Print 24 comment(s) - last by fteoath64.. on Mar 4 at 3:15 AM

An employee shows off one of SeaMicro's pricey Intel Atom-based servers.  (Source: Venture Beat)
Intel says that ARM cores may gain some ground in the laptop space but will not challenge it in servers

Facing an onslaught of ultra-power efficient, highly clocked, multi-core ARM CPUs from the likes of QualcommNVIDIA, and Texas Instruments, Intel, the world's largest CPU maker, has plenty to worry about.

But it's seemed remarkably unconcerned; stating that its smartphone chips will equal ARM chips in power performance and beat them in computing power by later this year.  Now Kirk Skaugen, general manager of Intel’s data center unit has gone on the record at a Morgan Stanley conference to say that Intel's server business is in no danger from ARM.

Like mobile devices, power efficiency is a key concern for servers.  Businesses keenly watch how to minimize the power they use per unit of computing power.

But Intel is convinced that businesses won't be interested in ARM servers, as it admits that its own efforts to put Atom systems-on-a-chip on servers was met with disappointment.  States Mr. Skaugen:

We’ve been out talking about Atom and servers for... And candidly, there hasn’t been a lot of interest in that architecture in a broad sense. I could see if you go out four to five years maybe 10% of the total market, give or take a couple percent, could be interested in such an architecture.

He backtracks a bit acknowledging that for some customers a low-power solution like Atom (or ARM) would be a good solution.  And he says it’s a tempting one to try to sell as the margins are bigger.  He explains:

What SeaMicro has done is they’ve put 512 Atoms into a 10U form factor. So if everybody in the world took a Xeon and bought an Atom because their servers were underutilized that would be a bad thing for Intel and our OEMs. That’s not what we hear from the customers on what they’re interested in doing. What they’re interested in doing is getting, for example, for dedicated hosting — let’s say they have $140,000 to spend, they’re wanting to know how many hosting nodes or how many customers they can host. So what SeaMicro has done is they’ve said, hey, I’m going to sell a $148,000 Atom server, they put 512 Atoms into a 10U and they say you can buy either I think 89 one-socket Xeons for the same price, 1U pizza box machines, so you can buy 89 1U’s or you can buy the single system which has 512 nodes in it…

Atom makes good margin for Intel; if the workload actually works that’s incredibly good.

So if some customers are interested, why does he see them picking Atom over ARM cores?  He says compatibility is the key issue:

Now what’s the challenge that ARM has in that same form factor? Well, it has an instruction set issue. So if you’re going to do hosting what application do you host? And what is an application porting effort — we did application porting with Itanium, it took us about 10 years. Hundreds and hundreds of millions of dollars to port about 14,000 applications. ARM has to port for hosting all those applications over. Second challenge is the A9 and the A15 as we know it are 32-bit processors. Microsoft only supports 64-bit operating systems today. So I’d encourage you to go ask Microsoft what their position is on 32-bit operating systems. But I think they’re pretty firm on their 64-bit. So it’s an instruction set issue as well as a 64-bit issue. Everything we do in servers for real servers will be 64-bit.

Of course, he's largely right on both points -- low computing power chips will have trouble gaining traction with many server customers, and there are major compatibility issues holding back ARM in both the laptop and server markets.  

But with Microsoft migrating its core code base to natively support ARM, that could be changing.  Holding up compatibility as a barrier to embracing a superior hardware product is valid to some degree, but in the long run such barriers are invariably eliminated.  Like it or not, Intel may have more of a fight on its hands in a number of markets than it's willing to admit, and its rosy outlook for the smartphone industry may be met with a dose of reality if it proves unable deliver the superior smartphone chips its promises to be shipping by the end of the year.

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

Atom... lol
By ChugokuOtaku on 3/1/2011 4:35:12 PM , Rating: 4
When you can't compete, criticize!

RE: Atom... lol
By TMV192 on 3/1/2011 4:58:13 PM , Rating: 1
And/or buy exclusivity agreements like they did the GoogleTV

RE: Atom... lol
By Jeffk464 on 3/1/2011 10:45:42 PM , Rating: 1
Exactly week, limited, and perfectly appropriate for smart phones. Which is more than can be said for the Atom.

RE: Atom... lol
By Mitch101 on 3/2/2011 11:23:39 AM , Rating: 2
I would say most limitation are still slightly GPU bound on the portable device end (Me Personally) but that will be short lived. Phones fine but tablets need just a touch more GPU power for me. For ARM to get into the laptop/desktop area it just needs an OS with more than tablets/phones currently deliver but its not far off.

Quad core ARM cpu's are just about reality and NVIDIA tegra 3 will be more power than needed on any phone and the perfect spot for tablets even if/when Apple's iPad3 ups the resolution with the retina display and its just evolutionary for them to get into laptops and desktops and with Microsoft going to support ARM this certainly threatens the x86 base of laptops/desktops.

Slap a quad core ARM with a Microsoft OS and a desktop GPU and you have a monster and actually an advantage because you dont need to back support 16 and 32 bit applications like current gen Windows does.

Why does Microsoft matter? Say what you will Microsoft makes awesome development tools for the programmer that makes it an easy platform to develop for which is crucial to adoption.

I can see why Intel is mouthing off. They have nothing for Phones. They dont have a good CPU for tablets and they defiantly dont have a GPU to leverage. But since Intel never existed in those markets they haven't lost anything. You have Microsoft backing ARM which is major and then ARM quads are plenty for Laptop and Desktops which is ARM starting to cut into Intel Marketshare. Dont forget AMD with fusion/llano cutting into the bottom end and sandy bridge. Finally if AMD bulldozer delivers Intel has competition across the board.

In a way Intel is a one trick pony with x86 and if ARM can become a competitor to x86 in the consumer segment of laptops and desktop then that leaves Intel with server segments. Hmm this could be interesting.

The more I think about it the more we could have a race between x86 and ARM in the near future on the consumer side.

RE: Atom... lol
By omnicronx on 3/2/2011 11:45:00 AM , Rating: 2
I would say most limitation are still slightly GPU bound on the portable device end (Me Personally) but that will be short lived.
Not really, the GPU is hardly the bottleneck, in fact GPU's such as the SGX540 which can push some 90 million+ pixels never come close to that in practice because of other system bottlenecks.

Even the gpu power of Nvidias Tegra lines have yet to be truly harnessed.

Some good points, some crazy
By SKiddywinks on 3/1/2011 6:48:08 PM , Rating: 1
I think he has a point in ARM having trouble making a large dent in Intel's server business (I'm sure in some cases ARM will be the better decision and will get some business, just not loads of it).

But I also think he is smoking some awesome shit if he thinks an x86 CPU is ever going to beat ARM in efficiency. Ok, maybe not ever (future manufacturing processes might make the x86 issues a moot point, and then you are just left with more performance than ARM), but I definitely do not see them beating ARM chips, which are designed from the ground up for smartphones, in anything resembling the near future.

RE: Some good points, some crazy
By sxr7171 on 3/1/2011 9:56:08 PM , Rating: 2
I don't think I know the real story, but what I've been led to believe is that both x86 as well as the Windows OS has taken on a lot of baggage over the years. Just a bunch of different ways to improve performance on what ultimately is based on the thinking of 15-20 years ago.

Smartphones are the future personal computers.

RE: Some good points, some crazy
By charrytg on 3/2/2011 5:26:22 PM , Rating: 2
Well regardless, I believe this article is more on the topic of competing in the server market, not the personal computer market, where smartphones may be the future, but servers will never be a part of. As far as servers go, I don't think intel has a lot to fear from ARM at the moment. But between ARM and Nvidia's GPGPU computing, they should at least keep an eye out to prevent losing opportunities in the server market.

I've heard people say that the x86 instructions cripple processors in terms of energy efficiency, but that was transistor sizes shrink, this is less of a problem as apparently the part of the chip that handles these instructions will take up less % of the chip each shrink. There may be more to the struggle intel is having with power efficiency, but I'm not qualified to say.

At the very least, intel seems to have a lead over others when it comes to what manufacturing node they use, so perhaps if they really wanted to make a big push against ARM, they may be able to, but perhaps they just aren't pouring enough money in to it. If they half-ass it, they may as well just be throwing away money since the competition is expirienced for the purpose.

By SimpleLance on 3/2/2011 1:41:13 AM , Rating: 2
I definitely do not see them beating ARM chips, which are designed from the ground up for smartphones

ARM was designed for PCs (Acorn Archimedes PC), but lost to X86 PCs. But being power-efficient, they survived in the embedded device market.

Now, the newest generation of ARM is so unlike the early ARM (ARM2). They have become more like Intel's new chips. Super scalar, pipe-lined, branch predictor (ARM8), large cache, SIMD, etc.

RE: Some good points, some crazy
By ET on 3/2/2011 2:40:56 AM , Rating: 2
I think that Intel can deliver on the low power promise. x86 compatibility will always add some baggage, but the more complex the chip becomes (and high end ARM ones are becoming quite complex), the less effect that added baggage should have. Intel has the thousands of engineers and the advanced processes to be able to compete effectively.

That said, I think that x86 will just not have a real advantage in the mobile market in the long run. Low power x86 is still useful these days for Windows compatibility on low power devices, but that compatibility doesn't mean much for mobile phone apps, and currently Windows isn't that great for tablets, so it'd be hard to convince people to go that way.

By theapparition on 3/1/2011 5:47:01 PM , Rating: 2
Why do I foresee this guys quote at the bottom of a Daily Tech page?

Right between the ones where Balmer said there's no chance of Apple gaining mobile marketshare and the Seagate exec saying his product lets you store more porn.

RE: Hmmmm
By tygrus on 3/1/11, Rating: -1
RE: Hmmmm
By sprockkets on 3/1/2011 7:39:46 PM , Rating: 2
The Atom was also limited to 32bit. In the future they could make a 64bit Atom or 64bit ARM or a 64bit MIPS.

Desktop atoms have always been 64 bit capable.

RE: Hmmmm
By JumpingJack on 3/2/2011 12:11:43 AM , Rating: 3
ARM has a magic Architecure
By vignyan on 3/2/2011 10:27:16 AM , Rating: 2

ARM definitely has a great low-power design techniques. While the ISA has some advantages over x86 ISA, its not too drastic. Even ARM has to convert their instructions into micro ops. I am not sure about the coherence and consistency models they are planning to support for the multi-core processors.

I am sure when they come to implementing features close to what Intel can offer, I see them short of performance/watt as compared to Intel.

I am not sure why people also think that multi-core is magical. Each core still has to be powerful enough to contribute to a good multi-core processor. Otherwise, you get a Phenom quad cores that are easily beat by Intel's dual cores.

RE: ARM has a magic Architecure
By silverblue on 3/2/2011 2:38:03 PM , Rating: 2
Let's also remember Intel's fishy compilers which sniff out your CPU's vendor ID and, if it's not GenuineIntel, throws it the least optimal codepath. So, it's not solely a case of AMD having less capable processors.

Whether they still do this, I've no idea, however that in itself is an obstacle to overcome.

By JKflipflop98 on 3/3/2011 9:39:10 PM , Rating: 2
Use AMD's compilers. Oh that's right they're even slower.

Intel is probably right.
By chromal on 3/1/2011 5:42:16 PM , Rating: 3
Intel says that ARM cores may gain some ground in the laptop space but will not challenge it in servers

I don't see the ARM architecture pushing Intel out of our shop any time soon. Our Xeon 5400 and 5500 series servers run heavy virtualization and database workloads.

The answer to low watt servers isn't low watt low power hardware, it's instead increased service density in physical environments that comes from the oversubscription of hardware via virtualization in high-performance environments.

By HoosierEngineer5 on 3/2/2011 8:55:37 AM , Rating: 2
I dare you to say that three times in a row...

Just talk
By Shadowmaster625 on 3/2/2011 8:20:05 AM , Rating: 2
If intel thought there was a market for atom servers, they would be making a 32 core atom chip.

RE: Just talk
By wordsworm on 3/2/2011 9:40:58 AM , Rating: 2
I think that Intel's guess is probably accurate. I went over to ARM to look at what their projections were. They predict that they'll have 0% of the server market in 2015. So, unless the projections of both Intel and ARM are wrong, then what Intel said was not just posturing.

Porting? What porting?
By Taft12 on 3/1/2011 5:25:36 PM , Rating: 2
we did application porting with Itanium, it took us about 10 years. Hundreds and hundreds of millions of dollars to port about 14,000 applications. ARM has to port for hosting all those applications over.

I'm not quite sure what he's talking about here. ARM has an instruction set issue? MS is porting their OS (yes, I know -- I'll believe it when I see it too) and Linux which runs on even more servers worldwide was ported to ARM many years ago.

Back when Itanium was introduced, much of the code that ran on servers was C and C++ using architecture-specific libraries. Is that what he means about porting effort?

Today server-side code is written in PHP, Java, Python, Perl, Ruby... Languages that don't depend on the underlying architecture. Over in the MS universe, it's .NET -- again, no porting problem if MS ports its OSes to ARM.

Someone show me what I'm missing.

By atlmann10 on 3/2/2011 10:51:19 AM , Rating: 2
Yeah right when Intel shows something significant in the Mobile processor market that outperforms to a degree an ARM device they have room to talk. As of right now they have basically failed at doing that although they have seemingly been trying for a couple of years with no results that are at least better to any amount.

By fteoath64 on 3/4/2011 3:15:37 AM , Rating: 2
The bottom line is really the number of "work units" per watt of power that should be most important compare to the cost per core. So the metric should be workUnits per watt per dollar. It does not matter how many cores you have as long as one you process all the incoming transactions in the given time (ie acceptable response time).
So the type of server design also depends on the type of workloads as benchmarks have clearly indicated to date.
Large companies with Virtualization farms actually maximize their hardware by shifting workloads over certain time-of-day. This means that Atom will have limitations for certain types of workloads, until we see some measurements in this area, it is difficult to determine any result.
This is "BruteForce Cores vs ManySmall Cores" and the subsequent I/O subsystems are just as important to ensure efficiency.

"My sex life is pretty good" -- Steve Jobs' random musings during the 2010 D8 conference

Copyright 2015 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki