Intel Arc Pro B70

“Or you could get a 9700 32GB or an Arc B70 for well under $2k, today.”

“Intel B70 with enough VRAM to run Q6 and some context on llama.cpp”

“currently get around 22 t/s which isn't great but isn't terrible either”

“You're better off with a single card, like intel arc pro b70 with 32 gigabytes of VRAM unless you need something external.”

“you can buy an Intel Arc B70 for $1000. There are definitely ways to get going for less.”

“You can buy an Arc Pro B70 with 32 GB of GDDR6 for $1k right now. Intel made exactly the thing you want.”

“it's somewhat slower than the 3090, when using a model small enough to fit entirely in nvram, but can fit models the 3090 can't”

“SYCL ecosystem versions. Absolutely horrendous and everything is hundred commits behind.”

“runs medicore on an Intel Arc B70 ($1000 + cost of a PC plus lots of fiddling”

“The B70 can run both models comfortably side-by-side so it makes better use of time and resources.”

“running qwen3-5-9b-q4-k-m and qwen3-6-27b-q6-k simultaneously on an Intel Arc Pro B70 with a lot of success”

“they're the 32gb cards that are actually affordable and worth getting”

“this is what I've been asking for since the 16GB Arc cards from earlier gens were so cheap”

“proper SR-IOV without hassle. Arc pro cards make for nice graphical acceleration devices for vms”

“At this speed, people end up paying more on electricity than api calls.”

“the more B70 you have, the slower the inference it gets(due to terrible software atm)”

“It lacked software support the for the primary target application, running LLM.”

“they reviewed the card for a ton of workloads nobody in their right mind would pick it for”

“the card fits perfectly here. Performance isn't great but it can just tug along in the background.”

“I'll build out with the Intel stack and keep the cache (and prompt processing speeds) happy.”

“32 GiB of RAM for under $1000, which makes it great for LLM inference”

“It's 32gb for people who can't go for scalped 5090s but have a 3090 budget.”

“Its performance is pretty unbalanced. If you're using it for the couple of things that it's good at, the TDP is competitive.”

“These seem amazing for hobbyist, but that TDP given the perf might be an issue deploying a lot of them”

“I'd much rather have an RTX 5090, an Arc B70 Pro, or an AMD AI PRO R9700”

“The VRAM capacity is definitely where it shines, rather than compute power.”

“Its not a performance monster, but it's quiet and relatively low power”

“The tokens/sec is nowhere near AMD or NVIDIA high end, but its unexpectedly kind of decent to use.”

“Llama.cpp with SYCL has thus far refused to work for me, and Vulkan is pretty slow.”

“For some workloads, the Arc Pro B70 actually does reasonably well when cached.”

“buy a new Intel Arc Pro B70. Use MoE models and offload to RAM for best bang for your buck.”

“Good to see that Intel learned to release product to more than just resellers.”

“I'm probably going to snag one of the Intel cards just for the SR-IOV and use with VM's”

“Good job Intel. Features like sr-iov however are just fantastic”

“They're explicitly not going after the gaming market so they don't need to be on the treadmill of constant driver updates”

“I'd pick this, because I can't just plug a Mac into a slot in my server”

“with those $2k you can have 2xB70, with 1.2Tb/sec and 64G Vram, on linux”

“They certainly look viable as replacements for my Tesla P40 for virtual workloads.”

“Being able to keep infrastructure on Linux is a big advantage.”

“that's a big enough gap to be motivating for a bunch of applications”