510.226.7366 | sales@exxactcorp.com |
Loading

Workstation Graphics and HPC Solutions  


Exxact Corp - Professional Graphics, Consumer Graphics, GPU Computing, Visual Computing, Other Products
EXXACT Peer-to-Peer GPU Solution
Not having full peer-to-peer (P2P) capability between all of the accelerators in a system can be a major setback for those running applications that rely heavily on performance and efficiency in communication between accelerators. Fortunately, Exxact's Peer-to-Peer GPU Solutions are designed to configure up to 10 GPU accelerators on a single PCIe root hub. Many current GPU-centered configurations only provide maximum performance between pairs of GPUs in a multi-socketed system. When this happens, applications requiring communication between arbitrary GPUs show signs of a performance delay. Our technology allows messages to be directly exchanged between each GPU, this allows for maximized performance for applications with heavy GPU P2P communications. Paired with NVIDIA Tesla K80 or the new Tesla M40, our technology incorporates maximum peering and scaling designed to provide the fastest results for fields such as Deep Learning.
Exxact P2P Solution Highlights:
• Full P2P communication between 4x, 8x, or 10x PCIe devices
• 4x, 8x, or 10x NVIDIA Tesla M40/M60 or NVIDIA GTX Titan X or GTX 1070/1080 Pascal GPUs
• NVIDIA-Qualified Driver & NVIDIA® CUDA® Toolkit 7.5
• Pre-installed standard Ubuntu 14 w/ Caffe, Torch, Theano, BIDMach, cuDNN v3, OpenCV v3.0,
  and CUDA Toolkit 7.5
Model GPU/CPU Key Features  
Spectrum TXR410-0032R

NVIDIA GeForce GTX GPU


NVIDIA Tesla GPU


Intel Xeon Processor

• 4x NVIDIA GeForce GTX Titan X or GTX 1070/1080 Pascal or Tesla M40 GPUs
• 32GB DDR4 Memory (Up to 512GB)
• 1x Intel® Core™ i7-5960X Processor
• Optional Google TensorFlow Software Library
• Tower / 4U Rackmount Convertible
Learn More
Spectrum TXN003-128N

NVIDIA GeForce GTX GPU


Intel Xeon Processor

• 4x NVIDIA GeForce GTX Titan X GPUs or GTX 1070/1080 Pascal GPUs
• 64GB DDR4 Memory (Up to 128GB)
• 1x Intel® Core™ i7-5960X Processor
• Optional Google TensorFlow Software Library
• ATX Cube Workstation
Learn More
Quantum TXR413-1500R

NVIDIA Tesla GPU


Intel Xeon Phi Coprocessor


Intel Xeon Processor

• Supports two Intel Xeon E5-2600 v3/v4 Haswell/Broadwell Processors
• Supports up to 3TB DDR4 System Memory
• Up to 10x Double Wide GPUs on a Single Root Complex (P2P)
• Supports 10x NVIDIA Tesla K80/K40/K10/ P100/P40 GPUs or Intel Xeon Phi Coprocessors
• 24 x 2.5" hot-swappable HDD bays
Learn More
Quantum TXR413-1536R-DP2

NVIDIA Tesla GPU


Intel Xeon Phi Coprocessor


Intel Xeon Processor

• Supports two Intel Xeon processor Scalable family
• Supports up to 3TB DDR4 System Memory
• Up to 10x Double Wide GPUs on a Single Root Complex (P2P)
• Supports 10x NVIDIA Tesla K80/K40/K10/ P100/P40 GPUs or Intel Xeon Phi Coprocessors
• 24 x 2.5" hot-swappable HDD bays
Learn More
 
Exxact P2P Deep Learning DevBox Clusters

Exxact's Turn-Key Deep Learning DevBox clusters are built, designed, and delivered ready to go. The fl¬exibility of the Exxact clustering platform allows users to confi gure the GPU solution to meet their specifi c requirements, and also remain con fident they can upgrade and enhance their system to address the demands of the future.

 

 

 

 

 

 






 



1x Headnode
• 2x Intel Xeon E5-2600 v3/v4 CPUs
• 4U Server, Up to 768GB DDR4
• 24x 3.5" Hot-swap SAS3/SATA3 drive bays




 




6x Quantum TXR550-512R
• 8x NVIDIA Tesla M40 GPUs
• 1x Intel Xeon E5-2600 v3/v4 CPU
• 5.5U Server, Up to 512MB DDR4
• Full P2P communication between PCIe devices



 


 



Ethernet Switch 
• 24-Ports
• 10/100/1000 Mb/s

 
Infiniband Switch 
• 36x QSFP non-blocking ports
• 36x EDR 100 Gb/s
• 7 TB/s aggregate switch throughput
• 130ns switch latency
         
 

 


Intel® Omni-Path Card / Infiniband Card
• FDR / EDR
Single or Dual-Port

   
 
» See More Exxact P2P Deep Learning DevBox Clusters Models
The Exxact GPU Solution Advantage
NVIDIA Elite Solution Provider:
As an Elite Solution Provider with NVIDIA, Exxact Corporation works closely with the NVIDIA team to ensure seamless factory development and support. We pride ourselves on providing value-added service standards unmatched by our competitors.
Scalable to Your Changing Needs:
Exxact offers multiple form factors should your computing needs change down the road, minimizing growing pains due to unpredictability.
In-House Engineering Support and Design:
Exxact GPU Workstations are built by in-house engineers and individually customizable for peak performance tailored to solving your unique and complex computing challenges.
3 Year Warranty:
Each GPU Workstation is engineer built and meticulously tested for precision and performance. We stand behind our products by offering 3-year warranty on our GPU Workstations.
Onsite Support:
Besides remote assistance, Exxact also offers on-site technical support including 8x5 NBD, 24x7 NBD, 24x7x4hr.