3 ms·

I am not convinced that Intel can win here. They seem to not succeed with home grown GPU tech and other big bang approaches. Now if they were to acquire decen

I am not convinced that Intel can win here. They seem to not succeed with home grown GPU tech and other big bang approaches. Now if they were to acquire decent GPU tech then I would bet on them. Just the homegrown route seems to not work out for them. I suspect part of the reason is the long time frames for dev of this tech. I suspect it is at least 2 years for this to see the light of day. That is forever in this space. Intel failed with Larrabee and itanium. Maybe this will go better?

mtgx 9y ago
It looks like Raja will lead the development of machine learning-focused GPUs. Isn't this Intel basically admitting that their Xeon Phi, Nervana, and Altera GPU efforts to win the machine learning market are all a dead-end? How many machine learning strategies is Intel going to try? Does it even know what it's doing? Spending billions of dollars left and right on totally different machine learning technologies kind of looks like it doesn't, and it's just hoping it will get lucky with one of them. And even if you think that's not a terrible strategy to "see what works", there's still the issue that they need to have great software support for all of these platforms if they want developer adoption. The more different machine learning strategies it adopts, the harder that's going to be for Intel to achieve.
- ironchef253 9y ago
  Intel needs a CEO that skates to where the puck is going to be, not where it was three goals ago.
  - keganunderwood 9y ago
    That's scary of that's what is required of a ceo. You would either need to be an oracle and predict where the industry will go or you'd need to make the industry go the direction you're taking the company.
- stochastic_monk 9y ago
  I do love AVX512 and FMA instructions for CPU code with lots of branching. But that's not definitely going to cut it for the pure almost solely linear algebra workloads of most deep learning. I do have high hopes for their memristor initiative, but that's got to be years out. The GPU move is smart for Intel.
- dirtyaura 9y ago
  I think there is more to AI and ML than deep learning in which TPUs and GPUs are obvious choice. But I bet that branching instructions (various variants of search) still play a big role when you go beyond classifiers to reinforcement learning etc. so there is need for other architectures beyond GPUs.
deepnotderp 9y ago
Titanium was really more of a failing of core technology than a failing of execution whereas Larrabee was much more of a failing of execution than core tech, and the issue with Larrabee was really the fact that software GPU functions were not only slower in speed, but even slower to develop (!).
- bhouston 9y ago
  Interesting. I guess I view homegrown tech just really risky if you try a new approach. Success isn't assured here unless you just mostly replicate with a just small little spin on top. Maybe they are building upon AMD's core tech based on that other licensing deal? If so I would bet on them succeeding.
- blattimwind 9y ago
  Itanium was an all-around failure, including idea, execution, management, and ecosystem.
- andreiw 9y ago
  I thought Itanic was just a really smart diversionary tactic to kill off HP-PA (HP), MIPS (SGI) and Alpha (DEC/Compaq) development. It worked, and it didn't matter that Itanic was itself on borrowed time. The later is further proof that if you sell to dinosaurs, you won't survive the Big One (in this case, x86 growing up courtesy of AMD and Arm spending 15 years washing the footing underneath both by moving personal computing to mobile devices). This should be a big warning sign to the OpenPower guys. You need to start small and scale/price up, not the other way around.
andreiw 9y ago
Intel also utterly failed in the mobile space, IoT things and are setting themselves up for failure in edge computing. What's one more thing to fail at?

mtgx 9y ago

It looks like Raja will lead the development of machine learning-focused GPUs. Isn't this Intel basically admitting that their Xeon Phi, Nervana, and Altera GPU efforts to win the machine learning market are all a dead-end? How many machine learning strategies is Intel going to try? Does it even know what it's doing? Spending billions of dollars left and right on totally different machine learning technologies kind of looks like it doesn't, and it's just hoping it will get lucky with one of them. And even if you think that's not a terrible strategy to "see what works", there's still the issue that they need to have great software support for all of these platforms if they want developer adoption. The more different machine learning strategies it adopts, the harder that's going to be for Intel to achieve.

ironchef253 9y ago
Intel needs a CEO that skates to where the puck is going to be, not where it was three goals ago.
- keganunderwood 9y ago
  That's scary of that's what is required of a ceo. You would either need to be an oracle and predict where the industry will go or you'd need to make the industry go the direction you're taking the company.
stochastic_monk 9y ago
I do love AVX512 and FMA instructions for CPU code with lots of branching. But that's not definitely going to cut it for the pure almost solely linear algebra workloads of most deep learning. I do have high hopes for their memristor initiative, but that's got to be years out. The GPU move is smart for Intel.
dirtyaura 9y ago
I think there is more to AI and ML than deep learning in which TPUs and GPUs are obvious choice. But I bet that branching instructions (various variants of search) still play a big role when you go beyond classifiers to reinforcement learning etc. so there is need for other architectures beyond GPUs.

deepnotderp 9y ago

Titanium was really more of a failing of core technology than a failing of execution whereas Larrabee was much more of a failing of execution than core tech, and the issue with Larrabee was really the fact that software GPU functions were not only slower in speed, but even slower to develop (!).

bhouston 9y ago
Interesting. I guess I view homegrown tech just really risky if you try a new approach. Success isn't assured here unless you just mostly replicate with a just small little spin on top. Maybe they are building upon AMD's core tech based on that other licensing deal? If so I would bet on them succeeding.
blattimwind 9y ago
Itanium was an all-around failure, including idea, execution, management, and ecosystem.
andreiw 9y ago
I thought Itanic was just a really smart diversionary tactic to kill off HP-PA (HP), MIPS (SGI) and Alpha (DEC/Compaq) development. It worked, and it didn't matter that Itanic was itself on borrowed time. The later is further proof that if you sell to dinosaurs, you won't survive the Big One (in this case, x86 growing up courtesy of AMD and Arm spending 15 years washing the footing underneath both by moving personal computing to mobile devices). This should be a big warning sign to the OpenPower guys. You need to start small and scale/price up, not the other way around.

andreiw 9y ago

Intel also utterly failed in the mobile space, IoT things and are setting themselves up for failure in edge computing. What's one more thing to fail at?