Will Bobcat be the home run AMD is looking for?

[exar333]

I think if you really want to use the benchmark performance disparity between Bobcat and Atom to speak towards the decision making that went into the creation of Atom you really need to take it to a deeper level of consideration.

Intel didn't/doesn't necessary want bobcat performance at bobcat ASP's and bobcat gross-margins...and until bobcat is actually released Intel does get to continue to make Atom gross-margins based on Atom performance.

Also Intel didn't/doesn't necessarily want bobcat performance at bobcat ASP's to cannibalize their existing low-end and higher-margin cpu products.

AMD doesn't really have a choice, Intel gets to have its cake and eat it to (profits...) whereas AMD continues to make decisions with the aspiration of simply getting out of the business of being a non-profit charity.

The reason why Intel made Atom at the executive level is different than the reason AMD pursued the creation of Bobcat. One is measured-risk at creating new markets and new market opportunity, the other is a counter-measure (reactionary) to the former.

I said the same of Intel's move to adopt 64bit x86 after AMD moved to it in order to create/operate the 64bit x86 server marketspace. They both do it, the rationalizations for them doing it have to be put into context though.

Comparing performance of Atom and Bobcat does not speak to the context of the decision making behind the creation of Atom IMO.

Isn't Intel working on a out-of-order version of the Atom to be released potentially next year? I thought I heard that. If true, that will put the Atom in an interesting position against Bobcat. AMD will have the GPU edge (likely) but Intel will have the general CPU advantage.

Also, what will happen when Intel finally moves Atom to 32nm?

 

[Cogman]

I will reserve judgment until I see battery life results.

What are you talking about, it consumes less power than the atom at idle and consumes a comparable amount with load. Battery life is going to be phenomenal. Only the OEMs could really screw it up by limiting the size of the batteries.

Not to mention its single threaded performance is through the roof when compared to the atom. For a netbook, this is a good thing.

Honestly, for the market it is targeting (squarely at the atom) it is an excellent contender. Low power + good enough performance = a win.

 

[Martimus]

Yeah. Much of it is AMD correcting problems Intel made at the basic design level when they built Atom, since Intel went first (like our earlier argument).

It is true that Atom was designed to meet a different market than Bobcat, but Intel was not able to meet that market so they created a new one. Bobcat appears to be a much better product for the market that Intel created (netbooks), however I have a hard time seeing it shrink to the market that Intel was actually targeting for Atom (Small form factor devices like Smart Phones). At the same time, I am not sure that Atom will ever get there either, since it still consumes quite a bit of power, and battery technology has barely advanced at all in the last 20 years.

 

[IntelUser2000]

Isn't Intel working on a out-of-order version of the Atom to be released potentially next year? I thought I heard that. If true, that will put the Atom in an interesting position against Bobcat. AMD will have the GPU edge (likely) but Intel will have the general CPU advantage.

Also, what will happen when Intel finally moves Atom to 32nm?

I don't think they'll go make out of order Atom in the next gen but sometime at least a generation after that.

2011-Saltwell, 32nm
2012-Saltwell derivative, 32nm
2013-OoOE Saltwell successor on 22nm?

The 1st gen 32nm Atoms will be around only 1 or 2 quarters earlier than 28nm Bobcat. Maybe not even for the Netbook chips.

 

[frostedflakes]

Really dissapointed in the Bobcat performance. The CPU was pretty dismal...it barely beat the in-order Atom in most of the benches (both were clocked around 1.6ghz). The GPU performance was pretty good, but when the platform is already so low-end, who cares about getting 22fps instead of 15fps on really crappy resolution and detail? The games you actually want to play with a low-end GPU (like SC2 or Civ5) ran abyssmally on the Bobcat.

Obviously a lot of people care about GPU performance, just look at the demand for ION-based platforms. Not so much the gaming aspect, but a better GPU can provide a smoother Aero experience and supports features like H264/VC-1 decoding acceleration, which Pineview does not. And being able to run old/semi-old games is a nice bonus, although I can't imagine why anyone would want to game on a netbook.

 

[psoomah]

To my knowledge Bobcat is Stars+40nm shrink+APU+maybe a little tweaking and of course shrinking.

The Stars architecture was extensively tweaked.

 

[Idontcare]

Yeah. Much of it is AMD correcting problems Intel made at the basic design level when they built Atom, since Intel went first (like our earlier argument).

It is true that Atom was designed to meet a different market than Bobcat, but Intel was not able to meet that market so they created a new one. Bobcat appears to be a much better product for the market that Intel created (netbooks), however I have a hard time seeing it shrink to the market that Intel was actually targeting for Atom (Small form factor devices like Smart Phones). At the same time, I am not sure that Atom will ever get there either, since it still consumes quite a bit of power, and battery technology has barely advanced at all in the last 20 years.

When it comes to Intel's decisions you really do have to view it from the gross-margin perspective.

It's not about whether or not Intel can do it...i.e. the question isn't "can they shrink atom's power-consumption to fit the smart-phone market they are targeting", the question being asked inside Intel at the budget meetings and project review meetings is "can we do xyz and still be reasonably confident we'll command >50% gross margins for 2-4 yrs afterwards?"

Everytime, without fail IIRC, that Intel has pulled back from a market or product lineup in the past 10yrs it has been for gross-margin reasons, not for technology issues and not for lack of capability (R&D at the right price).

Regardless what Atom can become or can do today, if it drags down gross-margins or is expected to not enrich existing gross margins over the next 2-4 yrs then it won't be pursued.

The last time they entered and then bailed on the mobile phone market it was for margins, same as their HDTV efforts around the same timeframe.

Look at gross margins for the discreet GPU market, it's no wonder Larrabee was scrapped as a discreet GPU product and is now being targeted at the much higher gross-margin HPC (tesla) marketspace.

 

[Cogman]

Look at gross margins for the discreet GPU market, it's no wonder Larrabee was scrapped as a discreet GPU product and is now being targeted at the much higher gross-margin HPC (tesla) marketspace.

Honestly, with that in mind, it is a wonder that they started the Larrabee project in the first place. The were never going to have a commanding lead in the discrete GPU market. Not like their integrated GPU market.

 

[maniac5999]

I think the big difference between Anand's article and a lot of the others are the other competitors in the test. Anand compared Ontario to Core i3 and i5 desktop parts. These obviously aren't Onterio's competitors, even calling the cheapest i3 mobile parts 'Onterio competitors' is a stretch, because as anand said 'This is the chip you’ll find in $400 nettops and notebooks in the $400 - $500 range.' Go to Newegg. See what you can buy for under $500. There isn't currently a single i3 machine available.
The numbers look a LOT better when you compare it to Atom units. Look at this article: http://hothardware.com/Reviews/AMD-Zacate-E350-Processor-Performance-Preview/?page=1 Ontario generally surpasses the performance of a Dualcore Atom + ION 2 while using about as much power as a 1st generation Atom w/o ION. for basic performance and light gaming I suspect that it'll be a homerun, although I too would like to see more CPU performance (ideally it would have enough CPU performance to at least match my current netbook.)

 

[Martimus]

When it comes to Intel's decisions you really do have to view it from the gross-margin perspective.

It's not about whether or not Intel can do it...i.e. the question isn't "can they shrink atom's power-consumption to fit the smart-phone market they are targeting", the question being asked inside Intel at the budget meetings and project review meetings is "can we do xyz and still be reasonably confident we'll command >50% gross margins for 2-4 yrs afterwards?"

Everytime, without fail IIRC, that Intel has pulled back from a market or product lineup in the past 10yrs it has been for gross-margin reasons, not for technology issues and not for lack of capability (R&D at the right price).

Regardless what Atom can become or can do today, if it drags down gross-margins or is expected to not enrich existing gross margins over the next 2-4 yrs then it won't be pursued.

The last time they entered and then bailed on the mobile phone market it was for margins, same as their HDTV efforts around the same timeframe.

Look at gross margins for the discreet GPU market, it's no wonder Larrabee was scrapped as a discreet GPU product and is now being targeted at the much higher gross-margin HPC (tesla) marketspace.

All decisions are cost sensitive. Even in my industry, where most seem to think that we will spend whatever it takes to get the desired performance.

I doubt that Intel is any different in that regard than most companies, with the exception being that they require higher margins than say AMD to consider a path worth traveling.

My point was that it appears to be more trouble than its worth to try to shrink the x86 architecture to meet <0.1W power consumption used in most mobile devices. Plus those margins likely won't be there by the time they do, if they manage to accomplish that feet. I don't so much see Atom as being abandoned, but I could see Intel diverging from their goal of small form factor electronics and redesigning it to better meet the Netbook / Tablet market it currently services.

 

[Gigantopithecus]

I had a Samsung NF310 in my Amazon cart, ready to purchase pending the results of the Brazos/Zacate performance previews. I took it out of my cart. I will await Zacate-based solutions now.

These are the numbers that matter to me:
Support for 4gb DDR3.

Load power is good. Idle power is great. This is based on an ES as well, so I expect the power usage on production chips will drop. My netbook does not spend much time under load. It spends most of its time idling.

Wow. Awesome! I *hate* laptops that run hot.

Looks like Zacate >>> Atom for those of us who produce rather than consume with our netbooks.

The bottom line is that single core Atom-based netbooks no longer provide 'good enough' computing for me. I played with a dual core Atom netbook a few days ago and it is 'good enough.' However, dual core Zacate-based netbooks will be better all-around, from performance to endurance to heat production. I see no real reason at this point to recommend Atom-based ultraportables to friends until Intel refreshes its Atom platform. I'm also eager to start building Zacate-based nettops for friends. It will be nice to deliver 'good enough' desktop computing for $300 or less including the display.

 

[cotak13]

I don't really see apple jumping at all. I find that they have been extremely loyal with their partnerships in the past.

I almost spit my tea into my Mac Pro reading that. That was before. Today Apple's as agonistic about what parts they use as any other company out there. You just have to look at the last iMac refresh to see they use whatever works for them.

 

[Gigantopithecus]

I think the big difference between Anand's article and a lot of the others are the other competitors in the test. Anand compared Ontario to Core i3 and i5 desktop parts. These obviously aren't Onterio's competitors, even calling the cheapest i3 mobile parts 'Onterio competitors' is a stretch, because as anand said 'This is the chip you’ll find in $400 nettops and notebooks in the $400 - $500 range.' Go to Newegg. See what you can buy for under $500. There isn't currently a single i3 machine available.

Yeah I wondered about that too. I really could not care less how Zacate fares compared to outdated desktop parts. Why is Zacate being compared to a Core i5? That's like providing metrics showing how much faster a 911 Turbo is compared to a Prius...

I care about how Brazos stacks up against Nile - where are those numbers? (Aside from the power consumption chart on the first page.) Where are the comparisons to *any* Atom parts aside from the D510, which is not available in a single netbook you can buy from Newegg? The D525 would be a more appropriate comparison.

 

[cotak13]

Yeah I wondered about that too. I really could not care less how Zacate fares compared to outdated desktop parts. Why is Zacate being compared to a Core i5? That's like providing metrics showing how much faster a 911 Turbo is compared to a Prius...

I care about how Brazos stacks up against Nile - where are those numbers? (Aside from the power consumption chart on the first page.) Where are the comparisons to *any* Atom parts aside from the D510, which is not available in a single netbook you can buy from Newegg? The D525 would be a more appropriate comparison.

Check the other review sites. Tom's has the D525 with ion.

 

[veri745]

The Stars architecture was extensively tweaked.

I hope that by "extensively tweaked", you mean "trashed and replaced from the ground up", because bobcat is in no way related to stars

 

[Nemesis 1]

sorry i looked an read it but i don't see what you talk about. could you elborate on what you think is skewed?

Skewed was a little strongly worded. But I don't believe the 800x600 didn't belong . Both should have used all 3 res. the same . Just gives the appearance of skewing. But still those are not bad results . But they are useless results none the same As , zacata will be up against Oaktrail . At 32nm with NO pcie . Comparing to 2core SB would be just silly. Except price . Oaktrail will surprize many and thats what its going against.

http://www.electronista.com/articles/10/11/01/chips.geared.for.tablets/

http://www.myce.com/news/intel-oak-trail-cpus-unveiled-for-netbooks-tablets-30313/

 

[cbn]

What are you talking about, it consumes less power than the atom at idle and consumes a comparable amount with load.

But don't the Desktop Atoms lack certain power saving features of the Netbook atoms?

IIRC IntelUser2000 has brought this up in some of his other posts.....I think one of the things he mentioned was EIST.

 

[cbn]

http://www.tomshardware.com/reviews/amd-fusion-brazos-performance,2790-2.html

Based on our use, it seems that AMD may see the biggest power savings by playing with the GPU. At the moment, it seems to be operating at max clocks all of the time.

Comments/Opinions on this?

 

[Idontcare]

Honestly, with that in mind, it is a wonder that they started the Larrabee project in the first place. The were never going to have a commanding lead in the discrete GPU market. Not like their integrated GPU market.

The man who convinced them this was possible has left the company.

I've told this story before, when Intel announced they were heading into mobile phone market as well as the HDTV markers (both of which TI heavily played in) we knew internally Intel made a calculated mistake and that they'd leave before landing a customer.

Most people outside TI probably thought "their goes TI!" when Intel announced they were going to compete with us...but we knew the gross-margin story of those markets and we knew Intel wouldn't/couldn't seriously get into competing there.

So 2 yrs later when they announced they were selling XScale and simply scrapping their HDTV project altogether it may have been a surprise to investors but generated no surprise internally at TI.

I'd be surprised if AMD or Nvidia ever took Intel's Larrabee effort seriously, internally, because they knew the gross-margin story for their markets better than Intel did and they knew that while Intel might create a performance competitive part there was no chance in hell that Intel would sell it at an ASP that commanded less than 50&#37; gross-margins and that meant it would seriously lag in price/performance comparisons and thus never gain traction.

We'll see if Atom ever ends up in a smartphone. There are so many players in that market who are willing to live with 30-40% gross margins that it will be a spectacular coup if Intel fields a product that commands 60% gross margins.

To put it differently, by virtue of the margins they command as an effective monopoly in the x86 space they have painted themselves into a corner when it comes to their options for diversifying their revenue stream. There is a reason they spun-off NAND flash as IM Flash.

 

[Rezist]

I find the CPU performance pretty weak, but the single core performance is quite good so it may lead to a better user experience. (windows/webpages)

 

[Idontcare]

I find the CPU performance pretty weak

Compared to what though? An Atom or Via Nano?

Technologically I am impressed that AMD was able to squeeze this kind of power-consumption footprint out of TSMC's bulk-silicon without the benefit of HK/MG.

If bobcat is this good on 40nm bulk w/poly-si gates just imagine how good it could have been on Intel's HK/MG 45nm or GloFo's SOI+HKMG 32nm!

As a process technologist I am impressed at what has been accomplished given the process-tech handicap that team-bobcat was given.

 

[CTho9305]

You seriously think this is going to be a P4 vs K8 situation? Based on what exactly? SB will wipe the floor with Zacate in terms of CPU performance, this is clear. The only uncertainty will be SB's GPU performance vs Zacate's GPU performance.

Wouldn't it make more sense to compare Sandy Bridge to Llano or some non-Fusion higher-end platform?

You contradict yourself here. You stated that you will recommend to your friends who buy low end notebooks the option with the best GPU. If they buy low-end notebooks and only do light tasks, then there is *no point* in even trying to go for the best GPU option. The IGP in this situation is a bottleneck only if you do more than light gaming.

He said his friends only do light gaming. You just said the IGP works for light gaming. Was there some other contradiction? Personally I thought the Modern Warfare 2 benchmarks on Zacate looked pretty good, so we're squarely past the "Farmville" level of gaming (27fps is squarely in the realm of playable) and even Starcraft 2 is passable on Zacate (8fps in the large-battle test is not smooth, but in a lot of online play I've experienced lag that routinely slows the game down to that range... your 4GHz Nehalem doesn't help when you're being limited by the person you're playing against).

Even if Zacate is more competitive on the GPU side than SB and in terms of TDP, Ivy Bridge will erase whatever advantage would exist.

A 22nm high-end processor that's a follow-up to a still-unreleased processor will outperform 40nm low-end processor that will be available at CES in January?

If we're going to look a year out, I think it's much more interesting to compare against ARM's upcoming A15, which is very architecturally impressive (3-wide, out of order loads, 128 bit FPU? Compare that to STARS!) and won't be saddled with any x86 penalties. If ARM actually hits the frequencies they're claiming (2.5 GHz), and someone puts one in a netbook/laptop form factor with a decent graphics coprocessor, I'd buy it. Nvidia is probably in a good position to design an SOC for that market, since they have both GPU and ARM experience (Tegra). x86 ate the old-school RISC vendors' lunches by accepting lower margins for approximate performance parity; ARM looks poised to do the same thing to the x86 vendors.

Given the hype for Bobcat, and also considering Bobcat is out-of-order while Atom is in-order, personally I don't find the performance all that amazing.

...

Yes per-clock Bobcat easily performs better than Atom, and this was expected. However when looking at multi-threaded performance, and looking at overall performance, the gap is not quite so big. Atom's HT helps quite a bit, and although it was not tested here, the top Atom (D525) runs at a higher clock than the top Bobcat. We should see dual-core Atoms hitting over 2Ghz next year as well.

A processor about half the size of Atom that is architecturally more powerful and outperforms it (sometimes dramatically) isn't impressive? Tough crowd!

Further to that, the conclusion there is vastly different vs. the Anandtech review.


I find the Anandtech article bizarre at best.

Some of my friends wondered if there was anything shady going on w.r.t. Anand's review... I think it's more likely that he's just been influenced by the large number of members in the forum who keep wanting to compare against Nehalem. My guess is that most people here tend to be interested in high performance over extreme mobility / cost savings. If his audience wants him to compare against 80mm^2 i5 processors with ??mm^2 chipsets, it's understandable that he'd cater to them.

The idle power consumption is great. At load though, Zacate's consumption is nothing special.

Don't forget to compare the performance at that power level.

 

[cbn]

It will be nice to deliver 'good enough' desktop computing for $300 or less including the display.

Good point. I am really interested to see how far AMD will be able to push into new markets (ie, poorer countries) with these APUs.

Hopefully MS has incentives to help them. For example, with Windows 7 starter connecting an atom netbook to a separate monitor or TV is not possible. (With atom's GPU limiting resolutions and video playback this is probably a non-issue, but with Ontario's GPU the situation likely changes).

 

[grimpr]

http://www.tomshardware.com/reviews/amd-fusion-brazos-performance,2790-2.html

Originally Posted by tom's hardware
Based on our use, it seems that AMD may see the biggest power savings by playing with the GPU. At the moment, it seems to be operating at max clocks all of the time.


Comments/Opinions on this?

Presumably, it will be enabled with the 11.1 Fusion Catalyst drivers from AMD and further bios and platform optimizations.

 

[cbn]

If we're going to look a year out, I think it's much more interesting to compare against ARM's upcoming A15, which is very architecturally impressive (3-wide, out of order loads, 128 bit FPU? Compare that to STARS!) and won't be saddled with any x86 penalties. If ARM actually hits the frequencies they're claiming (2.5 GHz), and someone puts one in a netbook/laptop form factor with a decent graphics coprocessor, I'd buy it. Nvidia is probably in a good position to design an SOC for that market, since they have both GPU and ARM experience (Tegra). x86 ate the old-school RISC vendors' lunches by accepting lower margins for approximate performance parity; ARM looks poised to do the same thing to the x86 vendors.

Yep, Cortex A15 is one interesting chip (from my layman's point of view).

But how will MS respond to these processor developments?

Will MS try to increase their x86 OS sales in more countries (using Bobcat) or do they think their ARM based OS stands a better chance for that type of expansion (even if Google is considerably ahead of them in development)?