Intel Details More Arc Graphics A-Series Hardware Specifications

[Mike Frett]

You do understand, do you not, that Windows has over 75% market share?

Do you understand why that is? It's called a monopoly. Any threat they see, they seek out and destroy it. It's past time to end this practice from Microsoft.

[qarium]

Yeah. I think it's likely to remain dominant for decades.
But if their monopoly does get broken, WebAssembly might be the tool.

their monopole already goes down right now.

to release a valve steam deck with linux pre-installed would be impossible if the microsoft monopole would be still intact.

also google web based Google Docs to be successfull would be impossible if the microsoft monopole would be still intact.

now we need these adobe products to be repleaced and then we see big movement.

[jaxa]

More details: It's over

https://www.techradar.com/news/intel-arc-gpus-could-be-canceled-already

[qarium]

More details: It's over
https://www.techradar.com/news/intel...nceled-already

i am an enemy of intel but this honestly sounds irrational.
it was clear from the start that it needs multible generations to overcome the hard work need to be successfull.

also the a380 on the feature side had a good start agaist 6400/6500 because of aV1 decode and encode support:::
but it is clear that intel can not compete in highend with 16gb vram and also not in midrange 8-12gb vram...

i don't unterstand why the need of highend and midrange ? they could stay at low-end like the a380 make battlemage a low-end card to
just port it to 5nm maybe go from 6gb vram to 8gb vram increase performance a little and make it a laptop dGPU with some PCIe cards for developers.

but maybe its already over for intel like ryzen7000 will kill of any hope for intel who knows.

but to cancel their gpu ambition after only 2 gpus (arc a380 was the second one) is just insanity.

intel did buy AMD GPU IP for the ARC series but its not clear why they did not license a complete gpu die...

[jaxa]

i am an enemy of intel but this honestly sounds irrational.
it was clear from the start that it needs multible generations to overcome the hard work need to be successfull.

also the a380 on the feature side had a good start agaist 6400/6500 because of aV1 decode and encode support:::
but it is clear that intel can not compete in highend with 16gb vram and also not in midrange 8-12gb vram...

i don't unterstand why the need of highend and midrange ? they could stay at low-end like the a380 make battlemage a low-end card to
just port it to 5nm maybe go from 6gb vram to 8gb vram increase performance a little and make it a laptop dGPU with some PCIe cards for developers.

but maybe its already over for intel like ryzen7000 will kill of any hope for intel who knows.

but to cancel their gpu ambition after only 2 gpus (arc a380 was the second one) is just insanity.

intel did buy AMD GPU IP for the ARC series but its not clear why they did not license a complete gpu die...

https://www.youtube.com/watch?v=cZr_LWAlDkg

Companies will even kill profitable but not profitable enough divisions to focus their efforts elsewhere. Intel's "AXG" division is losing money:

https://www.theregister.com/2022/07/...l_q2_earnings/
https://www.tomshardware.com/news/in...-5-billion-usd

The leak does not suggest that the cancellation will happen immediately, but that consumer discrete after Battlemage is toast (Celestial/Druid), and Battlemage might have a low-end release like you're suggesting, or some discrete laptop chips.

Low-end Intel GPUs do look good on paper. Iris Xe "DG1" was a beta test, but the A380 has a nice price point (for this market), 6 GB of VRAM vs. 4 GB on 6500XT and many older cards in the market, and better decode/encode support. The driver situation has made them into a laughingstock, and apparently the many years of integrated graphics experience/drivers did not translate over to drivers for discrete cards with dedicated VRAM. It still seems possible that the situation turns around and the card ages like FineWine. Or maybe it will age like FineMilk, as Kaby Lake-G did.

One aspect of the disaster is that they did not get the GPUs out in time to benefit from historically high GPU prices. Now they are competing (anemically, only A380 is out and drivers still suck) with GPUs selling under MSRP instead of up to triple MSRP, a flood of used mining GPUs, and incoming RDNA3 and Lovelace.

The A580 (8 GB), A750 (8 GB), and A770 (8/16GB) will still come out, and they could be pretty good at the right price points and with driver fixes.

Intel integrated graphics will continue to improve, and if they want to get my attention, all they have to do is put their biggest iGPU each generation on a desktop SKU that competes with the likes of the Ryzen 7 5700G and future Rembrandt/Phoenix/Strix APUs. That would be good enough performance for many people. This might be easier to do going forward as Intel moves to chiplets/"tiles" with Meteor Lake.

[coder]

maybe it will age like FineMilk, as Kaby Lake-G did.

That was only a tactical move and to cover a market gap. There was never any indication that it marked the start of a trend, or that it would be followed by further products of the sort.

[Michael_S]

their monopole already goes down right now.

to release a valve steam deck with linux pre-installed would be impossible if the microsoft monopole would be still intact.

also google web based Google Docs to be successfull would be impossible if the microsoft monopole would be still intact.

now we need these adobe products to be repleaced and then we see big movement.

Microsoft still owns somewhere around 88% of the consumer desktop operating system market. It's still a monopoly.

Good alternatives to Windows and Office existed 20 years ago, too. Existence doesn't matter by itself, market share does.

And replacing Windows and Office with Android and ChromeOS and Google Docs trades one evil monopoly for another. We need competition outside Microsoft, Google, Apple, and Amazon.

[qarium]

Companies will even kill profitable but not profitable enough divisions to focus their efforts elsewhere. Intel's "AXG" division is losing money:
https://www.theregister.com/2022/07/...l_q2_earnings/
https://www.tomshardware.com/news/in...-5-billion-usd
The leak does not suggest that the cancellation will happen immediately, but that consumer discrete after Battlemage is toast (Celestial/Druid), and Battlemage might have a low-end release like you're suggesting, or some discrete laptop chips.
Low-end Intel GPUs do look good on paper. Iris Xe "DG1" was a beta test, but the A380 has a nice price point (for this market), 6 GB of VRAM vs. 4 GB on 6500XT and many older cards in the market, and better decode/encode support. The driver situation has made them into a laughingstock, and apparently the many years of integrated graphics experience/drivers did not translate over to drivers for discrete cards with dedicated VRAM. It still seems possible that the situation turns around and the card ages like FineWine. Or maybe it will age like FineMilk, as Kaby Lake-G did.
One aspect of the disaster is that they did not get the GPUs out in time to benefit from historically high GPU prices. Now they are competing (anemically, only A380 is out and drivers still suck) with GPUs selling under MSRP instead of up to triple MSRP, a flood of used mining GPUs, and incoming RDNA3 and Lovelace.
The A580 (8 GB), A750 (8 GB), and A770 (8/16GB) will still come out, and they could be pretty good at the right price points and with driver fixes.
Intel integrated graphics will continue to improve, and if they want to get my attention, all they have to do is put their biggest iGPU each generation on a desktop SKU that competes with the likes of the Ryzen 7 5700G and future Rembrandt/Phoenix/Strix APUs. That would be good enough performance for many people. This might be easier to do going forward as Intel moves to chiplets/"tiles" with Meteor Lake.

its impossible to be successfull after only 2 generations DG1/SG2... and it is impossible to be successfull without marketshare ...

they did try to be in low-end and midrange and highend at the same time but they are to small to handle this if they fokus only on 1 chip at a time they would do better. they can not compete in midrange and highend for a long time. but many buy the a380 only because amd 6400/6500 ahs no AV1 decode...

"Now they are competing (anemically, only A380 is out and drivers still suck) with GPUs selling under MSRP instead of up to triple MSRP, a flood of used mining GPUs, and incoming RDNA3 and Lovelace."

all these used mining gpus do not have AV1 decode ... and RDNA3/Lovelace will be in highend market.

it is simple they have to accept that they are 3. class gpu manufacturer for a long time ... but they did try to be 1. class manufacturer what is bullshit.

amd did claim they make opensource linux drivers in 2007 and they did nearly need 10 years do make this happen.
this alone shows that intel need to think in longer periods they need 10 years to become 1. class gpu manucaturers.

[qarium]

The A580 (8 GB), A750 (8 GB), and A770 (8/16GB) will still come out, and they could be pretty good at the right price points and with driver fixes.
.

i found intel Arc a770 benchmarks:

https://gpu.userbenchmark.com/Compare/AMD-RX-Vega-64-vs-Intel-Arc-A770/3933vsm1850973

my Vega64 is 55% faster than the a770

[jaxa]

all these used mining gpus do not have AV1 decode ... and RDNA3/Lovelace will be in highend market.

We care about AV1 decode (A380 has encode too). Most people don't. AV1 decode should be fairly common soon (in Ryzen 7000 CPUs, for example).

People will gravitate towards RDNA2 and Ampere instead of Arc, simply because of the initial bad reviews. AMD and Nvidia will target high-end first with RDNA3 and Lovelace, leaving their current generation at lower tiers, which are more than enough to compete with everything Intel has. Then they will launch mid/low in 2023.

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.