Support NeoGAF

[LordOfChaos]

http://www.datacenterknowledge.com/...rm-servers-it-expands-its-data-center-network



Very interesting stuff. A lot of people used to the dominance of Intel had a lot of arrogance on this topic, saying ARM couldn't scale to such uses, but there was never anything in the ISA itself limiting such expansion, it's clean and sane and scalable. The licensing setup is also preferable for more companies to enter and build their own large cores around it.

Cloudfare says they're so much more efficient that it would be worth it even if Intel gave their processors for free, for their server use. Fascinating stuff.

 

[theclaw135]

It does seem like Intel's aging architecture is going to bite back eventually.

 

[Ancient Dragon]

I don't understand much of the technical terms but having better security is something I'll always support.

 

[shira]

ARM is some crazy efficient shit.

 

[ickythingz]

I feel as though Intel has been dropping the ball over the past few years. I could see a future where Intel goes the way of the dodo. There is nothing in their pipeline right now that I am aware of that is exciting or interesting. Also, If Nvidia keeps up what they are doing Intel could very well be useless sooner than later.

 

[LordOfChaos]

I feel as though Intel has been dropping the ball over the past few years. I could see a future where Intel goes the way of the dodo. There is nothing in their pipeline right now that I am aware of that is exciting or interesting. Also, If Nvidia keeps up what they are doing Intel could very well be useless sooner than later.

Feels like four years ago when Intel was making plans for future R&D and capital investment under the assumption that their only competition would always be their own installed base of n-1 generation CPUs. Under that assumption, why depress the earnings report by working hard on the next big things when you can milk it.

Then ARM revved up, and AMD too now has a uArch that largely converged with Intel and will be hard to shake off for years.

 

[theclaw135]

Isn't x86 still 100% completely binary, assembly, and even raw instruction set compatible with the 8086?

 

[LordOfChaos]

Isn't x86 still 100% completely binary, assembly, and even raw instruction set compatible with the 8086?

Only on a 32 bit system, 16 on a 64 gets complicated

https://www.xylos.com/blog/en/2012/04/legacy-16-bit-applications-on-64-bit-operating-systems/

 

[blu]

Only on a 32 bit system, 16 on a 64 gets complicated

https://www.xylos.com/blog/en/2012/04/legacy-16-bit-applications-on-64-bit-operating-systems/

You can still run 16-bit protected-mode code on an amd64. What gets complicated is real-mode code, but that's not a matter of ISA anyway. So yes, an intel etched today and running a 64-bit OS is still 8086-compatible by ISA.

 

[camelCase]

Lol killa watt

 

[Shifty]

Makes sense for a company that's all about about distributing load across a worldwide network of servers. If you care more about uptime than pure computational grunt, it looks like ARM is the way.

 

[DonMigs85]

Even Apple is rumored to be making their own ARM-based processor for future Macs. The days of x86 could be numbered, but it'll probably last over 10 more years

 

[Shifty]

The days of x86 could be numbered, but it'll probably last over 10 more years

ARM needs to close the processing power gap in order to really succeed x86. As of right now I think there's only one 'desktop' ARM machine that you can buy and it's more of a mini PC.

10+ years seems like a sensible figure.

 

[Horsemama1956]

arm doesn't scale up and x86 doesn't scale down. In this case arm does what they need. If it was more computing power they needed they wouldn't be making the switch.

Seems odd to almost attack an ex partner though. Just admit it's a cost savings thing.

 

[LordOfChaos]

Even Apple is rumored to be making their own ARM-based processor for future Macs. The days of x86 could be numbered, but it'll probably last over 10 more years

I'm excited for this. They go head to head in only one area right now, fanless, and A11/A10X fare so well against Core M that they're often pushing past the fanless Intel range into the actively cooled U series. What they could do with 15, 28, 45 actively cooled watts will be interesting to see.

Question is though how much they invest in making models for lesser selling models. 12", 13" Pros, up to 15", those are a given. What about the iMac Pro? Will they make a replacement for an 18 core Xeon W for a successful but niche machine?

And if Apple whole bananas the silicon, for gods sake they can't just forget to refresh it for 5 years...

arm doesn't scale up and x86 doesn't scale down. In this case arm does what they need. If it was more computing power they needed they wouldn't be making the switch.

Why do you think ARM doesn't scale up? What in the ISA would limit it?

Lack of big iron ARM cores are a matter of market demands for now, there's no inherent limitation to the ISA scaling up to big x86 cores.

 

[DonMigs85]

The IPC and Geekbench scores for Apple's designs are already insane. They can actually rival some midrange desktop Skylake and above processors. Just think what could be done with a laptop-sized chip.

 

[Musky_Cheese]

Shots Fired?

 

[LordOfChaos]

The IPC and Geekbench scores for Apple's designs are already insane. They can actually rival some midrange desktop Skylake and above processors. Just think what could be done with a laptop-sized chip.

It's why I'm bullish on the Apple ARM switch. They only have fanless sub 5W designs and those already creep past Core M into U series range. Now imagine them designing around 15W, 28W, 45W...

No guarantee the advantage would scale when hitting higher power, but it allows so much more freedom, we already see that Apple is perfectly happy to spend on die area because they know each SoC is sold with a high profit margin device wrapped around it and in large volumes, so they could go larger in the same power draw and win that way, maybe bring the big.little along with little cores handling OS stuff, throw more die area at their custom GPU than Intel does on IGPs (where they're already beating Core M class GPUs), etc etc.

Plus, by 2020 we should be at a fab node density convergence, so the timing is good. Probably why they chose it.

My main question is how much they invest into lower selling models. The Macbooks all sell well, but would they make a massively multicore ARM chip just for the iMac Pro, would they make dedicated GPU silicon as well, etc.

 

[Panajev2001a]

ARM needs to close the processing power gap in order to really succeed x86. As of right now I think there's only one 'desktop' ARM machine that you can buy and it's more of a mini PC.

10+ years seems like a sensible figure.

Right now, iPhone X, Apple has a phone with their fully custom CPU getting close to Mac Book Air performance. Take that, allow voltage and frequency to scale in a much bigger and better for heat dissipation enclosure and with a much much bigger battery... and let's allow it to sink in... .

 

[Panajev2001a]

It's why I'm bullish on the Apple ARM switch. They only have fanless sub 5W designs and those already creep past Core M into U series range. Now imagine them designing around 15W, 28W, 45W...

No guarantee the advantage would scale when hitting higher power, but it allows so much more freedom, we already see that Apple is perfectly happy to spend on die area because they know each SoC is sold with a high profit margin device wrapped around it and in large volumes, so they could go larger in the same power draw and win that way, maybe bring the big.little along with little cores handling OS stuff, throw more die area at their custom GPU than Intel does on IGPs (where they're already beating Core M class GPUs), etc etc.

Plus, by 2020 we should be at a fab node density convergence, so the timing is good. Probably why they chose it.

My main question is how much they invest into lower selling models. The Macbooks all sell well, but would they make a massively multicore ARM chip just for the iMac Pro, would they make dedicated GPU silicon as well, etc.

I think for a laptop kind of scenario even taking their top end iPhone or iPad Pro CPU and push frequency higher would work: they can afford to even get the voltage a tiny bit higher to give them frequency headroom (they have so much more battery to play with and a fan to help dissipate heat in an already well conducting aluminium body... that gives them a lot of headroom IMHO).