[Hexus] ARM compares A72 vs Intel Broadwell-Y

[Systems analyst]

Some 12 billion ARM chips were shipped in 2014, and 3.5 billion in the first quarter of this year.
ARM expect to have 20% of the server market by 2020.

 

[Systems analyst]

ARM and partners, and Intel are in competition; there is a titanic struggle underway, it seems to me. This is evidenced by Intel's mobile losses of 4.2 billion USD in 2014, and losses of over 3 billion USD in 2013. Their PC/mobile group may make much reduced profits this year, due to a downturn in the PC market, and continuing mobile losses. Intel are also trying to enter the nascent IoT market; this is a second area of competition between Intel, and ARM and partners.

 

[imported_ats]

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).

Your numbers are pretty much immaterial. No one is making an intentioned choice to buy 90+% of those ARMs and no one is making much actual money from those ARMs. The vast majority of those fall into the category as cheaper than free.

 

[CHADBOGA]

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

What was the form factor these ARM chips were in when tested against the Core M?

 

[NTMBK]

What was the form factor these ARM chips were in when tested against the Core M?

Strapped to a heatsink/fan on a prototype board, of course Certainly can't judge until we get actual hardware, I agree.

 

[imported_ats]

Some 12 billion ARM chips were shipped in 2014, and 3.5 billion in the first quarter of this year.
ARM expect to have 20% of the server market by 2020.

No, ARM thinks that they might have 20% of the server market by 2020 if everything goes absolutely according to plan. AKA, they want 20%. What they'll get...

And no one actually expects them to have 20%...

 

[ShintaiDK]

Strapped to a heatsink/fan on a prototype board, of course Certainly can't judge until we get actual hardware, I agree.

The ARM chip doesnt even exist outside simulation yet

 

[Systems analyst]

No, ARM thinks that they might have 20% of the server market by 2020 if everything goes absolutely according to plan. AKA, they want 20%. What they'll get...

And no one actually expects them to have 20%...

ARM work with many partners; I assume that their expectation is a joint effort at seeing into the future.

 

[dahorns]

The ARM chip doesnt even exist outside simulation yet

Who knows, that simulated chip may have been running on an Intel Xeon

 

[TuxDave]

Who knows, that simulated chip may have been running on an Intel Xeon

I'm pretty sure they designed the ARM chip on a Xeon.

 

[Systems analyst]

The ARM chip doesnt even exist outside simulation yet

see
http://www.anandtech.com/show/9057/mediatek-at-mwc-2015-a72-already-in-silicon

 

[Madpacket]

One thing to consider is many more companies developing chips or SoCs using ARM tech where in x86 land you have only two players left. This in itself gives ARM an advantage over x86 where they'll be a better fit for emerging markets, quicker turnaround for IOT etc.

I don't think we'll see the day where ARM on the desktop replaces X86 but that market will pale in comparison to the rest of the markets.

 

[Sweepr]

see
http://www.anandtech.com/show/9057/mediatek-at-mwc-2015-a72-already-in-silicon

ARM is using a hyphothetical 16nm FF+ 4C Cortex A72 @ 2.5GHz in the comparison.

 

[erunion]

Some 12 billion ARM chips were shipped in 2014, and 3.5 billion in the first quarter of this year.
ARM expect to have 20% of the server market by 2020.

It'd be interesting to see what ARMs internal projections are now.

20% was before the demise of calxeda and sea micro.

 

[Systems analyst]

ARM is using a hyphothetical 16nm FF+ 4C Cortex A72 @ 2.5GHz in the comparison.

True, but the Mediatek chip exists in silicon, and that is incorporated into a board for the tablet form-factor. Shintai made a blanket statement, contradicted by Anandtech's article.

 

[Systems analyst]

It'd be interesting to see what ARMs internal projections are now.

20% was before the demise of calxeda and sea micro.

Not so. See the latest ARM roadshow presentation, at www.arm.com

 

[erunion]

Not so. See the latest ARM roadshow presentation, at www.arm.com

Whether ARM cares to update its public number or not, the prediction does predate the demise of Calxeda.

Like I said, it be interesting to see ARMs current internal estimate.

 

[soresu]

Considering how far ahead ARM has been collaborating with TSMC on developing nodes, I would be surprised if they havent already had test wafers of A72 from TSMC 16nm nodes since they announced it as Maia last year. I remember hearing about the Cortex A57 tests (20nm) much more than a year ahead of production, and ARM seem to be improving their time to market from announcement too. As others have said 28nm versions are already in testing - likely including SD 618/620, so 16nm versions are likely only waiting for yield maturity rather than basic chip design.

Hopefully the 16nm processes are good, because Kirin 950 sounds very good on paper (if probably lacking high core count for Mali T880).

Did anyone notice the new version of Qualcomms Hexagon DSP on the list for Hot Chips? Very likely to be in SD 820 I assume.

 

[kimmel]

True, but the Mediatek chip exists in silicon, and that is incorporated into a board for the tablet form-factor. Shintai made a blanket statement, contradicted by Anandtech's article.

The thread is about the numbers presented in the slides which were simulations not the silicon. The health of the silicon in the mediatek chips is unknown.

 

[Systems analyst]

Whether ARM cares to update its public number or not, the prediction does predate the demise of Calxeda.

Like I said, it be interesting to see ARMs current internal estimate.

ARM's 2013 Q4 roadshow says that they expect 10% of the server market in 2017.
ARM's 2014 Q4 roadshow says that they expect 20% of the server market in 2020.

 

[FwFred]

...

 

[ShintaiDK]

True, but the Mediatek chip exists in silicon, and that is incorporated into a board for the tablet form-factor. Shintai made a blanket statement, contradicted by Anandtech's article.

The MediaTek chip isnt produced using 16FF+ nor running at 2.5Ghz while not being a quadcore A72 as ARM. The chip ARM uses in its compare is simply a chip that doesnt exist yet. And who knows if we even gonna see one with such specs.

 

[simboss]

Your numbers are pretty much immaterial. No one is making an intentioned choice to buy 90+% of those ARMs and no one is making much actual money from those ARMs. The vast majority of those fall into the category as cheaper than free.

ARM results can hold the comparison with Intel's legendary margins, and they do so by selling "for free".
This + the fact people don't care about the CPU in their hands anymore should tell you a lot about how much Intel is threatened. CPU design is not something you can charge 100s of $ anymore (at least from a consumer perspective), you are lucky if you can charge 20$ for the finished product.

 

[ShintaiDK]

ARM results can hold the comparison with Intel's legendary margins, and they do so by selling "for free".
This + the fact people don't care about the CPU in their hands anymore should tell you a lot about how much Intel is threatened. CPU design is not something you can charge 100s of $ anymore (at least from a consumer perspective), you are lucky if you can charge 20$ for the finished product.

The A8 already cost Apple 37$ just in pure manufactoring without R&D. In terms of Samsungs SoC in the S6 its 29.5$, again without R&D. Just add a 50% margin on top and see where you land.

ARM isnt really so cheap as its often portraited when performance matters. It also shows on Qualcomm for example.

 

[imported_ats]

ARM results can hold the comparison with Intel's legendary margins, and they do so by selling "for free".
This + the fact people don't care about the CPU in their hands anymore should tell you a lot about how much Intel is threatened. CPU design is not something you can charge 100s of $ anymore (at least from a consumer perspective), you are lucky if you can charge 20$ for the finished product.

No, you can still charge 100s and 1000s of dollars for CPU design just like you always have been. The reality is that for 90+% of the devices containing ARM CPUs it doesn't matter what is in them. They are only ARM because ARM is cheaper than everything else. Many of the devices are using cores that are almost free at this point. They are using cores that were designed a decade or more ago.

That's actually a market that ARM is very much at risk of losing in the future. Cause when the value of the CPU isn't there, you use the cheapest thing possible. As people start of open source RTL for RISC-V like designs and the tool flows start to support them, there will be little point in actually paying any licensing fees. And that part of the ARM market share will evaporate. And ARM would be massive fools if they don't see it coming.