BIOS IT’s latest GPU offering in partnership with Supermicro - the BIOS ANNA (Artificial Neural Network Accelerator) delivers maximum processing acceleration for the most compute intensive workloads in the smallest physical dimensions (1U form factor). Also on offer from BIOS IT is the NVIDIA® DGX-1™ - the world’s first purpose-built system for deep learning with fully integrated hardware and software that can be deployed quickly and easily and boasting revolutionary performance that significantly accelerates training time.

Both offerings make for powerful on-premise GPU solutions, ideal for HPC, Big Data, Finance and Deep Learning workloads – that is if your budget extends to accommodate your own physical GPU appliance. A more recent cost-effective approach that BIOS IT is seeing demand for from its customers, is access to GPU technology in the cloud. Thanks to its partnership with vScaler, a cloud infrastructure and services provider, BIOS IT can offer its customers just that.

 

BIOS IT have a range of on-premise and cloud based GPU solutions

To date BIOS IT’s GPU cloud platform incorporates K80 GPUs along with a number of popular Deep Learning frameworks, including Tensorflow, Caffe and Theano that can be leveraged on-demand. The team at vScaler are currently rolling out the inclusion of P100s which will offer a significant performance hike and of course save customers the initial capital expenditure of an on-premise GPU solution. For those customers that do desire an on-premise solution, BIOS IT can also offer burst capabilities into vScalers public cloud for when demand dictates, thus giving end users the option of a true hybrid approach.

Depending on the customer’s requirements, BIOS IT can provide a GPU accelerated solution to suit. The finer details of whether that means on-premise, or in the cloud; featuring K80s, M40 cards or P100s; incorporating frameworks for application specific workloads or not, all comes down to the customer use case and of course budget constraints.

Looking at the figures alone, the standard cost for P100-based on-premise hardware is approximately $9k per P100 GPU card included, tacked on to the price of the physical server. This means a 50% hike on the cost of a K80 based system (which historically has been approximately $6k per card). However, P100s offer 60% performance improvement over K80 – which should be incentive enough to justify the cost difference. With its cloud offering starting at $0.55 per/hour for a K80 based system (approx. $4800 over 12 months), and without the costs of the physical hardware, this already offers BIOS IT customers a much more cost effective route, with similar (relative) savings being promised once the P100 cards are in situ.

Contact your BIOS IT sales representative today to discuss your GPU requirements.