News Intel GPUs - Battlemage rumoured cancelled (again)

[NTMBK]

Raja Koduri replied on the Arc dGPU cancelling story:

https://twitter.com/x/status/1569150521038229505

I mean, they want to sell their current stock. They will deny everything and say everything is fine, right up until it isn't.

 

[KompuKare]

I mean, they want to sell their current stock. They will deny everything and say everything is fine, right up until it isn't.

Well, one take could be: It has now been officially denied [again], so it's going to happen then!

With fixed software the cards might be competitive at the rumoured prices but in terms of perf/transistors they are rather poor so since Intel are relying on TSMC having higher costs than their competitors is a major thing. To break into a new market lower margins are a given, so it really depends on whether Intel have the sticking power. I consider them serial quitters, but they might surprise me.

 

[DrMrLordX]

One thing is certain: someone is going to wind up looking stupid once the dust is cleared.

 

[KompuKare]

One thing is certain: someone is going to wind up looking stupid once the dust is cleared.

Well, I wouldn't mind if it was us doubters!
A viable third player would be very welcome.

 

[Heartbreaker]

Well, I wouldn't mind if it was us doubters!
A viable third player would be very welcome.

It's OK to doubt. Because there is certainly some chance they could bail out of the discrete GPU business.

The problem isn't doubt, it's being a gullible believer of someone like MLID that spins phony tales for a living.

It's posting Intel GPUs are over because MLID says so....

MLID is in my Youtube blocklist, so whenever someone posts a MLID link as "proof", I just get a blank screen. I get a lot of blank screens from his followers posting links. And even after his tales turn out to be phony, they rush to post the next one...

 

[PingSpike]

I mean, they want to sell their current stock. They will deny everything and say everything is fine, right up until it isn't.

I don't really even see a denial there. It was more of a "I'm just asking questions" distraction tweet honestly.

 

[DrMrLordX]

Well, I wouldn't mind if it was us doubters!
A viable third player would be very welcome.

Ditto. It's hard to tell of whom we should be more skeptical: Intel or MLID? Generally I'd prefer that Intel carried on, at least through one more generation. But it's easy to see why they wouldn't.

 

[Stuka87]

Cancelling ARC does not mean getting out of the dedicated GPU market. It could very well mean they move on to the next generation cards (battlemage?) that are already well under way.

 

[eek2121]

One thing is certain: someone is going to wind up looking stupid once the dust is cleared.

MLID and other Youtubers do that every day. No loss there. The number of videos he and others delete to salvage their reputations would astonish you.

I have a pretty keen eye for leaks, and I've seen nothing that would lead me to believe it was cancelled. All of the cancelation leaks appear to originate from 2-3 sources that have no history of leaks of any kind, which further leads me to believe the original leakers are shorting Intel stock and attempting to manipulate the price. I could be wrong, of course, but I've seen it before.

Also, Intel would be rather short-sighted if they cancelled the entire line of products prior to seeing how well they sell.

 

[DrMrLordX]

Cancelling ARC does not mean getting out of the dedicated GPU market. It could very well mean they move on to the next generation cards (battlemage?) that are already well under way.

Arc is their entire consumer dGPU line. Alchemist is the specific product that currently comprises Arc. Unless they pull support for Alchemist early AND rebrand their entire effort, there's no way for them to kill Arc and carry on with Battlemage.

 

[igor_kavinski]

All of the cancelation leaks appear to originate from 2-3 sources that have no history of leaks of any kind, which further leads me to believe the original leakers are shorting Intel stock and attempting to manipulate the price.

Intel would do well to release a blogpost by Raja or Pat, re-affirming their support for ARC's success no matter what and outlining what remains left to do in achieving their target and the approximate timeline.

 

[SteinFG]

Well, one take could be: It has now been officially denied [again], so it's going to happen then!

With fixed software the cards might be competitive at the rumoured prices but in terms of perf/transistors they are rather poor so since Intel are relying on TSMC having higher costs than their competitors is a major thing. To break into a new market lower margins are a given, so it really depends on whether Intel have the sticking power. I consider them serial quitters, but they might surprise me.

Let's say it's 7000 dollars per wafer. with 147x ACM-G10 dies per wafer, this is about 48$ per die.
Even if they can shrink down the chip by ~30%, that's ~200 chips per wafer, putting the raw cost to 35$ per die, just a 13$ difference.
I think you overestimate the effect die area plays.

 

[mikk]

If someone wants to check out the A730M in 45 games:

https://www.notebookcheck.com/Test-Medion-Erazer-Major-X10-Laptop-Premiere-fuer-die-Intel-Arc-A730M.647085.0.html

https://www.notebookcheck.com/Intel-Arc-A730M-Grafikkarte-Benchmarks-und-Spezifikationen.613268.0.html

 

[Aapje]

Intel would do well to release a blogpost by Raja or Pat, re-affirming their support for ARC's success no matter what and outlining what remains left to do in achieving their target and the approximate timeline.

It's never 'no matter what.'

 

[igor_kavinski]

It's never 'no matter what.'

That's what they need because their competitors have a huge head start.

 

[Aapje]

That's what they need because their competitors have a huge head start.

No, what they need is a sellable product. Not even necessarily at a decent price, but right now they made a ton of cards that they cannot seem to sell, which means all the costs, but without the income. That is just not sustainable.

 

[psolord]

Arc is their entire consumer dGPU line. Alchemist is the specific product that currently comprises Arc. Unless they pull support for Alchemist early AND rebrand their entire effort, there's no way for them to kill Arc and carry on with Battlemage.

Maybe someone from Intel, told that idiot m without lid, "they are cancelling alch" and he understood "arc".

 

[igor_kavinski]

right now they made a ton of cards that they cannot seem to sell

I don't see that ton of cards anywhere

Most likely, they have asked the retailers to hold off until they release some "final" driver that gets someone higher up's seal of approval, possibly Raja, or even someone higher up than him.

 

[igor_kavinski]

Also, Intel would be rather short-sighted if they cancelled the entire line of products prior to seeing how well they sell.

I wonder why they didn't come up with the bright idea of creating some sort of campaign with the title "Reserve your ARC. Now or Never!". That should drum up some interest. They should especially do that for the limited editions.

 

[Aapje]

I don't see that ton of cards anywhere

Exactly my point.

They planned to release millions of alchemist cards in 2022, yet the sales are clearly close to zero. The management seemed to think until very late that everything was going to work out fine with the drivers. You have to be very optimistic to think that they don't have at least hundreds of thousands of chips and other card components in a warehouse.

 

[Aapje]

I wonder why they didn't come up with the bright idea of creating some sort of campaign with the title "Reserve your ARC. Now or Never!". That should drum up some interest. They should especially do that for the limited editions.

They already had a minor PR mess up when they gave away Arc cards to people, but weren't able to actually ship them for more than 6 months after people won them. And that was only a handful of people. Imagine hundreds of thou...well, thousands of reservations of people all getting increasingly angry and complaining everywhere.

 

[maddie]

Let's say it's 7000 dollars per wafer. with 147x ACM-G10 dies per wafer, this is about 48$ per die.
Even if they can shrink down the chip by ~30%, that's ~200 chips per wafer, putting the raw cost to 35$ per die, just a 13$ difference.
I think you overestimate the effect die area plays.

I think its important to use a multiplier. If the GPU costs $X. then the final card MSRP approximates to (N)* $X. Using the raw $13 difference is unrealistic.

 

[igor_kavinski]

You have to be very optimistic to think that they don't have at least hundreds of thousands of chips and other card components in a warehouse.

My only worry is that they don't destroy those cards and recycle them for precious metals etc. That would be a retarded move.

 

[igor_kavinski]

Imagine hundreds of thou...well, thousands of reservations of people all getting increasingly angry and complaining everywhere.

That's why they won't do something like that again unless they know they can deliver. Someone stupidly (maybe Lisa, or Raja speaking on her behalf) gave an optimistic deadline for the driver's readiness which prevented the A500/A700 series cards from being shipped on time.

 

[maddie]

Intel would do well to release a blogpost by Raja or Pat, re-affirming their support for ARC's success no matter what and outlining what remains left to do in achieving their target and the approximate timeline.

This is a good point. Jon Peddie, not an industry novice and not normally considered a YT leaker, made his thinking clear in a recent MLID video.

Stock analysts and fund managers DO read his reports. Intel must know this, and yet are being ambiguous in their GPU statements. No clear message to the markets.

To me, this is clear evidence of an internal debate happening.

Some financial types are no doubt arguing for a stoppage of ARC. Years of division negative cash flow that otherwise could have been used to improve the balance sheet & stock price.

One needs a long term vision to stay with ARC. It will probably lose money for several years. How many Wall Street guys are that way?