI performance, as stated | "Up to 1 petaflop" of FP4 AI performance | No TOPS or FLOPS figure. Apple claims "over 4x the peak GPU compute for AI" of the previous generation |
| Operating system | Windows 11 | macOS (macOS 27 on the newest models) |
| Main AI software | CUDA, TensorRT, PyTorch-CUDA for Windows on Arm, Linux tools through WSL | MLX, Metal |
| Form factors | Laptops (14 to 16 inches) and compact desktops | 14-inch and 16-inch MacBook Pro |
| Starting price with this chip (US, as of November 18, 2026) | $3,799.99 for the Dell XPS 16 Creator Edition | $4,099 for the 14-inch MacBook Pro |

Core counts don't translate across companies. NVIDIA counts CUDA cores, which are small units, and Apple counts GPU cores, which are large blocks. 6,144 against 40 says nothing about which GPU is faster.

## CPU and GPU: 20 Grace cores vs 18 Apple cores

NVIDIA's [RTX Spark page](https://www.nvidia.com/en-us/products/rtx-spark/) lists two configurations of the N1X chip: a 20-core Grace CPU with a 6,144-core Blackwell GPU, and an 18-core CPU with a 5,120-core GPU. NVIDIA's [May 31 announcement](https://nvidianews.nvidia.com/news/nvidia-microsoft-windows-pcs-agents-rtx-spark) says the CPU and GPU are joined by NVLink-C2C, NVIDIA's chip-to-chip link, and that MediaTek helped design the CPU. NVIDIA doesn't publish a split between fast and efficient CPU cores.

Apple's [MacBook Pro tech specs](https://www.apple.com/macbook-pro/specs/) list the M5 Max with an 18-core CPU, made of 6 super cores and 12 performance cores, and either a 32-core or a 40-core GPU. Apple's [March 3 announcement](https://www.apple.com/newsroom/2026/03/apple-debuts-m5-pro-and-m5-max-to-supercharge-the-most-demanding-pro-workflows/) says the M5 Max is built from two 3nm dies joined into one chip, and puts a Neural Accelerator, a block for the matrix math AI models use, in every GPU core.

## Memory: same 128GB, different speed

Both chips share one pool of memory between CPU and GPU, which is what lets a laptop load a model that wouldn't fit on a graphics card. Both stop at 128GB. On RTX Spark, 128GB needs the 20-core configuration; the 18-core one tops out at 64GB. On the M5 Max, 128GB needs the 40-core GPU.

Bandwidth, how fast the chip can read that memory, is where they split. Apple publishes 614GB/s for the 40-core M5 Max. NVIDIA hasn't published a figure for RTX Spark. ASUS lists "up to 300GB/s" for its ProArt GR1X desktop, and Dell lists LPDDR5X at up to 9,400 MT/s on the XPS 16, without a bandwidth number.

That gap matters for chatbots and coding agents run at home. When a model writes text, it reads its weights from memory for every token, so bandwidth sets a ceiling on how fast words come out. Raw compute matters more for reading a long prompt, training and image generation. No maker has published token-per-second results that compare the two chips on the same model.

## AI performance: FP4 petaflop vs no number

NVIDIA states "up to 1 petaflop" of AI performance for RTX Spark. Its product page and its October 16 release say that figure is at FP4, a 4-bit number format that trades precision for speed. A chip quotes a bigger number at lower precision, so figures at different precisions can't be compared directly.

Apple gives no FLOPS or TOPS figure for the M5 Max on its spec page or in either announcement. It says the M5 Max delivers "over 4x the peak GPU compute for AI compared to the previous generation," and that pros can run language models "with hundreds of billions of parameters entirely on device." Apple's machine learning team reported [up to 4x faster time to first token](https://machinelearning.apple.com/research/exploring-llms-mlx-m5) on M5 than on M4 when running MLX, its open-source framework.

Neither claim has an independent benchmark behind it on the makers' pages, and the two can't be converted into each other.

## Software: Windows on Arm and CUDA vs macOS, MLX and Metal

RTX Spark machines run Windows 11 on an Arm CPU. That has two sides. On the plus side, NVIDIA says RTX Spark supports CUDA natively, offers Linux developer tools through WSL (Windows Subsystem for Linux), and brought the first PyTorch-CUDA build for Windows on Arm.

On the minus side, apps built for Intel and AMD chips run through emulation unless a native Arm version exists. NVIDIA's November 18 post says Adobe Illustrator's new native Arm version runs up to 30% faster than its emulated one, and lists native versions of Premiere, Photoshop, InDesign and a public beta of Substance 3D Painter. Check your own apps before you buy.

The M5 Max runs macOS. For AI, Apple's stack is Metal for the GPU and MLX for models, and Apple says MLX needs macOS 26.2 or later to use the Neural Accelerators. CUDA does not run on a Mac, so CUDA-only code has to be ported or run elsewhere.

## Which machines use each chip

RTX Spark comes from several PC makers. NVIDIA's [October 16 release](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-RTX-Spark-PCs-Available-Today--A-New-Beginning-for-Windows-PCs/default.aspx) lists these laptops:

- ASUS ProArt P14 and P16
- Dell XPS 16 (sold as the XPS 16 Creator Edition)
- HP OmniBook Ultra and OmniBook X 14
- Lenovo Yoga Pro 9n
- Microsoft Surface Laptop Ultra
- MSI Prestige N16 Flip AI+ and Prestige N16

NVIDIA's November 18 post says compact desktops from ASUS, Dell and MSI are on sale, naming the ASUS ProArt GR1X and the MSI EdgeMesa N AI+. Acer, GIGABYTE, HP and a Microsoft Surface RTX Spark Dev Box are listed as coming soon. On November 18, ASUS's GR1X page still showed a "Notify Me" button and Dell's AI PC page listed its Creator Edition Desktop as "available soon," so stock may lag the announcement.

The M5 Max comes from Apple alone. It ships in the 14-inch and 16-inch MacBook Pro, both with the Liquid Retina XDR display and in the new MacBook Pro with PROLED XDR Display that Apple [announced on October 28](https://www.apple.com/newsroom/2026/10/apple-unveils-groundbreaking-macbook-pro-with-proled-xdr-display/) ([our coverage](/news/apple-touchscreen-macbook-pro-proled)).

## RTX Spark vs M5 Max prices

US prices as of November 18, 2026, from the makers' own stores.

| Machine | Configuration | Price |
| --- | --- | --- |
| Dell XPS 16 Creator Edition (RTX Spark) | 18-core CPU, 5,120-core GPU, 32GB, 512GB | $3,799.99 |
| Dell XPS 16 Creator Edition (RTX Spark) | 20-core CPU, 6,144-core GPU, 32GB, 1TB | $3,999.99 |
| Dell XPS 16 Creator Edition (RTX Spark) | 20-core CPU, 6,144-core GPU, 64GB, 1TB | $4,799.99 |
| Dell XPS 16 Creator Edition (RTX Spark) | 20-core CPU, 6,144-core GPU, 128GB, 2TB | $6,999.99 |
| 14-inch MacBook Pro, Liquid Retina XDR | M5 Max, 32-core GPU, 36GB, 2TB | $4,099 |
| 16-inch MacBook Pro, Liquid Retina XDR | M5 Max, 32-core GPU, 36GB, 2TB | $4,399 |
| 14-inch MacBook Pro, PROLED XDR | M5 Max, 32-core GPU, 36GB, 2TB | $4,499 |
| 14-inch MacBook Pro, PROLED XDR | M5 Max, 40-core GPU, 48GB, 2TB | $5,099 |
| 16-inch MacBook Pro, PROLED XDR | M5 Max, 32-core GPU, 36GB, 2TB | $4,799 |
| 16-inch MacBook Pro, PROLED XDR | M5 Max, 40-core GPU, 48GB, 2TB | $5,399 |

NVIDIA said on October 16 that RTX Spark laptops start at $2,599, with prices varying by region; Dell's cheapest RTX Spark laptop costs more than that. Apple's store pages show starting configurations only, so the price of a 128GB M5 Max isn't in the table. ASUS and Dell haven't posted prices for their RTX Spark desktops, and MSI's US store refused automated requests, so its price couldn't be checked.

For the same 128GB, the one figure that can be compared today is Dell's: $6,999.99 for a 128GB XPS 16. Compare it with a 128GB MacBook Pro built in Apple's configurator before you decide.

## Which one to pick

Pick RTX Spark if your AI work depends on CUDA, you need Windows, or you want a small desktop that runs agents all day. Pick the M5 Max if you live in macOS, use MLX or Mac creative apps, or want the higher memory bandwidth Apple publishes. Wait for independent tests if you care about tokens per second: neither maker has published numbers that settle it.

## Frequently asked questions

### Is RTX Spark Arm?

Yes. RTX Spark PCs run Windows on Arm: NVIDIA lists native Windows on Arm versions of Adobe apps for RTX Spark, and apps without an Arm version run through emulation. MediaTek, an Arm chip designer, worked with NVIDIA on the 20-core Grace CPU.

### How much does an RTX Spark laptop cost?

NVIDIA said on October 16, 2026 that RTX Spark laptops start at $2,599, with prices varying by region. As of November 18, 2026, Dell's XPS 16 Creator Edition costs $3,799.99 with 32GB and $6,999.99 with 128GB.

### How does RTX Spark compare with the AMD Ryzen AI Max+ 395?

AMD lists the Ryzen AI Max+ 395 with 16 Zen 5 x86 cores, Radeon 8060S graphics with 40 graphics cores, up to 128GB of LPDDR5x-8000 and up to 126 overall TOPS. RTX Spark is an Arm chip with 20 CPU cores and 6,144 CUDA cores that NVIDIA rates at up to 1 petaflop of FP4; the two figures use different measures and can't be compared directly.

### What about RTX Spark vs M5 Pro?

Apple's M5 Pro tops out at 64GB of unified memory and 307GB/s of bandwidth, against 128GB for RTX Spark's 20-core configuration. As of November 18, 2026, the 14-inch MacBook Pro with M5 Pro starts at $2,499 with the Liquid Retina XDR display and $2,899 with the PROLED XDR display.

## Sources

- [RTX Spark compact desktops available now](https://blogs.nvidia.com/blog/rtx-spark-desktop-local-ai-available-now/), NVIDIA
- [NVIDIA RTX Spark](https://www.nvidia.com/en-us/products/rtx-spark/), NVIDIA
- [NVIDIA and Microsoft Reinvent Windows PCs for the Age of Personal AI](https://nvidianews.nvidia.com/news/nvidia-microsoft-windows-pcs-agents-rtx-spark), NVIDIA
- [NVIDIA RTX Spark PCs Available Today](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-RTX-Spark-PCs-Available-Today--A-New-Beginning-for-Windows-PCs/default.aspx), NVIDIA
- [MacBook Pro tech specs](https://www.apple.com/macbook-pro/specs/), Apple
- [Apple debuts M5 Pro and M5 Max to supercharge the most demanding pro workflows](https://www.apple.com/newsroom/2026/03/apple-debuts-m5-pro-and-m5-max-to-supercharge-the-most-demanding-pro-workflows/), Apple
- [Apple unveils groundbreaking MacBook Pro with PROLED XDR Display](https://www.apple.com/newsroom/2026/10/apple-unveils-groundbreaking-macbook-pro-with-proled-xdr-display/), Apple
- [Buy MacBook Pro with Liquid Retina XDR](https://www.apple.com/shop/buy-mac/macbook-pro), Apple
- [Buy MacBook Pro with PROLED XDR Display](https://www.apple.com/shop/buy-mac/macbook-pro-proled-xdr), Apple
- [Exploring LLMs with MLX and the Neural Accelerators in the M5 GPU](https://machinelearning.apple.com/research/exploring-llms-mlx-m5), Apple Machine Learning Research
- [Dell XPS 16 Creator Edition](https://www.dell.com/en-us/shop/laptop-computers/spd/xps16dx16263), Dell
- [Windows PCs and Workstations for Agentic AI](https://www.dell.com/en-us/shop/computer-pcs/lp/ai-development-portfolio), Dell
- [ProArt GR1X Mini PC](https://www.asus.com/displays-desktops/mini-pcs/proart-mini-pc-series/proart-gr1x-mini-pc/), ASUS
- [AMD Ryzen AI Max+ 395](https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-plus-395.html), AMD
