Tianshu Zhixin Zhikai MR-V100 32GB
Iluvatar Corex MR-V100 (Zhikai-100) is an inference accelerator optimized for high-throughput, low-latency workloads.
Unlike the BI training series, MR-V100 is specifically designed for multi-tenant environments with advanced resource sharing and isolation features between tasks — critical for cloud AI services with multiple clients.
Delivers 96 TFLOPS FP16 and 192 TOPS INT8. The built-in video decoder supports 128 simultaneous 1080p streams.
Low 150 W TDP enables single-slot configurations.
Widely used for trained AI model deployment, video analytics, intelligent chatbots, and multi-user cloud AI services.
VRAM
32
FP32
24
TDP
150
Interface
PCIe 4.0x16
Pre-order
Product status
Memory Specifications
VRAM
32
VRAM Type
HBM2e
Memory Bandwidth
410
Interface
PCIe 4.0x16
Interconnect Type
Inter-chip Interconnect
Interconnect Speed
200-400
Computing Performance
FP64 Vector-
FP32 Vector24
FP16 Vector96
TF32 Tensor-
FP/BF16 Tensor96
FP8 Tensor-
INT8 Tensor192
Graphics & Thermal
Pixel Rate (GPixels/s)
-
Texture Rate (GTexels/s)
-
TDP
150
Cooling Type
Пассивное
Form Factor
PCIe
Video Encoding
16x1080p30fps
Video Decoding
128x1080p30
Physical Dimensions
Slots
2
Length
- mm
Height
- mm
Width
- mm
Architecture
Architecture
CoreX
Cores
-
Price
On request