6 ms·

So after years of an effective Intel monopoly I'm really glad to see AMD is back in a way that I don't think I've seen since the Athlon64/Opteron days. Back the

by cletus 7y ago

So after years of an effective Intel monopoly I'm really glad to see AMD is back in a way that I don't think I've seen since the Athlon64/Opteron days. Back then it was AMD who pushed the x86-64 instruction set when Intel was claiming EPIC was the future (ha). At this point, Intel's move to 10nm processes is an embarrassment. I'm sure it's a difficult problem but historically Intel has been reasonably good at planning process advancements but in the case of 10nm they've been off by years. I believe the original goal was 2017? And we're still not there yet. I would dearly love to see an honest postmortem of this and see what went wrong. Who made promises they'd miss by so much, why, what the issues were and so on. The last PC I built (because apparently I still do that, even though it annoys me no end) has an Intel 9700 in it. At the time that was probably the best choice. 6 months later and it would no doubt have been a Ryzen. I hope AMD keeps this up as we need the competition.

lend000 7y ago
They got comfortable with Moore's Law, which in fairness, had held for a long time, and continued to use the model that increasing feature density quadratically was a linear problem. Now, it turns out that once you get near the 10nm gate size range, the difficulty diverges from linear to exponential (and perhaps even higher eventually as some hard limit is approached). That, and Intel isn't an innovative company anymore. Now they are a process company riding on their manufacturing dominance and x86 market share. It looks a lot like Apple under Tim Cook, except add another decade since there was innovative leadership (Andy Grove). They are a few consultants removed from IBM at this point.
- thethethethe 7y ago
  > It looks a lot like Apple under Tim Cook, except add another decade since there was innovative leadership (Andy Grove) Ehh this is debatable. Apple has put out a few products that have completely changed the market under Cook's tenure. AirPods have introduced a new head phone paradigm. Apple Watch is waaaay ahead of the competition. The iPhone X made full screen phones mainstream and introduced UI gestures that were copied by Android. Sure there have been some missteps cough butterfly keyboard cough but I’d say overall, they are still producing interesting products that define a large part of the consumer tech market
  - Polylactic_acid 7y ago
    I'm not really sure what AirPods introduced. We had wireless headphones before. They might be the nicest ones and the most popular but I'm not sure how they created a new headphone paradime
    - thebean11 7y ago
      To be fair I haven't tried too many high end wireless earbuds, but switching between the noise cancellation and transparent mode on the AirPods pro was my first "wow" moment using a piece of technology in a long time.
      - Polylactic_acid 7y ago
        Thats a feature that is available on most high end headphones. I get that the airpods might be quite refined but they are nothing brand new.
        thebean11 7y ago
        The transparent mode is? That's definitely (good) news to me, I have an android phone so I'd like to avoid airpods
    - SparkleBunny 7y ago
      Latency! Lack thereof, actually. I bought an Ipad just to be able to use them. My second Apple product. The first was a Macintosh SE. That says something.
  - mntmoss 7y ago
    What I think Cook misses that Jobs got, and made for more exciting releases, is the idea of a totally integrated service. The iPod's victory was also a victory for iTunes. The iPhone was also the App Store. And when Jobs left, those kinds of distinct pairings did too. They are hard to concieve of, and to execute on. In contrast, the AirPods and Apple Watch are more straightforward "make 'em smaller" incremental moves. The engineering work is leading in many respects, but it doesn't upend a market. And Intel does have a history that was like Apple's in parts. A big part of their advantage as the PC market heated up was in marketing an entire nomenclature of what the platform could be and to provide comprehensive path-of-least-resistance solutions around that, ensuring that the industry fell in line around their technical lead rather than IBM or some competitor. Those bones are still there in parts of the company - Intel chipsets are pretty well regarded for dependability(seeing Windows crash because of Intel drivers is a very rare event) and they've been good at getting the corporate office to standardize on them - but increasingly the platform is getting defined around mobile and server needs, which are a more competitive space generally. Intel doesn't get to call the shots on 5G, for example - and huge data center customers are in the business of optimizing the system end-to-end to provide the most efficient general computing resource possible; everything they touch commoditizes, and they will put their foot down if they smell enterprise contract crap.
erichocean 7y ago
I wonder if AMD came up with the Epyc name just to troll Intel.
- KMag 7y ago
  For those who missed the joke, EPIC was Intel/HP's name for the Itanium ("Itanic") style of instruction set that attempts to moves most of the work for parallel instruction scheduling out of silicon and into the compiler. (And, HP later asserted that EPIC refers specifically to the Itanium instruction set, not just Itanium-like instruction sets.) It wasn't a terrible idea at a high level, but the implementation was terribly complex and power-hungry, needing huge caches to compensate for its low instruction density. They also bet heavily on compiler advances that never materialized or materialized later than expected. I can imagine an alternative history where Intel took the EPIC idea, but went more conservatively and focused on minimizing complexity and the total number of transistors, designing in such a way that allowed for complex power-hungry HPC optimizations later on, but didn't depend on them for the initial roll-out. This would have resulted in a lot of the potential of the idea for HPC being left on the table, but may have allowed them to have more initial success in the server market and would potentially allowed them to scale down to cell phones more easily than trying to scale Atom down to cell phones.
ksec 7y ago
>I believe the original goal was 2017? It depends how you define "Original". If it was the initial Tick Tock roadmap, 22 and 14 were Late, otherwise it was 2015. But then since both were 6 months late it was expected to be 2016, counting from initial 14nm launch. We are now 4 years later and 10nm is barely working and yield. Although there were lots of promise during investor meetings of more 10nm products this year, it seems Intel wants to move pass 10nm as early as possible to 7nm and regain their lead by 2023 with 5nm. But judging Intel's recent record I am a little skeptical of their claims.
paulie_a 7y ago
Intel has been boring for the last 15 years. They incrementally boost things but rarely do any innovation. Amd lb for lb has been whooping intel in most aspects since the 90's. Sure you can buy intel at 500dollars a chip or buy and at 200 a chip that is slightly less powerful. Hell Intel's i7 has finally caught up to the and fx octo 12 years later. Meanwhile the ryzen 8-16 core absolutely crushes anything intel has to offer regardless of price.
- ksec 7y ago
  >Intel has been boring for the last 15 years. Intel may have been boring for the past 5 years, but suggesting they have not innovate since 2005 completely ignore the majority of tick tock execution they achieved.
  - paulie_a 7y ago
    Name one innovation that has actually mattered Oh in 2005 or so they made rdram a verryyyyyyy slow thing Bang for buck you can go back to 386 dx, And has always been killing intel. Let's see I could buy an intel chip for a couple Grand and it cracks 10k on CPU cpubenchmark Or buy a thousand dollar amd chip and get 40 cpu bench mark I can't understand why anyone would ever buy intel And as for stock prices intel has definitely been boring. Amd is kicking the shit out of intel It's extremely predictable. Amd loses because of stupid shorttraders And they get burned. Meanwhile amd had risen 2000+ percent in five years. Intel not so much Edit for the downvoters please prove me wrong on anything. Amd is the winner Intel has been unimpressive for decades. Pick your horse because and is three Laos ahead

by cletus 7y ago

lend000 7y ago

They got comfortable with Moore's Law, which in fairness, had held for a long time, and continued to use the model that increasing feature density quadratically was a linear problem. Now, it turns out that once you get near the 10nm gate size range, the difficulty diverges from linear to exponential (and perhaps even higher eventually as some hard limit is approached). That, and Intel isn't an innovative company anymore. Now they are a process company riding on their manufacturing dominance and x86 market share. It looks a lot like Apple under Tim Cook, except add another decade since there was innovative leadership (Andy Grove). They are a few consultants removed from IBM at this point.

thethethethe 7y ago
> It looks a lot like Apple under Tim Cook, except add another decade since there was innovative leadership (Andy Grove) Ehh this is debatable. Apple has put out a few products that have completely changed the market under Cook's tenure. AirPods have introduced a new head phone paradigm. Apple Watch is waaaay ahead of the competition. The iPhone X made full screen phones mainstream and introduced UI gestures that were copied by Android. Sure there have been some missteps cough butterfly keyboard cough but I’d say overall, they are still producing interesting products that define a large part of the consumer tech market
- Polylactic_acid 7y ago
  I'm not really sure what AirPods introduced. We had wireless headphones before. They might be the nicest ones and the most popular but I'm not sure how they created a new headphone paradime
  - thebean11 7y ago
    To be fair I haven't tried too many high end wireless earbuds, but switching between the noise cancellation and transparent mode on the AirPods pro was my first "wow" moment using a piece of technology in a long time.
    - Polylactic_acid 7y ago
      Thats a feature that is available on most high end headphones. I get that the airpods might be quite refined but they are nothing brand new.
      - thebean11 7y ago
        The transparent mode is? That's definitely (good) news to me, I have an android phone so I'd like to avoid airpods
  - SparkleBunny 7y ago
    Latency! Lack thereof, actually. I bought an Ipad just to be able to use them. My second Apple product. The first was a Macintosh SE. That says something.
- mntmoss 7y ago
  What I think Cook misses that Jobs got, and made for more exciting releases, is the idea of a totally integrated service. The iPod's victory was also a victory for iTunes. The iPhone was also the App Store. And when Jobs left, those kinds of distinct pairings did too. They are hard to concieve of, and to execute on. In contrast, the AirPods and Apple Watch are more straightforward "make 'em smaller" incremental moves. The engineering work is leading in many respects, but it doesn't upend a market. And Intel does have a history that was like Apple's in parts. A big part of their advantage as the PC market heated up was in marketing an entire nomenclature of what the platform could be and to provide comprehensive path-of-least-resistance solutions around that, ensuring that the industry fell in line around their technical lead rather than IBM or some competitor. Those bones are still there in parts of the company - Intel chipsets are pretty well regarded for dependability(seeing Windows crash because of Intel drivers is a very rare event) and they've been good at getting the corporate office to standardize on them - but increasingly the platform is getting defined around mobile and server needs, which are a more competitive space generally. Intel doesn't get to call the shots on 5G, for example - and huge data center customers are in the business of optimizing the system end-to-end to provide the most efficient general computing resource possible; everything they touch commoditizes, and they will put their foot down if they smell enterprise contract crap.

erichocean 7y ago

I wonder if AMD came up with the Epyc name just to troll Intel.

KMag 7y ago
For those who missed the joke, EPIC was Intel/HP's name for the Itanium ("Itanic") style of instruction set that attempts to moves most of the work for parallel instruction scheduling out of silicon and into the compiler. (And, HP later asserted that EPIC refers specifically to the Itanium instruction set, not just Itanium-like instruction sets.) It wasn't a terrible idea at a high level, but the implementation was terribly complex and power-hungry, needing huge caches to compensate for its low instruction density. They also bet heavily on compiler advances that never materialized or materialized later than expected. I can imagine an alternative history where Intel took the EPIC idea, but went more conservatively and focused on minimizing complexity and the total number of transistors, designing in such a way that allowed for complex power-hungry HPC optimizations later on, but didn't depend on them for the initial roll-out. This would have resulted in a lot of the potential of the idea for HPC being left on the table, but may have allowed them to have more initial success in the server market and would potentially allowed them to scale down to cell phones more easily than trying to scale Atom down to cell phones.

ksec 7y ago

>I believe the original goal was 2017? It depends how you define "Original". If it was the initial Tick Tock roadmap, 22 and 14 were Late, otherwise it was 2015. But then since both were 6 months late it was expected to be 2016, counting from initial 14nm launch. We are now 4 years later and 10nm is barely working and yield. Although there were lots of promise during investor meetings of more 10nm products this year, it seems Intel wants to move pass 10nm as early as possible to 7nm and regain their lead by 2023 with 5nm. But judging Intel's recent record I am a little skeptical of their claims.

paulie_a 7y ago

Intel has been boring for the last 15 years. They incrementally boost things but rarely do any innovation. Amd lb for lb has been whooping intel in most aspects since the 90's. Sure you can buy intel at 500dollars a chip or buy and at 200 a chip that is slightly less powerful. Hell Intel's i7 has finally caught up to the and fx octo 12 years later. Meanwhile the ryzen 8-16 core absolutely crushes anything intel has to offer regardless of price.

ksec 7y ago
>Intel has been boring for the last 15 years. Intel may have been boring for the past 5 years, but suggesting they have not innovate since 2005 completely ignore the majority of tick tock execution they achieved.
- paulie_a 7y ago
  Name one innovation that has actually mattered Oh in 2005 or so they made rdram a verryyyyyyy slow thing Bang for buck you can go back to 386 dx, And has always been killing intel. Let's see I could buy an intel chip for a couple Grand and it cracks 10k on CPU cpubenchmark Or buy a thousand dollar amd chip and get 40 cpu bench mark I can't understand why anyone would ever buy intel And as for stock prices intel has definitely been boring. Amd is kicking the shit out of intel It's extremely predictable. Amd loses because of stupid shorttraders And they get burned. Meanwhile amd had risen 2000+ percent in five years. Intel not so much Edit for the downvoters please prove me wrong on anything. Amd is the winner Intel has been unimpressive for decades. Pick your horse because and is three Laos ahead