Where the fuck is Intel Arc?

Where the fuck is Intel Arc?
I want to try their raytracing on loonix

Attached: 1662179559054.png (473x560, 485.92K)

Other urls found in this thread:

intel.com/content/www/us/en/download/726609/intel-arc-graphics-windows-dch-driver.html
twitter.com/AnonBabble

If they can get close to this with ANV they might have an interesting product for

Attached: 038qADx1zLG18BLkrwfbOmI-3.fit_lim.size_845x.png (845x470, 269.75K)

RRRRRRRRRRRRRRRRRRRRRRRAAAAAAAAAAAAAAAAAAAAAAAAAAAAAPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPEEEEEEEEEEEEEEEEEEEEEEEE

Attached: sexual thoughts.jpg (210x341, 17.11K)

yes, i am the manager.
what do you want?

i already told you, i want my novelty incel gpu

I think they're scrapping it, at least thats what mooreslawisdead and LTT are saying

no haram today boys

Attached: 1629127998029.jpg (1873x1405, 763.57K)

sexo

that guy we hired from amd was all potato, so product ended up being shit.
also we fired too many of our driver developers last layoff and we are having problems getting them back.
we decided to reduce the workload on the few devs we have, so dx11 and down is all done in dx12.
so if older games crash, thats the reason.
everything is in limbo, nobody is listening to me.

I feel bad for Raja, Vega was nice for compute workloads but AMD dropped HBM and kneecapped their own compute efforts (where the fuck is ROCm for consumer RDNA2 cards?).

Really curious about what went wrong with Arc (if anything), and if they can salvage it as a compute accelerator instead. Actually on the topic of accelerators, both AMD and Intel own the major FPGA companies now. Why haven't we seen CPU chips with on-board FPGAs?

>where the fuck is ROCm for consumer RDNA2 cards
it's already here, on fedora just install rocm-opencl from the official repos

Attached: lanadelray.webm (576x1024, 1023.16K)

>Why haven't we seen CPU chips with on-board FPGAs?
Because that's retarded, unless you have a very specific use case. And if you have a very specific use case, you need an ASIC. That's why the AS in ASIC stands for "application specific".

FPGAs are big, expensive and inefficient, but extremely versatile. That's why they're used in the design phase of development for ASICs. CPUs, in that regard, can also be considered ASICs. Accelerators are also ASICs.

Haven't you noticed that most devices that use ASICs are either used for development purposes or, if they're used in final devices, are built in very low quantities? Low enough that they probably don't meet the minimum order at a fab?

>Accelerators are also ASICs
That's kind of the point though. Instead of one ASIC for AV1, one for AV1 decode, one for h.264 encode, one for h.264 decode, one for aes whatever crypto etc, you can just have an FPGA on the CPU and load up whatever software is needed. For example AV1 encode is a waste of die for 99% of users, and for the 1% of users that do need it they're only using it maybe 5% of the time. You could just have one FPGA handle all these weird things that benefit greatly from an ASIC. As a bonus you can upload new algorithms with microcode.

>it's already here, on fedora just install rocm-opencl from the official repos

Only for RX 5xx series cards brah.

Attached: presidentMakron.jpg (827x965, 91.43K)

intel.com/content/www/us/en/download/726609/intel-arc-graphics-windows-dch-driver.html

Intel Arc 31.0.101.3277 graphics driver

i ran stable diffusion on my own rx 6800 with rocm-hip

Attached: sneeed.gif (346x312, 3.45M)

That's dumb. One FPGA will be larger, more inefficient and slower than all of those combined. Just look at FPGA die areas.

but you'd also not be limited to just those functions, any software vendor could ship bitstreams to accelerate whatever software or algorithm they have

Attached: gallery_item_03.jpg (1056x1404, 284.95K)

>i ran stable diffusion on my own rx 6800 with rocm-hip

I was under the impression that rocm was pretty much abandonware. Would be nice if AMD went HAM on it and beat out Nvidia but they're ten years too late on it.

who knows, there's plenty of corporate in not being vendor-locked and chiplet gpus might make amd very interesting in the hpc space, it's definitely good to see that they're improving their software

God, you're dumb. Would you pay double for your laptop CPU that lasts 30 minutes just to be able to decode new stuff when HEVC becomes obsolete in 10 years?

Sounds pretty cool. But I have a green GPU and plan to use pic related when next term comes around.

Attached: ProgrammingMassivelyParallelProcessors_4e.jpg (600x740, 64.17K)