Statements (52)
| Predicate | Object |
|---|---|
| gptkbp:instance_of |
gptkb:computer
|
| gptkbp:architecture |
gptkb:Cerebras_architecture
|
| gptkbp:competes_with |
gptkb:NVIDIA_GPUs
|
| gptkbp:cooling_system |
liquid cooling
|
| gptkbp:designed_by |
gptkb:Cerebras_Systems
|
| gptkbp:designed_for |
AI workloads
|
| gptkbp:development |
several years
|
| gptkbp:enables |
large-scale AI training
|
| gptkbp:features |
1.2 trillion transistors
400,000 AI cores |
| gptkbp:form_factor |
wafer-scale
|
| gptkbp:has |
dedicated AI accelerators
|
| gptkbp:has_achieved |
low latency
high throughput |
| https://www.w3.org/2000/01/rdf-schema#label |
Cerebras Wafer Scale Engine
|
| gptkbp:includes |
high-speed interconnects
|
| gptkbp:is |
a breakthrough in computing technology
a product of innovation a unique product in the market a scalable architecture a game changer for AI research a high-performance computing solution a leader in AI hardware technology a powerful tool for researchers a significant advancement in AI hardware a single chip solution a solution for complex computations a tool for accelerating machine learning |
| gptkbp:is_designed_for |
AI workloads
|
| gptkbp:is_notable_for |
largest chip ever made
|
| gptkbp:notable_users |
gptkb:Alibaba
gptkb:Microsoft gptkb:Google gptkb:Tesla gptkb:Open_AI |
| gptkbp:performance |
up to 100 petaflops
|
| gptkbp:power_consumption |
20 k W
|
| gptkbp:price |
several million dollars
|
| gptkbp:provides |
scalability
|
| gptkbp:ram |
40 GB on-chip memory
|
| gptkbp:released_in |
gptkb:2020
|
| gptkbp:service |
yes
|
| gptkbp:size |
7 nm process technology
|
| gptkbp:successor |
gptkb:Cerebras_WSE-2
|
| gptkbp:supports |
deep learning frameworks
multi-node configurations |
| gptkbp:target_market |
data centers
|
| gptkbp:technology |
wafer-scale integration
|
| gptkbp:used_for |
gptkb:Artificial_Intelligence
|
| gptkbp:uses |
custom silicon
|
| gptkbp:bfsParent |
gptkb:Cerebras_Software
|
| gptkbp:bfsLayer |
6
|