Business, Enterprise, Graphics

Dell Jumps on GPGPU Bandwagon


Overkill VR is a fast paced VR action game

History is definitely repeating itself. Back in early 2000s, AMD introduced Opteron and big OEMs were afraid of adopting AMD given close ties to Intel. But, IBM and HP took the risk and profited from it, while Dell was on the backburner.

GPGPU modules in this image show none other than Tesla M1050, not the M2050 which Dell touts in specs. Definitely a shortage of cards, even for a photo shoot.This time around, with GPGPU revolution at hand, Dell doesn’t want to miss this boat. First Top tier solutions provider to answer IBM’s GPGPU-enabled iDataPlex dx360 M3 is well-known computing giant from Red Rock, TX. Dell launched PowerEdge C410x, a 3U PCI Express expansion chassis that can house up to 16 Tesla M1050, M2050 and M2070 cards.

Unlike IBM’s product, Dell’s PowerEdge C410X is actually a 3U PCI Express expansion chassis that can house 16 PCIe cards. The cards on offer are Tesla M1050 4GB, M2050 3GB and M2070 6GB [to start shipping this quarter], but Dell also mentions non-GPGPU cards such as PCIe-based SSD storage and ultra-fast network cards – when they pass Dell’s certification process.

When it comes to computational power, Dell can pack up to  16.48 TFLOPS of computing power [single precision, 8.24 TFLOPS for DoublePrecision]. Given that nVidia revised [downspecc’ed] Tesla M2050 and M2070, it is easy to calculate that one module will deliver 1.03 TFLOPS SP i.e. 0.51 TFLOPS in Double precision mode. The performance is naturally, tied to power consumption and the strictest testing standard that Dell uses for their data center boxes.

In any case, good to see IBM and Super Micro getting some competition.