US What Intel Arc covers
Intel Arc is the brand Intel uses for its graphics hardware. An Intel Arc GPU can be a discrete card or module with its own graphics memory, or the integrated graphics inside an Intel Core Ultra processor. Intel launched the first discrete Arc GPUs in 2022, its first serious entry into dedicated graphics, and has since extended the name across its processors.
Arc appears in two places. As a discrete GPU, it is a separate card or MXM module with its own memory, used where integrated graphics are not enough. As integrated graphics, it is part of the processor itself and shares system memory. Both use the same drivers and software, so an AI model built with OpenVINO can run on either without being rewritten.
Arc generations and families
Intel names each Arc architecture after a fantasy class, in alphabetical order, and the letter carries into the product names.
-
//
Arc A-series, Alchemist (2022). The first discrete generation, from the entry-level A310 and A380 to the A750 and A770, plus mobile chips and embedded versions with an E suffix.
-
//
Arc B-series, Battlemage (2024). The second generation, starting with the B570 and B580 cards, with faster Xe2 cores and better efficiency.
-
//
Arc Pro. Professional cards with certified drivers and more memory for visualization and AI, from the Arc Pro A-series to the Arc Pro B-series.
-
//
Integrated Arc graphics. Built into Core Ultra processors from Meteor Lake onwards, rising to Arc B-series graphics in Panther Lake.
Discrete and integrated Arc compared
The two kinds of Arc share an architecture and software stack, but differ in memory, power and how much work they can take on.
| Feature | Discrete Arc GPU | Integrated Arc graphics |
|---|---|---|
| Where it sits | Separate PCIe card or MXM module | Inside the Core Ultra processor |
| Memory | Own GDDR6 memory, typically 4 to 24 GB | Shares system RAM |
| Power | Tens of watts upwards | Within the processor’s power budget |
| AI hardware | Xe-cores with XMX matrix engines | Xe-cores; XMX on newer generations |
| Video | Hardware AV1, HEVC and H.264 encode and decode | Same media engines |
| Best for | Heavier inference, many streams, multiple displays | Light inference, displays, video decoding in compact systems |
For the general trade-off between the two, see our guide to integrated vs discrete graphics. In both cases the CPU keeps running the operating system and application, and hands parallel work to the GPU.
What Arc GPUs do in edge systems
Gaming drives most interest in Arc, but its media and AI features make it useful in industrial and edge computers too.
-
//
AI inference. XMX engines speed up the matrix maths in neural networks. Through OpenVINO, the same model can run on the CPU, the Arc GPU or the NPU, and be moved between them.
-
//
Video encoding and decoding. Arc was among the first GPUs to encode AV1 in hardware, alongside HEVC and H.264, which suits recording and streaming many camera feeds.
-
//
Multiple displays. Discrete Arc cards drive several high-resolution screens for control rooms and signage.
-
//
Long-life embedded parts. Embedded Arc versions, such as the A370E, are offered for longer availability than consumer cards.
Choosing between integrated Arc, discrete Arc and an NPU
On a Core Ultra system there are up to three places to run an AI model: the integrated Arc graphics, the NPU and, where fitted, a discrete Arc GPU. The NPU suits small models that run all the time at low power, such as presence detection or speech. Integrated Arc suits moderate vision models on a few camera streams, with the bonus of hardware video decoding on the same chip. A discrete Arc GPU is the step up for larger models, more streams or higher frame rates, at the cost of extra power and a chassis built to cool it.
Because OpenVINO targets all three from one model, a project can start on integrated graphics and move to a discrete card later without changing the application. Our GPU vs NPU guide compares the two accelerator types in more detail.
Intel Arc in our range
Arc appears across our catalog in both discrete and integrated form. Options vary by model, so check the product page or ask us which builds carry which GPU.
-
//
Arc A370E. An MXM option in our Sintrones ABOX-5221 GPU computers, with 8 Xe-cores and 4 GB GDDR6.
-
//
Arc Pro B-series. Arc Pro B50 to B70 cards are among the GPU options in our dual-GPU edge AI computers.
-
//
Arc B390. Integrated graphics in the Core Ultra X7 358H and 368H processors in our NUCs, boards and edge AI computers.
-
//
Arc 140T. Integrated graphics in the Core Ultra 7 255H used in our NUC PCs.
Quick answers on Intel Arc
Intel Arc: common questions
Is Intel Arc a good GPU?
For its class, yes. Arc offers strong media encoding and good value in the entry and mid-range. Driver support has improved steadily since launch, and it works well with OpenVINO for AI.
Is Intel Arc integrated or dedicated?
Both. Arc is the name for Intel’s discrete GPUs and also for the integrated graphics in Core Ultra processors.
Does Intel Arc support CUDA?
No. CUDA is NVIDIA’s platform. Arc uses Intel’s oneAPI and OpenVINO, and supports common frameworks such as PyTorch through Intel’s extensions.
What is Intel Arc graphics in a laptop?
The integrated GPU in a Core Ultra chip. It shares system memory and handles displays, video and light AI work without a separate graphics card.
Can Intel Arc run AI models?
Yes. XMX engines accelerate inference, and OpenVINO deploys models to the Arc GPU, the CPU or the NPU from the same code.
Our industrial GPU computers include builds with Arc options, and our Intel Core Ultra computers carry integrated Arc graphics. Our Intel Core Ultra guide covers the processors in more depth. We fit and configure the GPU, drivers and OpenVINO runtime as standard, so the system is ready to run your models when it arrives.