Ai inference vision controller · AI Inference Edge Appliance
Hikrobot MV-VAC
- Warranty
- 36 months
Inference workloads and variants
Conventional vision, edge-learning CPU processing and the majority of deep-learning algorithms define the workload scope of MV-VAC. The appliance is offered around two Core processor configurations: an i5-8500 with 6 cores and 6 threads, or an i5-12400 with 6 cores and 12 threads. Both pair general processing with a 1070D AI inference accelerator architecture.
Use the controller where one edge system must execute conventional algorithms alongside learned inference. The i5-8500 configuration accepts two DDR4 S-DIMM slots at up to 16 GB each for 32 GB total and uses one 1070D card. The i5-12400 raises each slot to 32 GB for 64 GB total and is associated with two 1070D cards, providing the stated platform split for heavier concurrent workloads.
Selection is between the two listed MV-VAC processor configurations. Choose the i5-8500, 6-core, 6-thread path when its 32 GB memory ceiling and single accelerator match the application. Move to the i5-12400, 6-core, 12-thread path when the configuration needs up to 64 GB and two accelerators. Each 1070D card carries 8 GB GDDR4X and is rated at 70 TOPS for INT8/UINT8.
Overview
A vision controller with a hardware neural accelerator: a 1070D board rated at 70 TOPS sits alongside a Core i5 or i7 processor. It is needed where the solution is built on deep learning — defect detection on complex surfaces, classification, in-line recognition — and the CPU alone no longer holds the required cycle time.
‼️ The vendor publishes no separate datasheet for these six configurations. The list below comes from the 2026Q3 price list; we request the full specification from the vendor for a specific project.
Available versions
One device, different revisions. We ship the current one by default; if a specific revision is already installed on your line, name it in the request and we will match it.
| Part number | PN | CPU | Storage | AI accelerator | |
|---|---|---|---|---|---|
| MV-VAC3504H-128G01 | 312900015 | Intel® Core™ i5-8500, 6 cores, 6 threads | 128 GB M.2 2280 SATA SSD, and 2 TB 2.5-inch 7200 RPM Enterprise HDD | 1070D, 70 TOPS @ INT8/UINT8 | |
| MV-VAC3504X-128G01 | 312900013 | Intel® Core™ i5-12400, 6 cores, 12 threads | 128 GB M.2 2280 SATA SSD, and 2 TB 2.5-inch 7200 RPM Enterprise HDD | 1070D, 70 TOPS @ INT8/UINT8 | |
| MV-VAC3704H-128G01 | 312900014 | Intel® Core™ i7-8700, 6 cores, 12 threads | 128 GB M.2 2280 SATA SSD, and 2 TB 2.5-inch 7200 RPM Enterprise HDD | 1070D, 70 TOPS @ INT8/UINT8 | |
| MV-VAC3704X-128G01 | 312900012 | Intel® Core™ i7-12700, 12 cores, 20 threads | 128 GB M.2 2280 SATA SSD, and 2 TB 2.5-inch 7200 RPM Enterprise HDD | 1070D, 70 TOPS @ INT8/UINT8 | |
| MV-VAC5524-256G02 | 312900016 | Intel® Core™ i5-12400, 6 cores, 12 threads | 256 GB M.2 2280 SATA SSD, and 2 TB 3.5-inch 7200 RPM Enterprise HDD | 1070D × 2, 70 TOPS @ INT8/UINT8 | |
| MV-VAC5724-256G02 | 312900017 | Intel® Core™ i7-12700, 12 cores, 20 threads | 256 GB M.2 2280 SATA SSD, and 2 TB 3.5-inch 7200 RPM Enterprise HDD | 1070D × 2, 70 TOPS @ INT8/UINT8 |
Specifications
| Certification | CE-EMC |
| Algorithms | Supports all conventional algorithms, all edge learning CPU algorithms, and the majority of deep learning algorithms |
| Memory slots | Supports DDR4 S-DIMM memory slot × 2 (Max. memory of each slot is 16 GB, total max. memory is 32 GB.) (Core™ i5-8500, 6 cores, 6 threads) / Supports DDR4 S-DIMM memory slot × 2 (Max. memory of each slot is 32 GB, total max. memory is 64 GB.) (Core™ i5-12400, 6 cores, 12 threads) |
| Ai accelerator | 1070D card for AI inference edge appliance; Graphics memory size: 8 GB GDDR4X; Graphics memory frequency: 3200 MHz; Graphics memory bandwidth: 102.4 GB/s; Power: 25 W; Compute capability: 70 TOPS @ INT8/UINT8 (Tensor Core); 35 TFLOPS @ BF16/FP16 (Tensor Core);; 18 TFLOPS @ TF32 (Tensor Core);; 9 TFLOPS @ FP32; Graphics memory interface: PCIe Gen4 X4 interface (16 GB/s) (Core™ i5-8500, 6 cores, 6 threads) / 1070D card for AI inference edge appliance × 2; Graphics memory size: 8 GB GDDR4X × 2; Graphics memory frequency: 3200 MHz; Graphics memory bandwidth: 102.4 GB/s; Power: 25 W; Compute capability: 70 TOPS @ INT8/UINT8 (Tensor Core);; 35 TFLOPS @ BF16/FP16 (Tensor Core);; 18 TFLOPS @ TF32 (Tensor Core);; 9 TFLOPS @ FP32; Graphics memory interface: PCIe Gen4 X4 interface (16 GB/s) (Core™ i5-12400, 6 cores, 12 threads) |
Selection and applications
Conventional vision algorithm execution. Edge-learning CPU inference. Deep-learning inference with 1070D acceleration.
Minimum kit: The minimum compute kit is MV-VAC in the selected Core configuration, DDR4 S-DIMM memory within its stated ceiling and the corresponding 1070D accelerator-card arrangement.
FAQ
Which algorithm classes are supported by the MV-VAC appliance?
MV-VAC supports all conventional algorithms, all edge-learning CPU algorithms and the majority of deep-learning algorithms. This scope allows the controller to host mixed conventional and learned inference workloads.
How do the two listed Core processor configurations differ?
The Core i5-8500 configuration has 6 cores and 6 threads. The Core i5-12400 configuration also has 6 cores but provides 12 threads, and it is paired with the larger memory and accelerator arrangement.
What is the maximum DDR4 memory in each MV-VAC variant?
Both variants provide two DDR4 S-DIMM slots. The i5-8500 path allows 16 GB per slot and 32 GB total; the i5-12400 path allows 32 GB per slot and 64 GB total.
How many 1070D inference cards are assigned to each platform?
The i5-8500 configuration uses one 1070D card. The i5-12400 configuration uses two, each with 8 GB GDDR4X, a 3200 MHz memory frequency and 102.4 GB/s memory bandwidth.
What compute formats and ratings does the 1070D provide?
The 1070D is rated at 70 TOPS for INT8/UINT8, 35 TFLOPS for BF16/FP16, 18 TFLOPS for TF32 and 9 TFLOPS for FP32. Its connection is PCIe Gen4 X4 at 16 GB/s.
What power value applies to the installed AI accelerator?
The 1070D accelerator card is specified at 25 W. The i5-8500 configuration contains one card, while the i5-12400 configuration contains two cards with the same per-card compute ratings.