The Beastly Liquid-Cooled Multi-GPU Comino GRANDO RM-S Workstation
A ~$40,000 Monster In A Rack Case
The Comino GRANDO RM-S is a serious AMD powered workstation, with a 64 core Ryzen Threadripper PRO 3995WX on an ASUS Pro WS WRX80E-SAGE SE WIFI. That board is populated by no less than four NVIDIA RTX A6000s, 32GB of Gigabyte ECC DDR4-2600 and both an AORUS 2 TB NVMe Gen 4 M.2 SSD and a Samsung 7.68TB PM1733 2.5″ U.2 Enterprise SSD. The system is also cooled by a custom watercooling loop that covers the CPU and all four GPUs.
They tested this monster against a relatively normal computer as well as what TechPowerUp refers to as a Supercomputer. It does deserve that name, with it’s 20 Xeon Gold 6148 CPUs, 3.84TB DDR4 2666 and 16 NVIDIA Tesla V100s; which would run you a bit more than $40K.
The 3DMark performance might not amaze you, but for the scientific tests the Comino GRANDO RM-S more than measures up. It sits much closer to the Supercomputer than Intel probably likes. It never matches the performance but it certainly comes close, and is something a decent sized company or university could actually afford.
Comino impressed before with the consumer-grade liquid-cooled Otto, and we now step up a few notches to the workstation/server class GRANDO RM-S. It features multi-GPU and workstation-class hardware, cooled efficiently for high performance under non-stop work loads ranging from scientific applications to game development.
More Tech News From Around The Web
- High-end be quiet! PC Build – with 5800X3D & RX 6950 XT @ KitGuru
- ASRock DeskMeet B660 Small Form Factor PC @ Tweaktown
- Lenovo ThinkStation P340 Tiny 1L PC with NVIDIA GPU @ Serve The Home
- PCSpecialist Odin i7 Ti Air @ Kitguru
- Intel NUC 12 Extreme “Dragon Canyon” SFF PC @ Tweaktown
- Cyberpower Infinity X123 Pro Prebuilt @ Kitguru
“It sits much closer to the Supercomputer than Intel probably likes.”
Intel likely has no concerns: these systems are GPGPU hosts, the CPUs are only there to keep the GPUs fed. Not surprising as you’re spending 3-4x as much on GPUs as the CPU in this device (wholly half the total machine cost is the GPUs, the other half is every other component).