Get all your news in one place.
100’s of premium titles.
One app.
Start reading
TechRadar
TechRadar
Wayne Williams

Can't afford Nvidia's expensive AI accelerators? Then consider this 10.8Kw server cluster with 32 Intel GPUs and 768GB VRAM

Sparkle C741-6U-Dual 16P.
  • Sparkle’s dense PCIe 5.0 design offers lower-cost entry into AI and data research
  • Up to 32 Intel GPUs deliver 768GB VRAM for scalable AI computing
  • 10,800W redundant power system and advanced cooling enable continuous heavy workloads

Taiwanese graphics card manufacturer Sparkle has taken the wraps off its new dense, scalable GPU server which could offer an alternative to costly AI accelerator systems like Nvidia’s H100, H200, and B200 data center GPUs.

The C741-6U-Dual 16P is designed to house up to 16 Arc Pro B60 Dual graphics cards, providing a total of 32 Intel GPUs and as much as 768GB of combined VRAM.

Each Arc Pro B60 Dual features two Battlemage BMG-G21 GPUs, and when fully equipped, the system will deliver a whopping 81,920 GPU cores.

Advanced cooling

The server supports both 24GB and 48GB GPU models, paired with 4th or 5th Generation Intel Xeon Scalable processors.

It uses Intel’s C741 chipset and supports up to 32 DDR5 memory slots, allowing 256GB per module for intensive workloads.

Sparkle has built the system around a 6U chassis with strong power and cooling capacity.

The top configuration uses five 2,700W titanium power supplies, reaching a combined 10,800W output, while the smaller version includes four 2,400W units rated at 7,200W.

With as many as 15 fans installed, the system has clearly been built to manage heat effectively during continuous workloads.

Each GPU connects to the CPU through a PCIe 5.0 x8 interface, either directly or via a Microchip switch, maximizing data bandwidth while avoiding potential bottlenecks caused by shared links.

There’s support for dual M.2 slots and several NVMe and SATA bays for local storage.

Although there’s no way of knowing at this stage how the Arc Pro B60 will perform in large-scale inference or training tasks, the flexibility offered by the hardware could appeal to researchers and developers seeking affordable parallel compute power.

Unsurprisingly, Sparkle hasn’t revealed pricing for the C741-6U-Dual 16P but potential buyers can make a purchase inquiry on the website.

Via Videocardz

(Image credit: Sparkle)

Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!

And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.

You might also like

Sign up to read this article
Read news from 100’s of titles, curated specifically for you.
Already a member? Sign in here
Related Stories
Top stories on inkl right now
One subscription that gives you access to news from hundreds of sites
Already a member? Sign in here
Our Picks
Fourteen days free
Download the app
One app. One membership.
100+ trusted global sources.