News Intel GPUs - Battlemage rumoured cancelled (again)

[jpiniero]

That isn't really a sustainable position if your competition is good for everything including gaming.

They are not perfectly fine to compete with larger future iGPU/APU parts, from AMD and potentially more on the Windows-ARM side.

WoA is going to be a bust, and Strix Halo is going to be pretty niche even by AMD standards.

If it took AI hype to realize that they are wasting their time with dGPUs (esp when it seems unlikely they would have the capacity internally to fab them anytime soon)... you can see why Intel is in such trouble.

 

[sdifox]

They didn't have the courage to even do 64-bits. AMD forced their hand.

Not exactly. Intel wanted to kill off x86 and move everyone to Itanium. And Itanium was going to stay proprietary, so they could kill off everyone else.

 

[igor_kavinski]

Thank God for Itanic. It sank as expected.

 

[Aapje]

One the one hand, we got saved from one-vendor hell by AMD licensing x86. On the other hand, the industry might have switched to a more open or fully open ISA if that hadn't happened.

Imagine an alternate history where x86 was dead and Intel, AMD, Qualcomm, Samsung and Nvidia were all fighting over the desktop with an open ISA like RISC-V, or a much more freely available ISA like ARM.

 

[sdifox]

One the one hand, we got saved from one-vendor hell by AMD licensing x86. On the other hand, the industry might have switched to a more open or fully open ISA if that hadn't happened.

Imagine an alternate history where x86 was dead and Intel, AMD, Qualcomm, Samsung and Nvidia were all fighting over the desktop with an open ISA like RISC-V, or a much more freely available ISA like ARM.

I can't believe you said Intel and Open ISA in the same sentence.

 

[moinmoin]

One the one hand, we got saved from one-vendor hell by AMD licensing x86. On the other hand, the industry might have switched to a more open or fully open ISA if that hadn't happened.

Without AMD as a second source 64bit x86 would never have happened and Intel may still be pushing IA-64, likely killing x86 (still using RDRAM) along the way. Not sure what would be the major ISA in that kind of universe, but I'd be afraid that the relatively open PC platform would be replaced by one or several much more closed ones.

 

[DAPUNISHER]

Who knows what would have been different. Perhaps IBM and Apple would have taken way more desktop and notebook market share and the ipad and iphone would not have happened when they did. Or even by Apple. Altering how this century has played out from a microprocessor standpoint. Maybe we would all still be using some form of CrackBerry?

On topic: If Tom/MLID was correct all those many months back about Intel massively scaling back and basically killing off the dGPU division? That's the worst timeline. ☠️ There will be no one to blame but PC gamers though. Every time there is a ARC video the comments are full of well wishing and positive sentiments. Cheering for Intel to bring new competition. But none of them ponied up and bought the cards. We were told we should not treat Chipzilla like a startup, they have all the monies. But here we are. The CPU side was already in trouble and the stock prices taking a beating. When is the last time you read a post anywhere about how Intel has funds to throw at whatever they feel like? That was the rhetoric the first go around when Tom kicked the ant mound.

Most of us here cited how quickly Intel bails on projects. How much they reminded us of the old IBM. Raj and Ryan leaving were described as rats fleeing a sinking ship. We all said nawwww! Now you have to wonder. Is B series performance lagging behind the competition too much? The cost to produce too high to reasonably expect to recoup?

Hopefully we see dGPUs and not just IGP from B and C series. Otherwise it's a bad look.

 

[Aapje]

But none of them ponied up and bought the cards.

This is false. Some did, including on this forum. In fact, you were one of them. Did you forget already?

But I personally know better charities than 50 billion dollar companies and I try to not buy stuff that I won't use anyway. It's not our fault that Intel made such a mess of it.

 

[DAPUNISHER]

This is false. Some did, including on this forum. In fact, you were one of them. Did you forget already?

But I personally know better charities than 50 billion dollar companies and I try to not buy stuff that I won't use anyway. It's not our fault that Intel made such a mess of it.

You have dry sense of humor. Took me a good bit to get familiar with your posting style so that it comes through the way it should.

And of course you are not wrong. But it isn't that simple IMO. If PC gamers really want choice they have to be the change they want to see happen. Platitudes mean Jack crap and Jack left town. Our battle cry has been vote with your wallet! going back to the last century. All talk and no action. It's the sad state of the PC gaming community in the 2020s.

 

[Mahboi]

On topic: If Tom/MLID was correct all those many months back about Intel massively scaling back and basically killing off the dGPU division? That's the worst timeline.

I don't think MLID was wrong. He may shoot in the dark or outright invent a lot of his leaks, but he isn't completely irrational in his analyses.
ARC suffered from all the Intel problems: overly expensive to make, late, inefficient, trying to hold 25 plates all at once and juggle them, all the while not having the basics covered.
You can look at Meteor Lake and find the same things: silicon interposer, tiles, LP cores, capable iGPU, everything and the kitchen sink to make it amazing. But the ring is slow, the LP cores keep being insufficient and have to offload, and the core is just not very good. Meanwhile AMD keeps redrawing the same old I/O die since literally Zen 2 and puts all the cheapest stuff into the uncore, but the cores are so strong that they're still winning.

Basics mastered then advanced stuff. Intel lost the basics. I don't think they'll recover unless some extreme breakage happens.

 

[igor_kavinski]

Good combo deals at the bottom: https://www.newegg.com/p/pl?d=a770

Just wish the 12900KS weren't paired with the awful ACER card coz you need to depend on Acer for driver updates which they aren't doing a good job of: https://community.intel.com/t5/Inte...river-Includes-a-FIRMWARE-update/td-p/1514606

I think I'm gonna go with the LUNA. Hope it doesn't sell out. I think I need to put my money at stake. It might be what's needed for Intel to release something new. I mean, it would be so normal for Intel to release BM just when I get the A770, wouldn't it?

 

[igor_kavinski]

Basics mastered then advanced stuff. Intel lost the basics. I don't think they'll recover unless some extreme breakage happens.

It's corporate apathy at the very top. They (Pat and everyone else calling the shots) know that they will get a nice severance package before Intel sinks. They are just doing the bare minimum to keep the company afloat.

 

[Mahboi]

That's not my opinion of Pat.
But I do think that most of the heads at Intel are just trying to serve their own department over anything else, and when the head has lost control (long before Pat), it's very difficult to get everyone to come clean and work together.
It'll kill Intel eventually.
Zen 2 should've been the big alarm bell ringing.
Zen 5 is the bell falling and crushing them entirely.
If they haven't moved enough then, that's it. Just the years until it peters out and becomes a govt supported shell of what it was.

 

[deasd]

Good news. Author fixed his comment, only mobile discrete Battlemage is dead, desktop part still alive.

https://twitter.com/x/status/1788811592353906728

MLID is correct? NOOOOOOO!

 

[igor_kavinski]

Zen 5 is the bell falling and crushing them entirely.

How can you be sure of that?

 

[Mahboi]

How can you be sure of that?

Is there any room left for doubt?
Intel is literally facing 80% failures/instability in 14900ks.

Even if Zen 5 is "only" 25% better, they're out of the race.

 

[igor_kavinski]

Good news. Author fixed his comment, only mobile discrete Battlemage is dead, desktop part still alive.

A770M 16GB wasn't much of anything anyway. Would've been great if it had launched with excellent drivers.

 

[igor_kavinski]

Even if Zen 5 is "only" 25% better, they're out of the race.

They should've died after Zen 3 too. I think they are just too big to fail.

 

[Mopetar]

The demise of Intel is vastly overstated. They might be taken out behind the woodshed next generation due to a confluence of AMD having a great new product and Intel being unable to juice theirs quite as much, but it's nowhere near as bad as what AMD crawled through during the Bulldozer era when they weren't only behind their competition, but also behind themselves.

Intel had always had commitment problems where they'll throw a lot of money at something hoping to succeed and then retreat from the market when it wasn't immediately as successful as their heyday x86 monopoly days. They'll eventually find their footing again, but as a company they do need to broaden out a bit. GPUs are a part of that.

 

[GodisanAtheist]

Starting to fall into the cult of execution here. Its not about being the best necessarily, its more about not being really bad anywhere and just banging the execution like a precision drummer.

The mindshare and trust consumers will give you is worth all the riches in the world.

 

[Aapje]

Just look at the MSI Claw why mobile Arc makes no sense right now. Perf/watt is not there.

 

[NTMBK]

In the consumer CPU space, Intel's foundries do give them one big advantage still- they can guarantee that they can deliver an ungodly number of chips, on time, to their clients. They don't have the same TSMC bottleneck that AMD (or Qualcomm for that matter) do, they aren't competing for wafers against AI chips from companies with money to burn. I think Intel will continue to ship a metric buttload of laptop chips in the years to come, even if they aren't keeping up with the latest and greatest.

Of course Alchemist and Battlemage are fabbed at TSMC, so Intel threw that out the window

 

[KompuKare]

Good news. Author fixed his comment, only mobile discrete Battlemage is dead, desktop part still alive.

Better than nothing but doesn't fill me with confidence in perf/watt of Battlemage.

 

[Hans Gruber]

Better than nothing but doesn't fill me with confidence in perf/watt of Battlemage.

Arc was on (7nm) N6. Battlemage is on N4 (5nm). N4 alone will increase the efficiency by 20%+. Nvidia will be on the nearly identical silicon with their 50 series cards.

 

[Aapje]

they aren't competing for wafers against AI chips from companies with money to burn.

AI chips are not limited by TSMC's chip producing capacity, but by CoWoS capacity and HBM supply.

This is why Nvidia is letting Intel do the packaging for some of the AI chips.