202 Compute

Installed accelerator fleet

The 202
Compute Network.

A transparent view of the physical GPU systems operating inside our distributed neocloud. The current fleet combines three dual-GPU pods under one managed operating layer.

Rent network capacity Add a system
202 / MANAGED POD OPERATIONAL
Front view of an enclosed dual-GPU pod operated by 202 Compute
3dual-GPU pods6 × RTX 5090192 GB fleet VRAM
THEORETICAL FP32628.8TFLOPS
AI ACCELERATION20,112AI TOPS
CUDA CORES130,560across fleet
GPU MEMORY192GB GDDR7
GPU POWER ENVELOPE3.45kW TGP

01 / Operational systems

Three pods. One managed fleet.

Each dual-GPU pod is independently serviceable and can be allocated according to workload fit. GPU memory remains local to each GPU and is not represented as one unified pool.

1 / 3Drag or scroll horizontally to rotate the fleet

POD 01

Dual-RTX 5090 compute pod

OPERATIONAL
2installed GPUs

Multi-GPU capacity or two parallel 32 GB workers.

FP32
209.6 TFLOPS
AI
6,704 TOPS
VRAM
2 × 32 GB
CUDA
43,520 cores
GPU TGP
1,150 W
POD 02

Dual-RTX 5090 compute pod

OPERATIONAL
2installed GPUs

Multi-GPU capacity or two parallel 32 GB workers.

FP32
209.6 TFLOPS
AI
6,704 TOPS
VRAM
2 × 32 GB
CUDA
43,520 cores
GPU TGP
1,150 W
POD 03

Dual-RTX 5090 compute pod

OPERATIONAL
2installed GPUs

Multi-GPU capacity or two parallel 32 GB workers.

FP32
209.6 TFLOPS
AI
6,704 TOPS
VRAM
2 × 32 GB
CUDA
43,520 cores
GPU TGP
1,150 W

02 / Capacity profile

Installed capability, pod by pod.

Each chart uses the six-GPU fleet total as its full scale. Every pod contributes two RTX 5090 GPUs, or one-third of the installed network capacity.

THEORETICAL FP32

Peak shader throughput

628.8 TFLOPS
0314.4628.8 TFLOPS
Pod 01
209.6
Pod 02
209.6
Pod 03
209.6
Fleet
628.8
AI ACCELERATION

Published AI processing rate

20,112 AI TOPS
010,05620,112 TOPS
Pod 01
6,704
Pod 02
6,704
Pod 03
6,704
Fleet
20,112
GPU MEMORY

Installed GDDR7 capacity

192 GB
096192 GB
Pod 01
64 GB
Pod 02
64 GB
Pod 03
64 GB
Fleet
192 GB
HOW TO READ THESE NUMBERS

FP32 TFLOPS estimate peak non-Tensor shader throughput. AI TOPS are NVIDIA's published theoretical AI figure and use a different operation and precision basis; the two values are not additive. Figures are installed-hardware maxima, not guaranteed workload performance, utilization, uptime, or immediately available rental capacity. Actual results depend on software, precision, thermals, power, and workload design.

Review NVIDIA RTX 5090 specifications ↗

ACCESS THE NETWORK

Put the fleet to work.

Tell us your workload, GPU count, software stack, utilization pattern, storage needs, and preferred start date.

Request network capacity