Intel Details More Arc Graphics A-Series Hardware Specifications

[qarium]

As to your overall point, yes, Intel should stay the course and slowly build up their discrete market share and make driver improvements. But it appears they will quit instead. RIP in peace.

what i do not unterstand is intel struggle to enter the gpu market and the opensource people (Libre-Soc) also try to enter the gpu market

why not team up ? why not try a full opensource design without closed source firmware.

[jaxa]

what i do not unterstand is intel struggle to enter the gpu market and the opensource people (Libre-Soc) also try to enter the gpu market

why not team up ? why not try a full opensource design without closed source firmware.

They are already in the GPU market... by having shipped literally hundreds of millions of iGPUs. The top ones on mobile are reasonably capable too.

I don't understand how they could fumble GPU drivers so badly, given the iGPU experience. But I guess adding VRAM makes it a different ball game.

[qarium]

They are already in the GPU market... by having shipped literally hundreds of millions of iGPUs. The top ones on mobile are reasonably capable too.
I don't understand how they could fumble GPU drivers so badly, given the iGPU experience. But I guess adding VRAM makes it a different ball game.

yes they are in the gpu market like forever but people do not see intel as a gpu company.

gamers almost always get bad experience with their intel iGPU and then they go and buy nvidia and amd...

what i mean is that intel become visible as a gpu company. not just people who do microsoft office and no games buy intel iGPU ...

[jaxa]

yes they are in the gpu market like forever but people do not see intel as a gpu company.

gamers almost always get bad experience with their intel iGPU and then they go and buy nvidia and amd...

what i mean is that intel become visible as a gpu company. not just people who do microsoft office and no games buy intel iGPU ...

To belabor the point: the top 96 execution unit Xe iGPUs in mobile chips can deliver a "competent" gaming experience. With better results on older titles and lower resolutions.

Alder Lake and Raptor Lake desktop CPUs only include up to 32 execution units. But Meteor Lake might increase that to 96 EUs on desktop, 192 EUs on mobile, also with some level of raytracing support. Arrow Lake is expected to go up to 320-384 EUs (mobile).

I'd like to see Intel put their best mobile graphics on a desktop part, or just repackage a mobile die into a socketable desktop part (they already use mobile chips in NUCs). They can use that to compete with AMD desktop APUs (successors to the Ryzen 7 5700G). Chances are that it would be fine for 1080p gaming. One of the many upscaling techniques such as FSR, XeSS, or TSR (Unreal Engine) could help with 1440p/4K.

[coder]

what i do not unterstand is intel struggle to enter the gpu market and the opensource people (Libre-Soc) also try to enter the gpu market

why not team up ? why not try a full opensource design without closed source firmware.

LOL. Because they have completely different aims.

Intel is never going to give away all of its IP, and it's not going to tie itself to the old-ass process nodes that folks like Libre can afford to use in small-batch SoCs.

Meanwhile, the Libre people have their eyes set on "just enough" performance. They also need it to be cheap, small, and simple.

[coder]

I don't understand how they could fumble GPU drivers so badly, given the iGPU experience. But I guess adding VRAM makes it a different ball game.

My other thought is that scaling up the hardware shifts the bottlenecks and might expose the need for more driver optimizations.

To belabor the point: the top 96 execution unit Xe iGPUs in mobile chips can deliver a "competent" gaming experience. With better results on older titles and lower resolutions.

Ah, but maybe increasing the quality settings engages some API features that haven't received much attention from Intel's driver team, over the years.

[qarium]

LOL. Because they have completely different aims.
Intel is never going to give away all of its IP, and it's not going to tie itself to the old-ass process nodes that folks like Libre can afford to use in small-batch SoCs.
Meanwhile, the Libre people have their eyes set on "just enough" performance. They also need it to be cheap, small, and simple.

i know i know... but well you know thats stupid it is really stupid. they better aim to a total defeat und complete faulure than to team up with the libre-soc people.

you see the old thinking of intel is killing them. they did try to go the opensource driver route with closed source firmware...

but for that the market already has AMD gpus... why not try to eliminate the closed source firmware ? many people would to love that.

[arQon]

As you mentioned DX is more important than Vulkan since Microsoft bought out practically all the Vulkan studios which are now using DX.

It never gets talked about, but this is exactly the same playbook they used when they *launched* DX, after 5 made it actually usable. At the next GDC, MS came out as a publishing house with a VERY large checkbook, and were keen to hoover up our stuff (along with dozens of other studios and hundreds of indies) - but only if we switched from GL to DX. They made the same offer / condition to basically everyone there, and in 3 days that was it: they'd moved almost the entire industry to DX forever.

[arQon]

My current employer is a Big Company, where the entire IT organization basically pretends Linux doesn't exist, even though most of our products are now Linux-based embedded devices or run on Linux-based cloud instances.

If I didn't know where you lived, I might think you were at my last employer. I had to figure out how to transform the company's certificates just to be able to use the network at all - so that I could push code for the @#$%ing Linux system that we were developing...