Skip OMAP4 and go directly to OMAP5?

[DaMummy]

damn exophase, way to push windows onto us...

 

[Linux-SWAT]

@exophase


when he said, (thanks AMD) I think he was being sarcastic for giving x86 an extended life lease.

:mellow:


Well, no !


AMD pushed so far with their K6-K7 plus x86_64... intel were urged to respond. P4 failure, back to P3 into core architecture, etc... etc...


All good for consumers, thanks to AMD.

 

[Phawx]

@exophase


when he said, (thanks AMD) I think he was being sarcastic for giving x86 an extended life lease.

:mellow:


Well, no !


AMD pushed so far with their K6-K7 plus x86_64... intel were urged to respond. P4 failure, back to P3 into core architecture, etc... etc...


All good for consumers, thanks to AMD.

Oh, alright then. Microsoft isn't going anywhere anytime soon.

 

[larry]

So, do ARM Cortex-A15 and ARM Cortex-M4 have the same exact instruction set? Or would programs have to either target one or the other? (Or both, I guess.)


It'd be nice if the Linux scheduler could bounce programs between CPU cores based on how much time they wanted. This hypothetical future Pandora could then start programs on the low-power M4s and only transfer them to the high-power cores when they became CPU-bound. Obviously that won't work if the cores have different instruction sets.

 

[Exophase]

WinCE is a joke.
WP7 ? Where are WP7 users ? _Never_ saw one for real.

WinCE/Mobile shipped in a great multitude of devices. Its time has long passed but it served its purpose. Maybe you think it was a joke but the industry didn't agree.


As for WP7, I didn't realize that worldwide failure was determined not by sales numbers but by some guy's anecdotal account of "I've never saw a WP7 user."

Wow, you seems to ignore wintel history : they made the market. They force users to buy their products, OS and office suite. Beeing a best seller, especially in a dictatorship form, doesn't means the product is better. In fact microsoft OS'se are probably the worst ever created.

It's a little difficult to argue with someone who claims Intel and MS forced people to buy their products and are a "dictatorship." The "wintel" relationship as you put it is no more than historically de facto, and MS's move to ARM should illustrate this (not to mention Intel's move from Windows on say, Moorestown)


And I didn't think we were comparing things, I just think that new version after version continuing to sell is indicative of market acceptance which is enough to say they're not just "crappy old products." And the market HAS had offerings of alternatives over and over again and they've continually failed to take off. You'd have to be really delusional to think that Linux has no user perception disadvantages compared to Windows.

AMD made the nowadays PC. If Intel were alone, your next PC will be a P4.
ARM kicks butts, whether it is 32, 64 or 128 ^^, but we will see soon.

Yeah, right, it's AMD's CPUs - which as of now are vastly inferior in the high end - which has forced Intel to innovate so we should be thanking AMD and not Intel. And yet even though AMD has not been competitive since Core 2 came out we still saw Nehalem, Sandy Bridge, we'll still see Ivy Bridge, we'll still see node shrinks every ~2 years well ahead of everyone else in the industry.. but yeah, you keep believing that AMD is the only reason Intel left P4. Here's a thought - if Intel stops making better processors they can't sell the end user better processors.


It's really amazing that anyone can be so arbitrarily supportive of AMD and so arbitrarily against Intel when both companies are making x86 processors.

Again, we'll see soon.

I guess? So how do you want to do this, in a year when AMD hasn't picked up an ARM arch license I can say "I told you so"?

Err, wrong statement : PC sales are doing well. As consumer are forced to buy w7 with a PC... w7 sales are good B)
Doing a complex OS doesn't means doing a good OS. I don't care about a so-called complex OS.


And think about this : porting windows to another architecture... Okay, in ten years, maybe... But in ten years, nobody will remember microsoft, as for DEC now.

Ah, more of this "consumers are forced to buy Windows" argument. This isn't 10 years ago - mind you, the argument wasn't great then either. If you want to blame someone blame OEMs, but if you really want to blame someone blame consumers for not buying the non-Windows PCs that are put on the market. Nonetheless, you can't say Windows 7 isn't doing better than Vista - which is a perfect example of what happens if you release a Windows that doesn't do well. OEMs don't even upgrade to it. Believe it or not, the market DOES have to have a justifiable reason to move to new software.


So you think it'll take 10 years for MS to release their next version of Windows? Wow, you really are delusional. Then again, it'd help if you had any supporting argument for a thing you said, as opposed to saying things almost entirely arbitrarily.

Linux supports ARM since a very long time : it's a no match. I don't need to argue/autopilot.

Oh, being "20 years ahead" means supporting ARM. Right. Newsflash: Windows has been on other platforms aside from x86. It isn't actually that hard to port it. I don't know why you think it is. It isn't there now because the market demand wasn't there, and even now MS is taking a big gamble that it WILL be there because right now it isn't. Interesting that this makes it 20 years behind, here I thought it'd only take 10 years for MS to port Windows..

Where the heck did you read i hate kinect ?

*chortle*


Probably the part where you insulted it for no reason.


Next you'll be asking where I hear you hate MS in general, right?

Almost everyone knows WP7 is a failure but you... Even microsoft knows it...

Uh huh.


Me: 1.5 million people bought it in 6 weeks, IMO that makes it not a failure.


You: NUH UH EVERYONE KNOWS IT IS, EVEN MICROSOFT LOL

Again, where did you read i judge a merit on stock activity ?


Microsoft is sinking : "Got the facts" B)

Well gee, I figured that you brought up Gates and Ballmer selling stocks in an attempt to show stock performance as an indicator of company health. No idea what I was thinking there, I must be crazy.


I'll work on getting the facts, you work on presenting some to me.

 

[Exophase]

So, do ARM Cortex-A15 and ARM Cortex-M4 have the same exact instruction set? Or would programs have to either target one or the other? (Or both, I guess.)


It'd be nice if the Linux scheduler could bounce programs between CPU cores based on how much time they wanted. This hypothetical future Pandora could then start programs on the low-power M4s and only transfer them to the high-power cores when they became CPU-bound. Obviously that won't work if the cores have different instruction sets.

No, they're not exactly the same instruction set - Cortex-A15 is ARMv7a with some extensions AFAIK and Cortex-M4 is ARMv7me or something. The mobile profile with DSP instructions.


Even if they were ISA compatible you can't schedule threads on it because it doesn't have an MMU. I wouldn't even necessarily expect them to have the same view of main memory, if any view at all. Besides, I think you might be overestimating the clock speed of these things, we're talking probably under 100MHz, and IPC that's maybe around ARM9 level on a good day.

 

[Batou456]

Wow, you seems to ignore wintel history : they made the market. They force users to buy their products, OS and office suite. Beeing a best seller, especially in a dictatorship form, doesn't means the product is better. In fact microsoft OS'se are probably the worst ever created.

Microsoft's operating systems have routinely been more secure and capable then Macintosh's if prone to memory issues. That the combination of Linus' Law, corporate investment, and OSS Linux user modifying things to make it work the way they want with GNU has resulted in the dominating super computer/server/embedded OS is a statement on the strengths of that approach more then Microsoft's failure.

Yeah, right, it's AMD's CPUs - which as of now are vastly inferior in the high end - which has forced Intel to innovate so we should be thanking AMD and not Intel. And yet even though AMD has not been competitive since Core 2 came out we still saw Nehalem, Sandy Bridge, we'll still see Ivy Bridge, we'll still see node shrinks every ~2 years well ahead of everyone else in the industry.. but yeah, you keep believing that AMD is the only reason Intel left P4. Here's a thought - if Intel stops making better processors they can't sell the end user better processors.

It's really amazing that anyone can be so arbitrarily supportive of AMD and so arbitrarily against Intel when both companies are making x86 processors.

Vastly inferior?

A i3-530 Dual Core with Smart Cache rates about 3100 in 3DMark06 while a Athlon II X2 260 under my testing gets around 2900 with the clock moved from 200 to 225MHz and thus 3.6GHz. These being relevant as most things don't use more then two cores, unlike the synthetic benchmarks. A i7-920 gets around 5000 making it inferior per core given how 3dMark06 works, which isn't unexpected. In general we're talking about a roughly 10% difference in synthetic benchmarks for twice the cost or more for the Intel option. That's with better silicon thanks to SOI and high-k metal gate, a process advantage, and the gimmick that is Intel's SMT technology with its associated 10-15% performance add in the rare things with can take advantage of understrength fake cores instead of telling it to F-off. Given the better silicon and process should allow things to physically be closer and thus speed up, it's actually rather uninspiring.


If you're talking supercomputers Tianhe-1A is up front most of the FLOPs is from nVidia Tesla GPGPU Accelerators, and it's only doing 2.5 PetaFLOPs to Jaguars Opteron driven 1.75 which should pass Tianhe-1A's current rating in the next changeover.


If we're talking in general AMD has depopulated their clock circuit reducing waste heat dramatically and thus allowing higher clocks, integrated power gating, and switched to SOI high-k metal gate silicon to start with Llano and Bulldozer in April when the revised Intel Sandy Bridge chipsets are projected to hit market. If the complete overhaul with this new approach to handling multiple threads and integration of Accelerators work out as planned Intel's not in a good position to respond anymore then they were when they lost the Gigahertz or 64 bit race. A complete overhaul early in multicore has helped to carry them but the margin it bought them is slim now, and isn't going to help them verse Bobcat and Llano in the -book form factors that represent a surprisingly significant part of their sales. Market acceptance of APUs in this space has also been good so far based on reviews of products on sites like Amazon in general being favorable.

 

[Exophase]

Vastly inferior?

A i3-530 Dual Core with Smart Cache rates about 3100 in 3DMark06 while a Athlon II X2 260 under my testing gets around 2900 with the clock moved from 200 to 225MHz and thus 3.6GHz. These being relevant as most things don't use more then two cores, unlike the synthetic benchmarks. A i7-920 gets around 5000 making it inferior per core given how 3dMark06 works, which isn't unexpected. In general we're talking about a roughly 10% difference in synthetic benchmarks for twice the cost or more for the Intel option. That's with better silicon thanks to SOI and high-k metal gate, a process advantage, and the gimmick that is Intel's SMT technology with its associated 10-15% performance add in the rare things with can take advantage of understrength fake cores instead of telling it to F-off. Given the better silicon and process should allow things to physically be closer and thus speed up, it's actually rather uninspiring.

3DMark06 isn't a CPU benchmark, and your comparison is between two low end CPUs. This isn't a matter of perf/price, where you could probably construe an argument (ie, how AMD has survived), it's a question of who has a more efficient architecture.


"Most things not using more than 2 cores" is itself a massive strike against AMD, because core count (6 vs 4) is the only advantage they have in the desktop space. In terms of IPC they're typically way behind, and they don't have some kind of peak frequency advantage like P4 did. Single threaded performance suffers greatly. It's not usually 10% per clock, it's closer to 30+%, and in real world situations more than benchmarks (indeed, you were the one who used a benchmark..). By the way, IPC has nothing to do with process advantage in this case... Intel had much better IPC when both were still on 45nm. Hyper threading isn't a gimmick - it often gives quite a bit more than 10-15% you claim (particularly on Atom, but that's another ball of wax) - but even if that were the extent of its improvements it's well worth the the transistor real estate which is fairly minimal. I don't know why you're calling them "fake cores", as Intel has never called them cores.

If you're talking supercomputers Tianhe-1A is up front most of the FLOPs is from nVidia Tesla GPGPU Accelerators, and it's only doing 2.5 PetaFLOPs to Jaguars Opteron driven 1.75 which should pass Tianhe-1A's current rating in the next changeover.

No I'm not talking about heavily throughput oriented tasks, that's a very different market. I was talking about Intel vs AMD in the desktop space and who is forcing whom to innovate, nothing else.

If we're talking in general AMD has depopulated their clock circuit reducing waste heat dramatically and thus allowing higher clocks, integrated power gating, and switched to SOI high-k metal gate silicon to start with Llano and Bulldozer in April when the revised Intel Sandy Bridge chipsets are projected to hit market.

No I.. what? AMD is playing catch up with Intel on every one of those fronts. That's good for AMD, but you're missing the point - this wasn't an attack against them as a company,s o there's no need to try to make such a defensive counterpoint to show where they'll be in the future. You might want to look back more on why I said what I did, which is to illustrate that despite the K8 derived platform growing old Intel still kept releasing on tick-tock like usual and are doing a good job with chip architecture. Please, look at the context here. I'm refuting the claim that if it weren't for AMD my next CPU would be a Pentium 4.

If the complete overhaul with this new approach to mhandling multiple threads and integration of Accelerators work out as planned Intel's not in a good position to respond anymore then they were when they lost the Gigahert or 64 bit race. A complete overhaul early in multicore has helped to carry them but the margin it bought them is slim now, and isn't going to help them verse Bobcat and Llano in the -book form factors that represent a surprisingly significant part of their sales.

So which is it, most things don't benefit from > 2 cores or Intel will get trampled by Bulldozer's proficiency in the > 4 core market? Make up your mind. Bobcat is a totally different market and has nothing to do with this; depending on how much more important GPUs and compute become (or don't) the same might apply to Llano. At the very least recognize that none of that has to do with who has the higher performing CPU core.

 

[Eridger]

I've thought that for a while also, when a Beagle Board can do all the basic computing tasks that most users would want to do with a few watts of power when an Atom uses multiple times that and at a slightly slower pace then something is up.


....


This is why Pandora2 needs a nice digital output, I'd use it as my main PC at home it if could output 1080p(the rez of my 24" monitor).

Oh, absolutely. If the Pandora 2 shipped with anything close to an OMAP5 with the Pandora's trademark battery life and gaming controls, I'd use it as my desktop for sure. Then when I'd need to go anywhere I'd just pull out all the cables and off I go.


I'd easily slap down a few hundred bucks for that.


And as for the microsoft thing... They're not going anywhere, simple as that. They'll still be major in the (in my opinion unlikely) event that x86 does become obsolete. As much as I don't like them or their software it doesn't change the fact that they'll still be around for a long time to come.

 

[Batou456]

3DMark06 isn't a CPU benchmark, and your comparison is between two low end CPUs. This isn't a matter of perf/price, where you could probably construe an argument (ie, how AMD has survived), it's a question of who has a more efficient architecture.

3dMark06 is considered a CPU benchmark last I checked, particularly as almost every reputable review site uses it like that. The particular tests for that involve rendering something entirely with CPU in a way that realizes the full benefit of multicore. It has a specific CPU Score independent of the others, which is what I'm using. Hence four instead of five digits.

Your injunction of it being low end is nonsensical as i3s have Smart Cache, Hyperthreading, all the other performance add features, and rate higher per core then their larger brethren. They're limited in that they're dual core, not feature sets. This strikes a lie to your injunction about wanting to discuss architecture, as architecture has nothing to do with core count. Just like SOI high-k metal gates differentiated at the 45nm process, and thanks to their various benefits should allow things to be closer and has nothing to do with micro-arch and everything to do with silicon processing.


That you don't recognize the Dual verse Quad issue and associated lag in programmer acceptance doesn't help, particularly in a begging the question that treats it like some insane novel concept, instead of a well established area of discussion independent of manufacturer.

"Most things not using more than 2 cores" is itself a massive strike against AMD, because core count (6 vs 4) is the only advantage they have in the desktop space. In terms of IPC they're typically way behind, and they don't have some kind of peak frequency advantage like P4 did. Single threaded performance suffers greatly. It's not usually 10% per clock, it's closer to 30+%, and in real world situations more than benchmarks (indeed, you were the one who used a benchmark..). By the way, IPC has nothing to do with process advantage in this case... Intel had much better IPC when both were still on 45nm.

SOI high-k equals silicon bulk now for current leakage and related issues effecting how close together circuits are? You're being a shill to fail to acknowledge that and act like 45nm was equivalent silicon. Your conclusion also deviate from even Anandtech who tend to attribute Intel's advantage in single thread more to Turbo then anything.

Hyper threading isn't a gimmick - it often gives quite a bit more than 10-15% you claim (particularly on Atom, but that's another ball of wax) - but even if that were the extent of its improvements it's well worth the the transistor real estate which is fairly minimal. I don't know why you're calling them "fake cores", as Intel has never called them cores.

Which is why a lot of end users turn it off, right? Given your higher estimates for hyperthreading correspond to supposedly higher performance numbers in general I think we can leave it at that.

No I'm not talking about heavily throughput oriented tasks, that's a very different market. I was talking about Intel vs AMD in the desktop space and who is forcing whom to innovate, nothing else.

In which case you're nothing but a shill as you've demonstrated no innovation by Intel, or even why they're a better choice for most applications as related to investment, as any reasonable analysis should always include. Particularly given Intel's love of socket demarcation and high prices, and hence higher system upgrade costs. Tools are tools, status symbol bragging right nonsense has little place and the more expensive brand would always win that anyways as quality is not a factor there. This isn't the 1990s where the CPUs are so limited programmers are forcing people to need to upgrade to play new games. Indeed Crytek's new Cryengine 2 actually has lower system requirements then the original Crysis Engine from 2007.

Innovation is development of new ideas and AMD despite being an ant next to Intel has remained the x86 innovator company. A multicore design overhaul in 2006 that finally gave Intel advocates something to point at why they pay a premium, and way too often less motherboard associated features, is not innovation. Accelerated Processing Units taking a page from the development of Personal Supercomputers to bring more power to the home user and another approach beyond AMD's SMP or IBM's SMT to the multithreading problem, are. Acting like Bulldozer is just SMP as you are is beyond daft.

 

[craigix]

I see it like this:


10 years ago the man in the street, if he had a Computer, he almost certainly had to have Windows and an X86 based device.


Now? I know dozens of people who have neither. The only place they would use Windows would be at work, and that's changing too.


There is a huge shift coming and Windows will become irrelevant.


A few years ago I had a meeting with a parts supplier, and he brought along someone who worked with Google. He told me this long


story about how google was planning to take over devices with their new OS (he was saying to me TVs, Cars, Music devices). At the time I really could not see it happening, but here we are Android appearing everywhere and people trust the name too.


I just can't see Microsoft coming back from this. Their phones have been huge failures. The man in the street is looking at iPads or Blackberries.


When OMAP4/5 (etc.) laptops start to hit with 10-20 hour battery time the tide will turn even harder.


And I highly suspect Apple will lead the way with those Laptops, I bet more people ask if it can run Angry Birds than ask if it can run Windows.

 

[Exophase]

3dMark06 is considered a CPU benchmark last I checked, particularly as almost every reputable review site uses it like that. The particular tests for that involve rendering something entirely with CPU in a way that realizes the full benefit of multicore. It has a specific CPU Score independent of the others, which is what I'm using. Hence four instead of five digits.

Okay, I didn't know you were using the CPU score. But just the same, ONE benchmark is far from enough to form a conclusive comparison and any review site would be ridiculed for trying to pass that.

Your injunction of it being low end is nonsensical as i3s have Smart Cache, Hyperthreading, all the other performance add features, and rate higher per core then their larger brethren.

Injunction..? You should.. probably look that word up..


If the lowest end of Intel's mainstream desktop x86 line rates higher than the higher end then there's a problem. Overclocking can change things but it's not like these processors have zero binning performed on them..

They're limited in that they're dual core, not feature sets. This strikes a lie to your injunction about wanting to discuss architecture, as architecture has nothing to do with core count. Just like SOI high-k metal gates differentiated at the 45nm process, and thanks to their various benefits should allow things to be closer and has nothing to do with micro-arch and everything to do with silicon processing.

Right, so you're honestly claiming that the only difference between an i3, i5, and i7 is core count. Intel must be doing something wrong here. Look, NO process technology advancements for these CPUs have an impact on IPC, what they have an impact on are clock speed and core count. And Intel is winning in IPC per core big time.

That you don't recognize the Dual verse Quad issue and associated lag in programmer acceptance doesn't help, particularly in a begging the question that treats it like some insane novel concept, instead of a well established area of discussion independent of manufacturer.

Uh, what? I have no idea what you just said. You were the one who was saying that > dual core is rarely being utilized.

SOI high-k equals silicon bulk now for current leakage and related issues effecting how close together circuits are? You're being a shill to fail to acknowledge that and act like 45nm was equivalent silicon. Your conclusion also deviate from even Anandtech who tend to attribute Intel's advantage in single thread more to Turbo then anything.

Read what I said about IPC. At any rate, even if GloFo is catching up in process advantage Intel is still marching forward with 22nm and will easily outpace them there again so that's going to be all but negated beyond the scant months after Bulldozer and before Ivy Bridge is out. And seriously, shill? I'm a paid decoy for Intel now, right? Please start using words more responsibly.

Which is why a lot of end users turn it off, right? Given your higher estimates for hyperthreading correspond to supposedly higher performance numbers in general I think we can leave it at that.

A lot of end users turn off hyperthreading? If so they're making a big mistake. I think you may not realize that the impact of hyperthreading has changed a lot from when it was first introduced on P4.

In which case you're nothing but a shill as you've demonstrated no innovation by Intel, or even why they're a better choice for most applications as related to investment, as any reasonable analysis should always include. Tools are tools, status symbol bragging right nonsense has little place and the more expensive brand would always win that anyways as quality is not a factor there. This isn't the 1990s where the CPUs are so limited programmers are forcing people to need to upgrade to play new games. Indeed Crytek's new Cryengine 2 actually has lower system requirements then the original Crysis Engine from 2007.

Oh hey there's that word again. Do I really need to write you an essay about why I think Sandy Bridge is innovative? Go read David Kanter.


Once again you're completely ignoring the point that I'm somehow being forced to repeat over and over again. Linux-SWAT claimed that if it weren't for AMD Intel would have stopped making faster CPUs, and I've shown that they've made faster CPUs in spite of a lack of real competition on this front from AMD for a few years now. You are apparently deeply offended by this and are now going on tirades about how much of a shill I am for thinking it's important to begin with (which I didn't even say)


Stick with context for fuck's sake. Get off your fanboy pedestal. You've clearly got such a chip on your shoulder against Intel or such a love for AMD or god knows what that you've dragged me into an argument for daring to compare Intel favorably against AMD in making a pretty isolated point.

Particularly given their love of socket demarcation. Innovation is development of new ideas and AMD despite being an ant next to Intel has remained the x86 innovator company. A multicore design overhaul in 2006 that finally gave Intel advocates something to point at why they pay a premium, and way too often less motherboard associated features, is not innovation. Accelerated Processing Units taking a page from the development of Personal Supercomputers to bring more power to the home user and another approach beyond AMD's SMP or IBM's SMT to the multithreading problem, are. Acting like Bulldozer is just SMP as you are is beyond daft.

And yet everything I said about the K8 descendants falling behind Intel over the last few years. Athlon64 and AthlonX2 were several years ago now. Are you really that offended that Intel managed to move ahead of AMD? Something that AMD themselves are pretty blatantly admitting by ditching the architecture, the same way Intel blatantly admitted that Netburst was a dead end? APUs are great, but they have nothing to do with CPU innovation, or at least not the type I'm talking about (and Intel released the first real desktop class IGP integrated on a CPU anyway... may well have been AMD's idea but they're taking too long to market with it, or do you deny that Llano is taking longer to come out than it was supposed to?). JUST THE SAME, this wasn't supposed to be about Bulldozer, Brazos, or even Llano, it's about Intel's ability to keep making better CPUs during an era where AMD stopped being the kind of threat they once were. And if you think AMD was making Intel anxious with their desktop offerings over the last few years then I don't really know what to tell you. Likewise, AMD lost most of their server market for a reason.


And I never once said Bulldozer is "just SMT" or anything even minutely resembling that and I am at a complete loss as to how you've interpreted such a thing. I've barely said anything about Bulldozer at all, in fact, aside from that it's more throughput/core optimized than latency optimized. Whose posts are you reading dude?

 

[Batou456]

I just can't see Microsoft coming back from this. Their phones have been huge failures. The man in the street is looking at iPads or Blackberries.

I don't see Microsoft ceasing to exist even after the current move by many world governments to Linux and OSS, although they are probably going to shrink quite a bit. They became what they did to a large extent because they were good at iterative improvement verse an established competitor, ala their taking away Word Perfect's crown, and this guy who managed the Office 2007 and Windows 7 projects clearly has the makings to help them keep it up. There's also the issue of a lack of unified API in the Linux world and resistance to open source and associated issues with drivers related to kernels.

When OMAP4/5 (etc.) laptops start to hit with 10-20 hour battery time the tide will turn even harder.

And I highly suspect Apple will lead the way with those Laptops, I bet more people ask if it can run Angry Birds than ask if it can run Windows.

Based on the reception of Bobcat APUs I think the demand is ultimately more for a lightweight device with reasonable power, then supreme battery life. Apple kind of already is leading the way on that with their tablets, and Intel fussing with nVidia chipset wise probably isn't helping matters.

 

[Exophase]

When OMAP4/5 (etc.) laptops start to hit with 10-20 hour battery time the tide will turn even harder.

OMAP5 laptops with 15+ hour battery life won't happen. At least not in something that the market will call a "laptop." The larger you scale in size the less of an advantage the SoC has and the more your screen


and SSD and fast and wide multi-channel memory takes up the overall power budget. They'll still be differentiated but not as much.

And I highly suspect Apple will lead the way with those Laptops, I bet more people ask if it can run Angry Birds than ask if it can run Windows.

I dunno, I think we could stand to have a market not stolen by Apple. Then again, I don't know if they're going to be so willing to purge their entire Mac line.


I for one think it's going to be a sad day for gaming when Angry Birds is all anyone cares about anymore.

 

[Batou456]

Okay, I didn't know you were using the CPU score.

In which case you don't know shit about the benchmark. A console in the first rating is 6-8k, my inexpensive box is 13k.

Injunction..? You should.. probably look that word up.

If the lowest end of Intel's mainstream desktop x86 line rates higher than the higher end then there's a problem.

You really are an ignoramus about desktop space aren't you? The TDP of Dual Cores have never been half that of quad core and hence higher performance per core. It's not exactly complicated, and it's mentioned in any reasonable assessment of the two.

And yet everything I said about the K8 descendants falling behind Intel over the last few years. Athlon64 and AthlonX2 were several years ago now. Are you really that offended that Intel managed to move ahead of AMD? Something that AMD themselves are pretty blatantly admitting by ditching the architecture, the same way Intel blatantly admitted that Netburst was a dead end?

Projecting much? You're the one making a supreme issue of it, going after people, and derailing a thread so you can stand on a soapbox and try to act big while not having anything to back you up. And really what's with this tirade about K8 and its descendant as if they're a mistake? They developed the current standard, this whole ditching them dialogue reads like a Intel advocate tantrum over pent up frustration over Intel's ineffectual response prior to the Core overhaul. I suppose Pentium Pro was a mistake because it was eventually replaced?

APUs are great, but they have nothing to do with CPU innovation, or at least not the type I'm talking about (and Intel released the first real desktop class IGP integrated on a CPU anyway... may well have been AMD's idea but they're taking too long to market with it, or do you deny that Llano is taking longer to come out than it was supposed to?). JUST THE SAME, this wasn't supposed to be about Bulldozer, Brazos, or even Llano, it's about Intel's ability to keep making better CPUs during an era where AMD stopped being the kind of threat they once were.

A certain lawsuit had a not so minor influence on what window they shot for, but by all means keep telling yourself that had no influence and that evil usurper AMD could never possibly compete with Intel again. When you're down to disputing a definition because you wanted to use the word a buzzword devoid of its meaning, you clearly don't have much. Innovation is intentionally divided from iterative development, hence mere iterative overhauls never have and never will classify as innovative.

And if you think AMD was making Intel anxious with their desktop offerings over the last few years then I don't really know what to tell you.

Again with pointless emotional nonsense.

Likewise, AMD lost most of their server market for a reason.

The 2010 IDC report I assume you're referencing, given early 2010 reports are opposite, doesn't indicate losing "most of their server share." All it indicates is their sale of new servers is slightly down pending the release of new products during a quarter. Given how you characterize, and fail to reference this, thank you for lying.

And I never once said Bulldozer is "just SMT" or anything even minutely resembling that and I am at a complete loss as to how you've interpreted such a thing. I've barely said anything about Bulldozer at all, in fact, aside from that it's more throughput/core optimized than latency optimized. Whose posts are you reading dude?

So now SMP aka Symmetric Multiprocessor is SMT aka Simulatenous Multithreading? You're all mixed up, Jack.

 

[Exophase]

So wait, you were that dude who was insulting me over and over in the thread where I was talking about ARM and Atom and shit right? You sound really familiar. Okay, let's get this over with.

In which case you don't know shit about the benchmark. A console in the first rating is 6-8k, my inexpensive box is 13k.

Yes, and? Me not knowing shit about 3DMark2006 ratings has surprisingly little to do with most of what I've said. Give an exhaustive set of benchmarks to illustrate your points on Phenom II vs Core i7, although in all fairness that should probably include Sandy Bridge if you want to stick to the latest from both sides. You don't really have to give your own, you can find one of the many websites out there that have done this already.

You really are an ignoramus about desktop space aren't you? The TDP of Dual Cores have never been half that of quad core and hence higher performance per core. It's not exactly complicated, and it's mentioned in any reasonable assessment of the two.

No, I would say I have a functional knowledge about how desktop microarchs work, but even if I didn't, it's pretty universally agreed upon that Nehalem (never mind Sandy Bridge) is well ahead of "K10.5" as it's often referred to, I'm hardly alone in this sentiment. But boy, you do like to throw those insults around..

Projecting much? You're the one making a supreme issue of it, going after people, and derailing a thread so you can stand on a soapbox and try to act big while not having anything to back you up. And really what's with this tirade about K8 and its descendant as if they're a mistake? They developed the current standard, this whole ditching them dialogue reads like a Intel advocate tantrum over pent up frustration over Intel's ineffectual response prior to the Core overhaul. I suppose Pentium Pro was a mistake because it was eventually replaced?

I'm "projecting"? It fascinates me when people use that term like you are, I really must be an ignoramus because I have no idea what it is I must be projecting. I am flattered that you consider me solely responsible for derailing the thread, but I must share at least some of that honor with you.


I would contend that in no small terms has Pentium Pro been replaced to the extent that Netburst has or K8 will be. I can follow an evolutionary path from P6 to Pentium M/Core to Core 2 to Nehalem to Sandy Bridge and forward in the same manner I can from K7 to K8 to K10 to K10.5. P4 doesn't fit in that first path and as far as I'm concerned Bulldozer (and Bobcat) don't fit in the latter. So yes, I consider it scrapped. I didn't call it a mistake (in fact, I don't consider "P4" a mistake per se, although definitely more so of one than the K8 path was), I called it a "dead end", meaning that a lot of it had to give in to overhauling.

A certain lawsuit had a not so minor influence on what window they shot for, but by all means keep telling yourself that had no influence and that evil usurper AMD could never possibly compete with Intel again. When you're down to disputing a definition because you wanted to use the word a buzzword devoid of its meaning, you clearly don't have much. Innovation is intentionally divided from iterative development, hence mere iterative overhauls never have and never will classify as innovative.

I didn't realize I had been telling myself that that certain lawsuit had no influence on the delay. But I suppose by that line of thinking you should keep telling yourself that no engineering challenges, be them design related or process maturity related had any contribution. And I really didn't know you were the authority on what can be considered "innovative" vs "iterative", I guess it's not as subjective as I thought.

Again with pointless emotional nonsense.

I.. uh.. you're calling "And if you think AMD was making Intel anxious with their desktop offerings over the last few years then I don't really know what to tell you." emotional nonsense? Why are you doing that?

The 2010 IDC report I assume you're referencing, given early 2010 reports are opposite, doesn't indicate losing "most of their server share." All it indicates is their sale of new servers is slightly down pending the release of new products during a quarter. Given how you characterize, and fail to reference this, thank you for lying.

Ah yes, I'm a liar now. Yes, of course I'm referring to the IDC report. Fortunately AMD server share numbers are easy to google for so I don't have to lie. It's still unfortunate to have to do this much work for someone as rude as you are, but ah well.


AMD server share:


2005: 16.4% http://www.theregister.co.uk/2006/01/26/amd_q4_2005_market_share/


2006: 23.06% http://www.geek.com/articles/chips/intel-gains-slight-share-over-amd-in-2007-20080123/


2007: 22.26% "" ""


2008: 14.3% http://venturebeat.com/2008/11/12/amd-takes-a-blast-at-intel-with-new-server-processors/


2009: 10.2% https://www.trefis.com/company?hm=AMD.trefis&driver=0581#


2010: 6.5% "" ""


These numbers are not all from the same quarter - THAT is a little difficult. If you have issue with them please go ahead and give more accurate numbers and don't just continue slandering me. If you can't do this much I'm definitely not talking to you anymore, consider yourself warned (not that I think you care)


Assuming you don't come back saying they're total shit: it may well be the case that they've lost share due to the market anticipating Bulldozer (as well they should), but there's no way that you can look at this and not determine that they've lost a major competitive edge over the last several years. And these numbers illustrate what I've been trying to say about the K8 path vs the Core 2 path.

So now SMP aka Symmetric Multiprocessor is SMT aka Simulatenous Multithreading? You're all mixed up, Jack.

Oops, I misread what you wrote. Sorry, I'll correct my response:


And I never once said Bulldozer is "just SMP" or anything even minutely resembling that and I am at a complete loss as to how you've interpreted such a thing. I've barely said anything about Bulldozer at all, in fact, aside from that it's more throughput/core optimized than latency optimized. Whose posts are you reading dude?


.. and say, on the off chance that anyone else is reading, and given that I know how many technically inclined people do read these forums, does anyone have any insight to chime in with? Because these one on one bullying matches get a little lonely.

 

[Phawx]

When OMAP4/5 (etc.) laptops start to hit with 10-20 hour battery time the tide will turn even harder.

OMAP5 laptops with 15+ hour battery life won't happen. At least not in something that the market will call a "laptop." The larger you scale in size the less of an advantage the SoC has and the more your screen


and SSD and fast and wide multi-channel memory takes up the overall power budget. They'll still be differentiated but not as much.

I don't know man. Apple has been putting bigger and bigger batteries in their devices. I think the Macbook Air is like 70% battery (exaggerating, but it's got to be close to that number).

The fan can get axed in that pic, the pcb will be WAY smaller, could easily shove 3 more batteries in that space. Or just manufacture one HUGE ass battery and embed the SoC somewhere in the monitor.


Right now that is 35 watt hour. But it's awfully gap filled. If they could just make the entire base a battery, you could probably pull off 70 watt hour. And def be in the 12-15 hour range on USE time.

 

[Exophase]

Right now that is 35 watt hour. But it's awfully gap filled. If they could just make the entire base a battery, you could probably pull off 70 watt hour. And def be in the 12-15 hour range on USE time.

Guess I should clarify, I mean you won't see it with the typical laptop batteries. A 70 Wh one could do it maybe sorta (still dunno if the screen would really use 2W at that size), but that's a really tall order to fill for a laptop battery. I don't think you'll see anyone bothering with something like that. That said, 12+ hours is well past the useful threshold for long lasting battery life, outside of press release material for SoCs.

 

[Tom`]

.. and say, on the off chance that anyone else is reading, and given that I know how many technically inclined people do read these forums, does anyone have any insight to chime in with? Because these one on one bullying matches get a little lonely.

I'm not technically inclined enough to address specifics, but I've seen the benchmarks, and I can offer a consumer's point of view. I think it's fairly incontrovertible that Intel's CPUs are superior to AMD's in terms of performance per clock at least at the high end. I'm not sure how AMD lost the massive advantage they had a few years ago (before Intel came up with the Core architecture), but I'd go with the i7 950 over the latest six-core AMD CPUs, especially since the 950 is quite a bit cheaper at Microcenter.


With that said, I don't really like how Intel segments the market with their multiple socket types, but I guess that's more marketing than anything else.

 

[Phawx]

Wasn't there also the problem that some compilers purposely crippled AMD cpu's? Wasn't this why Intel paid AMD 1.5 billion?


In either case, I will easily say that my Intel based system feels much better than my AMD based system. And I am an AMD fanboy. But I got a really good deal on a Q6600 in 2007 and I am still rocking that system.


Hyperthreading: I've never actually seen any performance increase with HT enabled on my Atom. But I never seen a battery penalty with HT enabled either, so I've always kept it on.


x86 hate: I think the biggest problem I have with x86 is the power envelope. Atom is pretty craptastic and is extremely power hungry. AMD's APU's look nice but they are still power hungry (comparatively to ARM/ImgTec). The psp2 showed off some really impressive shit and I think that SoC is only taking 5-8watt (I'll have to fact check). The benchmarks on AMD's Zacate chip look nice and I know that's available *now* so it *might* be able to save even more power with a smaller process sometime, but still that guy is taking 18watts. I guess the major reason I don't like x86 is it always seems like solutions are band-aids and not real solutions, whereas ARM was designed from the ground up to be power friendly.