IBM Cloud Enhances GPU Capabilities
To support customers’ HPC and AI workloads efficiently, IBM Cloud™ GPU offerings provide 17.3 percent better performance per dollar compared to AWS.
Why IBM for GPUs?
The value of choosing IBM Cloud for your GPU requirements rests within the IBM Cloud enterprise infrastructure, platform and services. You get direct access to one of the most flexible server-selection processes in the industry, seamless integration with your IBM Cloud architecture, APIs and applications, and a globally distributed network of modern data centers at your fingertips.
Where GPUs can help you
Gaming
Whether you develop online games, build game platforms or create gaming innovations, IBM Cloud can help you deliver gaming experiences that perform and engage.
Financial services
With smart and secure solutions from IBM Cloud, financial institutions of all sizes can operate like nimble startups.
Scientific research
Scientists, researchers and doctors can solve complex molecular modeling calculations faster and analyze massive amounts of seismic data.
Healthcare
IBM Cloud helps healthcare innovators thrive in a hybrid, multicloud world. The journey to the cloud is easy with IBM’s robust suite of advanced data, AI tools, and HIPAA and GxP-enabled data sets.
IBM versus the competition
When it comes to GPU servers, IBM offers a wider variety of options than competitors do.
IBM versus AWS
IBM Cloud bare metal servers with GPUs deliver better performance per dollar on five TensorFlow machine-learning models than Amazon Web Services servers.
IBM versus Google Cloud
IBM Cloud offers GPUs on both bare metal and virtual servers. The only GPU options from Google Cloud are virtual server instances, as Google Cloud doesn’t offer bare metal instances.
IBM versus Alibaba
Both IBM Cloud and Alibaba Cloud offer GPU options on virtual machines, but IBM is unique in that they offer bare metal GPU as an option.
Virtual server GPU options
NVIDIA Tesla V100
GPU: 1 x Volta V100
Memory: 16 GB HBM2
NVIDIA Tensor Cores: 640
NVIDIA CUDA Cores: 5120
Memory bandwidth: 9000 GB/sec
Hourly
Up to 2 GPUs per virtual server
Up to 16 vCPU and 120 GB RAM
AI, deep learning and HPC
Starting at:
USD 3.06/hr.
Monthly
Up to 2 GPUs per virtual server
Up to 16 vCPU and 120 GB RAM
AI, deep learning and HPC
Starting at:
USD 2,233.80/mo.
NVIDIA Tesla P100
GPU: 1 x Pascal GP100
Memory: 16 GB HBM2
NVIDIA CUDA Cores: 3584
Memory bandwidth: 720 GB/sec
Hourly
Up to 2 GPUs per virtual server
Up to 16 vCPU and 120 GB RAM
Rapid scale and provisioning
Starting at:
USD 1.95/hr.
Monthly
Up to 2 GPUs per virtual server
Up to 16 Cores and 120 GB RAM
Rapid scale and provisioning
Starting at:
USD 1,279.18/mo.
Bare metal server GPU options
NVIDIA Tesla 32 GB V100
GPU: 1 x Volta GV100
Memory: 32 GB HBM2
NVIDIA Tensor Cores: 640
NVIDIA CUDA Cores: 5120
Memory bandwidth: 9000 GB/sec
Monthly
Dual Intel Xeon
E5-4110
Starting at:
USD 519/mo.
Dual Intel Xeon E5-4110 (2.10 GHz)
16 Cores
32 GB RAM
1 x 1 TB SATA
20 TB BW*
Monthly
Dual Intel Xeon
E5-5120
Starting at:
USD 634/mo.
Dual Intel Xeon E5-5120 (2.20 GHz)
28 Cores
32 GB RAM
1 x 1 TB SATA
20 TB BW*
Monthly
Dual Intel Xeon
E5-6140
Starting at:
USD 754/mo.
Dual Intel Xeon E5-6140 v4 (2.30 GHz)
36 Cores
32 GB RAM
1 x 1 TB SATA
20 TB BW*
Spectral MD transforms burn care with advanced imaging and machine learning
Diagnosing the severity of burns is hard; even the most experienced doctors frequently make errors. Harnessing the power of GPUs for cloud servers on IBM Cloud bare metal servers, Spectral MD, Inc. increased the speed of its DeepView Wound Imaging System, enabling more accurate diagnoses that drive better outcomes for patients and lower costs for healthcare providers.
“ Thanks to the computational power of the GPU processors, we can complete in less than one day machine learning tasks that previously would have taken us up to 10 days. ”
— Jeffrey Thatcher, PhD Chief Scientist, Spectral MD
Client success stories
What is a GPU?
Alex Hudak, Offering Manager for IBM Cloud, defines a graphic processing unit (GPU) as "the extra brain power that the CPU can't do on its own." Watch as Alex explains the core basics of GPUs for cloud servers, why you would need them and how they are being used.
Get started
Run your most intense enterprise workloads with GPUs for cloud servers for unparalleled performance. Sign up now and your first USD 200 of applications and servers is complimentary.
Footnote
*20 TB bandwidth included in US, Canada and EU data centers; 5 TB bandwidth included in all other data centers. New prices and offers may not be combined with any other current or future discounts.