[Hexus] ARM compares A72 vs Intel Broadwell-Y

[Fjodor2001]

Also, 16 core ARM vs x86 Xeon price is what at similar performance?

 

[Lonyo]

*For multi-threaded workloads Core M WILL be thermally limited and not able to reach maximum target frequency

So there. Fact.

 

[videogames101]

yeah, IPC will certainly be close with final RTL, but clockspeeds and power? hell no

 

[Redentor]

No its not. The X-Gene showed exactly what was wrong. It was severely lacking hardware performance and performance/watt.

PayPal will use X-Gene for its own servers
http://www.theregister.co.uk/2015/04/29/aookied_micro_q4_2015_results/

 

[ShintaiDK]

PayPal will use X-Gene for its own servers
http://www.theregister.co.uk/2015/04/29/aookied_micro_q4_2015_results/

You can also find previous users of ARM servers. It just didnt change anything because any useful metrics are in those ultra niche scenarios.

The numbers in the article says everything.

Gopi's optimism needs to be tempered a little by the company's results: consolidated net revenue for the quarter was US$37.0 million, producing a GAAP net loss of $15.1 million. For the year, the company can point to $156m of revenue and a $55m loss.
Not nice numbers, to be sure, but Gopi sees lots of upside now that he can point to PayPal and others buying X-Gene. Indeed, the company said it's now shoved 10,000 of the CPUs out the door.

 

[Redentor]

You can also find previous users of ARM servers. It just didnt change anything because any useful metrics are in those ultra niche scenarios.

The numbers in the article says everything.

Everything starts from small numbers.

Even Mobile market.

 

[Guest1]

Also, 16 core ARM vs x86 Xeon price is what at similar performance?

More like 16 core ARM to 16 core Atom servers since the performance is roughly the same.

 

[soresu]

Its a bit of a strange comparison considering that it lacks full system details, such as GPU architecture and configuration which can vary so wildly in ARM based chips, depending on the process and chip house preference.

I think they may be gunning for some of the same high end tablet and chromebox type form factors that Intel is currently occupying.
If the Android app on Chrome initiative continues to improve, this could actually favor A72 significantly because of the high share of native ARM code in Android apps and games. On the small chance that nVidia ever make their own chromebook/box, how likely is it that their 'TegraZone' exclusives (Crysis 3/Half-Life 2) will play well on an intel chip when compatibility with translated ARM code is still not so hot?

I could see the next Chromebook Pixel having an ARM processor if they continue to improve and refine their designs.
​

 

[PPB]

So there. Fact.

And that quote leaves me with this tought. Intel obviously knew sustained workloads would be Core M's bane. OEMs knew this too. And still insist on some design decision I cant understand. Even the new Zenbook could use just a tiny blower to help the temps (I know it could help diverting that awful noise some ultrabook made while processing that reminded me of coil whine, being in absolute silence and hear that whinny sound is really bad).

This leaves me thinking 4.5W TDP was an stretch all along for these chips and the TDPup of 6.0W according to ark.intel would have been a better target for cooling capacity: at least it would have forced into using active cooling which I think is still mandatory for these CPUs. We arent quite there with fanless ultrabooks and the Zenbook's 85-90c under gaming is only showing of this.

 

[ShintaiDK]

Like ARM doesnt throttle. But I guess the RTL to the rescue again

 

[AtenRa]

but that given the current state of the 14nm processes on which this would be manufactured, there's simply no way for them to know how power-sippy the final, mass-produced products will be.

Why is that ??? TSMC 16nm FF+ specs are finalized and if im not mistaken there are already ARM Tapeouts on the TSMC 16nm FF. Also mass production of TSMC 16nm FF starts in Q3 2015. So at this point products power consumption will be known.

And Samsung/GloFo 14nm FF process is already in mass production and a lot of ARM tapeouts have already been out even from the GloFo plant.

 

[Tuna-Fish]

Why is that ??? TSMC 16nm FF+ specs are finalized and if im not mistaken there are already ARM Tapeouts on the TSMC 16nm FF. Also mass production of TSMC 16nm FF starts in Q3 2015. So at this point products power consumption will be known.

It takes more than that. Power consumption cannot be reliably predicted from the specs, and it's variable enough that single-run chips are not necessarily representative of the average that mass production yields.

The final clocks and power consumption won't be known until well after mass production has started. Accurate power consumption numbers for chips that can be sold in bulk takes statistics over thousands of chips.

 

[B-Riz]

Why do people think ARM is some magic bullet that is going to "kill" the Intel / AMD x86 market?

It is a RISC instruction set (http://en.wikipedia.org/wiki/ARM_architecture), nothing more, nothing less.

We live in an x86 world (ARM gets to have cell phones, whoop de do), and not much, if anything is going to change its dominance.

http://research.cs.wisc.edu/vertical/papers/2013/hpca13-isa-power-struggles.pdf

"Key Finding 11: It is the microarchitecture and design methodologies
that really matter."

Guess What? Intel is still king boss of pushing new nodes and micro-architectures for cpu's that actually do work for people.

(AMD has contributed much also, but their smaller scale and recent stumbles have left Intel to dominate these past years.)

If ARM RISC could have toppled CISC x86, it would have done it by now...

 

[ShintaiDK]

Another A72 article up:
http://techreport.com/review/28189/inside-arm-cortex-a72-microarchitecture

The Core M in question is also in the form of the Dell Venue Pro.

 

[soresu]

As to the argument about ARM not being a magic bullet, I do agree with this - it will certainly not displace x86/64 where it is strongest (Windows/legacy win32). But new markets and form factors have already and continue to come into play. Chrome OS devices are one of them as I mentioned (the NaCl based ARC may be more favourable to native ARM code given its market share in certain programs, e.g. Drastic DS emulator).
Another form factor is Android or other mobile OS that does the Ubuntu thing and gives us a full desktop when plugged into a USB-C dock.
ARM may not be the magic bullet that hits x86 from the front, but from the side as other platforms built on it eat into x86 Windows market share.
This is in my opinion a major reason why MS is pulling out all stops to get people to migrate to Windows 10, because the transition to higher performance ARM designs coupled with increasingly capable Android and Chrome platforms are threatening their own market share.

On the other hand, something like VISC will probably come along and disrupt everything anyway. Wouldnt that be great? A hypervisor running full ISA compatible Android when in mobile mode, and switching to Windows full fat when connected to video dock - thats dual boot id pay for.

 

[B-Riz]

As to the argument about ARM not being a magic bullet, I do agree with this - it will certainly not displace x86/64 where it is strongest (Windows/legacy win32). But new markets and form factors have already and continue to come into play. Chrome OS devices are one of them as I mentioned (the NaCl based ARC may be more favourable to native ARM code given its market share in certain programs, e.g. Drastic DS emulator).
Another form factor is Android or other mobile OS that does the Ubuntu thing and gives us a full desktop when plugged into a USB-C dock.
ARM may not be the magic bullet that hits x86 from the front, but from the side as other platforms built on it eat into x86 Windows market share.
This is in my opinion a major reason why MS is pulling out all stops to get people to migrate to Windows 10, because the transition to higher performance ARM designs coupled with increasingly capable Android and Chrome platforms are threatening their own market share.

On the other hand, something like VISC will probably come along and disrupt everything anyway. Wouldnt that be great? A hypervisor running full ISA compatible Android when in mobile mode, and switching to Windows full fat when connected to video dock - thats dual boot id pay for.

All in all, I hold the view that ARM compliments the x86 ecosystem.

It is never going to compete with it directly and will never replace it.

MS already won corporate computing, now they just want to lock down emerging markets / countries with low disposable income that have lots of people needing / wanting an inexpensive device. (That coincidentally COULD run old x86 software; bam, old software / games is still viable on that full Win 8.1 tablet)

 

[xthetenth]

Another A72 article up:
http://techreport.com/review/28189/inside-arm-cortex-a72-microarchitecture

The Core M in question is also in the form of the Dell Venue Pro.

Venue Pro's one of the weaker Core M implementations, isn't it?

 

[ShintaiDK]

Venue Pro's one of the weaker Core M implementations, isn't it?

Yep. Else ARM wouldnt use it

 

[dahorns]

Venue Pro's one of the weaker Core M implementations, isn't it?

According to the Anandtech Core-M review, yes. Also note, they are running against the 5y10c, which is capped at 2 ghz. They aren't comparing it to the 5y70 or 5y71, which turbo up to 2.9 ghz.

 

[TuxDave]

Another A72 article up:
http://techreport.com/review/28189/inside-arm-cortex-a72-microarchitecture

The Core M in question is also in the form of the Dell Venue Pro.

In the "Maximizing Throughput Density" slide on page 2, they seem to suddenly lose their ability to measure power and reverted back to ARK TDP numbers. It's especially convenient since SpecINT shouldn't come anywhere close to TDP.

 

[teejee]

Why do people think ARM is some magic bullet that is going to "kill" the Intel / AMD x86 market?

It is a RISC instruction set (http://en.wikipedia.org/wiki/ARM_architecture), nothing more, nothing less.

We live in an x86 world (ARM gets to have cell phones, whoop de do), and not much, if anything is going to change its dominance.

In my world Intel is a niche product (laptop, desktop, workstation and server) and ARM is the dominating ISA. 2013 about 10 billion CPUs with ARM cores was sold, that is about 20 times more than Intel. The growth is in new areas where Intel has almost no control or market share. So Intels dominance for CPUs in general does not even exist today, only in a few small segments (but very profitable segments though).

 

[Enigmoid]

In my world Intel is a niche product (laptop, desktop, workstation and server) and ARM is the dominating ISA. 2013 about 10 billion CPUs with ARM cores was sold, that is about 20 times more than Intel. The growth is in new areas where Intel has almost no control or market share. So Intels dominance for CPUs in general does not even exist today, only in a few small segments (but very profitable segments though).

The two segments are not remotely comparable. Intel ships high performance chips. ARM sells the vast majority of its chips for coffee makers and such.

Its like saying that company A dominates the medical industry because they sell 100 million bandages yearly while company B, selling 100 MRI machines is a bit player. The two companies sell to radically different market segments and have very different business strategies.

 

[NTMBK]

Venue Pro's one of the weaker Core M implementations, isn't it?

It's also the closest to a typical ARM tablet- 11" chassis, slim, no fan.

 

[teejee]

The two segments are not remotely comparable. Intel ships high performance chips. ARM sells the vast majority of its chips for coffee makers and such.

Its like saying that company A dominates the medical industry because they sell 100 million bandages yearly while company B, selling 100 MRI machines is a bit player. The two companies sell to radically different market segments and have very different business strategies.

But what happens now is that they intersect in certain segments, like tablets, home servers and micro servers.
And many companies envy Intels high profits. And x86 is not an option for them (except AMD).
For example ARM v8.1 is tailor made for servers (no SOCs available yet).
Intel is very strong, especially in process/production but competition from many directions is still tough.

 

[xthetenth]

It's also the closest to a typical ARM tablet- 11" chassis, slim, no fan.

And very dubious ability to dissipate heat in the long term. However yes, it is close on form factor even though I'd argue the issues are more peculiar to the implementation than the form factor.