ARM Profit Craters

[blackened23]

So Samsung and Apple should just dump ARM to become slaves to Intel solely because Intel has a better mobile chip.

Brillant! Both of them will be contacting you shortly for a top CXO position for your excellent business talents.

That's precisely why mobile manufacturers have used ARM for so many years - they had the better IP which led to more efficient chips. Efficiency is paramount for ultra portable devices, and intel did not deliver;ARMH had the better IP for efficiency. Until now.

Guess what. That's changing. And much to intel' haters chagrin, there will be a market for it if intel creates a better product with better performance/efficiency , period. And it will be quite funny to see ARMH's hubris catch up to them "Intel will never beat us in efficiency". Quite funny indeed.

 

[Arachnotronic]

The ISA with regards to Android is a non issue actually, the real showstopper is over at the business side of things where Intel hits a brickwall at every corner and the most consumers simply doesn't care about SoCs. The foundry business is a no-go either, even Apple isn't that stupid to bet their fate on a foundry with conflicting interests.

I think it's quite the opposite - Intel hasn't had a competitive SoC in this space yet, which is why Qualcomm has owned the merchant market without exception.

If people didn't care about SoCs, then Qualcomm would not be investing massive amounts of R&D to have performance leadership in this space, and it wouldn't be pressuring the foundries to get FinFETs into manufacturing ASAP.

 

[blackened23]

I think it's quite the opposite - Intel hasn't had a competitive SoC in this space yet, which is why Qualcomm has owned the merchant market without exception.

100% this. Intel hasn't had a competitive SOC, which is why everyone used ARMH' IP for their SOCs. This is changing. By all accounts intel will have better performance (not a surprise, they ALWAYS had way better performance) and will have better efficiency. This will shake things up and send ARMH back to the drawing board, and they will have to spend some of their cash for R+D to spice the game up.

Heck, the ULV Haswell is about as efficient as the best ARM SOCs. Now someone tell me why Bay Trail (in terms of efficiency) won't be similar with that in mind?

 

[Arachnotronic]

100% this. Intel hasn't had a competitive SOC, which is why everyone used ARMH' IP for their SOCs. This is changing. By all accounts intel will have better performance (not a surprise, they ALWAYS had way better performance) and will have better efficiency. This will shake things up and send ARMH back to the drawing board, and they will have to spend some of their cash for R+D to spice the game up.

Heck, the ULV Haswell is about as efficient as the best ARM SOCs. Now someone tell me why Bay Trail (in terms of efficiency) won't be similar with that in mind?

blackened23,

Nice post. Note that Haswell-ULT, while based on a 22nm SoC, still incorporates an on-package but off-die 32nm PCH. 22nm Silvermont not only has everything on a single chip, but it's also all 22nm, so there's some efficiency left on the table on the SoC side that gets added in.

But quite frankly, at the end of the day, process tech is what enables fundamental advances in micro-architecture/power/performance. Why do you think both the Radeon 7970 and GTX 680 are so close in performance and performance/watt? If Intel can leverage its process tech by bringing full SoCs based on the "big" core (not 2-chip kludges) as well as very good, leading-edge node Atom parts to this market, then I see no reason why it cannot have leadership in terms of compute and GPU performance/watt.

The main sticky areas I see for Intel is that Qualcomm has a notably better modem team, and it has more of the connectivity (WiFi, Bluetooth, etc.) in house and suitable for low power. It is my guess that while Intel will have a relatively easy time taking major share in tablets, phones may prove elusive for the time being due to these reasons.

Your thoughts?

 

[SiliconWars]

I will be completely upfront - I hate ARM hyperbole with a passion, and so it is with GREAT pleasure that I point out to the folks who were saying how "screwed" Intel was for reporting a 29% Y/Y profit drop (making "only $2B) that ARM's profits just took a complete nose-dive:

See that IFRS "profit before tax" line? ARM's profit before tax under IFRS (that is, they can't sweep share based compensation under the rug) is now LESS THAN A THIRD of what it was a year ago. The company made a whopping $23M this quarter *before* tax.

Guess that's what happens when you actually need to increase your R&D and marketing costs and try to hide it by paying everybody in overpriced stock to play in the big leagues, eh?

Intel17,

Let me educate you...

ARM had yet another great quarter. They took a large one-off charge that seriously affected the bottom line.

Revenue in Q2 2013 soared 26 percent year-over-year to £171.2 million ($264.3 million), while profit before tax was up 30 percent £86.6 million (in "normalised" terms).

http://forums.anandtech.com/showpost.php?p=35291265&postcount=151 :biggrin:

 

[Arachnotronic]

Intel17,

Let me educate you...

ARM had yet another great quarter. They took a large one-off charge that seriously affected the bottom line.

http://forums.anandtech.com/showpost.php?p=35291265&postcount=151 :biggrin:

SiliconWars,

Fair is fair, and I probably deserved that post

 

[SiliconWars]

100% this. Intel hasn't had a competitive SOC, which is why everyone used ARMH' IP for their SOCs. This is changing. By all accounts intel will have better performance (not a surprise, they ALWAYS had way better performance) and will have better efficiency. This will shake things up and send ARMH back to the drawing board, and they will have to spend some of their cash for R+D to spice the game up.

Intel will not have better performance. They have designed Atom for a lower performance target than they anticipated would be required, and graphically will be miles behind as usual. Efficiency wise they'll be ahead of S800 but probably not S600. The top end Bay Trail T is likely to be good competition for S600, in my opinion.

You only need to read the comments coming from each camp recently to see what the case is.

Krzanich at the Q2 call - "“Bay Trail really first and foremost we believe gives solid performance, solid battery life relative to the competition in price points and markets that were simply not in, in a big way today"

"Solid" performance, not class leading or devastating or anything like that - just "solid".

Some other Intel guy said that is was curious how Qualcomm were now talking about performance now that Intel was on equal or better power characteristics (can't find that link any longer).

If Intel was faster why not just dismiss it? Nobody at Intel is saying they have the performance lead.

Qualcomm are mostly talking about performance while saying power draw hasn't gone backwards.

 

[SiliconWars]

SiliconWars,

Fair is fair, and I probably deserved that post

Yep you did, but kudos for recognizing it.

 

[Idontcare]

What I DO know, is that Arm, give users of their Arm licences, the option to customize, as one of the benefits of Arm chips.

The option comes with a price, it isn't included in the basic licensing package.

ARM has quite a few tiers to their licensing structure, the one you are describing is the absolute top of the line "includes every option in the total package" tier.

ARM isn't ran by dummies. They know if you have the kind of money necessary to hire the headcount (1000's) for the time (years) needed to make your own customized ARM-compatible core then you have the money to pay for the absolute top-end licensing premium.

Regardless, your point still stands which is that neither AMD nor Intel offer this sort of customizable x86 core licensing option.

For instance, Apple couldn't buy a license to do to an Atom core what they can do to an A9 or A15 from ARM. Intel would want Apple to buy from Intel whatever Intel felt like designing and selling.

 

[Keysplayr]

http://techreport.com/review/25067/an-intro-to-all-things-arm

If anybody hasn't read this, they should. Quite interesting, and for the purposes of settling all arguments in this thread, quite informative.