- Sparkle’s dense PCIe 5.0 design gives lower-cost entry into AI and information analysis
- As much as 32 Intel GPUs ship 768GB VRAM for scalable AI computing
- 10,800W redundant energy system and superior cooling allow steady heavy workloads
Taiwanese graphics card producer Sparkle has taken the wraps off its new dense, scalable GPU server which may provide an alternative choice to expensive AI accelerator methods like Nvidia’s H100, H200, and B200 information heart GPUs.
The C741-6U-Twin 16P is designed to deal with as much as 16 Arc Professional B60 Twin graphics playing cards, offering a complete of 32 Intel GPUs and as a lot as 768GB of mixed VRAM.
Every Arc Professional B60 Twin options two Battlemage BMG-G21 GPUs, and when totally outfitted, the system will ship a whopping 81,920 GPU cores.
Superior cooling
It makes use of Intel’s C741 chipset and helps as much as 32 DDR5 reminiscence slots, permitting 256GB per module for intensive workloads.
Sparkle has constructed the system round a 6U chassis with robust energy and cooling capability.
The highest configuration makes use of 5 2,700W titanium energy provides, reaching a mixed 10,800W output, whereas the smaller model contains 4 2,400W models rated at 7,200W.
Signal as much as the TechRadar Professional e-newsletter to get all the highest information, opinion, options and steering your online business must succeed!
With as many as 15 followers put in, the system has clearly been constructed to handle warmth successfully throughout steady workloads.
Every GPU connects to the CPU by way of a PCIe 5.0 x8 interface, both straight or through a Microchip change, maximizing information bandwidth whereas avoiding potential bottlenecks attributable to shared hyperlinks.
There’s assist for twin M.2 slots and several other NVMe and SATA bays for native storage.
Though there’s no manner of figuring out at this stage how the Arc Professional B60 will carry out in large-scale inference or coaching duties, the pliability supplied by the {hardware} may attraction to researchers and builders in search of reasonably priced parallel compute energy.
Unsurprisingly, Sparkle hasn’t revealed pricing for the C741-6U-Twin 16P however potential consumers could make a purchase order inquiry on the web site.
By way of Videocardz
Follow TechRadar on Google News andadd us as a preferred source to get our professional information, opinions, and opinion in your feeds. Ensure that to click on the Observe button!
And naturally you can even follow TechRadar on TikTok for information, opinions, unboxings in video type, and get common updates from us on WhatsApp too.