Grifo: the most powerful and versatile supercomputer in the world


Pushing the limits of AI


The MACHINE for the age of elastic computing has arrived

The most powerful supercomputer in the world

Organizations of all kinds are incorporating AI into their research, development, product, and business processes. However, traditional compute infrastructures aren’t suitable for AI due to slow CPU architectures and varying system requirements for different workloads and project phases. We help organizations to overcome these limits and to succeed in a world that desperately needs the power of AI to solve big challenges.

We have designed the world’s first family of systems purpose-built for AI: GRIFO systems. 

Computing cores
Accelerators directly connected
PetaFlops AI-Rack

GRIFO is cluster-scale compute in a single system, single OS, and even single memory that holds up to 90 accelerators directly connected using a global shared memory to the CPU host.

More than 622’000 of computing cores in a single system. More than 5x faster than any system in the world.

And if you need more... you can combine multiple GRIFO together!

A new era: more than 5x faster
than any existing system.


(Ancient Greek: γρύψ, grūps; Classical Latin: grȳps or grȳpus) is a legendary creature with the body, tail, and back legs of a lion; the head and wings of an eagle; and sometimes an eagle's talons as its front feet. Because the lion was traditionally considered the king of the beasts, and the eagle the king of the birds, by the Middle Ages, the GRIFO was an especially powerful and majestic creature.

Why GRIFO is the perfect solution

GRIFO integrates many different disruptive technologies that span from performance optimized CPU node to record breaking interconnect technology, ultra low latency topology and integrated ultra efficient cooling systems designed to provide an optimal solution for hundreds of deeply coupled accelerators (GPUs and/or FPGAs). 

Exceptional Compute Power

Exceptional Compute Power

Elastic Network Fabric

Revolutionary Compassable Elastic Network Fabric

AI & Big Data Scale

AI & Big Data Scale on en embarassing New Level

Forward Compatibility

Forward Compatibility

All these features make GRIFO the most versatile system in the market:

GRIFO is able to host any king of accelerator including GPUs, FPGA, ASIC in any combination (up to 90) in a single compute node.

GRIFO supports any kind of virtualization and containerization without using special software (it’s a single monolithic server with multiple accelerators).

GRIFO can run any application without any code changes (CDA, Open CL, ...).

GRIFO is fully compliant with any GPUs including all the software stack available for NVidia.

Single massively accelerated supercomputing system

In a single rack units, using max system power of 29kW, the GRIFO packs the performance of a giant room full of servers (more than 10.000) resulting the single computer most powerful available now in the world.

With cluster-scale compute available in a single system, single OS, and even single memory, you can push your researches where you never imagine, exploring new frontiers, creating better products, or simulating what today you can only imagine at a fraction of the cost. Finding the insights hidden in oceans of data can transform entire industries, from personalized cancer therapy to helping virtual personal assistants converse naturally, predicting the next big hurricane or the next pandemic…

Cooling system


GRIFO enables users to gain key insights from massive amounts of data that were previously unmanageable. Given the critical research these systems are tasked to perform, these high-density clusters are expected to run at 100% utilization for sustained periods, making cooling performance critical.


GRIFO Direct Liquid Cooling (DLC) uses the exceptional thermal conductivity of liquid to provide dense, concentrated cooling to targeted areas.

The system is designed to be installed in traditional air-cooled data centers.

Reduction in fan speed
Total server heat load captured into liquid

Model Vector 10090s


Model Vector A10090


10090s VS A10090

  Model Vector 10090s Model Vector A10090 Model Vector A10090 sparsity
Vector Cores 460,8 622,08  
Tensor Cores 57,6 38,88  
GPU Memory 2,88 TBytes 3,6 TBytes  
GPU Main Bandwidth TB/s 102 TB/s 144 TB/s  
FP16   2,826 TFlops 7,02 TFlops N/A
TC AI 11.7 PFlops 28.0 PFlops 56.1 PFlops
BF16 TC (*) 11,7  TFlops 28.0 TFlops 56,16 TFlops
FP32 1,476  TFlops 1,755 TFlops N/A
TF32 TC  1,476  TFlops 14,04 TFlops 28,08 TFlops
FP64 738  TFlops 873 TFlops N/A
FP64 TC  738  TFlops 1,755 TFlops N/A
INT8 TC  5,85 TOps 56,16 TOps 112,32 TFlops


(*) This format is a truncated version of the 32-bit IEEE 754 single-precision floating-point format used to accelerate machine learning


a3cube Respect for your privacy is our priority

We use cookies to ensure that we give you the best experience on our website. Accept and continue to consent to the use of all cookies. If you want to learn more or give consent only to certain uses click here. You can consult our updated Privacy Policy and Cookie Policy at any time.