NVIDIA A10 24GB - a universal Ampere accelerator for servers and VDI
The A10 brings together under a single heatsink three types of workloads that are usually spread across different cards: neural network inference, virtual workstations and video processing. It offers 24 GB of GDDR6 with ECC and 600 GB/s of bandwidth on a 384-bit bus - enough to keep mid-size language and recommendation models in memory without falling back to host RAM.
The Ampere architecture provides 9216 CUDA cores, 288 3rd-generation Tensor cores and 72 2nd-generation RT cores. The Tensor units compute TF32, BF16, FP16, INT8 and INT4 with sparsity: 125 TFLOPS on TF32, up to 250 TFLOPS on FP16 and up to 500 TOPS on INT8. Peak FP32 performance is 31.2 TFLOPS.
- Memory: 24 GB GDDR6 ECC, 384-bit bus, 600 GB/s
- Compute: 9216 CUDA, 288 Tensor (3rd gen), 72 RT (2nd gen)
- Interface: PCIe Gen 4.0 x16, 1x 8-pin power, 150 W TGP
- Form factor: single-slot FHFL, passive cooling, no display outputs
- Media engine: 1 encoder + 2 decoders with AV1 decode
- vGPU: vWS, vPC, vApps, vCS profiles from 1 to 24 GB
The passive single-slot 150 W solution is designed for dense deployment: several A10 cards fit into a server with well-planned airflow without losing slots. Full NVIDIA vGPU support allows one card to be divided into profiles for CAD/CAM stations in VDI, and the media engine with two decoders offloads transcoding from the CPU during video analytics.
ETE.UA is the official PNY/NVIDIA supplier in Ukraine. The card comes with a distributor warranty, and an engineer will help calculate the number of A10 cards for your model's VRAM and compatibility with your server platform. Tel. +380 93,594 00 77.
Frequently Asked Questions
What power supply does this card need?
Auxiliary power connector: 1x8pin. With two of these cards in one system, size the supply with headroom and check the circuit rating.
Will this card fit my case?
The card takes 1 slots, length 267 mm. Measure the clearance from the rear panel to the drive cage before ordering: that is where a few millimetres usually go missing.
What model size fits into 24 GB of memory?
Roughly a 13B model at 4-bit, or a 7B model with a long context. The exact figure depends on context length: longer conversations consume more cache. Memory: 24 GB GDDR6.
Which GPU does this card use?
NVIDIA A10, 9216 compute cores. The GPU generation decides which compute formats are accelerated in hardware, and that affects inference speed more than raw memory capacity does.
What warranty is provided and what is in the box?
Warranty: 36 months. Ships in the manufacturer's original packaging. No accessories are included, the card only.
Not sure this card fits your system?
We will check compatibility with your platform, calculate the power draw for your configuration and tell you whether the memory is enough for your models.
The model you need is out of stock? ETE is listed among official PNY partners, so we bring professional NVIDIA cards to order with the manufacturer warranty.
Message us on Telegram or leave a request.
