Try our cookies Alza.cz a. s., Company identification number 27082440, use cookies and other data to ensure the proper functioning of the website and, with your consent, also, among other things, to personalize advertising and the content of our websites. By clicking on the “I understand“ button, you agree to the use of cookies and the transfer of data regarding the behavior on the website for displaying targeted advertising on social networks and advertising networks on other websites.
I understand Detailed settings Reject everything
Alzak icon
Catalogue
Intel Arc Pro B70 32GB - Graphics Card

Reviews Intel Arc Pro B70 32GB

Interested in reviews of Intel Arc Pro B70 32GB? It got 5.0 out of 5 stars from our customers. Find specific customer reviews of Intel Arc Pro B70 32GB below. We will appreciate if you also share your experiences with Intel Arc Pro B70 32GB after purchasing.
1,699 €
In stock > 5 pcs
Add to Cart More Information

User Reviews

Ondřej, Bílovice nad Svitavouflag
Rated 04/08/2026, variant Intel Arc Pro B70 32GB
Verified purchase
Since I have the Arc A770 and B580, this isn't exactly new to me. The weaker software support is noticeable in a lot of applications, especially in scientific computing or AI. It's just in a different optimisation space, mainly due to the smaller GRF, so software that is highly adapted for AMD or Nvidia often won't run as well on Intel Arc. For llama.cpp, the SYCL backend is slower than the Vulkan backend. It's also worth looking at pull request #24408, which brings a much more optimal implementation of GEMM and FA for Intel Arc on Vulkan. OpenVINO, on the other hand, isn't fully polished. For AI, it might be suitable for training your own models, etc., but probably not so much for pure LLM inference. Some architectures might benefit from the relatively high computing power, but a lot of algorithms will simply be limited by the GRF size. Overall, we'll have to wait a while longer to see its full potential realised.
32GB VRAM
Runs smaller models fairly comfortably at full precision.
The multimedia/codec support is unrivalled. It supports a whole range of professional profiles, but honestly, buy an Arc B580 if you want an accelerator for video (de/trans)coding. The support is the same.
Support for matrix/tensor operations, but on a smaller 16x8x8 format instead of the 16x16x16 as with Nvidia.
You get to avoid Nvidia drivers on Linux.
A "blower-style" fan that expels hot air out of the case, and the GPU is only two slots high.
Slower RAM
It's an Intel, so it has a different architecture from AMD or Nvidia. It has a smaller GRF, so the algorithms have to be adapted to utilise its full potential, because otherwise the cores are just waiting for data that would otherwise be held in the registers on an AMD or Nvidia.
As has been written before, the main problem is probably the unoptimised software. So there is definitely still potential here.
0 ×Report
Automatic translation
P-DC1-WEB06