MSI is taking the plunge into the miniscule desktop space with the new EdgeMesa N AI+ 1.3-litre mini-PC, based on the Nvidia RTX Spark. This is important because a ton of GPU power is now in a computer you can put on a desk or shelf. This offers the purchaser a real world product to get high end AI acceleration, but without having to buy a full-blown tower.
MSI EdgeMesa N AI+ mini-PC specifications and pricing details
This mini-PC is based on Nvidia RTX Spark, with MSI using a 20-core CPU. So much compute power is now stuffed inside this small aluminium chassis. It's made for people that need high-grade graphical power, but want to conserve space.
Specifications
- GPU Platform: Nvidia RTX Spark
- CUDA Cores: 6,144
- Memory: 64 GB or 128 GB LPDDR5X
- Storage: 1 TB or 4 TB PCIe Gen 5 SSD
- Case Volume: 1.3 litres
Inside the EdgeMesa N AI+ is 6144 CUDA cores for parallel processing and it features a vapour chamber cooling setup with dual fans. This machine can squeeze up to 140W worth of performance. There's either 64GB or 128GB of soldered LPDDR5X RAM. Storage is either 1TB or 4TB PCIe Gen 5 SSD.
We also have 10 Gigabit Ethernet, HDMI 2.1b and four USB Type-C (20Gbps each). There's Wi-Fi 7 and Bluetooth 5.4 too. We covered the MSI Prestige N16 Flip AI+ launches, which also use the RTX Spark. So we can see MSI is diving in on this new technology.
It starts at $4,699 USD (about $4,699) for the US, with the 64GB RAM and 1TB SSD version. The higher spec SKU, with 128GB RAM and 4TB storage, is $7,999 USD (about $7,999). These SKUs will be available globally, but international pricing and availability isn't yet known.
Source: NotebookCheck
seven thousand dollars for a desktop the size of a book is insane ngl
You are missing the point. It is not for gaming. It is for local inference on 128GB of unified memory. The 6144 cores are useless if you cannot feed them data.
Can that 140W thermal headroom handle sustained workloads without throttling the 20 core cpu? I worry about the vapour chamber density.
Unlikely. The chiplet interconnects will bottleneck before the GPU hits thermal limits. Expect significant power capping after ten minutes of heavy load.