Question Intel's new x86 instruction sets: APX and AVX10

[A///]

I'm assuming that this is a literary reference that went over my head?

Possibly but I was two bottles deep into a vintage rioja.

 

[SpudLobby]

It's not like that at all. SVE's point is that it is designed in such a way that you can run the exact same codepath on CPUs with wildly different vector lengths, and other than running a different amount of iterations, get the same result.

AVX10 256/512 are not like that at all, the only convenience is that you can easily query for vector length, everything else is the job of the developer (or, more likely, the compiler). They are deranged if they think most devs are going to emit code that does anything other than supports some minimal common subset.

This is effectively giving up on the 64-byte vector length for client CPUs, while retaining the predication and other useful features of AVX-512. I am mildly disappointed, but really, the wide vectors were the least interesting and important part of AVX-512 anyway. A world with AVX10.2 256-bit is still much, much better than AVX2, and if they now start supporting it across the product line and do so consistently, then after about a decade, consumer stuff can finally start targeting it.



Also look at the liabilities column. They have the most cash on hand and highest assets ever, because they took $11B of long-term debt last quarter, and are sitting on most of the cash. (This is not a criticism of this action, it was actually quite shrewd of them, as their ability to loan money at decent terms might be worse in the future, so might as well take out all the debt now and park it in treasuries until it's needed.)

Also total assets is not a good measure of a healthy company, because a lot of assets are very illiquid and might not be worth nearly as much as their book value if they had to be sold, and in the meantime they are just burning capital through depreciation.

I'm not an Intel doomer. I think there are a lot of smart people there, including people I know personally, and many of them are working really hard to turn the ship around and fix their problems. But it's really deluded to look at Intel's current financials and think that they are a healthy company. They really, really aren't. They have tens of billions of dollars invested in old, depreciating assets, their current product line is nowhere near good enough to regain their old position, and their manufacturing, the old crown jewel of the company, is still either lagging or at least hasn't proven it has caught up to the competition.

Yeah, I also want to note that the odds Intel regains it's old position are almost certainly 0. It will not happen. It cannot structurally happen due to the demand for compute and the revival of the only other (VIA excluded) firm with an X86 license, foundry costs (which Intel previously got around and was able to use vertical integration to their advantage, but fixed costs are too high now), and worse yet, the growth of mobile and slow death of the X86 stronghold on general purpose compute.


Maybe if ET hands them alien technology they can manufacture at scale, or Apple and AMD cave in overnight, TSMC and Samsung and their American facilities are hit by missiles, Arm folds, AND Intel decides IFS isn't for them after that and they want full vertical integration back - then yea, maaaybe we could return to the previous state.


But even that last part is dubious due to the economics. Intel had a very, very unique situation with its CPU design teams, the vertical integration in foundry and competent foundry nodes and what was tacitly an X86 monopoly given AMD poor performance since about 2000 or whatever and especially since Nehalem.

It's over for the design units best years, which I also want to note was largely about two things: DC CPU's and consumer CPU's esp laptops. Altera management is atrocious. They missed mobile. They blew the modem unit. DCAI is a joke and they missed that too despite throwing much more than AMD has at it - I suspect AMD will end up #2 now, because they don't have the sclerosis Intel does and Lisa is well aware of DC AI compute trajectories + AMD's issues. Intel also sucks still with GPU's end of the day and basically failed there too. Intel were sloppy about new design increments on CPUs until recently. MobilEye may actually be the best thing you could say about them in terms of breaking into a new market. Historically, even pre-10NM disaster and AMD revival, there were salient signs of deep rot. You could smell it, the sloth and all. It's so, so perplexing people act like Intel only has one issue with CPU area and e.g. process or dGPU's - no, the mismanagement, greed, sloth were evident running years back.

Overall, missing/losing mobile, GPUs and AI are their biggest sins. X86 I think was bound to return to either a competitive equilibrium OR end up less relevant in the event AMD totally died out.

The foundry even though I am skeptical I am more optimistic on because the moat is much more obvious and their competence is arguably more obvious - further, the demand trajectory for silicon bodes well for them in the next 5-10 years, as do geopolitics even in perfectly fine optimistic extrapolations of current trends (e.g. no war in the pacific).

 

[soresu]

Maybe if ET hands them alien technology they can manufacture at scale, or Apple and AMD cave in overnight, TSMC and Samsung and their American facilities are hit by missiles, Arm folds, AND Intel decides IFS isn't for them after that and they want full vertical integration back - then yea, maaaybe we could return to the previous state.

Woah there with the shovel I think they're buried already 😂

 

[rainy]

Worst struggle in history=second highest amount of money they had in the last 12 years...(and highest total assets EVER)
Maybe you should explain what you consider struggle for a capitalistic company that is trying to make money.

That's pretty hilarious argument, maybe you should look at their long-term debt as it was mentioned already:
https://www.macrotrends.net/stocks/charts/INTC/intel/long-term-debt

 

[Exist50]

Great sales are usually dependent on the average dope that knows next to nothing about PC hardware and can easily be led around by the nose by sales staff in retail.

Said average consumer would probably not even notice that AVX512 was gone (if they even knew it was there in previous generations), but the more discerning enthusiasts would and hesitate to buy as a result.

Also I would argue that with the sheer effort over years they put into pushing AVX512 software development that they still definitely shot themselves in the foot - now when they introduce new instruction sets devs are going to be significantly more gunshy about implementing them without monetary compensation from Intel directly to that effect.

No meaningful amount of people because of an ISA checkmark on a specs sheet. Rocket Lake was a failure despite having AVX512, and Alder Lake did well without it. Nor has there been some spike in Zen 4 adoption because of AVX either. Likewise for Skylake-X, and even Haswell if you go far enough back.

People buy CPUs based on how they perform in their workloads. For the average web + office PC, no one really cares about fancy vector support. Productivity might care a bit more, but has been blunted by the addition of stuff like AVX2 VNNI. And if you're doing large scale CPU ML, you might care quite a bit (hence why Intel's kept it around in servers). But let's not reverse cause and effect. ISA additions are only as useful as the performance they provide in user software.

 

[soresu]

People buy CPUs based on how they perform in their workloads

People buy all sorts of products for different reasons - CPUs and other PC hardware included.

My old university housemates simply bought Intel because it was Intel regardless of the reviews/perf/specs etc - and they had no viable excuse of lacking general PC knowledge either like my nan does.

Sometimes they would buy new PC stuff simply because it was new - like a football team supporter buying the shirt designs for a new season - and then they would complain endlessly that they had no money for takeout food 😂

 

[Exist50]

People buy all sorts of products for different reasons - CPUs and other PC hardware included.

Sure, there might be some handful of people who want specific ISA available to play around with, or who have a use case disproportionately affected by certain ISA (e.g. emulation). But those sales do not meaningfully contribute to the bottom line, nor do they factor into product design. I'm reminded of this XKCD: https://xkcd.com/1172/

 

[soresu]

Sure, there might be some handful of people who want specific ISA available to play around with, or who have a use case disproportionately affected by certain ISA (e.g. emulation)

Not what I meant - your assertion was that sales are purely driven by performance in a given workload or workloads.

Mine was that even enthusiasts can buy stuff for dumb reasons and that those numbers are probably not as insignificant as you might like to think they are 😅

 

[Exist50]

Not what I meant - your assertion was that sales are purely driven by performance in a given workload or workloads.

Mine was that even enthusiasts can buy stuff for dumb reasons and that those numbers are probably not as insignificant as you might like to think they are 😅

I think we're on the same page. I acknowledge that people can have all sorts of niche reasons to buy a product, both logical and illogical; I just don't think those contribute meaningful numbers of sales. Heck, even on this forum, for how many people do you think ISA is a first order determinant?

 

[soresu]

for how many people do you think ISA is a first order determinant?

ISA in what sense?

When I see ISA I think x86/ARM/RISC-V.

Or do you mean the instruction set extensions?

 

[Exist50]

ISA in what sense?

When I see ISA I think x86/ARM/RISC-V.

Or do you mean the instruction set extensions?

Yeah, meant in that context. Like, the kind of ISA extensions that're a performance difference, more than a compatibility one.

 

[soresu]

Like, the kind of ISA extensions that're a performance difference, more than a compatibility one

The greatest irony is that probably out of all software easily accessible it's likely that Intel's own code like Embree and the SVT codecs have the most AVX512 code 😂

That being said not for nothing, but Embree is finally going GPU compute with v4.

Whether that is a result of their dGPU ambitions, removing AVX512 from ADL or a combination of both I don't know, but food for thought that Intel may well be eyeing a future less dependent on AVX anyway.

 

[Panino Manino]

I know nothing, but this seems like a mess.
Makes me think that Intel should go to AMD and discuss developing together a new architecture with sane vector instructions.

 

[aigomorla]

I know nothing, but this seems like a mess.
Makes me think that Intel should go to AMD and discuss developing together a new architecture with sane vector instructions.

that would be like the creation of a hydrgen car... uses hydrogen fuel, to power a electric motor.
lol...

 

[itsmydamnation]

I know nothing, but this seems like a mess.
Makes me think that Intel should go to AMD and discuss developing together a new architecture with sane vector instructions.

The current ones are perfectly sane. technical debt accrues over time, that happens in any product. unless you explicitly drop support for old features. The only real miss-step is AVX-512 not aligning to intel's actual plans.
Everyone still wants alot of the "old" features , we just dont want the now very old features. So if they keep deprecating really old things as they add new things i dont see what the deal is.

 

[H433x0n]

Genuine Question: How does every topic on this forum seemingly always pivot to talking about how dead Intel is / will be?

This was ostensibly about ISA expansion at one point.

 

[tamz_msc]

Genuine Question: How does every topic on this forum seemingly always pivot to talking about how dead Intel is / will be?

This was ostensibly about ISA expansion at one point.

It's a meme at this point.

 

[JoeRambo]

Typical to these forums, so much gloom
The reality is that 99+% of users would benefit from APX real soon after it is out and they've got hw. Don't forget that most people only browse web and watch videos and it would only take browser vendor to improve their JS VM and code generation and push update.
They have already insane JS optimizers that would really love additional registers and some of those new instructions.

The key takeaway for me is -> Intel did this in least painful way, actually caring about implementation, thinking about how to make the least pain for OS to support APX. It's good evolution for x64 that will carry us to 2040+ or so, just like original x64 did for last ~20 years.

Personally i think they've achieved their goals with APX.

 

[TheELF]

I don't know what he was thinking, but I think they are struggling since they have lost more marketshare, especially in servers than ever. And the layoffs. If you adjust those numbers for inflation, I am sure it will also change the picture. But marketshare affects future income.

Market share of AMD fell back to normal immediately after the pandemic, sure server is the highest it ever was for amd but it's still low enough for it to be a non issue for intel.

AMD and Intel CPU Market Share Report: Recovery on the Horizon (Updated)

Hope is on the horizon, mostly.

www.tomshardware.com

That's pretty hilarious argument, maybe you should look at their long-term debt as it was mentioned already:
https://www.macrotrends.net/stocks/charts/INTC/intel/long-term-debt

Perspective...the debt increase is pretty small compared to other times, and especially during a time that they build FABs like nobodies' business having such a low increase in debt is pretty impressive.

Genuine Question: How does every topic on this forum seemingly always pivot to talking about how dead Intel is / will be?

This was ostensibly about ISA expansion at one point.

Changes in how a CPU works can affect how well a CPU is perceived which can affect sales which does affect how well a company does.
TL;DR
Terrible ISA = dead company.

 

[Exist50]

Changes in how a CPU works can affect how well a CPU is perceived which can affect sales which does affect how well a company does.
TL;DR
Terrible ISA = dead company.

Frankly, anyone conflating the technical merits of an ISA with its success in the market has not learned from history. People have been complaining about x86 for literal decades at this point. Has not stopped it from being one of, if not the most commercially successful ISAs around. We see some of this same discussion both for and against RISC-V today. The success or failure of an ISA is about the business model and performance of the companies behind it far more than any technical merit.

 

[rainy]

what was tacitly an X86 monopoly given AMD poor performance since about 2000 or whatever and especially since Nehalem.

I don't know how old you are but it sounds funny because 2000-2006 was AMD best period (K7/K8 architecture) before Zen happened.

 

[rainy]

Perspective...the debt increase is pretty small compared to other times, and especially during a time that they build FABs like nobodies' business having such a low increase in debt is pretty impressive.

Changes in how a CPU works can affect how well a CPU is perceived which can affect sales which does affect how well a company does.
TL;DR
Terrible ISA = dead company.

Nice art of manipulation, graphs below are not that pretty./s

 

[Tuna-Fish]

Market share of AMD fell back to normal immediately after the pandemic, sure server is the highest it ever was for amd but it's still low enough for it to be a non issue for intel.

The issue is that their revenue is in the dumps, despite having such a high market share, AKA = their ASP is low because they are competing with AMD. Last quarter, they had a gross profit of $4B. The company is structured around gross profits of $9B+. The last time their opex was below $4B was in 2011. Unless they dramatically increase profit, they will need to lay off about a third of the company. On top of the layoffs they have already had.

Perspective...the debt increase is pretty small compared to other times, and especially during a time that they build FABs like nobodies' business having such a low increase in debt is pretty impressive.

Yes, it is not bad to take debt. But the issue is, you were crowing about them having a lot of cash on hand, while the only reason that happened was that they took a lot of debt. They lost a lot of money in the business, and then took debt to maintain cash levels.

Again, not an Intel doomer. I think they will recover to be a healthy company again. But their latest quarter was not a good one. If you want to be positive on Intel, don't try to lie about their last quarter, picking random numbers and trying to twist them into looking good, despite the fact that they had the worst quarterly loss in their history. Instead, concentrate on the good things they have in their pipeline: For example, they are going for backside power well before TSMC. If they succeed, this will probably be enough to get them a reliable process advantage until TSMC catches up to it.

 

[Doug S]

The reality is that 99+% of users would benefit from APX real soon after it is out and they've got hw. Don't forget that most people only browse web and watch videos and it would only take browser vendor to improve their JS VM and code generation and push update.
They have already insane JS optimizers that would really love additional registers and some of those new instructions.

And why does that matter, other than higher scores in the "how fast is your browser" thread?

Interpreted/JITted code is not used for anything performance critical, and that's where it will take years before APX has any impact. So great, if you buy an APX CPU your browser will run faster, but if a web site is slow on brand new Intel or AMD CPUs it is broken.

Getting APX to speed up a broken slow website is like having a sewer line full of roots and saying "hey no problem I got a new toilet with a stronger flush, that'll help push the turds through better!" Once the website is fixed or you have the roots augured out, neither APX nor your new toilet make any difference.

 

[JoeRambo]

And why does that matter, other than higher scores in the "how fast is your browser" thread?

Sir, i bet you composed this forum post on ZX Spectrum, right? Not much else to say about the rest of Your post.