Have you ever hit a screw with a hammer because you didn’t have a screwdriver at hand?
No? Me neither.
Each job requires using the right tool for the best results. In high-performance computing or for “big data” and hosting applications, the main job is maximizing throughput.
In this case, you need the right processor for the right job: an NVIDIA Tesla GPU to provide ultra-high throughput for parallel tasks, and a high-performance Sandy Bridge CPU for single-thread jobs.
The new Dell PowerEdge C8000-series systems provide both, along with a host of other tools optimized for a range of accelerated computing needs.
At launch, the servers support Tesla M2090s, the only available GPU option. Best of all, later they will support the forthcoming Tesla K20 GPUs based on Kepler, the world’s highest performance, most energy-efficient compute architecture.
These new Dell systems are ideal for a new class of mainstream, high-volume enterprise customers who need the power and flexibility to handle the next generation of compute-intensive apps.
Integrated Design Ideal for HPC, “big data” and Hosting
Each 4U chassis houses up to 4 C8220X double-wide sleds with two Tesla GPUs and two CPUs in each node – that’s up to 80 GPUs in a single server rack.
Another standout feature is the “shared infrastructure,” which integrates the latest Sandy Bridge CPUs, NVIDIA GPUs and storage in a single chassis. With this high level of integration, users will benefit from the high performance density — an impressive 5.3 teraflops double-precision per 4U chassis in GPUs alone.
The chassis design also enables flexibility. Customers can mix and match to tailor their configuration to their target workloads.
The PowerEdge C8000 series will initially ship with Fermi-based Tesla M2090 GPUs. Future systems will offer the next-generation Tesla K20 GPUs – the same GPUs that will power supercomputers such as the world-leading Titan system at Oak Ridge National Labs, the Blue Waters system at the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign, and the Stampede system at the Texas Advanced Computing Center.
If you’re running compute-intensive apps in a mixed-workload environment, what GPU-CPU density and other features would you like to see in your next server?
For more on Kepler, Tesla K20 GPUs and GPU computing follow @NVIDIATesla.