New Intel CEO wants company to move faster refine later

[Vesku]

http://www.dailyherald.com/article/20130622/business/706229966/

' "He wants to see Intel move faster. That's very clear," Justin Rattner, the company's chief technology officer, said Monday. "We've been legitimately accused of trying to get everything perfect before it comes to market." '

Regrets about the slow Atom development?

 

[Sheep221]

Intel is far from perfect, despite the desktop CPUs are better than AMDs offerings for a long time now, they sill don't arrive with any performance increases. The Atom is shit, their mobos are also crap, the IGPs are beaten by AMD and the only thing that is really cool now are haswell mobile processors(which didn't debut yet anyway). The desktop CPUs also haven't moved from quad core for several years already, it's obvious the new guy noticed the lack of innovation.

 

[Magic Carpet]

Intel is far from perfect, despite the desktop CPUs are better than AMDs offerings for a long time now, they sill don't arrive with any performance increases.

It's not that bad. Imagine 2001 when Intel reduced IPC and doubled power consumption. It's the very same company, love it or hate it

 

[Ayah]

From a marketing point of view, it's the correct idea. New product = Uninformed people will buy the newest and the "best" (since new > old).

From the point of performance/reliability, I'm always iffy at faster development cycles, usually means something will be rushed, unless the content per cycle is reduced proportionally to the cycle time reduction.

 

[blackened23]

Intel is far from perfect, despite the desktop CPUs are better than AMDs offerings for a long time now, they sill don't arrive with any performance increases. The Atom is shit, their mobos are also crap, the IGPs are beaten by AMD and the only thing that is really cool now are haswell mobile processors(which didn't debut yet anyway). The desktop CPUs also haven't moved from quad core for several years already, it's obvious the new guy noticed the lack of innovation.

 

[dealcorn]

I did not save the links to several articles on Seeking Alpha by a retired Intel manufacturing engineer who helped me interpret where Bryan is coming from. Intel thinks it's Finfet transistors are a big deal that will permit Intel to dominate SoC efficiency. Intel sees no evidence insightful micro architecture design will permit a competing fab to match Intel efficiency. It took Intel 5 years or more to produce Finfet because there were many, sequential steps that can not be shortcut while maintaining efficiency. The secret sauce, good stuff is outside public view and each competitor will need to actually do much of the hard work themselves. There is no possibility any competitor will catch up in the next 12 months.

Bryan thinks Intel has a substantive, transitory, competitive advantage with a shelf life somewhere between fresh fish and canned goods. The only way to screw it up is if some self referential jerk delays releasing product to polish obscure, niche interfaces. Bryan is saying grab market share now and we will clean up in a year at 14 nm.

Competitors may share a different world view.

 

[VirtualLarry]

I remember IDC talking about how Intel "perfects" its process nodes now, which is why we don't see CPUs released with lower clock speeds, and then released with higher clock speeds six months later on the same node. Thus, they are getting all they can from the process upon release.

If I'm reading the details from the guy from Intel properly, isn't that code-speak for saying that they will go back to the old way of process mfging, that they will release products based on a new (but immature) process, and refine the process as they go along? So possibly they will release 14nm sooner than later, but upon an immature 14nm, which will get better as they go along. Will be interesting to see.

 

[ShintaiDK]

I remember IDC talking about how Intel "perfects" its process nodes now, which is why we don't see CPUs released with lower clock speeds, and then released with higher clock speeds six months later on the same node. Thus, they are getting all they can from the process upon release.

If I'm reading the details from the guy from Intel properly, isn't that code-speak for saying that they will go back to the old way of process mfging, that they will release products based on a new (but immature) process, and refine the process as they go along? So possibly they will release 14nm sooner than later, but upon an immature 14nm, which will get better as they go along. Will be interesting to see.

It would be 10nm if anything manufactoring related. 14nm starts production in Q4.

 

[Idontcare]

After reading the link in the OP, I have got to believe that something got lost in translation here. And that mistranslation has unfortunately become the topic of discussion.

I cannot believe any CEO would actually espouse the philosophy that their company needs to rush shittier buggier products to market instead of refining their QRA procedures to speedup the iterative loop that goes into creating reliable stable products.

As presented in the context captured by the link, the CEO is being made out to appear as if he is reckless and willing to risk creating Intel's own TLB bug snafu that AMD experienced, or another FPU bug that Intel experienced...and I just don't buy that any CEO would be so reckless.

So I have to conclude the new CEO's comments were completely taken out of context and are being misrepresented to create a story where there is none.

 

[Idontcare]

I remember IDC talking about how Intel "perfects" its process nodes now, which is why we don't see CPUs released with lower clock speeds, and then released with higher clock speeds six months later on the same node. Thus, they are getting all they can from the process upon release.

If I'm reading the details from the guy from Intel properly, isn't that code-speak for saying that they will go back to the old way of process mfging, that they will release products based on a new (but immature) process, and refine the process as they go along? So possibly they will release 14nm sooner than later, but upon an immature 14nm, which will get better as they go along. Will be interesting to see.

That's exactly what I get out of it too, and I hope that is all they are getting at.

Unfortunately the way the article is presented it is made to sound like they will be shipping earlier (buggier) steppings of the chips themselves. Which is something that I just can't fathom a reasonable CEO would condone.

But maybe that is exactly where Intel is going? Push crap out because what else are people going to buy? (someone else's buggy crap?)

 

[mrmt]

I remember IDC talking about how Intel "perfects" its process nodes now, which is why we don't see CPUs released with lower clock speeds, and then released with higher clock speeds six months later on the same node. Thus, they are getting all they can from the process upon release.

I don't think this is an issue with their manufacturing tech, as it is already faster than all the industry. It should be a matter of design. Intel took 4 years to bring Silvermont to bear, that's an unacceptable high lead time that cost the company a huge opportunity on mobile.

 

[Idontcare]

I don't think this is an issue with their manufacturing tech, as it is already faster than all the industry. It should be a matter of design. Intel took 4 years to bring Silvermont to bear, that's an unacceptable high lead time that cost the company a huge opportunity on mobile.

To which I would argue that Intel needed to add more resources to the project so that its release timeline could be expedited.

Standard project management triangle stuff.

If you want to improve on the development schedule then you either pare back the scope (reduce the feature set), or you pare back the quality (considered to part of the scope) or you increase the development cost (add resources).

The new CEO is saying "cut quality!"...or so he is being made out as having said.

The CEO's comment (taken out of context?) implies the CEO would have preferred the silvermont team to just rush buggy silicon to market and deal with the fall-out from that afterwards.

If this really is the attitude of the new CEO then AMD just caught a huge break. This is the equivalent to a Ford CEO going on record in advance of releasing the Pinto and saying "people may die because we've cut quality, but it is a calculated risk and we are willing to take the gamble with our customers, lets hope they don't mind".

 

[VirtualLarry]

If this really is the attitude of the new CEO then AMD just caught a huge break. This is the equivalent to a Ford CEO going on record in advance of releasing the Pinto and saying "people may die because we've cut quality, but it is a calculated risk and we are willing to take the gamble with our customers, lets hope they don't mind".

Just thinking about Intel's server market share. If that is correct about the new CEO, then ARM-based servers may make huge inroads, if Intel scales down their server validation efforts, in an effort to save costs. (Or possibly, AMD can grow their server market again.)

 

[VirtualLarry]

But maybe that is exactly where Intel is going? Push crap out because what else are people going to buy? (someone else's buggy crap?)

But one reason that Intel charges a premium, and that people pay it, is because Intel is known as a "gold standard" for reliability. That's why the Pentium FPU bug was such a big fiasco, if not financially, but in the press.

 

[ShintaiDK]

To which I would argue that Intel needed to add more resources to the project so that its release timeline could be expedited.

Standard project management triangle stuff.

If you want to improve on the development schedule then you either pare back the scope (reduce the feature set), or you pare back the quality (considered to part of the scope) or you increase the development cost (add resources).

The new CEO is saying "cut quality!"...or so he is being made out as having said.

The CEO's comment (taken out of context?) implies the CEO would have preferred the silvermont team to just rush buggy silicon to market and deal with the fall-out from that afterwards.

If this really is the attitude of the new CEO then AMD just caught a huge break. This is the equivalent to a Ford CEO going on record in advance of releasing the Pinto and saying "people may die because we've cut quality, but it is a calculated risk and we are willing to take the gamble with our customers, lets hope they don't mind".

I think the issue is different. Simply the startline for new projects. When the original Atom was finished, there wasnt even any idea to start the next Atom uarch. It was a wait and see (how much it sold) game. perfect for Intel is also in terms of margins and financials.

Hence why it took ~5 years to get a new uarch. While the smartphone segment ran crazy.

 

[mrmt]

If you want to improve on the development schedule then you either pare back the scope (reduce the feature set), or you pare back the quality (considered to part of the scope) or you increase the development cost (add resources).

I think we're talking here is about a smaller scope, in the sense that an imperfect product now is better than a Silvermont 4 years down the road, when the market has changed and there is an awful lot of big established players there.

Check the speech itself, it's around 10:00

http://www.bloomberg.com/video/inte...-big-thing-summit-6AjF6cArSSak8vxTwl7Obw.html

Ed: His choice of words wasn't a happy one, but this tends to happen when you have technical guys speaking in marketing events.

 

[Khato]

I think we're talking here is about a smaller scope, in the sense that an imperfect product now is better than a Silvermont 4 years down the road, when the market has changed and there is an awful lot of big established players there.

Correct. I can see how the article can be interpreted otherwise if you want to, but there's a pretty decent nudge in the right direction immediately after the quoted bit:

Rattner, who spoke at the Bloomberg Next Big Thing Summit in Half Moon Bay, Cali., said that the company is going to respond to a mobile market where products are introduced then rapidly updated.

Last I checked the ARM SoC camp doesn't release buggy silicon. They just limit the scope in each of their releases. aka, they don't re-design the wheel every chance they get >.>

 

[Vesku]

Listening to the video I'm getting a "these mobile chip companies don't have the same execution problems as AMD" vibe. Definitely talking about the chips and not the manufacturing process.

 

[Exophase]

I remember IDC talking about how Intel "perfects" its process nodes now, which is why we don't see CPUs released with lower clock speeds, and then released with higher clock speeds six months later on the same node. Thus, they are getting all they can from the process upon release.

If they still did this they'd start new generations with lower peak performance than the last generation. Like Llano.

 

[Idontcare]

If they still did this they'd start new generations with lower peak performance than the last generation. Like Llano.

And Prescott

 

[RadiclDreamer]

The Atom is shit.

As much as I love the recent intel cpus, you are 100% correct on atom, it is a smoking turd

 

[Magic Carpet]

As much as I love the recent intel cpus, you are 100% correct on atom, it is a smoking turd

Intel's bad apple

 

[BrightCandle]

In hindsight Intel missed the mobile opportunity. That does not mean they need to move any faster than a new architecture every 2 years, it just means that they need to expend resources towards that market with the same vigor they have done the higher power performance parts. I can't believe the argument is a reduction in quality or faster released products that aren't tested correctly, that doesn't make any sense because they are only a couple of products away from death if they start producing buggy hardware.

 

[OBLAMA2009]

so with the last few products, theyve had major bugs at launch and he wants to launch even crappier stuff?

 

[dealcorn]

From a products and niches perspective, Baytrail is the model. Contrast how much Snapdragon 800 outperforms Saltwell to the projections of how Baytrail outperforms Saltwell. I get that Snapdragon 800 has double Baytrail graphics and half Baytrail's efficiency (twice as efficient versus 4.4-4.7 times as efficient). If you want an Android gaming tablet and have access to a plug, Baytrail is not a contender. Intel cedes that niche at 22 nm. However, like Natalie Portman in The Professional, the Android gaming tablet niche is not yet mature.

Some folks do not want a dead battery to kill their access to SpongeBob SquarePants @1080p when they need him. They like Baytrail. Gender, age, and whatever will affect consumer preferences. I think SpongeBob claims more than 75% of the market versus less than 25% for gamers. If Intel clearly dominates the biggest niche, Bryan may see that an OK first step even if Baytrail can not access portions of the market.

I understand Anand's statement that Intel's mobile team needs a "graphics wakeup call" but from a strategic perspective Byran might differ. Classic Intel graphics trash talk lacks traction in the mainstream market if Baytrail delivers better than 1080p media consumption with Angry Birds on the side. However, any hint of failure on efficiency leadership will be dealt with harshly by mobile consumers. Better graphics cost both efficiency and precludes access to price sensitive segments of the market because more transistors cost more money. Baytrail is strategically targeted to deliver mainstream market share.

Like Natalie, the Android gaming tablet market may mature and remain attractive. It's heresy, but Atom needs a viable Android gaming tablet to clean up the market. I hear Bryan requesting/demanding some spin on IRIS Pro for Atom at 14 nm. If it has the tablet market's best performance that is cool. However, Baytrail may not even need the world's best graphics. Some mobile gamers lack ready access to a plug. They are low hanging fruit Bryan wants to harvest today, while Finfet benefits are incontestable.

The fact that everyone in this space claims efficiency leadership evidences the broad recognition that efficiency is a/the killer issue in the mobile SoC space. Right or wrong, Bryan thinks Intel wins efficiency and wants to claim the prize now. He does not want distractions.