Shop
Hardware catalog
Select a product to review specifications and submit an order request.
18 · All
12- Genuine H100
H100 PCIe 80GB
Genuine Hopper PCIe accelerator for dual-slot servers.
- Architecture
- NVIDIA Hopper (GH100)
- GPU Memory
- 80GB HBM3
- Memory Bandwidth
- Up to 2 TB/s
Order →AED 114,000
AED
- SXM5
H100 SXM5 80GB
SXM5 Hopper module for HGX / DGX-class NVLink fabrics.
- Architecture
- NVIDIA Hopper (GH100)
- GPU Memory
- 80GB HBM3
- Memory Bandwidth
- 3.35 TB/s
Order →AED 128,500
AED
- NVL
H100 NVL 94GB
Dual-GPU PCIe bridge with 94GB HBM3 per GPU for large LLMs.
- Architecture
- NVIDIA Hopper NVL
- GPUs
- 2× H100 Tensor Core
- GPU Memory
- 94GB HBM3 per GPU (188GB total)
Order →AED 191,000
AED
- 8× H100
HGX H100 8×
Full NVLink domain — eight H100 SXM5 GPUs for cluster training.
- GPUs
- 8× NVIDIA H100 SXM5 80GB
- Total GPU Memory
- 640GB HBM3
- Interconnect
- NVLink 4.0 + NVSwitch
Order →AED 1,046,000
AED
- HBM3e
H200 SXM 141GB
Hopper successor with 141GB HBM3e for LLM training and inference.
- Architecture
- NVIDIA Hopper (H200)
- GPU Memory
- 141GB HBM3e
- Memory Bandwidth
- 4.8 TB/s
Order →AED 158,000
AED
- NVL
H200 NVL
Dual H200 PCIe board with 141GB HBM3e per GPU for large models.
- Architecture
- NVIDIA Hopper NVL
- GPUs
- 2× H200 Tensor Core
- GPU Memory
- 141GB HBM3e per GPU (282GB total)
Order →AED 285,000
AED
- Blackwell
B200 SXM 192GB
Blackwell training GPU with 192GB HBM3e for next-gen clusters.
- Architecture
- NVIDIA Blackwell
- GPU Memory
- 192GB HBM3e
- Memory Bandwidth
- 8 TB/s class
Order →AED 245,000
AED
- Ampere
A100 80GB PCIe
Proven Ampere accelerator still widely used for AI training and inference.
- Architecture
- NVIDIA Ampere (GA100)
- GPU Memory
- 80GB HBM2e
- Memory Bandwidth
- 2 TB/s
Order →AED 48,500
AED
- SXM4
A100 80GB SXM4
SXM4 Ampere module for DGX / HGX-class NVLink training nodes.
- Architecture
- NVIDIA Ampere (GA100)
- GPU Memory
- 80GB HBM2e
- Memory Bandwidth
- 2 TB/s
Order →AED 54,500
AED
- Inference
L40S 48GB
Ada Lovelace inference and graphics accelerator for AI servers.
- Architecture
- NVIDIA Ada Lovelace
- GPU Memory
- 48GB GDDR6
- Memory Bandwidth
- 864 GB/s
Order →AED 42,000
AED
- Workstation
RTX 6000 Ada
Workstation Ada GPU for local AI, design, and simulation.
- Architecture
- NVIDIA Ada Lovelace
- GPU Memory
- 48GB GDDR6 ECC
- Memory Bandwidth
- 960 GB/s
Order →AED 38,500
AED
- Unavailable
DGX Spark
Personal Grace Blackwell AI computer — currently unavailable.
- Chip
- GB10 Grace Blackwell Superchip
- GPU
- NVIDIA Blackwell
- CPU
- 20-Core Arm Grace
Out of stockAED 14,650
AED
2- 192GB HBM3
Instinct MI300X
CDNA 3 accelerator with 192GB HBM3 for open ROCm stacks.
- Architecture
- AMD CDNA 3
- GPU Memory
- 192GB HBM3
- Memory Bandwidth
- 5.3 TB/s
Order →AED 72,500
AED
- 256GB HBM3e
Instinct MI325X
Next-gen Instinct with 256GB HBM3e for bigger KV caches.
- Architecture
- AMD CDNA 3 enhanced
- GPU Memory
- 256GB HBM3e
- Memory Bandwidth
- Up to 6 TB/s class
Order →AED 109,500
AED
1
3- Top config
Frontier 4019
Higher-density Blackwell system for larger local workloads.
- Chip
- 2× GB10 Grace Blackwell
- GPU
- NVIDIA Blackwell
- CPU
- 40-Core Arm
Order →AED 153,400
AED
- Pro
Frontier 4018
Compact Blackwell workstation for local inference and fine-tuning.
- Chip
- GB10 Grace Blackwell
- GPU
- NVIDIA Blackwell
- CPU
- 20-Core Arm
Order →AED 76,700
AED
- Best value
Frontier 4010
Entry Blackwell system with 64GB unified memory.
- Chip
- GB10 Grace Blackwell
- GPU
- NVIDIA Blackwell
- CPU
- 10-Core Arm
Order →AED 31,500
AED