Intel Atom Z600

[Exophase]

Anand showed Sunspider numbers against NEW Android phones with ICS, but surprise surprise, apparently Intel only showed scores from competitors not running ICS who are thus far behind in V8 performance (vs the latest available for ARM).. and I'm still hearing in forum after forum that "Javascript is a good litmus test on phones" despite scenarios like this being presentable. With Anand's numbers I thought Intel was probably gaming the benchmarks by comparing one thing that's known to be behind in software: with Intel's numbers I KNOW they're gaming the benchmarks by comparing one thing that's know to be WAY behind in software, and superseded! I don't know why no one is reporting on this. When nVidia tried pulling those farcical Tegra 3 scores vs Core 2 people reported, people did not put up with it. This is the same damn thing and no one cares, instead we're getting pretty much unanimous claim that Medfield's Atom is a superior uarch than Cortex-A9.

Seriously, comments like this "If the reference platform was clocked at 1.6GHz, and when it was underclocked to a hypothetical 1GHz, performance in Sunspider should still be around 2x that of the A9 based OMAP4." make me want to slap Charlie. He should know much better than writing this trash. Not to mention "Saying that Intel has noticeably better CPU performance isn’t going out on a limb here." Even if it's true in some context, it's worse than useless information. It's very misleading.

 

[Erinyes]

Im a bit late to the thread i guess..but i really expected Medfield to come with a SGX 545 like Cedartrial instead of the now ancient SGX 540. By the time Medfield launches, it will have been almost two years since Hummingbird with the SGX 540 was available. And especially considering that they should have been aiming for Win 8 tablets with Medfield as well (since SGX 540 is not DX9 certified). I guess thats now down to its successor, which should sport a dual core Atom with a DX9 class GPU. The DX9 limitation narrows down the options, it'll have to be either one of SGX 544/545/554. It remains to be seen if Intel is able to release this chip out in time for Win 8.

And like others have stated already, i expect the real chip to watch out for will be next year with 22nm. A dual core Atom with OooE and possibly a Rogue GPU wont be too shabby!

 

[french toast]

Anand showed Sunspider numbers against NEW Android phones with ICS, but surprise surprise, apparently Intel only showed scores from competitors not running ICS who are thus far behind in V8 performance (vs the latest available for ARM).. and I'm still hearing in forum after forum that "Javascript is a good litmus test on phones" despite scenarios like this being presentable. With Anand's numbers I thought Intel was probably gaming the benchmarks by comparing one thing that's known to be behind in software: with Intel's numbers I KNOW they're gaming the benchmarks by comparing one thing that's know to be WAY behind in software, and superseded! I don't know why no one is reporting on this. When nVidia tried pulling those farcical Tegra 3 scores vs Core 2 people reported, people did not put up with it. This is the same damn thing and no one cares, instead we're getting pretty much unanimous claim that Medfield's Atom is a superior uarch than Cortex-A9.

Seriously, comments like this "If the reference platform was clocked at 1.6GHz, and when it was underclocked to a hypothetical 1GHz, performance in Sunspider should still be around 2x that of the A9 based OMAP4." make me want to slap Charlie. He should know much better than writing this trash. Not to mention "Saying that Intel has noticeably better CPU performance isn’t going out on a limb here." Even if it's true in some context, it's worse than useless information. It's very misleading.

Yep, everyone has just took it on board as gospel that Atom now pawns A9= Intel dominates smartphones...
EXCEPT Intel doesn't ship any Morestown or medfield despite pulling this nonsense before, both in 2010&2011, what happned to that LG phone then?
This from Tomshardware is even more ridiculous;

''When you are done reading this article, you will agree that Intel is going to overtake Qualcomm in three years''

http://www.tomshardware.com/reviews/medfield-krait-smartphone-mobile-soc,3117.html

I have been telling everyone that will listen that the clock speeds, android version are not equal, Intel doesn't include any baseband let alone 4g, Medfield apparently doesn't include power management either.

Thats without talking about Krait&A15, and what ever apple has cooked up. people who are smoking what Intel are smoking are going to get a nasty bite on the backside when the next gen show up.

 

[tangey]

The DX9 limitation narrows down the options, it'll have to be either one of SGX 544/545/554. It remains to be seen if Intel is able to release this chip out in time for Win 8.

They were demoing clovertrail at CES on a tablet, and I assume that is PowerVr graphics too, so the chip seems to be reasonably well on in development.

 

[Deleted member 13524]

I saw that article as more of a historic narration that ends with an opinion rather than an objective and purely informative article.

There's something you can't deny though: Medfield is ancient.
It's using a CPU core that's been pretty much unchanged during the 4 years it's been in the market, it uses a GPU that's been used in smartphones for 2 years and is also built on a process that Intel has used for over 2 years.

There's a huge room for improvement on the Atom line-up, the fact that it's actually competitive with high-end smartphones while using such old tech is kind of impressive.
Remember Medfield was first supposed to come out somewhere in 2010.

I don't believe it'll be Intel's top-end model for smartphones for long.

Furthermore, he justifies the "3 years" not only with the fact that, by then Intel will be using state-of-the-art process technology to make Atom but also with the fact that Qualcomm has lost a lot of important people in their graphics department.


Intel has design wins, that was the hardest barrier they faced. Now that the elephant isn't in the room anymore and they are part of the market, they probably feel more secure about using their "alien technology" to improve the Atom lineup.

 

[Silent_Buddha]

Heh, all this talk about comparing Medfield to X ARM CPU reminds me of the old days of all the computing talk revolving around comparing Intel's CPU arch to X company's RISC based CPU arch and how Intel was doomed to failure, and how in only 2-3 years RISC CPU's would dominate the computing landscape because they were so much more powerful and efficient. This was back in the early to mid 90's. And continued through the late 90's...

Heck at one point it was virtually unthinkable that Intel would eventually dominate the server industry with x86 CPUs.

One thing I learned from back then is that you should never count Intel out of the picture.

Regards,
SB

 

[Raqia]

Who wants to bet that Intel's new core is another repurposed, existing design? Maybe

Silvermont : Silverthorne :: Pentium m : Pentium?

The Pentium m was an out of order core that pulled out all the stops for power consumption, and when built on a 22nm process as opposed to a 90nm process, these cores would be down-right tiny and fit 4 to a die + GPU easily.

They might have a real winner if they pair something like this w/ a PowerVR Series 6 GPU; I'd certainly bite if someone built a netbook from an SoC like this and included features like an IPS screen.

 

[Laurent06]

Heh, all this talk about comparing Medfield to X ARM CPU reminds me of the old days of all the computing talk revolving around comparing Intel's CPU arch to X company's RISC based CPU arch and how Intel was doomed to failure, and how in only 2-3 years RISC CPU's would dominate the computing landscape because they were so much more powerful and efficient. This was back in the early to mid 90's. And continued through the late 90's...

Heck at one point it was virtually unthinkable that Intel would eventually dominate the server industry with x86 CPUs.

One thing I learned from back then is that you should never count Intel out of the picture.

Following your thought, should we assume Intel will always dominate every market it enters? Frightening to say the least.

But this time it's very different: Intel could enter the server market because it was able to provide good enough performance for a good price. But the smartphone market is almost the exact contrary: how low are they willing to price their chips? And how flexible will they be? Phone makers want diversity (both in terms of functionality and in terms of HW providers), how much will Intel provide them?

So indeed Intel shouldn't be counted out, but assuming they'll succeed because they had successes in the past looks dubious.

 

[tangey]

I think Intel will succeed because they know they HAVE to. Everyone and his uncle is going mobile, ultra-mobile, or ultra-embedded. Ratio of volume of "traditional" processors to low cost processors is getting worse and worse at an accelerating rate. They can NOT afford NOT to be in this segment.

I think their public decision to advance their mobile segment as quickly as possible, to get onto the same fabrication cycle as the traditional segment, is testiment to the fact that they have realised this. Succeed of course does not necessarily mean dominate in the same fashion as they did in the past with mainstream processors.

The question is simple. Do you think, if Intel wholeheartedly commits to the mobile platform on the basis that's its crucial to the company's success, that they can do it ?

 

[Laurent06]

The question is simple. Do you think, if Intel wholeheartedly commits to the mobile platform on the basis that's its crucial to the company's success, that they can do it ?

The answer is obvious: they have the cash and the engineers to succeed, but if they fail to deliver on time (if Silvermont is late) and if the board doesn't accept lower margins, they'll fail.

 

[Deleted member 13524]

The answer is obvious: they have the cash and the engineers to succeed, but if they fail to deliver on time (if Silvermont is late) and if the board doesn't accept lower margins, they'll fail.

Heresy! Intel doesn't fail, it's Intel.
If they wanted, they could even make high-end discrete graphics ca..

Oh wait...

 

[tangey]

Heresy! Intel doesn't fail, it's Intel.
If they wanted, they could even make high-end discrete graphics ca..

Oh wait...

That's irrelevant in this context. They also failed to get any significant foothold with their own ARM based Xscale line. But at their respective times, they were both minor sideshows to their meat&drink income.

I genuinely think there is now a lot riding on Intel making the mobile division work.

 

[Laurent06]

I genuinely think there is now a lot riding on Intel making the mobile division work.

Yes, they have no choice. But again the question is how much are they willing to change their business model, and will the board let them do the required changes and see the dividends lower?

 

[tangey]

Yes, they have no choice. But again the question is how much are they willing to change their business model, and will the board let them do the required changes and see the dividends lower?

Indeed, but if Intel can get the key criteria down to those that you stated, and away from technical issues (power, X86 relevance etc), then it'll be an amazing transformation.

And clearly the suits aren't stupid, they'll be looking ahead and will surely see a point where dividends will be lower if they DON'T act.

 

[tangey]

I wonder is there any validity in this Fudzilla article:-

http://fudzilla.com/processors/item/25575-intel-working-on-quad-core-tablet-atom

22nm Dual-core medfield H2 2012 ?

Dual-core medfield with SGX54x graphics this year was stated in the Anadtech article (nothing about 22nm)

a 22nm dual-core atom with SGX54xMP2 by end of year would really excite the market, both in how competitive it would be, and also as a statement of intent by Intel.

I'll file it under exciting, but highly speculative.

 

[Ailuros]

There are too many "ifs" and "buts" in most of your posts here gentlemen for Intel to truly succeed in the small form factor market that I personally have a hard time to see any convincing argument so far.

tangey,

Not that one is related to the other since there are lightyears of differences and completely different factors, but where's Knights corner in terms of hw and when exactly had it been promised to be ready? If the discussion here would be about CPUs it would be truly a totally different chapter, but for other markets where Intel let's say doesn't "specialize" yet in as much as CPUs I'm raising an eyebrow until execution is where it should be.

Under that light I'm willing to believe that 22nm Medfield might have a MP2 GPU; I'm just not holding my breath that they'll execute as they expect themselves. If they do, which isn't unlikely then more power to them. But so far a couple of 2nd thoughts aren't unjustified.

 

[french toast]

Well i will bet that even the staunchest ARM supporter will concede that Intel has many of the cards going forward, if Intel had put some actual effort behind medfield and got it out for mid 2011, then ARM would have been seriously worried.

This year Medfield is nothing but an experiment just for Intel to say they can build a fully functioning smartphone, with acceptable batterylife..no mean feat in itself..but they still havn't delivered yet....

If they can push forward 22nm this year it could really shape the landscape up nicely for the next couple of years, because if 22nm out of date Atoms can compete well against 28nm krait/A15, then ARM is in big trouble going forward.
Inversly if they get that 22nm out and it gets spanked then Intel may not recover in the smartphone race for a long time.

 

[Ailuros]

Well i will bet that even the staunchest ARM supporter will concede that Intel has many of the cards going forward, if Intel had put some actual effort behind medfield and got it out for mid 2011, then ARM would have been seriously worried.

And what guarantees that it didn't launch earlier because of lack of more resources as you seem to imply and not simply a very specific desing decision that was rated as adequate for Medfield's real timeframe?

If they can push forward 22nm this year it could really shape the landscape up nicely for the next couple of years, because if 22nm out of date Atoms can compete well against 28nm krait/A15, then ARM is in big trouble going forward.
Inversly if they get that 22nm out and it gets spanked then Intel may not recover in the smartphone race for a long time.

Slight but important difference being that Intel doesn't compete directly with ARM CPU IP but with all of ARM's licensees not being exactly small as semiconductor manufacturers. It's no coincidence that Intel in the longrun sees in Qualcomm for instance a much larger competition threat than in AMD or anyone else.

How likely do you think is it that Intel starts licensing CPU IP to other semiconductors or other semiconductors to abandon SoC development and buy SoCs from Intel?

 

[Exophase]

How likely do you think is it that Intel starts licensing CPU IP to other semiconductors or other semiconductors to abandon SoC development and buy SoCs from Intel?

They were supposed to be licensing Atom macros for SoC usage on TSMC. Word is that third parties showed zero interest. My guess is that either perf/W was much lower or the licensing price made the total price unattractive compared to a similar SoC from Intel.. which would make the licensing costs really high compared to what it costs to get an ARM hard macro.

 

[Ailuros]

They were supposed to be licensing Atom macros for SoC usage on TSMC. Word is that third parties showed zero interest. My guess is that either perf/W was much lower or the licensing price made the total price unattractive compared to a similar SoC from Intel.. which would make the licensing costs really high compared to what it costs to get an ARM hard macro.

Same could go for other semis hypothetically buying SoCs from Intel. Intel's so far SoCs haven't been cheap from what I recall.