RUMOURS : Intel Arc (and Optane) cancelled.

[Saracen999]

Just to be clear - stressing the "rumour" part of this,

The cancellation of Optane seems to be nailed on, and not a huge surprise, but Arc? If, and I'll repeat one last time, IF that's true, it is bad news, and not just for Intel. I am not a fan of any pretty expensive market segment (consumer 'gaming' GPUs, in this case) being dominated for years by a duopoly .... or especially what might be argued (* see note) to be a monopoly, and a single "me too" wannabe.

I want to see Arc released and at least put some market pressure on nVidia from the 800lb computing gorilla in the room that is Intel. Yet, those rumours suggest serious hardware issues with Alchemist (Arc first round) that can't be solved with drivers and that while Intel's marketing machine will push HARD on Arc, if for no other reason than to shift existing stock and contracted manufacture, Arc Round 2 (codename Battlemage) has been cancelled.

If true, buying any level of Arc needs to be a carefully considered choice. It MIGHT still be a decent option, at the right price, for some use cases and for now and a couple of years, but it looks like ongoing support, if indeed it is cancelled, might be .... erm, lacking in intensity.

This whole situation is especially pernicious as, whether true or not, it is going to undermine the propensity most buyers will have to buy into Arc. It certainly halts me in my tracks, unless/until it's clearly demonstrated (and not just denied) to be untrue.

Why pernicious? Because if Intel say "it's untrue", will that be enough? Not for me, and probably many others, because I'm thinking "Well, if they have stiock and are committed to manufacturing contracts, they would say that, wouldn't they. Should I chance it anyway? Hell, no".

Also pernicious because, due to the above logic, if I were, for example, a PR company wanting to conduct a product undermining campaign, this sort of "rumour" is a damn near perfect way to do it, if done carefully is virtually untraceable, and Arc is seriously wounded whether this is all true or not.

That's especially the case since :-

- it's pretty detailed and extensive, and
- is from a source not likey to be blithely floating unsubstantiated rubbish, and
- is allegedly based on multiple Intel-internal documents and sources.

My personal conclusion, therefore, is it's either true (whether denied by Intel or not and, so far, comments seem to be carefully phrased to imply it's untrue without categorically stating that in a way that later can be demonstrably seen to have been lies), or the product of a careful, detailed and probably expensive campaign to undermine Arc before it gets off the ground. My inclination is to suspect the rumours are true, and to not consider taking a chance on Arc, either way. In which case, even if it is a nasty tricks campaigned, it has worked on me. I'd bet I'm not the only one going "Woah ...." and stepping back, nervously.

In fairness, personally, if I do my contemplated desktop build at all, I'm probaby going a bit too high-end for Arc to be a consideration anyway, in the time frame I'm looking at. But nevertheless, this is either a very embarrassing and monumental cockup by Intel, or a very nasty dirty tricks campaign, which begs the question "Who could, or would, do that?"

Certainly, if I were one of the Intel guys doing the press rounds in the last week or so and I knew about this (if it's true) I wouldn't have the front to show my face in public for years. If I were still in the press, I wouldn't believe a word they said, either. And if I were one of those guys and somehow didn't know about it, and was sent out to the press like a sacrificial goat tethered to a stake, I'd be hopping mad and spitting bullets.

All things considered, I think I'm going to be sitting back and watching this space carefully for a while. Popcorn, anyone? Toasted marshmellows, perhaps?




* Note - Contrary to common perception, a monopoly is not where only one competitor exists, but rather, where one is so dominant and they can affect market conditions, and prices, by their actions alone. A classic example was IBM up to, oh, the '80s or 90's. They had about 70% of the computer market, and the other 10 or 12 companies shared out the remaining 30% between them, some in pretty narrow niches. I'd suggest nVidia is in a similar dominant role now, given that they can (or could, until recently, and arguably still can) ramp up prices and get away with it, and if they significantly cut prices, much of AMDs market share would vanish.

[Jonj1611]

There was a couple of videos on youtube of arc being benchmarked, not sure if they were true or not but they didn't perform that well. Maybe a return to the drawing board?

To add fuel to the fire there are an awful lot of stories from the usual sources in the last 24 hours saying the same thing.

[kompukare]

Have to say I hadn't considered the conspiracy theory angle part, but for especially part of this:

That's especially the case since :-

- it's pretty detailed and extensive, and
- is from a source not likey to be blithely floating unsubstantiated rubbish, and
- is allegedly based on multiple Intel-internal documents and sources.

Well, this is Intel.
And ever since IBM gifted them the PC market*, Intel have been extremely cash rich, but often poor at executing and very often very poor at attempting to break into new market.

And this has never been due to cash:

1. Itanium - their big multi billion to force the market away from x86 to something where they had no competition.
2. Larrabee - their previous attempt to break into the (compute) GPU market
3. 5G Modems
4. Atom - strangled at birth and constantly crippled to not compete with Core. Part of the idea seems to have been that they needed to do something to keep old fabs churning along, and that something was Atom and chipset (at last there is a logical explanation why Intel chipsets seldom support more than one or two gen)
5. Contra revenue after the panicked that the tablet and phone market was all happening without them.
6. 5G Modems
7. Think there was also an attempt to break into networking at some stage - Intel NICs are established (but their WiFi driver support is overrated, IMO), but this was to grab the big money Cisco and Juniper were making and that part failed.

So that's a long list (and I might have missed some) of very expensive projects which Intel abandoned in each case after spending $billions on it.

And that, IMO, is why it is so easy to start a rumour that Intel are going to abandon ARC.

The possibility that Nvidia (or less likely AMD) started this rumour is interesting but Intel are already their worst enemy in this thing due to their history.

However, since getting super computer and other HPC projects without a GPU is getting almost impossible so the can't really abandon Compute GPU as they need that even without consumer ARC. Although I guess Fujitsu's Fugaku doesn't have any GPUs at all and does the vector stuff in the main CPUS - can't see a bloated x86 design doing that though - and AVX512 after all the hype can almost be added to above list at least at the consumer level since Alder Lake had it removed.


* And IBM's poor choice of going with x86 meant all through the 80s and longer programmers and users in IBM PC market had to deal 64KB memory segments, real and protected mode and all kind of nonsense which if IBM had picked Motorola 68000 with this 32 bit design and registers would never have been the case.

[kompukare]

There was a couple of videos on youtube of arc being benchmarked, not sure if they were true or not but they didn't perform that well. Maybe a return to the drawing board?

To add fuel to the fire there are an awful lot of stories from the usual sources in the last 24 hours saying the same thing.

For A380 there have been more than a few youtubers now as 3 German sites bought their own samples from China and reviewed them. It was pretty poor at best and very very inconsistent - lots of old games don't work at all and most likely never will.
Later TPU also did a review so that makes 4 professional reviews so far:
https://www.computerbase.de/2022-07/...0-review-test/
and a follow-up: https://www.computerbase.de/2022-07/...md-ryzen-test/
https://www.golem.de/news/arc-a380-i...07-166780.html
https://www.igorslab.de/en/intel-arc...sive-teardown/
and another follow-up: https://www.igorslab.de/en/intel-arc...-api-overhead/
(Warning Igor runs his articles through a machine translator.)
https://www.techpowerup.com/review/intel-arc-a380/

[Saracen999]

@kompuKare .... yup, agreed. Read that quickly (wife standing here, it's lunchtime, back later) but, yeah. It's Intel. The whole thing is utterly pausible. A good rumour/sneak attack needs to be to work, though. So I'm not sure that makes me any more or less inclined to believe it. My gut still says "probably true", and my bullpoop radar is beeping "too much smoke for there to be no fire at all".

Or it's a smoke granade.


Oh, and my reference to IBM, by the way, was referring more (much more) to the mainframe/mini market than PC which was, at best, nascent, in the period I was talking about, if not non-existent. Not disagreeing with you, just pointing out it wasn't what I was getting at.

[Apex]

Who actual knows but Intel.... I think you have already pointed at this but what if all this noise was the plan by team red / green, it certainly helps distract people and the fallout from it will mean people won't give ARC a chance.

[Saracen999]

My source, BTW, for this thread was the chat between Linus/Luke on this week's WAN show. Should've said.

[Saracen999]

Who actual knows but Intel.... I think you have already pointed at this but what if all this noise was the plan by team red / green, it certainly helps distract people and the fallout from it will mean people won't give ARC a chance.

Yup. If i was what I was after I might still look at Arc but IF it I was assured that whatever the issues were weren't going to bite me in the wotsit, AND the pricing was good enough to justify not just going red/green. Some of the video encoding stuff on Arc looks kinda promising. But by and large, as you said, many just won't take a chance. As Linus/Luke discuss, the delays there've already been, and the "rumours" also probably mean people will be cautious as hell about next-Gen Arc if/when it comes out, too.

[Apex]

My source, BTW, for this thread was the chat between Linus/Luke on this week's WAN show. Should've said.

Pauls Hardware covered it too...

Something does seem with hindsight to be off, the recent media blitz is NOT like Intel under any CEO and its not Like Pat hasn't worked there before....

[DanceswithUnix]

1. Think there was also an attempt to break into networking at some stage - Intel NICs are established (but their WiFi driver support is overrated, IMO), but this was to grab the big money Cisco and Juniper were making and that part failed.

To be fair the Intel NICs are pretty good, just horribly overpriced. Intel did start off with the DEC network hardware team who were world leading at the time. But otherwise, yeah Intel do have a habit of getting bored of their latest toys. I would throw at least i860 onto that list as another supposedly world beating Intel product that never took off.

But one thing to remember here is that Intel will be shipping integrated graphics for the foreseeable future, which puts an odd slant on the idea that stand alone GPUs might go away. If the current generation sucks though, then that could well make Intel pause their plans. They had a massive mountain to climb scaling from integrated to high end gamer setups, I would have been stunned if they pulled it off so quickly.

I'm mainly stunned at the idea that Optane wasn't profitable when it cost such a stupid amount per GB.

[AGTDenton]

Larrabee 2.0?
Intel have made a mistake in going for a broad spectrum of price points, low mid and high end cards. They should have focused on one area not 6 (if that rumour is true). Back in the 90s they often only had two cards available per manufacturer. Much easier to manage and grow

It is a shame there isn't a 3rd or even 4th player.
Matrox should come back into the business, mobile & low end segment, not everyone wants gaming capabilities. A lot of their tech was way ahead for things like multi monitors, long before ATi & nVidia were doing them.

[Saracen999]

For anyone interested in .... how do I put this .... what appears to be the best-informed (by direct internal Intel and board-partner leaks) and a pretty cogent opinion of what's going on, the whole picture, and how those leaks support it, is Moore's Law Is Dead, on YT.

https://www.youtube.com/watch?v=DH2s5HeZzs8

This is the channel that Linux/Luke (and Linus' body language when covering this was, to me, not happy at all) based their .... segment, on the WAN show on. MLID does a pretty good job of putting it all together. It's about 25 mins-ish, but if interested in trying to work out what's really going on with the allegedly deliberately misleading bullpoop Intel's graphics people are allegedly serving up, even to other Intel divisions and top brass, this is worth a watch. It's part technical, but it looks like a lot is more small-p political, and one inference seems to be (WAN show implied this) designed to delay the possible Arc cancellation and the canning of at least the consumer desktop discrete graphics gaming product line. Maybe a job-extender, as an interim measure. Who knows.

It seems, as per the story, there are hardware issues both with Alchemist and now Battlemage, and Intel's graphics division thought they could resolve them, by now, in drivers. They couldn't, and signs are, may never be able to, so "high end" SKUs with castrated performance are being hyped and repurposed as low to low-mid SKU's, for "later" this year, with maybe a "refresh" with fixed hardware next year. The inference seems to be those cards currently being hyped are hardware already manufactured and sitting in warehouses and if so, they have zero long-term future. Also, come upgrade time, if even half of this stuff is true, I don't fancy the effect on resale value when you want to upgrade from them.

Personally, I'm not planning on upgrading my next PC (assuming I do it at all) so that isnt a worry. But then, I'm not looking at that level anyway, and if I were, the likely lack of ongoing driver support most certainly is an issue for me.

So I can't help but conclude that even for me, Arc would be a risk too far at this point and for most people, it's a worse scenario than it would be for me. Unless I revise my hardware target to this level, AND the pricing on Alchemist (if they ever do actually hit the shelves) is really, really aggressive, and I mean basically firesale-type aggressive, they're a non-starter at this point, largely because of what appears to be, if the stories are true, the gross disparity between the picture we're being given by some in Intel, and the reality.

IMHO, for what it's worth, if the scenario that the graphics division have gone rogue and their very recent marketing drive is as deceitful as alleged, then Intel's CEO (Pat Gelsinger) really needs to bite the bullet and stomp on this, very publicly, or the stench will wash over everything other Intel groups do, too. There's that old adage in polititical malfeasance .... it's not the actual malfeasance that buries you, it's when the cover-up is exposed. That is what people won't forgive. In the extremely unlikely event Gelsinger knows what I think, or even less likely event that he cares a single jot, those are the really shark-infested waters, and where the public backlash feeding frenzy will be. Putting a stop to this, in a timely fashion will look good on him, but ending up tarred with the cover-up brush will not.

All assuming the rumours, stories and even leaks (for which other motivations are possible, if seeming unlikely) are true, of course. So far, we're very rich in such rumours, leaks, opinions and speculation but actual hard facts are pretty sparse.

[Saracen999]

...

It is a shame there isn't a 3rd or even 4th player.
Matrox should come back into the business, mobile & low end segment, not everyone wants gaming capabilities. A lot of their tech was way ahead for things like multi monitors, long before ATi & nVidia were doing them.

Yeah, it is, and I was a Matrox fan for years.

My bet is that a large part of the issue is that while loads of manufacturers was viable in the early days, as the graphics chip technology matured and evolved, the level of difficulty in progressing further went up exponentially and the costs of doing so did too. Even now, at the higher end (where margins are better) it's still a fairly limited market, and if you exclude miners, WAY more so still. That is, the ingredients needed to bake the cake are expensive, and the oven needed is extremely so, so unless there are very large numbers looking to buy very expensive cakes, there's only so much customer demand for cake to go around.

I don't think I can push that analogy any further than that.

[kalniel]

I think it's probably rubbish and this is an attempt at a smear by someone who stands to gain. We'll find out in a few weeks so not worth amplifying it IMHO.

[Saracen999]

I hope so, kal. But unless the guy with all the sources (who, of course, are anonymous or they'd no doubt be unemployed) is either corrupt or a complete moron unfamiliar with the Jornalism 101 notion of always confirming your sources, and double/triple confirming anonymous (though no to him) ones, I don't think there's any chance it's all smoke and mirrors. Among those sources are several AIB partners (and that narrows the list of possibilities a lot) and supposedly widely disseminated internal Intel documents, i.e. stuff that is no doubt company-confifential, but within Intel, the subject of widespread inter-division staff briefings.

Something is going on. Of course, it could be adding 2+2 and coming up with 13.7 recurring. But something's up. I guess more than that, only time will tell.

[kalniel]

Journalism 101 doesn't seem to apply to social media click baiters. This can be an advantage or disadvantage, but we'd be wise to be wary of applying old-school standards.