[HN Gopher] Intel's Immiseration
___________________________________________________________________
Intel's Immiseration
Author : rbanffy
Score : 96 points
Date : 2024-08-08 19:11 UTC (3 hours ago)
(HTM) web link (thechipletter.substack.com)
(TXT) w3m dump (thechipletter.substack.com)
| PaulHoule wrote:
| The biggest problem I've seen with Intel is that they are
| "getting high on their own supply" and the whole tech press
| enables them with the sole exception of Charlie Demerjian.
|
| They should pick a few random people out of the phone book and
| pay them handsomely for their advice, this way they might start
| to "see themselves as other see them". Most of all they need to
| recognize the phenomenon of "brand destruction" which has
| manifest in clueless campaigns such as "Ultrabooks", brands that
| should have been killed off years ago ("Celeron", sorry, low
| performance brands have a shelf life, there's a reason why Honda
| is still making the Civic but GM is not making the Chevette) and
| that they should write it in their bylaws that they are getting
| out of the GPU business permanently (they've made so many awful
| products it will take 15 years for people to get the idea that an
| Intel GPU adds value, instead it's this awful thing you have to
| turn off so that it won't screw up graphics in your web browser
| when you're using a Discrete GPU.) People might get their heads
| around the idea that an Intel CPU is a premium brand _if_ it had
| a GPU chiplet from a reputable manufacturer like AMD or NVIDIA.
|
| (Oddly when Intel has had a truly premium brand, as in SSDs,
| they've ignored it. Hardly anybody noticed the great 95% latency
| performance that of Intel SSDs because hardly anybody realizes
| that 95% latency is _what you feel when your computer feels
| slow_. Intel SSDs were one Intel product that I would seek out by
| name, at least until they sold their SSD division. Most people
| who 've run a lot of Intel SSDs swear by them.)
| is_true wrote:
| Does the new owner provides the same performance?
| christkv wrote:
| The CEO had only been in charge since 2021 bringing an engineer
| back to the helm. How long does it take to bring a big ship
| like Intel around and undo decades of internal rot? It will be
| interesting to see who is let go in the coming months.
| akira2501 wrote:
| > How long does it take to bring a big ship like Intel around
| and undo decades of internal rot?
|
| If it takes that long, then this is your actual problem, and
| not how many engineers are in the loop at the board level.
| soulbadguy wrote:
| >If it takes that long, then this is your actual problem
|
| Most IC design life cycle are 5+ years. Expecting to turn
| over in less is part of the problem intel is having
| currently.
| macintux wrote:
| I'm skeptical. Big company, design & manufacturing cycles
| on the order of years because chips are that sophisticated
| these days, I don't know what you'd expect to see within 3
| years of a new CEO.
| barkingcat wrote:
| to be honest, they should have done a layoff 3 years ago,
| as soon as Pat Gelsinger came onboard, and fired 2 entire
| layers of management.
|
| No engineers or support people, but just pure middle
| managers - fire 2 or 3 layers. if they did that 3 years
| ago, they would be in a better position now.
| code_biologist wrote:
| Though prior to Gelsinger, Youtuber and chip leaker / rumor
| monger "Moore's Law is Dead" mentioned that Jim Keller's
| stint at Intel was so short (April 2018 - June 2020) because
| of internal cultural toxicity. In particular, things like
| leaders of major groups in Intel trying to get employees of
| other groups fired, so as to make their own relative progress
| look better. Take with a pound of salt of course, but MLiD's
| sources have been pretty decent.
|
| You only resolve that kind of badness by firing a bunch of
| SVPs and VPs and Gelsinger hasn't done that, for many
| possible reasons.
| Miraste wrote:
| A lot of these are symptoms of the root cause: a bad product.
| Ultrabooks aren't a terrible concept; they're the Wintel
| version of Macbook Airs. Making a non-Apple Apple device is a
| fine business strategy, and Samsung has made plenty of money
| off of it. The problem was that Intel chips continually
| crippled them, making them hot and slow with no battery life.
| People don't like Celerons and iGPUs because they all run at
| the speed of molasses.
|
| Any of these branding decisions would have worked fine, if the
| products did. But no amount of marketing will fix bad
| engineering.
| soulbadguy wrote:
| bad product is a symptoms of a even deeper cause... Lack of
| competition. The decline of intel started long ago. The
| financial results here are just very very lagging indicator
| of that. The lack of competition combined with very agressive
| anti competitive practice allow them to survive on pretty bad
| product line. In a healthy market, competition would have
| force intel to improve well before the point we are now
| adrian_b wrote:
| While the Intel laptop CPUs may have been not good enough,
| that is not where Intel has lost money.
|
| The current financial results have shown decent profits for
| the consumer CPUs, which could have been greater if Intel had
| succeeded to produce more Meteor Lake CPUs. Intel could not
| produce as many Meteor Lake CPUs as the demand because their
| new Intel 4 manufacturing process has low production yields.
|
| While Intel lost money in their foundry, that was unavoidable
| due to the high expenses needed for catching up with TSMC.
|
| Where Intel had great losses was in server CPUs. The reason
| for these losses must have been that Intel had to grant very
| large discounts to the big buyers of server CPUs, in order to
| convince them to not buy superior AMD server CPUs.
|
| These losses in server CPUs were easily predictable, because
| even if Intel has succeeded to reach in time all the publicly
| announced milestones of their roadmap, their roadmap itself
| does not hope to reach parity with the AMD server CPUs before
| H2 2025 (and that assuming that AMD will not introduce Zen 6
| during 2025, which would move the target).
|
| It looks like Intel has made a PR mistake. Even if their
| public roadmap has shown all the time that they will continue
| to have big losses for at least one more year, that was
| obvious only for technical people, who could compare the
| future Intel products with the AMD products, because Intel
| has not said any word in their roadmap about what the
| competition will have at the same time. For non-technical
| people, the Intel press releases sounded more optimistic than
| they should have been, leading to disappointment caused by
| financial results that should not have been surprising.
| Tuna-Fish wrote:
| > The current financial results have shown decent profits
| for the consumer CPUs,
|
| Only after serious financial alchemy. The client group has
| a decent op margin, but the foundry group has an absolutely
| terrible one, and the client group sells products that they
| buy (internally) from the foundry group at arbitrarily set
| prices. What would the client margins be if the internal
| price they pay for their products was set so that the
| foundry breaks even?
| einpoklum wrote:
| > Ultrabooks aren't a terrible concept; they're the Wintel
| version of Macbook Airs.
|
| I'd say both are a terrible concept. I mean, it sells, but
| it's still terrible. A laptop being thin is not really such a
| virtue and people's life would be better with a decent-travel
| keyboard, better heat dissipation and more ports than with
| 5-10mm less laptop height.
| paganel wrote:
| I don't really care that much about brands nor laptops for
| that matter, but my M1 Air is really, really nice given the
| money I've paid for it. Why should I burden myself with
| carrying a big laptop thing when I can instead carry a
| Macbook Air that is much lighter and which does very good
| job?
| Almondsetat wrote:
| Who are you to say what would make people's lives better?
| PaulHoule wrote:
| An actual computer user who's probably at least as
| qualified as whoever made these decisions at Intel.
| Almondsetat wrote:
| I'm an actual computer user and I disagree. Now what?
| wtallis wrote:
| Heat dissipation and lots of ports are why I have a
| desktop. I don't want that from my MacBook Air. A fan isn't
| necessary or desirable for a machine that is primarily used
| for web browsing, watching movies, and light gaming.
| saulpw wrote:
| You know what I want for my desktop, is a battery, so my
| computer doesn't reboot when I lose power for 5 seconds.
| Yes I know about UPSes, but they're $100+ and unwieldy. I
| just need 30 seconds grace period to save my work!
|
| This is the main reason why I probably won't get a
| desktop again.
| askafriend wrote:
| That's wrong and the market has proven it. The M3 MacBook
| Air is the best computer for most people - period, not just
| best ultrabook or best laptop. There are almost no
| compromises for 99% of people's usage.
| soulbadguy wrote:
| > The M3 MacBook Air is the best computer for most
| people.
|
| Maybe, that it doesnt imply that it's the best because
| it's an ultrabook, or that a non-ultrabook version of the
| macbook air (not the macbook pro...) wouldn't have been
| better.
| soulbadguy wrote:
| There is nothing wrong about the concept of ultrabook. Just
| terrible implementation.
| impossiblefork wrote:
| But don't they need to get into the GPU business?
|
| Surely GPUs or similarly very parallel machines for things like
| ML training are very needed and will remain very needed. Seeing
| as firms such as Groq have done fine, surely Intel can do
| something of that sort without much difficulty?
|
| Since their GPU business has been unsuccessful perhaps they can
| go for whatever makes sense, as there's nothing of this sort
| that they can release that will compete with their own
| products.
| pradn wrote:
| The AI boom is, in theory, a godsend for Intel and AMD. You
| can focus on creating good tensor computation hardware,
| without having to worry about getting gamers on board. No
| need for "Game ready drivers" or compatibility with complex
| legacy graphics APIs.
|
| Of course, there's the elephant the room for general purpose
| tensor machines, which is CUDA - famously closely guarded
| Nvidia. But with the new wave of "above-CUDA" APIs like
| TensorFlow, PyTorch, and Keras, there's an opportunity to
| skip the CUDA layer altogether.
| Sparkyte wrote:
| AI doesn't exist with the right amount of collected data,
| we're going to hit a wall very soon where we just get bad
| AI data constantly because we are throwing data to the wall
| and hoping it sticks. Businesses should not be investing in
| AI right yet maybe 2-3 more years.
|
| The businesses that should are the startups and businesses
| who are helping define the model, like ChatGPT and
| Microsoft, but the adoption is still too early. If I was
| McDonald's or Wendy's and I wanted to use AI to help
| promote sales to customers I would need to be able to grab
| demographic data which may not be appropriate PII data.
|
| All of the lawsuits happening right now for data collected
| without permission of the data provider is going to all
| change the landscape.
| nullc wrote:
| I think there is still a good business to be made taking
| the AI stuff we have _right now_ and making it highly
| performant.
|
| Stable diffusion models are awesome and super useful. I
| don't necessarily mean the simplistic text->image stuff
| (though it's clearly useful)-- but denoising, in
| painting, animation from stills, 3d from stills.... style
| changes, etc.
|
| LLMs are at least somewhat useful.
|
| We can radically improve image and video compression with
| performant enough AI.
|
| Plenty of other applications of AI right now are already
| useful or would be useful if inference were made more
| efficient or more local (thus private). No more data
| needed.
|
| You don't need some grand vision of an AI future to get a
| total addressable market for high performance AI that is
| in the same ballpark as high performance GPUs for non-AI
| usage.
| Sparkyte wrote:
| But for every business the chase seems to fail in finding
| worth. I can already tell you from where I am working we
| are struggling because eventually data is too old to be
| good or too incomplete. We need at least 3-4 years of
| collecting and storing the information to make full use
| of AI.
| PaulHoule wrote:
| In practice it's the other way around.
|
| In practice AMD and Intel GPUs should be suitable for
| machine learning but the software story isn't good. I know
| I can buy a NVIDIA card and know it's a good investment
| because everything worth with CUDA and be training models
| in less than a day. If I went with some other brand I'd
| expect to put 6 man*months into figuring out the software
| story, and that's a lot more than the price difference in
| the cards.
|
| I've been wondering about the soundness of Intel's software
| strategy in that OneAPI ("One" is a bad smell in marketing
| speak, the only premium product in my mind that has "One"
| in the name is "Purina ONE" pet food, the XBOX ONE is an
| astonishing own goal of a name because you'd never get your
| mom to understand that an XBOX ONE is better than an XBOX
| or XBOX 360) is based on OpenCL and the one thing I know
| about OpenCL is that I don't know anyone who likes coding
| for it. The idea that you could code for FPGA and GPU out
| of the same API also seems delusional because FPGA is
| (mostly) about latency and GPU is about throughput. That
| is, FPGA can do certain small operations insanely fast and
| avoid the overhead of 16-bit math when 13-bit math will do,
| GPU is all about doing large operations in bulk.
| bryanlarsen wrote:
| > If I went with some other brand I'd expect to put 6
| man*months into figuring out the software story, and
| that's a lot more than the price difference in the cards.
|
| Several companies are spending >$10B annually on AI
| compute. There are a lot more than 6 man-months of
| savings in it for them...
| PaulHoule wrote:
| One firm saving $2B saves $2B, open source software that
| opens up the market for everyone is priceless.
|
| A company like OpenAI that is mostly concerned about pace
| might see going with the #2 GPU vendor is risky, even if
| it has the possibility to save money.
| ein0p wrote:
| The whole situation with Intel and AI is also baffling to
| me. They have an excellent product in this space - Gaudi2.
| Faster than A100 and very attractively priced. I've tried
| it, it works fine. Gaudi3 is also about to come out, and
| it's twice as fast as Gaudi2. Yet nobody is buying. I get
| why you wouldn't want this for training - it requires some
| minor code modifications and your Triton kernels are
| worthless. But for generative inference this is just what
| the doctor ordered.
| soulbadguy wrote:
| > generative inference this is just what the doctor
| ordered.
|
| Naive question, wouldn't you need a descent tool chain
| for inference as well ?
| mlsu wrote:
| The AI boom that lead to Nvidia's market position is still,
| 3 years on, almost entirely speculative. The only products
| that are currently driving value (i.e., that are providing
| consumer surplus, for which businesses and individuals are
| actually opening their checkbooks for) are paid chat-bots.
|
| Nvidia ate the whole pie because they were in the right
| place at the right time with shovels to sell. Getting into
| it now, in house, would be a huge bet that the pot of gold
| will be there when they get to the end of the rainbow --
| and they have a long way to go.
| Dalewyn wrote:
| Intel actually has it even worse because they completely
| failed to bring ARC to market in time. The product came
| out (very) late, underdelivered, and permanently damaged
| the brand reputation. Yes, the drivers and thus the cards
| have gotten better, but who's talking about them now?
| Noone.
|
| They say the opposite of love is indifference, and
| basically everything Intel has sold outside of CPUs and
| NICs have suffered market indifference. Actually, maybe
| even the NICs given everyone seems to prefer Realtek NICs
| over Intel NICs these days.
|
| Nvidia and AMD have it way better because if the market
| doesn't love them, you bet the market will hate on them
| instead of be indifferent.
| Sparkyte wrote:
| Intel has always been an echo chamber of fiscal earnings and
| cutting corners to appease investors. For the longest time they
| stagnated on Mac hardware improvements because it would cost
| money, they would not have lost Mac as a business customer if
| they continued to innovate.
| PaulHoule wrote:
| Also they should have had a policy of "not one transistor for
| the national labs" and "not one transistor for hyperscalers",
| particularly because anything they do to appease hyperscalers
| just saves them money that they'll spend on their custom
| silicon transition.
|
| There's something prestigious about HPC but the story of how
| commercial data processing went parallel in the oughts shows
| how out of touch the HPC community is.
|
| In the meantime Intel has botched the deployment of SIMD,
| slowplaying to the point where maybe 7 or 8 years from now
| some mainstream software might support AVX512, maybe.
| Somebody should be tarred and feathered for introducing the
| idea that features should be fused off for all but the
| highest paying customers. If the customer is not paying for
| the fused off functionality, the shareholders are. It sounds
| really ruthless and avaricious and might impress some
| business analysts as a form of vice signalling but it's pure
| waste.
| Sparkyte wrote:
| Right it is wasteful. The reason some businesses are so
| successful is that they look at the common denominators
| rather than trying to perpetuate worth. Fewer variants and
| don't wall off features. AMD did this for the longest
| period getting back on it feet, but I am seeing a failure
| of them continuing down this path. They need to stop making
| variants and focus on 3 or 4 sets of consumer processors.
| PaulHoule wrote:
| Yep, "too many SKUs" killed off the consumer HDD a few
| years faster than it had to die. If these companies had
| focus groups that reached out to ordinary people they'd
| realize it sounds absolutely incredible to most people
| that you need one hard drive for a 3-5 bay NAS, another
| for a 6-9 bay NAS, another if you are recording from
| video cameras.
|
| Even if there was a real benefit from optimizing firmware
| and saving a few cents with cheap washers for drives in a
| low vibration environment there is a much more certain
| cost in that you have to qualify the firmware, risk
| introducing errors, etc. I can imagine that Best Buy
| might have been able to stock one model of high capacity
| HDD but no way were they going to stock more colors of WD
| drive than there are on the rainbow flag. That lack of
| product sense locked manufacturers out of the retail
| market.
|
| (Makes me think of the cringe monthly hard drive roundup
| in Anandtech where an enterprise drive produced in huge
| numbers in a single SKU would usually be $80-100 cheaper
| than prosumer drives that, according to the spec sheet,
| consumed about 0.5W less and were about 3db quieter...
| And I have plenty of those enterprise drives still
| spinning after all these years, the only time I think
| about the noise is when the machine boots and the drives
| emit a throaty chirp that makes me think of a Ferrari
| spooling up.)
| mrandish wrote:
| _> " Somebody should be tarred and feathered for
| introducing the idea that features should be fused off for
| all but the highest paying customers. If the customer is
| not paying for the fused off functionality, the
| shareholders are."_
|
| This is very well put.
|
| While I've always understood the economic motivation for
| margin optimization and the technical reality that CPU
| wafers aren't very granular, fundamentally, it's a long-
| term losing idea to intentionally fab gates that don't add
| end-user value. While it may work in the shorter-term, when
| it becomes the plan of record (as opposed to fixing a one-
| gen design or product mix error) it signals a shift to
| prioritizing value extraction over value creation.
| oivey wrote:
| Even more simply, their goal stopped being making the
| best product at the best price. Features that were fused
| off could have made their products better for no extra
| manufacturing cost. Your business is broken when you view
| your products as being too featureful as a cost.
| jzb wrote:
| "they would not have lost Mac as a business customer if they
| continued to innovate"
|
| I'm not entirely sure this is true. I mean, I guess it
| depends on whether one expected Intel to be able to make not
| just a decent chip for PCs/laptops, but also one for phones &
| tablets. Once Apple started dabbling in its own chips for the
| iPhone and iPad, it seemed inevitable that they'd expand that
| to their macOS systems too. Apple has been a rough customer
| to please for chip designers/manufacturers, I'm not sure any
| company could've satisfied them with a general-purpose chip.
| hbn wrote:
| If they played their cards right as the top dog ~20 years
| ago, they could have perhaps ended up with the relationship
| with Apple that TSMC does today.
| talldayo wrote:
| 1. Intel didn't have big fabs 20 years ago, so they
| didn't have anything to offer Apple that would resemble
| TSMC's partnership.
|
| 2. There's no money in designing licensed CPU cores as a
| middleman for a company that drives margins so low
| they're associated with suicide nets and forced labor.
|
| 3. If TSMC is lost due to Chinese aggression (which is a
| non-zero chance), Apple is left fabless and has to choose
| between importing Samsung silicon at-cost or partnering
| with Intel.
| PaulHoule wrote:
| Might take them a few years to update their software to
| run on the Samsung stack.
|
| What are they going to do without all the auxiliary
| processors they build in?
|
| Or do they have a team which has their software working
| on other people's silicon in case their supply chain get
| blown up?
| talldayo wrote:
| Also, importantly, how hard will the US government be
| taxing imports and propping-up Intel? I think that an
| anti-Apple and pro-Apple administration would both want
| them to stop shipping jobs offshore. There's a tariff
| knob that can be adjusted until only Intel (or
| GlobalFoundries, lmao) are feasible choices, and if the
| US wants their Intel investment recouped then they may
| well push for that kind of deal.
|
| Personally, I find the "samurai's honor" shit where one
| company avoids another to be childish and stupid. If you
| make Apple's board choose between ending their grudge-
| match or blowing up their margins, I don't think they'll
| care much either. Or maybe I'm wrong, and the 2027 iPhone
| is manufactured with 24nm wafers intended for the Toyota
| Prius.
| AtlasBarfed wrote:
| Companies that are built on engineering eventually become
| subsumed by soul-sucking profit sucking bonus sucking middle
| management types that just plop themselves in their big huge
| hierarchy and destroy the company long-term with inertia and
| apathy.
|
| See also. Boeing Medtronic GE AT&t
| immibis wrote:
| I have an Arc A770 and it seems to work fine so far... though
| I've only had it for a month. There are some tolerable teething
| problems: I can't set the fan curve from Linux so I had to duct
| tape some extra fans, and it's incompatible with BIOS boot
| (requires UEFI) which should be considered acceptable in 2024.
| For presumably the same reason, there's no graphics output in
| early Linux boot until the GPU driver is loaded.
|
| The IGPU in my previous machine's i7-6700K (Skylake) was also
| just fine. Intel Graphics Media Accelerator, yeah that really
| sucked, but that was like 15 years ago?
| samstave wrote:
| Disclaimer: I used to work at Intel:
|
| Specifically, I was the second person on the planet to game on
| a Celeron Proc. How? Me and my best friend ran the Intel
| Developer Relations Group Gaming Test Lab in the later half of
| the 90s.
|
| Our job was to specifically test all PC games that were being
| developed against the Intel Celeron procs and the amd alts...
| with a focus on Subjective Gaming Performance of the Celeron as
| an "Optimized" feature leader whereby developers were given
| handsome marketing monies to optimize the games against latest
| SIMD instructions to allow for the games to be more performant
| on the Celeron.
|
| THe goal: Show the world that a $1,000 fully capable gaming PC
| was possible, in their budget, and _desirable_.
|
| ---
|
| The Issue at the time was the graphic bottlenecks -- all the
| pieces had yet to come to fore: AGP, Unreal, OpenGL, ~~NERDS~~
| NURBS!, Games, Graphix Chips (VooDoo, 3DFX, blah blah)
|
| Celeron should have died long ago - but certainly when the
| first, actual GPUs came along and did the heavy lifting.
|
| I have a lot of thoughts about Intels mess (Transmetta really
| fucked Intel up in the way an abusive step-relative would) and
| caused them to lose focus...
|
| Then, just the ridiculous amount of marketing over
| engineering....
|
| _(if anyone worked at intel in the DRG circles in SC5 - and
| has access to emails of that day - search my name and the
| thread I started asking why we cant just physically stack CPUs
| on top of eachother... (this was prior to the Voxel timeline)
| and was laughed at. it wasnt until several years after that I
| went on a hike with a labs head and found out about the 64-core
| text CPUs that were only coming out)_
|
| ---
|
| I was just having shower-thoughts about intels future as
| effectively computing _grout_ -- a mechanism to get data from
| real-world INTO NVIDIA GPUs and then displayed again real-
| world. And thats it. Thats the only thing Intel may be doing in
| the future - is grout to display the data components solely
| computed, as NVIDIA CEO stated himself "All compute will be
| done on NVIDIA chips" -- _delivery_ of the data through intel
| 's grout (minimally) then again delivered to an NVIDIA desktop
| GPU...
|
| Intel is like the maintenance staff of a home. NVIDIA is the
| architect, interior designer, party planner and end user.
|
| _(Intel was my first ever stock package: $70 with a $125
| option price. 10K shares. I left before ever vesting... it was
| the late 90s and I had to chase the dream
| intc:https://i.imgur.com/U9PWURv.png )_
| spookie wrote:
| Having a third competitor in the GPU is great, for both Intel
| and the consumer. If one starts blaming newcomers for their
| shortcomings then we are failing to see the bigger picture.
|
| When you talk about web browsers failing to display content
| because of the iGPUs I fail to even have heard of such a case.
| Maybe the performance is bad but that's it. Either way, that's
| more of a failure of the OS and not Intel. Windows is
| particularly biased to not use your dGPU on a laptop, but I
| blame the kernel for making badly calculated assumptions based
| on nothing but !={game, 3D modelling tool} then iGPU, and not
| the hardware.
| kjellsbells wrote:
| The classic moat metaphor that the OP article and others use
| needs to be fleshed out to match Intel's predicament.
|
| A moat protects a castle that adversaries want to take over. The
| presence of the castle defines what can and cannot be done with
| the surrounding landscape. But if the castle ceases to protect
| what people care about, or make meaningful additions to the
| environment, it becomes irrelevant and the presence of the moat
| makes no difference.
|
| Intel's problem isn't that competitors want to storm the castle
| and achieve domination over the landscape that x86 controls. It's
| that the competition have built their own castles on the other
| side of the river, and a lot of the peasants are tilling the
| lands around Castle ARM and Chateau NVIDIA.
|
| To put it another way, Intel thought the castle was "control of
| computing" and the moat was "leadership in x86" but irrelevance
| comes a little closer with each passing chip generation. It is
| fortunate for Intel that corraling an ecosystem into existence
| around the alternatives to x86 is an insanely difficult task, but
| it has been done with ARM, it has been done (albeit for a niche)
| with NVIDIA and it can be done again with whatever comes next.
| soulbadguy wrote:
| >Intel's problem isn't that competitors want to storm the
| castle and achieve domination over the landscape that x86
| controls.
|
| IMO it's both. While the importance x86 is declining, AMD is
| aggressively eating what ever part of it is left. I also think
| that in the long term, as intel and amd build better x86 chips,
| the value proposition of ARM will slowly fade in favor of
| something like risc-v
| apantel wrote:
| Asking because I don't know: what is the value proposition of
| ARM?
| selimnairb wrote:
| In PCs, ARM CPUs perform just as good or better than AMD64
| but have much better battery life. In the cloud, ARM CPUs
| are much cheaper (ca. 25% less) for the same or better
| performance.
| soulbadguy wrote:
| Not quite. I think we need to split the value of ARM the
| instruction from specific implementation.
|
| 1 - In term of pure efficiency, nothing magical about
| ARM. Looking at AMD latest strix laptop platform they are
| about on par with qualcomm new arm laptop chips.
|
| Apple M* CPU are still better. However, a lot of that
| efficiency is platform derived.
|
| 2 - The lower cost in the cloud is a function of the
| middle man being removed. Amazon is selling graviton
| cheaper simply because they don't have to pay the markup
| of AMD or Intel.
| adrian_b wrote:
| The value of ARM is that anyone with enough money can
| license Arm cores and incorporate them in their own
| products, which can be optimized for some custom
| applications.
|
| The level of customization possible with an x86 CPU is much
| less. You must buy a complete computer board or module and
| incorporate it in your product.
|
| While for custom applications it is easy to create a
| superior solution with Arm cores, for general-purpose
| computers it is hard to compete with the Intel and AMD
| CPUs. All the computers with Arm cores have worse
| performance per dollar than similar x86 computers. (For
| instance there was recently a thread on HN about a credit-
| card-sized computer with an Intel N100 CPU and with the
| same price or lower as a Raspberry Pi, but with a much
| higher performance.)
| klelatti wrote:
| You've missed out one of Arm's central value propositions
| which is power efficiency and the reason why it has more
| than 99% of the smartphone market.
| adrian_b wrote:
| There have been tons of proprietary CPU architectures
| with the same power efficiency as Arm. Only the x86
| architecture is an outlier that requires an unusually
| complicated instruction decoder, which may degrade a
| little the energy efficiency.
|
| Arm has eliminated most of the competing CPU
| architectures not by providing better power or energy
| efficiency, but by its business model of being licensable
| to anyone.
|
| The ARM ISA was somewhat better than MIPS and SPARC, both
| of which have been handicapped by including some
| experimental features that have been later proved to be
| bad ideas. However there have been many other RISC ISAs
| more or less equivalent with ARM. Only the business model
| has extracted ARM from the crowd, not its technical
| advantages.
| klelatti wrote:
| You're conflating the characteristics of the ISA with the
| value proposition offered by the designs using it. Not
| the same at all.
|
| Other historical architectures typically targeted higher
| performance.
|
| Arm specifically went after low-power applications, which
| continues today when we see the priorities in the design
| of Arm and x86 cores, for example.
| soulbadguy wrote:
| It's been said many time and the correlation between ISA
| and power efficiency have been debunked many time. ARM is
| power efficient because most ARM implementation are power
| efficient. Right now x86 AMD strix are about on par with
| qualcom x elites
| klelatti wrote:
| Agreed. That power efficiency is still a central part of
| the Arm value proposition though. Others are competing
| with real designs on this with Arm in laptops but not in
| - for example - smartphones.
| adrian_b wrote:
| Arm "specifically went after low-power applications" only
| in comparison with x86 or in comparison with a few other
| architectures restricted to workstations and servers,
| like DEC Alpha or Intel Itanium.
|
| Before 2000, there were at least a dozen CPU
| architectures that went for the same low-power
| applications as Arm. There were a lot of microcontroller
| or processor vendors and each one of them had at least
| one proprietary ISA, if not 2 or 3 or even more different
| proprietary ISAs.
|
| More than 20 years ago, I have redesigned various kinds
| of communication equipment, in order to replace many
| other kinds of CPUs, for example Motorola MC683xx or
| ColdFire or IBM PowerPC CPUs, with Arm CPUs.
|
| In none of those cases the Arm CPUs had a lower power
| consumption or any other technical advantage. In fact in
| all cases the Arm CPUs were technically inferior to the
| CPUs replaced by them, which has required the
| implementation of various hardware and software
| workarounds.
|
| There was only one reason why the Arm CPUs had been
| selected to replace their predecessors with different
| architectures, and that was the lower price. Their lower
| price was in great part due to the fact that there
| already were many competing vendors of Arm CPUs, so if
| you did not like one of them it was easy to replace it
| with another vendor.
| AlexDragusin wrote:
| Radxa X4 low-cost, credit card-sized Intel N100 SBC goes
| for $60 and up
| https://news.ycombinator.com/item?id=41007348
| hypercube33 wrote:
| AMD does offer custom x86 - see the steam deck, surface
| laptops and Xbox and PS4 and 5. Given there aren't a ton
| of small fish making custom parts they are excellent at
| what they are made for.
|
| AMD is pushing x86 to Apple ARM levels that keep power
| use low enough (best I've seen is 16 hour battery life on
| a device - I think MacBooks best this still) but
| performance per watt I haven't seen ARM really top
| charts. They are awesome and I want arm and risc-v to
| really shine in laptops but the only player on the PC
| side is Qualcomm who was told to destroy their only
| flagship by ARM.
| soulbadguy wrote:
| > AMD does offer custom x86
|
| Not the same thing. On X86 you have to pay AMD or intel
| to design something for you.
|
| In arm, you get to decide who design your chip or even
| have your own in house CPU design team.
| NortySpock wrote:
| (simplifying) ARM provides verified, tested, standardized,
| reasonably well designed chips (logic circuits) that your
| company can purchase a license for and then send that chip
| design / logic circuit to be etched on a wafer, cut,
| encapsulated, and soldered to a printed circuit board.
|
| Those ARM CPUs support a standard (but ARM-flavored)
| assembly programming language. (Formally: Instruction Set
| Architecture)
|
| Designing your own chip previously was risky because you
| might have logic or hardware bugs in your chip that were
| very hard to debug, and then you hope that someone will
| bother to write assembly code that works on your chip.
| Since you probably designed your own assembly language that
| co-evolved with your chip, those assembly code developers
| are going to be sinking a lot of time into understanding
| your chips and assembly code quirks to wring performance
| out of them.
|
| RISC-V standardizes a RISC-V flavored assembly code (ISA)
| and also provides some certification test packages to prove
| that "this particular chip design" can execute the RISC-V
| assembly language according to specifications.
| moffkalast wrote:
| AMD's stormed the castle with Ryzen long ago and planted their
| flag, but since everyone's still asleep at Castle Intel they
| haven't really noticed.
| causal wrote:
| The article quotes a mention of Intel getting into the foundry
| business - this seems like the most obvious good move to make,
| even if a little late, right?
|
| Being able to operate a general fab for chips seems far more
| important now than the design of the chips, since design has been
| somewhat commoditized by ARM. Any big downside for this move?
| barkingcat wrote:
| the downside is that the foundry process has to work properly,
| and by all reports, it hasn't started working properly yet.
| causal wrote:
| Only report I saw on it was that Intel said they have tested
| working 18A chips.
| sct202 wrote:
| They have to be able to mass produce on that node for it to
| save them. Samsung has been claiming technological
| leadership for that past few years with their 3nm node but
| nothing has come to market using it. IBM demoed a 2nm test
| chip in 2021, but has no ability to mass produce.
| BadHumans wrote:
| We are coming up on 7 years since the first Ryzen chip. In 7
| years AMD went from very behind to a little behind then on-par
| and finally now market leader. The fact Intel let this happen in
| such a short time frame is a bit mind boggling.
| layer8 wrote:
| AMD is not the market leader in any segment, by any
| approximation:
| https://www.theregister.com/2024/05/10/amd_gains_on_intel/
| BadHumans wrote:
| My phrasing is confusing and I apologize for that. I do not
| mean market leader in that there are more AMD CPUs than Intel
| CPU. I'm speaking about hardware performance.
| layer8 wrote:
| You are right for multi-core performance, but Intel has
| still a slight edge in single-core performance, and in idle
| power usage, which is important to many. Intel also has a
| stronger offering in the now-popular mini PC segment.
| adrian_b wrote:
| In server CPUs, Intel still has a larger market share than
| AMD, but judging for the published financial results, where
| most of the loss is in server CPUs, Intel has succeeded to
| keep that diminishing market share only by accepting a great
| loss caused by huge discounts, which has determined the
| action price fall, so perhaps trying so hard to retain the
| market share has not been an optimal decision.
|
| So AMD definitely leads over Intel from the POV of the
| profits obtained in the server CPU market segment.
|
| There are also various small markets where AMD leads
| comfortably over Intel by volume, e.g. Amazon currently sells
| much more AMD CPUs than Intel CPUs.
| nasdaq-txn wrote:
| Market leader in what? Intel's Q2 revenue is over double that
| of AMD's. Intel still controls well over 60% of the x86 space.
| Intel and AMD's most performant x86 offerings are fairly close
| to each other.
| BadHumans wrote:
| Performance. Intel still manages to squeak out some wins
| against AMD when it comes to single-threaded CPU task but in
| every other metric, they are chasing AMD.
| adrian_b wrote:
| Not only in performance, but also in profits, as shown by
| the Intel vs. AMD financial results. Especially in the
| server CPU market segment, where Intel has a diminishing
| market share and losses of billions, while AMD has an
| increasing market share and profits.
|
| The new Zen 5 has much better single-thread performance
| than any Intel CPU (e.g. the slower 5.5 GHz Zen 5 launched
| this week matches a 6.0 GHz Raptor Lake), so for a couple
| of months, until Intel launches Arrow Lake S, AMD will have
| much better single-thread performance. After that, Intel
| and AMD will be again at parity, with negligible
| differences in ST speed.
| debatem1 wrote:
| Intel's main problem is its people. Sorry to say.
|
| The management culture there is insanely siloed and toxic, full
| of scar tissue from previous periods of political turmoil and a
| few relentless sociopaths who absolutely will not stop climbing
| just because it's obviously damaging to the company.
|
| The engineering culture is full of secrets, moats, and the
| general sense that while making progress might be your job,
| stopping it is _everyone 's_ job. And there are always more of
| them than there are of you.
|
| Perhaps most fatally, responsibility for decisionmaking is
| ultimately delegated to the process of decisionmaking rather than
| the people involved. The resulting proliferation of checkboxes
| and box checkers have elevated dysfunctions like "the perfect
| being the enemy of the good" and "none of us is as dumb as all of
| us" into de facto mottos for large parts of the company.
|
| All of this happens against the backdrop of eroding mind and
| marketshare, a total lack of real product clarity, and a
| pervasive culture of bullshit that puffs politically convenient
| non-accomplishments into events worth of companywide applause
| while spinning, downplaying, or even hiding bad news of
| calamitous import.
|
| Intel has vast reserves of talent but without deep (IMO,
| impossible) cultural changes it will inevitably fail under its
| own weight.
| age1mlclg6 wrote:
| I had to create an account to say that I agree with this
| comment.
|
| I am a former Intel engineer who worked there for 8 years. I
| left Intel 2 years after the ACT mass layoff in 2016.
|
| The culture there is extremely toxic from top to bottom. The
| VP's and top level executives were out to get stocks and golden
| parachutes. The middle managers played politics. In addition,
| they also pulled a blind over leadership on many significant
| happenings. The engineers / technicians kept secrets as a way
| of keeping themselves relevant and valuable in case of layoff
| events.
|
| The main and sole money maker for Intel is getting quality
| silicon out of the door and that should be the number one
| objective for everyone. However, due to stack ranking, that is
| getting as many bullet points as possible in order to appear
| relevant when it comes annual review or mass layoff events,
| many people come up with bullshit projects that while on the
| surface look extremely impressive, have nothing relevant to
| pushing quality silicon out. The middle managers play favorites
| and push or highlight those bullshit projects.
|
| When they fly expensive executives around for the BUM's
| (Business Update Meetings) or to pitch some bullshit ideas, not
| many had the courage to ask the relevant questions. The few who
| asked them were met with canned and non-informative responses
| that only showed that these expensive suits had no awareness of
| what was going on externally. At the same time, we were too
| busy patting ourselves on the back, claiming that we were
| number one.
|
| We were too busy playing political games and competing
| internally that we forgot about the real external competitions,
| and then it was too late and they zoomed right past us.
|
| I doubt the CEO (past and present) or VP's care or even have an
| inkling on what has been going on in the culture.
| einpoklum wrote:
| I don't know, a lot of these explanations seem more like reading
| astrological maps in hindsight. Example:
|
| > _Intel passed on making the System on Chip for Apple's new
| iPhone, Instead Apple used an Arm designed processor in an SoC
| built by Samsung. Too late Intel realised the mistake._
|
| Mistake? AMD didn't even try to build SoCs to compete with ARM /
| Qualcomm / etc. They seem to be doing ok.
|
| Now, you could say "Well, throwing 10^10 USD on trying to build
| an SoC was the mistake" - maybe, but then it's the opposite
| mistake than the article claims.
| klelatti wrote:
| No it was a both a huge mistake and a symptom of underlying
| problems. Making smartphone SoCs is a huge profitable market
| that would have given Intel massive volumes to support R&D etc
| and slowed TSMC. The fact that they couldn't manage it with
| every apparent advantage is very telling.
| talldayo wrote:
| Feels like you're rewriting history, to me. Smartphone SoCs
| are _not_ hugely profitable, and a big point of the past 10
| years were Chinese manufacturers entirely ignoring ARM IP to
| increase profits. The only people that make money off ARM
| hardware are the export-officiated manufacturers that don 't
| pay per-unit licenses; which is exclusively Apple. Nobody
| else has a more permissive ARM architecture license, not even
| Nvidia.
|
| Intel at the time was fabless, would have needed to license
| or design a RISC architecture, and would have ended up just
| as squeezed as any other part of Apple's supply chain. And if
| Intel made a serious profit, Apple would have replaced them
| anyways. It's an all-risk-no-reward scenario.
| klelatti wrote:
| > Intel at the time was fabless
|
| ????????!
| soulbadguy wrote:
| > Making smartphone SoCs is a huge profitable market that
| would have given Intel massive volumes to support R&D etc and
| slowed TSMC.
|
| Not quite, and that's the core of the "innovator" dilemma.
| The reason why intel passed on making SOC, is the same why
| nvidia decided to pass on getting into xboxss and
| playstations. Those market have much lower margin than server
| side business.
|
| AMD, and TSMC can operate in those space because they are
| much more efficient companies, intel is not.
|
| Intel got addicted to fat server CPU margin, grew
| inefficient...
| talldayo wrote:
| > Intel couldn't break into smartphone SoCs with clear process
| leadership and the financial strength to invest heavily. Why
| should it be able to break into other competitive markets today
| when it no longer has those advantages?
|
| Maybe because you're looking at it backwards? There must be some
| conspiracy among Softbank investors to portray everyone that
| doesn't pay for ARM licenses as a Muppet. Intel has no
| competitive advantage manufacturing licensed RISC CPUs when their
| desktop and server markets are hot, and even when x86 dies it
| will still be less attractive than owning fabs. With how bad the
| Cortex designs continue to be and how much leeway ARM gives
| licensees like Apple, even businesses like Qualcomm can't be
| bothered to take the ISA seriously. It feels disingenuous to say
| ARM is the panacea when designing ARM cores would have left Intel
| in an even worse position.
|
| Look at Samsung; a company people on this site would describe as
| misery incarnate, but chances are they're typing their comment
| from a device using Samsung-fabbed silicon. They're not even that
| competitive either; they just offer a cheap alternative to TSMC
| that offers OEMs an economically-minded option for less dense
| components.
| klelatti wrote:
| > It feels disingenuous to say ARM is the panacea when
| designing ARM cores would have left Intel in an even worse
| position.
|
| Sorry, where does the article say Intel should design Arm cores
| in 2024?
| 42lux wrote:
| They just slept for a decade... I had an overclocked i7 3970k
| (2012) and there was no need to update for 8 years. The
| perfomance increase was always marginal. I finally pulled the
| trigger when AVX happened.
| scotty79 wrote:
| I'm buying PCs for myself for three decades and never chose Intel
| for my build (apart from the Pentium 133 in my first pc
| computer). Every time I considered it but alternative was just
| significantly cheaper option at power levels that interested me.
|
| The only Intel's I got were in laptops where I was interested
| only about the price of a whole PC that interested me for other
| reasons than CPU.
|
| Since Internet became a thing I choose CPU by downloading price
| and benchmark results, drawing price/power chart and choosing a
| point on a good edge of a point cloud that's not unreasonably
| expensive. It never happened to be Intel.
| projektfu wrote:
| I think referring to bankruptcy is excessive. Intel bulls might
| be going bankrupt, but the share price doesn't affect most
| companies unless their only reasonable way to raise money is
| through the stock market. The Intel bond yields are not too high
| yet.
|
| The question then becomes, is this a low (trading around book
| value) with recovery in the future, or is it the beginning of a
| long, dark time, leading to an inability to raise money down the
| line?
| theparanoid wrote:
| Intel's decline was obvious when I worked for them in 2010, their
| compensation package wasn't competitive with FANG and they
| consistently missed out or lost their better engineers.
| gumby wrote:
| The "paradox of the x86" is simply the classic Innovator's
| Dilemma. In fact the history of the last 20 years against ARM
| could be a case study right out of the book.
|
| Worse for Intel, AMD flubbed it time and time again, but now
| Intel is too weak to defend against a resurgent AMD.
|
| Meanwhile they are cutting headcount deeply yet only now
| suspended the dividend. Madness!
|
| They should take a page out of AMD's book and spin off fab. That
| new company can then be flooded with "strategic" government aid,
| and maybe the rest of intel can catch up (or not) but it would at
| least give the shareholders a better chance. Right now the
| combination is acting like two anchors tied together.
| sargun wrote:
| I feel like Intel has embarked upon a bunch of good ideas like
| SDI (https://www.intel.com/content/www/us/en/manufacturing/softwa
| ...), IoT / Edison (https://ark.intel.com/content/www/us/en/ark/p
| roducts/84572/i...), Silicon Photonics (https://www.intel.com/con
| tent/www/us/en/products/details/net...), and SDN
| (https://www.intc.com/news-events/press-
| releases/detail/640/i...).
|
| But, they manage to fail to capitalize on any of these. What's
| wrong with them?
___________________________________________________________________
(page generated 2024-08-08 23:02 UTC)