Intel Shifting Focus to Integrated GPUs: “Less Need for Discrete Graphics Going Forward”

Tsing

The FPS Review
Staff member
Joined
May 6, 2019
Messages
12,871
Points
113
Intel's future lineup of Arc graphics cards, including "Battlemage" (second gen), "Celestial" (third gen), and "Druid "(fourth gen), may comprise fewer models than what can be found in the current "Alchemist" series.

See full article...
 
as it should be as long as they keep at it. But maybe this mean they wont. Really 3 models is all that is needed. I dont know how many they had though.
 
I lost track but the lower they got the more I thought why as the gap shortened between iGPU and dGPU. I know people want budget cards but you can do a lot with a strong CPU with integrated graphics. Any time I buy a new gaming laptop one of the things I usually test is the iGPU just to see how it performs and the one my current, with a 13900HK, can play basic games fairly decently at 1080p.
 
Kinda sounds like Intel is throwing in the towel already. A 3rd player would be nice for keeping nVidia and AMD in check, but sadly Intel still hasn't really become a 3rd player in the GPU market yet. More like an interesting curiosity.

Uvilla said:
Really 3 models is all that is needed.
Click to expand...
I agree with this too. Just need low-end, middle-ground, and high-end SKUs.
 
Intel just doesn't have the money to do it, and they are firing people left and right, teams are disappearing. It's sad to see, to be honest, they had the potential to be a real discrete GPU player and would have been a good one, they were on the right track.
 
DrezKill said:
ree with this too. Just need low-end, middle-ground, and high-end SKUs.
Click to expand...
As long as this does not result in wasted die's it's fine, the abundance of models is usually due to silicon defects and lack of production capacities
 
Brent_Justice said:
Intel just doesn't have the money to do it, and they are firing people left and right, teams are disappearing. It's sad to see, to be honest, they had the potential to be a real discrete GPU player and would have been a good one, they were on the right track.
Click to expand...
Unfortunately a lot of Intel's decisions haven't paid off and throwing money at that market segment just really isn't in the cards for them right now.
 
I was concerned this would happen.

Intel only jumped in because of the insane profit margins during the crypto boom, and now that those are gone, they are out.

Which is sad. We really need 3-5 active competitors in the industry for it to function right.

We almost had three for the first time in over 20 years, then AMD announced they were no longer going to focus on the high end, Nvidia announced they are no longer a graphics company, and Intel does this...

The industry is really ****ed and completely broken when something for which there is so much demand just isn't worth it to these companies.
 
Zarathustra said:
Intel only jumped in because of the insane profit margins during the crypto boom, and now that those are gone, they are out.
Click to expand...
Need to send their MBAs to anti-ADD school, like most short-sighted companies
 
LazyGamer said:
Need to send their MBAs to anti-ADD school, like most short-sighted companies
Click to expand...

I don't think it was necessarily the WRONG thing to at least try, given the margins GPU's were fetching there for a while.

And many were preaching how crypto was the future. They probably expected the boom to stick around longer.

The justification was probably based on the fact that if things didn't work out they could always just scale it down to IGP levels and use it to better compete against AMD's increasingly performant APU's and it wouldn't be a complete waste.

Though I suspect that was held in the reserve in case they really went up shirts Creek, but here we are with Intel in a precarious situation. If they weren't in as much trouble as they are right now, I bet they would continue to market Arc discrete graphics products, but you know, they can't sell discrete graphics cards if they aren't there to sell them because they ran out of money.

So they need to right the ship first.

Maybe after that, they will come back to discrete GPU'e. Who knows.

If I fault the Intel Old Guard for anything, it is probably a bit of hubris, partly from a "We are a Intel, we can do anything" perspective and partly from a perspective where they never expected to find themselves in this situation.

Extreme times call for extreme measures. Intel has had some troubles since early 10nm woes (which we didn't know of yet)forced them off of the Tick-Tock model in 2016, but I don't think they ever expected to find themselves in the situation they are now, despite these difficulties.

They have always had an air of "we are the market" and inevitability about them, and that's how giants fall.
 
Last edited:
Zarathustra said:
If I fault the Intel Old Guard for anything, it is probably a bit of hubris, partly from a "We are a Intel, we can do anything" perspective and partly from a perspective where they never expected to find themselves in this situation.

They have always had an air of "we are the market" and inevitability about them, and that's how giants fall.
Click to expand...
IBM - Hold my beer!
BlackBerry - Hold my beer!
Any number of 'too big to fail!' 🤣
 
I suppose they're in not great shape as a company - but they're not even close to out of the game CPU-wise. Arrow-Lake has some game-killing issues that may or may not be resolved, yet Intel's cores are as stout as ever; Intel has what it takes to stay in the winning position, market-share wise, IMO.

But they can't have started down the GPU road without understanding that marketshare is fed by long-term investment and commitment. Just getting game devs to try to cater to their GPU architecture and simultaneously getting their own driver effort up to snuff was always going to be a challenge, we all knew this when they decided to jump in, right?

And honestly they're either on the cusp of something good, or they're so far off the mark with their next release (Battlemage IIRC) that they really simply cannot get the work done.
 
LazyGamer said:
I suppose they're in not great shape as a company - but they're not even close to out of the game CPU-wise. Arrow-Lake has some game-killing issues that may or may not be resolved, yet Intel's cores are as stout as ever; Intel has what it takes to stay in the winning position, market-share wise, IMO.
Click to expand...

I don't think they are out of the game, but you do realize that the last CPU family they launched (at least in the consumer space) that was problem free was 12th Gen Alder Lake, and that was 3 years ago.

They were so desperate to take the performance crown with Raptor Lake that they overvolted the CPU's too much with insufficient validation, and paid the price, and now Arrow Lake is both a disappointment performance-wise, and has some problems.

And we also learned that they were so desperate to maintain their lead that even back when they had very poor competition from AMD they did stupid unsafe stuff with SMT that resulted in Spectre/Meltdown.

I think they have been teetering on the edge and coasting on their reputation for longer than we realized, and the 10nm issue just set them back enough that the cracks started showing.

They seem to have a decent architecture when it comes to Integer/FPU, but by their own admission they are so far behind that they will never catch up on NPU/TPU (if that indeed does wind up becoming important for the consumer, which is still unclear) and they ceded the mobile market long ago when they their late to the party Atom based smartphone SoC's were canceled in 2016 (which in retrospect seems to coincide with the end of the Tik-Tok model when the realities of their 10nm problems started to set in)

They are in serious catch-up mode when it comes to silicon fab manufacturing, and if they can do so remains to be seen.

They could turn into another fabless semiconductor manufacturer relying on TSMC, and could probably do so successfully, if they can get over whatever issues are holding Arrow Lake back (possibly just an issue of Intel's design being optimized for their own manufacturing process, not for TSMC's, and them still building experience when it comes to manufacturing at TSMC) who knows.

All of this will take time and money though, and time and money are both in short supply.

It's not that their IP suddenly does not have value anymore. It's that they might very well run out of money before they can capitalize on it in a big way again.

In a way, they are in better shape than AMD was for in the past. The money and time issue is very similar, but AMD also didn't have a good modern architecture design until they pulled a rabbit out of a hat with the development of Zen which saved the company just in time.

Intel can pull it off, but they have to shed their sense of inevitability and comfort and adopt a sense of urgency if they are going to do it. I think AMD had an advantage there. They were always the scrappy underdog with a need to fight to get ahead. Intel has rested on their laurels for so long now that creating the kind of cultural change they need for this to happen is probably going to be very very difficult.

They need to exit "Ivory Tower" mode and enter "High Paced Startup/Underdog" mode and those kinds of attitudes don't easily change overnight.
 
Zarathustra said:
I don't think it was necessarily the WRONG thing to at least try, given the margins GPU's were fetching there for a while.

And many were preaching how crypto was the future. They probably expected the boom to stick around longer.

The justification was probably based on the fact that if things didn't work out they could always just scale it down to IGP levels and use it to better compete against AMD's increasingly performance APU's and it wouldn't be a complete waste.
Click to expand...

There is enough money atm in AI for GPU's that it would be silly to stop on that now, less SKU's does not mean none, if they can launch a couple of SKU's at a decent price/perf level while also working on more indutry focused progress there should be a chunk of the market available to them.

AMD is also making good money with their instinct line of cards and they also struggle to compete in the discrete GPU market and also did sometimes with less SKU's like their 5000 line yet they are still around.

.From what I seen, intel's drivers are pretty good, their Ray tracing performance in up there and if they don't do anything stupid Battlemage could be a small win.
 
Zarathustra said:
I think they have been teetering on the edge and coasting on their reputation for longer than we realized, and the 10nm issue just set them back enough that the cracks started showing.
Click to expand...
They were stuck on 14nm three or four years longer than planned and, from what I can tell, haven't been able to actually manufacture their CPUs to the level of their design capabilities since. This has also led to a volume bottleneck or so it seems.

Denpepe said:
There is enough money atm in AI for GPU's that it would be silly to stop on that now, less SKU's does not mean none, if they can launch a couple of SKU's at a decent price/perf level while also working on more indutry focused progress there should be a chunk of the market available to them.
Click to expand...
I think they're at the point where they can pull an AMD (or ATi) and just ship something, as long as there's some performance to be had and there's volume, whether that be aimed at productivity or AI/ML or gaming.
 
You must log in or register to reply here.
Become a Patron!
Back
Top