AMD made the point that FP is secondary with these chips. The cores are integer. There is FP hardware, but it's fairly weak.
?since most workloads are integer-heavy. And FP-intensive work is gradually moving to the GPU.
Quote:
Originally Posted by melgross
It's 50%, and it doesn't add 5% to the die. It's more like 20% per core. Intel's hyperthreading adds 5%, thats what you're thinking about.
No, it is 5% for the whole die, counting caches, memory controllers, etc. and ~12% for the second integer core on a module.
If you've looked at AMD's recent dies, the cores don't take up as much of the die compared to Intel's.
Quote:
Originally Posted by melgross
And we have 4 core Intel up against 4 core AMD, and six core Intel against 6 core AMD. Intel wins every time there, and not by a little.
Thuban won't be up against Gulftown, it's price (and TDP) is closer to Lynnfield/Gulftown. Next year, it'll be 8-core Zambezi, that is if AMD doesn't raise their prices.
Quote:
Originally Posted by melgross
The extra core does give what AMD CLAIMS is an 80% increase in INTEGER performance per core, as opposed to the known quantity of 30 to 40% improvement per core for hyperthreading. It also uses much more power, and needs more heat dissipation that Intel's chips.
The 80% number makes sense, as it's an entire second integer core. Two cores would give close to 100% performance increase in multithreaded situations.
SMT, CMT, and dual-core give various amounts of performance improvements for various increases in die size. AMD thought that CMT gave the most performance per core size increase compared to the other ways.
Bulldozer also features FP improvements. (Sandy Bridge also does.)
Quote:
Originally Posted by melgross
It's very funny, you know, because when Intel first began using the dual die per package, AMD was criticizing them for it.
That's true, but now things are reversed.
Quote:
Originally Posted by melgross
Thats totally wrong. However many cores are there, per core performance is as important as it ever was.
No, since more apps are moving to multithreaded. There are situations where single-threaded performance is important now and in the future. Now those apps are taken care of by more powerful cores (Intel), Turbo Boost and Turbo Core. In the future those apps will be taken care of by a few relatively large cores. In that respect Intel may have the advantage because their cores are already relatively large. However AMD may have the multithreaded FP advantage with Fusion.
Quote:
Originally Posted by melgross
Intel's ahead on the P/W spec, and is usually ahead on the P/$ as well.
Not at all in servers, and in the desktop they are usually behind in at least one if you are looking at multithreaded tasks.
Intel has the advantage with Clarkdale/Arrandale's integration, but that will be erased with Llano.
Quote:
Originally Posted by melgross
As usual, Intel will drop prices on its chips, putting pressure on AMD.
That will be when they are being pressured by AMD.
Quote:
Originally Posted by melgross
I'm willing to bet that Sandy Bridge will beat the pants off Bulldozer.
I doubt it. Not when Interlagos is expected to be over 33% increased performance over Magny-Cours for 33% more cores. So Intel will pretty much end up where they are this year, next year (and from the speculations I've heard that will be the best case scenario for Intel in servers).
Seriously I think people see the 13" machine as an excellent value. Part of that is because consumers are more informed about the significance of having a good GPU on a machine running Mac OS/X.
Quote:
Originally Posted by melgross
I don't think it's a problem for the devices that will be using these chips. It's overblown. Look at the 13" MBP. It was sold out at numerous locations the first day it went on sale. That's what matters.
Yes but it is being sold out with a GPU that is significantly faster than Intels. We aren't talking a little bit faster either, it is significantly faster with video acceleration and good OpenCL support to boot.
Now I would agree with you that many consumers don't care about the techie details. But they will respond to the idea of video acceleration.
Quote:
These will be used for consumers who don't care about another 25% of graphics performance. For those who do, AMD's on chip solution is no good anyway, and something along the lines of what Nvidia, or better yet, Apple did, will be far better than worrying about Intel's IP here.
The whole point here is that solutions that use multiple GPUs are only worthwhile if they can fit into the hardware in question. For the hardware that Apple seems to want to build Intels integrated GPU (in the CPU module) is no better than AMDs. In fact all signs indicates it is worst.
Quote:
It's really a non issue, except on the tech boards.
Actually I think it is a huge issue for Apple. They don't want to be seen as going backwards performance wise. Thus we get a much faster replacement for the 9400m and modest CPU improvements. All of this in an existing platform.
They need progressively better hardware for marketing reasons as it is hard to sell a machine that is slower than last years model. They also need that GPU for OpenCL which is a lot harder to market to consummers. OpenCL is a key to keeping the Mac competitive OS wise.
A side ways look at this would be the use of the GPU in the iPhone and iPad which is extensive. If Apple can get as much of Quartz running on the GPU on the Mac as it is on iPhone OS we could see major improvements in Mac responsiveness. In other words I expect Apple to continue to try to exploit the GPU as much as possible most likely leading to resolution independence and a few other long sought goals. Arrandale alone in such an environment would be a very low end solution.
First of all forget about trashing the optical drive. Most people want them.
Yes but not as internal drives.
Quote:
There's a survey at either Macsimumnews or Macnn, I forget which, that showed that 78% of those polled wanted optical drives. I believe it.
I haven't seen that poll but frankly most online polls are worthless as they don't represent the entire user community. In any event the lack of an internal drive does not imply a lack of support for Optical drives in the OS.
The best example here is Apples own MBA with no internal Optical. AIR is a machine that is rightly skewered for some of the design choices made with the computer but seldom is the lack of an Optical drive raised.
Quote:
I imagine that we'll see a version of the i3, or possible the i5 in a later MBP. I'm not worried about the graphics performance. See my above post.
If you aren't worried about GPU performance then I honestly think you are missing the direction that Apple and the industry are turning. Frankly with the somtimes sluggishness of 10.6.3 they need to use whatever resources they can to speed it up.
I have no doubt people want them. People wanted Floppy Drives, Serial and Parallel Ports, too, when Apple removed them in favour of USB and Optical Drives. What I don't think people realize is how often they, as consumers, are actually using these large power-hungry components in their modern notebooks.
Especially in smaller laptops where the Optical may take up a good portion of the available space.
Quote:
I do think that Apple will likely have to give up the CPUs with the Northbridge in the next revision, which in turn means the next 13" MBP will need to get a discrete CPU. Unless they pull some major engineering feat like it appears they did with the next generation iPhone logic board I think that finally means the ODD moving to the outside of the device.
Or they could simply go AMD and send Intel a no thank you message for Arrandale. AMD has some very good "integrated" GPUs that would go well in a notebook. By the time the next round of AMD mobile chips hit the street performance should be significantly better than the Core 2.
Quote:
Note that removing the ODD doesn't mean that ODDs can't be used or that they would be removed from all machines. The fact that Apple hasn't supported Blu-ray may not back up this eventual more to remove the ODD but it certainly doesn't hurt the argument.
The more I think about this the more I think that it should be a snap for Apple to add an Optical-less 13" MBP. All they would really need to do is to programtheir CNC machine to skip cutting the Optical slot and make a thinner bottom cover. Even if a slot is added to support another "disk drive" it should still be a snap. So Apple ends up with a thin MBP with a sister with a fat bottom.
Quote:
Just an idea: While they current MBPs take a 12.7mm HDD, dropping it to only take a 9.5mm HDD would shave 3mm off the case size. If they went to a dual drive system using 7mm drives that more than make up for the smaller HDD drives capacity. It's also the standard for SSDs and a size you can get HDDs in now. Of course this would require the removal of the ODD.
Yep an excellent idea! Due to the manufacturing process it should be extremely easy for them to do.
Quote:
Note that the largest capacity drive in the new MBPs is a 512GB HDD. Why didn't they go with 640GB, 750GB or even 1TB in the HDD. I think Apple will be pressing SSDs harder and harder and that this MBP revision is just an overdue stopgate for a major HW revision. The NAND nanometer size will show a 50% increase in capacity this year as well as a major price drop.
Unfortunately I think this just Apple being Apple. Considering recent history with laptop HHD manufactures it probably pays for Apple to go with drives with a track record. I've tried the bleeding edge myself for desktop drives and believe me getting burned on storage is no fun.
Quote:
I know we have never come close to seeing eye-to-eye on this, but I can't see how the ODD is a long term option for notebooks when it's doing nothing but holding back notebooks in some many areas while offering so little use to most notebook users.
in the context if smaller machines I agree 100%. As long as the laptop has plenty of ports an external drive is a OK solution. A laptop with 2 or more internal storage slots is even better. I say slots here because I really think that modern SSD need to be made available as small PCI Express PC cards. What is on the card doesn't matter, it can be flash or one of the new technologies. The idea is for very thin, that is less than 5mm, storage cards that are truely next gen.
…since most workloads are integer-heavy. And FP-intensive work is gradually moving to the GPU.
If you read the articles on this, you will see that AMD is stating that they BELIEVE that to be true. That doesn't make it true. And they're talking about server workloads, not the average for what most people do.
Quote:
AMD believes that 80%+ of all normal server workloads are purely integer operations.
No, it is 5% for the whole die, counting caches, memory controllers, etc. and ~12% for the second integer core on a module.
I'd like to see a link to that. They wouldn't be making a big deal about it if it were only increasing the dies size by 5% for the extra integer cores.
Quote:
AMD has come back to us with a clarification: the 5% figure was incorrect. AMD is now stating that the additional core in Bulldozer requires approximately an additional 50% die area. That's less than a complete doubling of die size for two cores, but still much more than something like Hyper Threading.
Same article
Quote:
If you've looked at AMD's recent dies, the cores don't take up as much of the die compared to Intel's.
I'd rather see a statement of umbers, not take a look at the pictures.
Quote:
Thuban won't be up against Gulftown, it's price (and TDP) is closer to Lynnfield/Gulftown. Next year, it'll be 8-core Zambezi, that is if AMD doesn't raise their prices.
It's possible. But with AMD's record, they may not get it out on time anyway, and then they will be competing with upgraded chips from Intel as has happened the past two years.
Quote:
The 80% number makes sense, as it's an entire second integer core. Two cores would give close to 100% performance increase in multithreaded situations.
They might give UP TO 80% in multithreaded situations. That's what AMD is saying. These aren't full cores, integer or otherwise. Read the article and a couple of others. You'll see.
Quote:
SMT, CMT, and dual-core give various amounts of performance improvements for various increases in die size. AMD thought that CMT gave the most performance per core size increase compared to the other ways.
Of course they did.
Quote:
Bulldozer also features FP improvements. (Sandy Bridge also does.)
It's got poor FP performance compated to Intel, and they knoe it.
Quote:
AMD claims that the performance benefit from the second integer core on a single Bulldozer module is up to 80% on threaded code. That's more than what AMD could get through something like Hyper Threading, but as we've recently found out the impact to die size is not negligible. It really boils down to the sorts of workloads AMD will be running on Bulldozer. If they are indeed mostly integer, then the performance per die area will be quite good and the tradeoff worth it. Part of the integer/FP balance does depend on how quickly the world embraces computing on the GPU however...
That pretty much says a lot. If Intel has the same core number, it will always beat these chips, easily.
Quote:
That's true, but now things are reversed.
Yes, funny that AMD would now use the inferior technology, and suddenly claim it's better.
Quote:
No, since more apps are moving to multithreaded. There are situations where single-threaded performance is important now and in the future. Now those apps are taken care of by more powerful cores (Intel), Turbo Boost and Turbo Core. In the future those apps will be taken care of by a few relatively large cores. In that respect Intel may have the advantage because their cores are already relatively large. However AMD may have the multithreaded FP advantage with Fusion.
Most apps we use still can't use more than two cores. That will be true for a while. And when they do, it's rarely more than four.
It will be years before software catches up. And that includes using the GPU for most of the FP, or vector math. The OS's are using it more, led by Apple, but programs are trailing very slowly.
Quote:
Not at all in servers, and in the desktop they are usually behind in at least one if you are looking at multithreaded tasks.
AMD has done well in HPC. It's an area that Intel has ignored for several years. Look at the power ratings in various Anandtech articles rating Intel and AMD. Intel is ahead virtually all the time when comparing comparable chis, and even some that are not so comparable.
Quote:
Intel has the advantage with Clarkdale/Arrandale's integration, but that will be erased with Llano.
I'm still not convinced.
Quote:
That will be when they are being pressured by AMD.
Thats very funny, because it's usually described as Intel being able to afford dropping prices, and they do it to put pressure on AMD who can't afford to.
Quote:
I doubt it. Not when Interlagos is expected to be over 33% increased performance over Magny-Cours for 33% more cores. So Intel will pretty much end up where they are this year, next year (and from the speculations I've heard that will be the best case scenario for Intel in servers).
Really?
Quote:
There's not much to talk about from a CPU standpoint with AMD in 2010, so AMD is heavily focused on 2011 and what it plans to do with its first on-die GPU in Llano. Bobcat and Bulldozer also make it back into the headlines as AMD is long overdue for another microprocessor architecture. Bobcat stands to be AMD's first competitive mobile architecture while Bulldozer may ensure AMD will be competitive at the high end.
I love the "may". It's AMD we're talking about here. Nothing's for sure.
Seriously I think people see the 13" machine as an excellent value. Part of that is because consumers are more informed about the significance of having a good GPU on a machine running Mac OS/X.
Yes but it is being sold out with a GPU that is significantly faster than Intels. We aren't talking a little bit faster either, it is significantly faster with video acceleration and good OpenCL support to boot.
What I've said here several times, is that it doesn't matter because Apple did a very elegant job. It's not how you get there, but whether you DO get there. Apple has.
Quote:
Now I would agree with you that many consumers don't care about the techie details. But they will respond to the idea of video acceleration.
They will respond to the machine doing what they want it to. They don't care how or why. The people buying the 13" MBP simply want to get the aluminum machine at a price that's not that much more than the plastic MB. Their desires aren't that much different as to use.
Quote:
The whole point here is that solutions that use multiple GPUs are only worthwhile if they can fit into the hardware in question. For the hardware that Apple seems to want to build Intels integrated GPU (in the CPU module) is no better than AMDs. In fact all signs indicates it is worst.
I don't think that that's ever a real problem. If Apple wasn't to do it, they can. They're often using Nvidia's which are better than either.
Actually I think it is a huge issue for Apple. They don't want to be seen as going backwards performance wise. Thus we get a much faster replacement for the 9400m and modest CPU improvements. All of this in an existing platform.
They need progressively better hardware for marketing reasons as it is hard to sell a machine that is slower than last years model. They also need that GPU for OpenCL which is a lot harder to market to consummers. OpenCL is a key to keeping the Mac competitive OS wise.
A side ways look at this would be the use of the GPU in the iPhone and iPad which is extensive. If Apple can get as much of Quartz running on the GPU on the Mac as it is on iPhone OS we could see major improvements in Mac responsiveness. In other words I expect Apple to continue to try to exploit the GPU as much as possible most likely leading to resolution independence and a few other long sought goals. Arrandale alone in such an environment would be a very low end solution.
I haven't seen that poll but frankly most online polls are worthless as they don't represent the entire user community. In any event the lack of an internal drive does not imply a lack of support for Optical drives in the OS.
The best example here is Apples own MBA with no internal Optical. AIR is a machine that is rightly skewered for some of the design choices made with the computer but seldom is the lack of an Optical drive raised.
Dave
You see, we've got three or four (at most) guys here who keep talking about getting rid of internal optical drives, but a much larger number who say no. You forget that. This poll is at least as valid as the numbers here when we have these discussions (of which, this hasn't become one yet). Not all online polls are bad. Polls such as this are often pretty good. There's no real reason for people to get excited about it, so they will vote with their mind, rather than with their heart.
You see, we've got three or four (at most) guys here who keep talking about getting rid of internal optical drives, but a much larger number who say no. You forget that.
Not really, because I'm looking at the small segment of the market covered by the 13" MBP which I consider to be an ultra mobile laptop. Here I honestly believe the users value mobility and long battery life over the ability to use an Optical drive.
One only needs to look at net books and other small laptops to see the trend here.
Quote:
This poll is at least as valid as the numbers here when we have these discussions (of which, this hasn't become one yet). Not all online polls are bad.
I'd have to say all are skewed. It would be pretty tough for them not to be. Think about a poll run on Appleinsider, the only people responding to the poll will be the small subset of Mac users that actually come here and also bother with the polls. That is an exceedingly small number of Apple customers.
Quote:
Polls such as this are often pretty good. There's no real reason for people to get excited about it, so they will vote with their mind, rather than with their heart.
They are good in the sense that they represent the small minority of people willing to go online to a fan site and then are willing to bother with polls. That doesn't always reflect what the market at large does.
IPad is a perfectly good example here. The thing sells well even though you wouldn't think so based on some of the polls taken on various sites. Wether it was the lack of Mac OS/X or some other feature (such as the paltry RAM in my case) people still are buying the unit in vast numbers. That may be related to early adopters and as such could roll off quickly. Or it could mean that the forums simply don't reflect the mass market.
In any event I still maintain that droping the internal Optical from the 13" MBP would allow Apple to configure the machine in a way that would make it even more attractive to the target market.
If you read the articles on this, you will see that AMD is stating that they BELIEVE that to be true. That doesn't make it true. And they're talking about server workloads, not the average for what most people do.
So you have no problem with HyperThreading?s large variety of performance increases (or even decreases) depending on the workload?
Intel says that dual-core mobile Sandy Bridge will be 20% faster than Arrandale. So is that what Intel BELIEVES to be true? And does that make Sandy Bridge?s ?real? performance increase over Westmere 10%?
Oh, and AMD BELIEVED that Magny-Cours would give 50%-60% more performance than Istanbul, they actually got 80%-119% more.
Quote:
Originally Posted by melgross
I'd like to see a link to that. They wouldn't be making a big deal about it if it were only increasing the dies size by 5% for the extra integer cores.
The OP there works for AMD so he knows the numbers.
Quote:
Originally Posted by melgross
Same article
See article.
Quote:
Originally Posted by melgross
I'd rather see a statement of umbers, not take a look at the pictures.
Measure them yourself. (Hint: One company?s cores are 50% bigger than the other company?s on the same process.)
It's possible. But with AMD's record, they may not get it out on time anyway, and then they will be competing with upgraded chips from Intel as has happened the past two years.[/quote]I?m assuming you?re fine with Q3 2011 for Sandy Bridge E?
Quote:
Originally Posted by melgross
They might give UP TO 80% in multithreaded situations. That's what AMD is saying. These aren't full cores, integer or otherwise. Read the article and a couple of others. You'll see.
I did. It says 2 integer cores, each of them more powerful than a K10 integer core. Hence my ?over 33% performance increase? comment on my last post.
And why not ?up to 35%? for HyperThreading? Because according to the link you posted, that?s exactly what it is.
Of course they did.[/quote]Exactly.
Quote:
Originally Posted by melgross
It's got poor FP performance compated to Intel, and they knoe it.
?because they will be offloading FP instructions to the GPU.
Quote:
Originally Posted by melgross
That pretty much says a lot. If Intel has the same core number, it will always beat these chips, easily.
They won?t have the same core counts at same prices or same TDPs. As I?ve shown, AMD will generally have higher core counts than Intel at equivalent prices/TDP. That?s been my main point here, and once I refuted your counterpoint you evade the situation entirely.
With Sandy Bridge, Intel?s not increasing core counts at all, except in two areas (high-end desktop and server), while AMD has slowly done so in all segments, especially servers. In servers AMD has double the cores of Intel, and that?ll continue next year. In desktops, if Thuban?s prices are any indication, Zambezi will be priced against 6-core, maybe 4-core Sandy Bridge. Llano brings quad-core down to even lower TDPs (as if 25 W Champlain won?t be enough).
So what ends up from there is that AMD can make a CPU comparable to (sometimes worse, sometimes better, depending on the segment) Intel?s in multithreaded situations by adding more cores, even though Intel spent all the work designing one or two new microarchitectures (Nehalem and Sandy Bridge). That?s AMD?s small-core strategy, whether accidental (because they didn?t have the resources to make a new microarchitecture) or deliberate. More cores for less money and lower TDP.
Intel has and will have single-threaded advantage, AMD will have multithreaded advantage (they already do in a number of cases), and AMD will have the GPU advantage. AMD took a big hit in 2006-2009, but from 2010 the tables are slowly turning.
AMD won?t have the fastest CPU in notebooks and desktops, even with Bulldozer. But for Apple, that?s not an issue because they aren?t using the fastest notebook and desktop CPUs anyway.
Quote:
Originally Posted by melgross
Yes, funny that AMD would now use the inferior technology, and suddenly claim it's better.
Doesn?t matter because now they have the advantage. Or how about this? Why didn?t Intel continue using their superior dual-die approach? 12-core Westmere would beat 12-core Magny-Cours, no questions there.
Quote:
Originally Posted by melgross
Most apps we use still can't use more than two cores. That will be true for a while. And when they do, it's rarely more than four.
How many of those apps use 100% of a core? I?d like more cores for video encoding and other intensive tasks thank you very much.
Plus with only 2 or 4 task threads the HyperThreading you mentioned won?t be useful on a number of Intel CPUs.
Quote:
Originally Posted by melgross
It will be years before software catches up. And that includes using the GPU for most of the FP, or vector math. The OS's are using it more, led by Apple, but programs are trailing very slowly.
That?s why Apple hasn?t used AMD CPUs yet. I am looking at the future. Many people who criticize AMD?s chips are looking at only the past or present. When software does catch up (in addition to the software that already use multiple cores and GPUs), AMD will be in a very good position. What about Intel?
AMD has done well in HPC. It's an area that Intel has ignored for several years. Look at the power ratings in various Anandtech articles rating Intel and AMD. Intel is ahead virtually all the time when comparing comparable chis, and even some that are not so comparable.[/quote]I?ve read them. Magny-Cours is close to Westmere in performance. And you?re missing price again. Magny-Cours is cheaper.
Quote:
Originally Posted by melgross
I'm still not convinced.
30 W quad-core with 480 shader GPU.
Quote:
Originally Posted by melgross
Really?
Go read the article I linked to. Over 33% more integer performance for 33% more cores. Bulldozer also increases FP performance so one module would be faster than two K10 cores. Oh, and for about the die area of one K10 core.
Not really, because I'm looking at the small segment of the market covered by the 13" MBP which I consider to be an ultra mobile laptop. Here I honestly believe the users value mobility and long battery life over the ability to use an Optical drive.
One only needs to look at net books and other small laptops to see the trend here.
You aren't seriously comparing the 13" MBP to a netbook?
Quote:
I'd have to say all are skewed. It would be pretty tough for them not to be. Think about a poll run on Appleinsider, the only people responding to the poll will be the small subset of Mac users that actually come here and also bother with the polls. That is an exceedingly small number of Apple customers. [/qote]
I don't agree. If there is just one poll, and nowhere else is this discussed, then I might doubt that one poll with nothing else to substantiate it.
But we've had a long history here of discussing the optical-less laptop, and most people here aren't yet ready for it other than perhaps for the Air.
Quote:
They are good in the sense that they represent the small minority of people willing to go online to a fan site and then are willing to bother with polls. That doesn't always reflect what the market at large does.
Well, again, it just agrees with what most people say here. At some point you just have to see the tide is against you.
Quote:
IPad is a perfectly good example here. The thing sells well even though you wouldn't think so based on some of the polls taken on various sites. Wether it was the lack of Mac OS/X or some other feature (such as the paltry RAM in my case) people still are buying the unit in vast numbers. That may be related to early adopters and as such could roll off quickly. Or it could mean that the forums simply don't reflect the mass market.
Again, I don't agree. It's a totally different kind of thing. I 20% of the buying public say they would buy something, that's pretty darned good. It would lead to tens of millions of sales. You're confusing the meaning of the different kinds of polls (or surveys).
[uote]
In any event I still maintain that droping the internal Optical from the 13" MBP would allow Apple to configure the machine in a way that would make it even more attractive to the target market.
Dave
It definitely be a different kind of machine. I can agree with that.
So you have no problem with HyperThreading?s large variety of performance increases (or even decreases) depending on the workload?
Intel says that dual-core mobile Sandy Bridge will be 20% faster than Arrandale. So is that what Intel BELIEVES to be true? And does that make Sandy Bridge?s ?real? performance increase over Westmere 10%?
Except for a short time several years ago, Intel has had an excellent record of beating its predictions of what its new chips could do. You can look at the reviews of them. But over most of its life, AMD has not. It's history. And the past two years have been very bad for AMD.
Quote:
Oh, and AMD BELIEVED that Magny-Cours would give 50%-60% more performance than Istanbul, they actually got 80%-119% more.
That link doesn't lead to any test that I could see. Could you be more specific with it?
Quote:
The OP there works for AMD so he knows the numbers.
You'll have to show the tests.
Quote:
Measure them yourself. (Hint: One company?s cores are 50% bigger than the other company?s on the same process.)
That's not really possible. But I am seeing conflicting numbers.
Quote:
It's possible. But with AMD's record, they may not get it out on time anyway, and then they will be competing with upgraded chips from Intel as has happened the past two years.
I?m assuming you?re fine with Q3 2011 for Sandy Bridge E? [/quote]
I'm concerned that AMD will again have the situation they had with Barcelona, and Phenom. You remember that fiasco?
Of course, no one is happy about Sandy Bridge being pushed back three months. It does give AMD some breathing time.
Quote:
I did. It says 2 integer cores, each of them more powerful than a K10 integer core. Hence my ?over 33% performance increase? comment on my last post.
Yes, I understand.
Quote:
And why not ?up to 35%? for HyperThreading? Because according to the link you posted, that?s exactly what it is.
You can say that. but testing has shown that most of the time it comes pretty close, and sometimes has exceeded it. Again, you can go to Anandtech.
Quote:
?because they will be offloading FP instructions to the GPU. [/qote]
Yes, we know that. It's what AMD hopes. But you also must have rad that they are taking a chance with this, because, so far, very few programs are taking advantage of the GPU. Even Adobe CS5 doesn't make much use of it, and of all the programs out there, it needs it really badly.
In addition, AMD can't count on knowing which GPU will be in any give computer. If it's a fast one that can e be used for this, then all may be well. But most computers manage with integrated graphics. It believe that's about 80% of all machines out there. They'll be getting no help from that GPU.
That means that machines that are being used for big spreadsheets (for example) that may need FP won't be getting any real FP punch at all.
Quote:
They won?t have the same core counts at same prices or same TDPs. As I?ve shown, AMD will generally have higher core counts than Intel at equivalent prices/TDP. That?s been my main point here, and once I refuted your counterpoint you evade the situation entirely.
I never evade anything. I've already stated that I don't agree with that either. Intel can afford to be very aggressive and keep AMD's prices down where their profits are cut short. AMD competes, ironically, best on the HPC high end, and on the low end. Everywhere else, they can't compete. So if you are talking about the cheapest chips, then, maybe, assuming that their stuff comes out when it should, and it's as good as they say, then, for a while, they'll be competitive at that point, with those chips.
Quote:
With Sandy Bridge, Intel?s not increasing core counts at all, except in two areas (high-end desktop and server), while AMD has slowly done so in all segments, especially servers. In servers AMD has double the cores of Intel, and that?ll continue next year. In desktops, if Thuban?s prices are any indication, Zambezi will be priced against 6-core, maybe 4-core Sandy Bridge. Llano brings quad-core down to even lower TDPs (as if 25 W Champlain won?t be enough).
AMD has had to because core for core, they can't compete.
Quote:
So what ends up from there is that AMD can make a CPU comparable to (sometimes worse, sometimes better, depending on the segment) Intel?s in multithreaded situations by adding more cores, even though Intel spent all the work designing one or two new microarchitectures (Nehalem and Sandy Bridge). That?s AMD?s small-core strategy, whether accidental (because they didn?t have the resources to make a new microarchitecture) or deliberate. More cores for less money and lower TDP.
They have no choice. They're getting clobbered core for core. They need more cores just to keep up, or make a small advance over Intel. But you keep forgetting that Intel drops the process several times over the lifetime of these chips, and AMD is forced to follow.
Quote:
Intel has and will have single-threaded advantage, AMD will have multithreaded advantage (they already do in a number of cases), and AMD will have the GPU advantage. AMD took a big hit in 2006-2009, but from 2010 the tables are slowly turning.
AMD won?t have the fastest CPU in notebooks and desktops, even with Bulldozer. But for Apple, that?s not an issue because they aren?t using the fastest notebook and desktop CPUs anyway.
Doesn?t matter because now they have the advantage. Or how about this? Why didn?t Intel continue using their superior dual-die approach? 12-core Westmere would beat 12-core Magny-Cours, no questions there.
How many of those apps use 100% of a core? I?d like more cores for video encoding and other intensive tasks thank you very much.
Plus with only 2 or 4 task threads the HyperThreading you mentioned won?t be useful on a number of Intel CPUs.
That?s why Apple hasn?t used AMD CPUs yet. I am looking at the future. Many people who criticize AMD?s chips are looking at only the past or present. When software does catch up (in addition to the software that already use multiple cores and GPUs), AMD will be in a very good position. What about Intel?
AMD has done well in HPC. It's an area that Intel has ignored for several years. Look at the power ratings in various Anandtech articles rating Intel and AMD. Intel is ahead virtually all the time when comparing comparable chis, and even some that are not so comparable.
I?ve read them. Magny-Cours is close to Westmere in performance. And you?re missing price again. Magny-Cours is cheaper.
30 W quad-core with 480 shader GPU.
Go read the article I linked to. Over 33% more integer performance for 33% more cores. Bulldozer also increases FP performance so one module would be faster than two K10 cores. Oh, and for about the die area of one K10 core.
That's all rehashing earlier stuff. Look, I hope that AMD stays around, and is competitive, but right now, they aren't. With Barcelona, they promised the world, 40% better performance per MHz, and what did they deliver 6 months late? Chips that not only were at lower speeds, but were 20% SLOWER per MHz!. This carried on to the Phnom as well, though not quite as badly.
This is what I'm concerned about. They will have to show that they won't repeat that. When Intel backs off a few months, the industry doesn't get concerned. It's usually because they'r looking back at AMD and don't see much competition, so they figure they can make some more money on current designs. When AMD is late, everyone is thinking that they are having major problems.
Let's see how AMD does and if they show any signs of closing the gap with Intel chips in performance.
I'm not seeing anything with the Phenom II that would make Apple want to switch. They really need a mobile contender, which needs to be more than just a good price for the performance but good power usage for the performance.
I'm not seeing anything with the Phenom II that would make Apple want to switch. They really need a mobile contender, which needs to be more than just a good price for the performance but good power usage for the performance.
They wouldn't want phenom II. They'd want a system on chip solution using fusion and AMD's far superior graphics technology.
You aren't seriously comparing the 13" MBP to a netbook?
No not exactly what I'm saying here is that it has become clear that people are willing to give up certain technologies to get an extremely mobile PC. An Optical drive being one thing people find that they don't need.
Quote:
It definitely be a different kind of machine. I can agree with that.
Think of it as a device that corrects the mistakes made on AIR! In fact that is a very good way to look at it. It would be a laptop where what can be deleted is, but the stuff that many of us believe is important is still there. The machine would stress good performance and battery life.
Given the new AMD six core chips are actually beating i7s at some workloads, and cost less than one third of Intel's prices, it may be the high end we see AMDs in. OS X is all about parallel computing, from threading, Grand Central, OpenCL, to distributed compilations. If apple can knock $700 (6 core) or $1400 (12 cores) off the price of the Mac Pro, wouldn't they?
I'd buy one in a second. 3D rendering is my thing.
Given the new AMD six core chips are actually beating i7s at some workloads, and cost less than one third of Intel's prices, it may be the high end we see AMDs in. OS X is all about parallel computing, from threading, Grand Central, OpenCL, to distributed compilations. If apple can knock $700 (6 core) or $1400 (12 cores) off the price of the Mac Pro, wouldn't they?
I'd buy one in a second. 3D rendering is my thing.
What an outstanding value. I think Apple would be crazy not to start delivering some AMD solutions next year with Llano. I like Intel as much as anyone but truth be told a lot of my work leads to the computer sitting here waiting on me. I think AMD certainly delivers a lot of processing power per buck. I'm content to let the "I gotta have the fastest computer available" people spend a premium on Intel based Macs. I'm more of a system balance kind of guy. Which is why I like SSD and will like faster GPU when they represent more than just increased fps in games.
Given the new AMD six core chips are actually beating i7s at some workloads, and cost less than one third of Intel's prices, it may be the high end we see AMDs in. OS X is all about parallel computing, from threading, Grand Central, OpenCL, to distributed compilations. If apple can knock $700 (6 core) or $1400 (12 cores) off the price of the Mac Pro, wouldn't they?
Yeah, I saw a review of the new 6-core AMD this morning and the i7-980 was 30% faster in some tests and I was ready to have the usual dismissive reaction but the price difference hit home. The chips were 30% slower only in the worst case but are under 1/3 of the price so as you say, Apple could build a 12-core Mac Pro $1400 cheaper than they could with Intel's chips (depending on which Intel ones they chose).
I'd much rather have an $1800 6-core AMD Mac Pro than an Intel one at $2500 when you're only going to see 30% improvement doing the long rendering stuff, which you're going to be waiting for anyway.
Plus, the 12-core AMD machine would be cheaper than a 6-core Intel so in terms of performance-per-dollar, the AMD offering will win in most cases.
I wonder if Apple will make the jump at this refresh. I expected the Mac Pro update to be out by now.
I really wish they'd make a smaller one. 12-core, 2/3rds the size and maybe 12kg instead of 18kg.
You aren't seriously comparing the 13" MBP to a netbook?
No, he's comparing them to the 13" CULV powered machines like the Thinkpad Edge, HP DM3, and others that use 1.2-1.6Ghz Intel and AMD CPUs and deliver performance halfway between a netbook and a full sized notebook like a Macbook Pro. What he neglects to account for is that the current Macbook Pro delivers better battery life than any of them while being around the same weight with significantly more power. Again, this is an example of the machine a user wants existing and being readily available from anyone but Apple.
Comments
AMD made the point that FP is secondary with these chips. The cores are integer. There is FP hardware, but it's fairly weak.
?since most workloads are integer-heavy. And FP-intensive work is gradually moving to the GPU.
It's 50%, and it doesn't add 5% to the die. It's more like 20% per core. Intel's hyperthreading adds 5%, thats what you're thinking about.
No, it is 5% for the whole die, counting caches, memory controllers, etc. and ~12% for the second integer core on a module.
If you've looked at AMD's recent dies, the cores don't take up as much of the die compared to Intel's.
And we have 4 core Intel up against 4 core AMD, and six core Intel against 6 core AMD. Intel wins every time there, and not by a little.
Thuban won't be up against Gulftown, it's price (and TDP) is closer to Lynnfield/Gulftown. Next year, it'll be 8-core Zambezi, that is if AMD doesn't raise their prices.
The extra core does give what AMD CLAIMS is an 80% increase in INTEGER performance per core, as opposed to the known quantity of 30 to 40% improvement per core for hyperthreading. It also uses much more power, and needs more heat dissipation that Intel's chips.
The 80% number makes sense, as it's an entire second integer core. Two cores would give close to 100% performance increase in multithreaded situations.
SMT, CMT, and dual-core give various amounts of performance improvements for various increases in die size. AMD thought that CMT gave the most performance per core size increase compared to the other ways.
Bulldozer also features FP improvements. (Sandy Bridge also does.)
It's very funny, you know, because when Intel first began using the dual die per package, AMD was criticizing them for it.
That's true, but now things are reversed.
Thats totally wrong. However many cores are there, per core performance is as important as it ever was.
No, since more apps are moving to multithreaded. There are situations where single-threaded performance is important now and in the future. Now those apps are taken care of by more powerful cores (Intel), Turbo Boost and Turbo Core. In the future those apps will be taken care of by a few relatively large cores. In that respect Intel may have the advantage because their cores are already relatively large. However AMD may have the multithreaded FP advantage with Fusion.
Intel's ahead on the P/W spec, and is usually ahead on the P/$ as well.
Not at all in servers, and in the desktop they are usually behind in at least one if you are looking at multithreaded tasks.
Intel has the advantage with Clarkdale/Arrandale's integration, but that will be erased with Llano.
As usual, Intel will drop prices on its chips, putting pressure on AMD.
That will be when they are being pressured by AMD.
I'm willing to bet that Sandy Bridge will beat the pants off Bulldozer.
I doubt it. Not when Interlagos is expected to be over 33% increased performance over Magny-Cours for 33% more cores. So Intel will pretty much end up where they are this year, next year (and from the speculations I've heard that will be the best case scenario for Intel in servers).
I don't think it's a problem for the devices that will be using these chips. It's overblown. Look at the 13" MBP. It was sold out at numerous locations the first day it went on sale. That's what matters.
Yes but it is being sold out with a GPU that is significantly faster than Intels. We aren't talking a little bit faster either, it is significantly faster with video acceleration and good OpenCL support to boot.
Now I would agree with you that many consumers don't care about the techie details. But they will respond to the idea of video acceleration.
These will be used for consumers who don't care about another 25% of graphics performance. For those who do, AMD's on chip solution is no good anyway, and something along the lines of what Nvidia, or better yet, Apple did, will be far better than worrying about Intel's IP here.
The whole point here is that solutions that use multiple GPUs are only worthwhile if they can fit into the hardware in question. For the hardware that Apple seems to want to build Intels integrated GPU (in the CPU module) is no better than AMDs. In fact all signs indicates it is worst.
It's really a non issue, except on the tech boards.
Actually I think it is a huge issue for Apple. They don't want to be seen as going backwards performance wise. Thus we get a much faster replacement for the 9400m and modest CPU improvements. All of this in an existing platform.
They need progressively better hardware for marketing reasons as it is hard to sell a machine that is slower than last years model. They also need that GPU for OpenCL which is a lot harder to market to consummers. OpenCL is a key to keeping the Mac competitive OS wise.
A side ways look at this would be the use of the GPU in the iPhone and iPad which is extensive. If Apple can get as much of Quartz running on the GPU on the Mac as it is on iPhone OS we could see major improvements in Mac responsiveness. In other words I expect Apple to continue to try to exploit the GPU as much as possible most likely leading to resolution independence and a few other long sought goals. Arrandale alone in such an environment would be a very low end solution.
Dave
First of all forget about trashing the optical drive. Most people want them.
Yes but not as internal drives.
There's a survey at either Macsimumnews or Macnn, I forget which, that showed that 78% of those polled wanted optical drives. I believe it.
I haven't seen that poll but frankly most online polls are worthless as they don't represent the entire user community. In any event the lack of an internal drive does not imply a lack of support for Optical drives in the OS.
The best example here is Apples own MBA with no internal Optical. AIR is a machine that is rightly skewered for some of the design choices made with the computer but seldom is the lack of an Optical drive raised.
I imagine that we'll see a version of the i3, or possible the i5 in a later MBP. I'm not worried about the graphics performance. See my above post.
If you aren't worried about GPU performance then I honestly think you are missing the direction that Apple and the industry are turning. Frankly with the somtimes sluggishness of 10.6.3 they need to use whatever resources they can to speed it up.
Dave
I have no doubt people want them. People wanted Floppy Drives, Serial and Parallel Ports, too, when Apple removed them in favour of USB and Optical Drives. What I don't think people realize is how often they, as consumers, are actually using these large power-hungry components in their modern notebooks.
Especially in smaller laptops where the Optical may take up a good portion of the available space.
I do think that Apple will likely have to give up the CPUs with the Northbridge in the next revision, which in turn means the next 13" MBP will need to get a discrete CPU. Unless they pull some major engineering feat like it appears they did with the next generation iPhone logic board I think that finally means the ODD moving to the outside of the device.
Or they could simply go AMD and send Intel a no thank you message for Arrandale. AMD has some very good "integrated" GPUs that would go well in a notebook. By the time the next round of AMD mobile chips hit the street performance should be significantly better than the Core 2.
Note that removing the ODD doesn't mean that ODDs can't be used or that they would be removed from all machines. The fact that Apple hasn't supported Blu-ray may not back up this eventual more to remove the ODD but it certainly doesn't hurt the argument.
The more I think about this the more I think that it should be a snap for Apple to add an Optical-less 13" MBP. All they would really need to do is to programtheir CNC machine to skip cutting the Optical slot and make a thinner bottom cover. Even if a slot is added to support another "disk drive" it should still be a snap. So Apple ends up with a thin MBP with a sister with a fat bottom.
Just an idea: While they current MBPs take a 12.7mm HDD, dropping it to only take a 9.5mm HDD would shave 3mm off the case size. If they went to a dual drive system using 7mm drives that more than make up for the smaller HDD drives capacity. It's also the standard for SSDs and a size you can get HDDs in now. Of course this would require the removal of the ODD.
Yep an excellent idea! Due to the manufacturing process it should be extremely easy for them to do.
Note that the largest capacity drive in the new MBPs is a 512GB HDD. Why didn't they go with 640GB, 750GB or even 1TB in the HDD. I think Apple will be pressing SSDs harder and harder and that this MBP revision is just an overdue stopgate for a major HW revision. The NAND nanometer size will show a 50% increase in capacity this year as well as a major price drop.
Unfortunately I think this just Apple being Apple. Considering recent history with laptop HHD manufactures it probably pays for Apple to go with drives with a track record. I've tried the bleeding edge myself for desktop drives and believe me getting burned on storage is no fun.
I know we have never come close to seeing eye-to-eye on this, but I can't see how the ODD is a long term option for notebooks when it's doing nothing but holding back notebooks in some many areas while offering so little use to most notebook users.
in the context if smaller machines I agree 100%. As long as the laptop has plenty of ports an external drive is a OK solution. A laptop with 2 or more internal storage slots is even better. I say slots here because I really think that modern SSD need to be made available as small PCI Express PC cards. What is on the card doesn't matter, it can be flash or one of the new technologies. The idea is for very thin, that is less than 5mm, storage cards that are truely next gen.
Dave
…since most workloads are integer-heavy. And FP-intensive work is gradually moving to the GPU.
If you read the articles on this, you will see that AMD is stating that they BELIEVE that to be true. That doesn't make it true. And they're talking about server workloads, not the average for what most people do.
AMD believes that 80%+ of all normal server workloads are purely integer operations.
http://www.anandtech.com/show/2881
No, it is 5% for the whole die, counting caches, memory controllers, etc. and ~12% for the second integer core on a module.
I'd like to see a link to that. They wouldn't be making a big deal about it if it were only increasing the dies size by 5% for the extra integer cores.
AMD has come back to us with a clarification: the 5% figure was incorrect. AMD is now stating that the additional core in Bulldozer requires approximately an additional 50% die area. That's less than a complete doubling of die size for two cores, but still much more than something like Hyper Threading.
Same article
If you've looked at AMD's recent dies, the cores don't take up as much of the die compared to Intel's.
I'd rather see a statement of umbers, not take a look at the pictures.
Thuban won't be up against Gulftown, it's price (and TDP) is closer to Lynnfield/Gulftown. Next year, it'll be 8-core Zambezi, that is if AMD doesn't raise their prices.
It's possible. But with AMD's record, they may not get it out on time anyway, and then they will be competing with upgraded chips from Intel as has happened the past two years.
The 80% number makes sense, as it's an entire second integer core. Two cores would give close to 100% performance increase in multithreaded situations.
They might give UP TO 80% in multithreaded situations. That's what AMD is saying. These aren't full cores, integer or otherwise. Read the article and a couple of others. You'll see.
SMT, CMT, and dual-core give various amounts of performance improvements for various increases in die size. AMD thought that CMT gave the most performance per core size increase compared to the other ways.
Of course they did.
Bulldozer also features FP improvements. (Sandy Bridge also does.)
It's got poor FP performance compated to Intel, and they knoe it.
AMD claims that the performance benefit from the second integer core on a single Bulldozer module is up to 80% on threaded code. That's more than what AMD could get through something like Hyper Threading, but as we've recently found out the impact to die size is not negligible. It really boils down to the sorts of workloads AMD will be running on Bulldozer. If they are indeed mostly integer, then the performance per die area will be quite good and the tradeoff worth it. Part of the integer/FP balance does depend on how quickly the world embraces computing on the GPU however...
That pretty much says a lot. If Intel has the same core number, it will always beat these chips, easily.
That's true, but now things are reversed.
Yes, funny that AMD would now use the inferior technology, and suddenly claim it's better.
No, since more apps are moving to multithreaded. There are situations where single-threaded performance is important now and in the future. Now those apps are taken care of by more powerful cores (Intel), Turbo Boost and Turbo Core. In the future those apps will be taken care of by a few relatively large cores. In that respect Intel may have the advantage because their cores are already relatively large. However AMD may have the multithreaded FP advantage with Fusion.
Most apps we use still can't use more than two cores. That will be true for a while. And when they do, it's rarely more than four.
It will be years before software catches up. And that includes using the GPU for most of the FP, or vector math. The OS's are using it more, led by Apple, but programs are trailing very slowly.
Not at all in servers, and in the desktop they are usually behind in at least one if you are looking at multithreaded tasks.
AMD has done well in HPC. It's an area that Intel has ignored for several years. Look at the power ratings in various Anandtech articles rating Intel and AMD. Intel is ahead virtually all the time when comparing comparable chis, and even some that are not so comparable.
Intel has the advantage with Clarkdale/Arrandale's integration, but that will be erased with Llano.
I'm still not convinced.
That will be when they are being pressured by AMD.
Thats very funny, because it's usually described as Intel being able to afford dropping prices, and they do it to put pressure on AMD who can't afford to.
I doubt it. Not when Interlagos is expected to be over 33% increased performance over Magny-Cours for 33% more cores. So Intel will pretty much end up where they are this year, next year (and from the speculations I've heard that will be the best case scenario for Intel in servers).
Really?
There's not much to talk about from a CPU standpoint with AMD in 2010, so AMD is heavily focused on 2011 and what it plans to do with its first on-die GPU in Llano. Bobcat and Bulldozer also make it back into the headlines as AMD is long overdue for another microprocessor architecture. Bobcat stands to be AMD's first competitive mobile architecture while Bulldozer may ensure AMD will be competitive at the high end.
I love the "may". It's AMD we're talking about here. Nothing's for sure.
Seriously I think people see the 13" machine as an excellent value. Part of that is because consumers are more informed about the significance of having a good GPU on a machine running Mac OS/X.
Yes but it is being sold out with a GPU that is significantly faster than Intels. We aren't talking a little bit faster either, it is significantly faster with video acceleration and good OpenCL support to boot.
What I've said here several times, is that it doesn't matter because Apple did a very elegant job. It's not how you get there, but whether you DO get there. Apple has.
Now I would agree with you that many consumers don't care about the techie details. But they will respond to the idea of video acceleration.
They will respond to the machine doing what they want it to. They don't care how or why. The people buying the 13" MBP simply want to get the aluminum machine at a price that's not that much more than the plastic MB. Their desires aren't that much different as to use.
The whole point here is that solutions that use multiple GPUs are only worthwhile if they can fit into the hardware in question. For the hardware that Apple seems to want to build Intels integrated GPU (in the CPU module) is no better than AMDs. In fact all signs indicates it is worst.
I don't think that that's ever a real problem. If Apple wasn't to do it, they can. They're often using Nvidia's which are better than either.
Actually I think it is a huge issue for Apple. They don't want to be seen as going backwards performance wise. Thus we get a much faster replacement for the 9400m and modest CPU improvements. All of this in an existing platform.
They need progressively better hardware for marketing reasons as it is hard to sell a machine that is slower than last years model. They also need that GPU for OpenCL which is a lot harder to market to consummers. OpenCL is a key to keeping the Mac competitive OS wise.
A side ways look at this would be the use of the GPU in the iPhone and iPad which is extensive. If Apple can get as much of Quartz running on the GPU on the Mac as it is on iPhone OS we could see major improvements in Mac responsiveness. In other words I expect Apple to continue to try to exploit the GPU as much as possible most likely leading to resolution independence and a few other long sought goals. Arrandale alone in such an environment would be a very low end solution.
[/QUOTE]
Apple is moving ahead.
Especially in the context of the smaller MBP!
Yes but not as internal drives.
I haven't seen that poll but frankly most online polls are worthless as they don't represent the entire user community. In any event the lack of an internal drive does not imply a lack of support for Optical drives in the OS.
The best example here is Apples own MBA with no internal Optical. AIR is a machine that is rightly skewered for some of the design choices made with the computer but seldom is the lack of an Optical drive raised.
Dave
You see, we've got three or four (at most) guys here who keep talking about getting rid of internal optical drives, but a much larger number who say no. You forget that. This poll is at least as valid as the numbers here when we have these discussions (of which, this hasn't become one yet). Not all online polls are bad. Polls such as this are often pretty good. There's no real reason for people to get excited about it, so they will vote with their mind, rather than with their heart.
You see, we've got three or four (at most) guys here who keep talking about getting rid of internal optical drives, but a much larger number who say no. You forget that.
Not really, because I'm looking at the small segment of the market covered by the 13" MBP which I consider to be an ultra mobile laptop. Here I honestly believe the users value mobility and long battery life over the ability to use an Optical drive.
One only needs to look at net books and other small laptops to see the trend here.
This poll is at least as valid as the numbers here when we have these discussions (of which, this hasn't become one yet). Not all online polls are bad.
I'd have to say all are skewed. It would be pretty tough for them not to be. Think about a poll run on Appleinsider, the only people responding to the poll will be the small subset of Mac users that actually come here and also bother with the polls. That is an exceedingly small number of Apple customers.
Polls such as this are often pretty good. There's no real reason for people to get excited about it, so they will vote with their mind, rather than with their heart.
They are good in the sense that they represent the small minority of people willing to go online to a fan site and then are willing to bother with polls. That doesn't always reflect what the market at large does.
IPad is a perfectly good example here. The thing sells well even though you wouldn't think so based on some of the polls taken on various sites. Wether it was the lack of Mac OS/X or some other feature (such as the paltry RAM in my case) people still are buying the unit in vast numbers. That may be related to early adopters and as such could roll off quickly. Or it could mean that the forums simply don't reflect the mass market.
In any event I still maintain that droping the internal Optical from the 13" MBP would allow Apple to configure the machine in a way that would make it even more attractive to the target market.
Dave
If you read the articles on this, you will see that AMD is stating that they BELIEVE that to be true. That doesn't make it true. And they're talking about server workloads, not the average for what most people do.
So you have no problem with HyperThreading?s large variety of performance increases (or even decreases) depending on the workload?
Intel says that dual-core mobile Sandy Bridge will be 20% faster than Arrandale. So is that what Intel BELIEVES to be true? And does that make Sandy Bridge?s ?real? performance increase over Westmere 10%?
Oh, and AMD BELIEVED that Magny-Cours would give 50%-60% more performance than Istanbul, they actually got 80%-119% more.
I'd like to see a link to that. They wouldn't be making a big deal about it if it were only increasing the dies size by 5% for the extra integer cores.
http://www.amdzone.com/phpbb3/viewto...st=0&sk=t&sd=a
The OP there works for AMD so he knows the numbers.
Same article
See article.
I'd rather see a statement of umbers, not take a look at the pictures.
Measure them yourself. (Hint: One company?s cores are 50% bigger than the other company?s on the same process.)
It's possible. But with AMD's record, they may not get it out on time anyway, and then they will be competing with upgraded chips from Intel as has happened the past two years.[/quote]I?m assuming you?re fine with Q3 2011 for Sandy Bridge E?
They might give UP TO 80% in multithreaded situations. That's what AMD is saying. These aren't full cores, integer or otherwise. Read the article and a couple of others. You'll see.
I did. It says 2 integer cores, each of them more powerful than a K10 integer core. Hence my ?over 33% performance increase? comment on my last post.
And why not ?up to 35%? for HyperThreading? Because according to the link you posted, that?s exactly what it is.
Of course they did.[/quote]Exactly.
It's got poor FP performance compated to Intel, and they knoe it.
?because they will be offloading FP instructions to the GPU.
That pretty much says a lot. If Intel has the same core number, it will always beat these chips, easily.
They won?t have the same core counts at same prices or same TDPs. As I?ve shown, AMD will generally have higher core counts than Intel at equivalent prices/TDP. That?s been my main point here, and once I refuted your counterpoint you evade the situation entirely.
With Sandy Bridge, Intel?s not increasing core counts at all, except in two areas (high-end desktop and server), while AMD has slowly done so in all segments, especially servers. In servers AMD has double the cores of Intel, and that?ll continue next year. In desktops, if Thuban?s prices are any indication, Zambezi will be priced against 6-core, maybe 4-core Sandy Bridge. Llano brings quad-core down to even lower TDPs (as if 25 W Champlain won?t be enough).
So what ends up from there is that AMD can make a CPU comparable to (sometimes worse, sometimes better, depending on the segment) Intel?s in multithreaded situations by adding more cores, even though Intel spent all the work designing one or two new microarchitectures (Nehalem and Sandy Bridge). That?s AMD?s small-core strategy, whether accidental (because they didn?t have the resources to make a new microarchitecture) or deliberate. More cores for less money and lower TDP.
Intel has and will have single-threaded advantage, AMD will have multithreaded advantage (they already do in a number of cases), and AMD will have the GPU advantage. AMD took a big hit in 2006-2009, but from 2010 the tables are slowly turning.
AMD won?t have the fastest CPU in notebooks and desktops, even with Bulldozer. But for Apple, that?s not an issue because they aren?t using the fastest notebook and desktop CPUs anyway.
Yes, funny that AMD would now use the inferior technology, and suddenly claim it's better.
Doesn?t matter because now they have the advantage. Or how about this? Why didn?t Intel continue using their superior dual-die approach? 12-core Westmere would beat 12-core Magny-Cours, no questions there.
Most apps we use still can't use more than two cores. That will be true for a while. And when they do, it's rarely more than four.
How many of those apps use 100% of a core? I?d like more cores for video encoding and other intensive tasks thank you very much.
Plus with only 2 or 4 task threads the HyperThreading you mentioned won?t be useful on a number of Intel CPUs.
It will be years before software catches up. And that includes using the GPU for most of the FP, or vector math. The OS's are using it more, led by Apple, but programs are trailing very slowly.
That?s why Apple hasn?t used AMD CPUs yet. I am looking at the future. Many people who criticize AMD?s chips are looking at only the past or present. When software does catch up (in addition to the software that already use multiple cores and GPUs), AMD will be in a very good position. What about Intel?
AMD has done well in HPC. It's an area that Intel has ignored for several years. Look at the power ratings in various Anandtech articles rating Intel and AMD. Intel is ahead virtually all the time when comparing comparable chis, and even some that are not so comparable.[/quote]I?ve read them. Magny-Cours is close to Westmere in performance. And you?re missing price again. Magny-Cours is cheaper.
I'm still not convinced.
30 W quad-core with 480 shader GPU.
Really?
Go read the article I linked to. Over 33% more integer performance for 33% more cores. Bulldozer also increases FP performance so one module would be faster than two K10 cores. Oh, and for about the die area of one K10 core.
Not really, because I'm looking at the small segment of the market covered by the 13" MBP which I consider to be an ultra mobile laptop. Here I honestly believe the users value mobility and long battery life over the ability to use an Optical drive.
One only needs to look at net books and other small laptops to see the trend here.
You aren't seriously comparing the 13" MBP to a netbook?
I'd have to say all are skewed. It would be pretty tough for them not to be. Think about a poll run on Appleinsider, the only people responding to the poll will be the small subset of Mac users that actually come here and also bother with the polls. That is an exceedingly small number of Apple customers. [/qote]
I don't agree. If there is just one poll, and nowhere else is this discussed, then I might doubt that one poll with nothing else to substantiate it.
But we've had a long history here of discussing the optical-less laptop, and most people here aren't yet ready for it other than perhaps for the Air.
They are good in the sense that they represent the small minority of people willing to go online to a fan site and then are willing to bother with polls. That doesn't always reflect what the market at large does.
Well, again, it just agrees with what most people say here. At some point you just have to see the tide is against you.
IPad is a perfectly good example here. The thing sells well even though you wouldn't think so based on some of the polls taken on various sites. Wether it was the lack of Mac OS/X or some other feature (such as the paltry RAM in my case) people still are buying the unit in vast numbers. That may be related to early adopters and as such could roll off quickly. Or it could mean that the forums simply don't reflect the mass market.
Again, I don't agree. It's a totally different kind of thing. I 20% of the buying public say they would buy something, that's pretty darned good. It would lead to tens of millions of sales. You're confusing the meaning of the different kinds of polls (or surveys).
[uote]
In any event I still maintain that droping the internal Optical from the 13" MBP would allow Apple to configure the machine in a way that would make it even more attractive to the target market.
Dave
It definitely be a different kind of machine. I can agree with that.
So you have no problem with HyperThreading?s large variety of performance increases (or even decreases) depending on the workload?
Intel says that dual-core mobile Sandy Bridge will be 20% faster than Arrandale. So is that what Intel BELIEVES to be true? And does that make Sandy Bridge?s ?real? performance increase over Westmere 10%?
Except for a short time several years ago, Intel has had an excellent record of beating its predictions of what its new chips could do. You can look at the reviews of them. But over most of its life, AMD has not. It's history. And the past two years have been very bad for AMD.
Oh, and AMD BELIEVED that Magny-Cours would give 50%-60% more performance than Istanbul, they actually got 80%-119% more.
http://www.amdzone.com/phpbb3/viewto...st=0&sk=t&sd=a
That link doesn't lead to any test that I could see. Could you be more specific with it?
The OP there works for AMD so he knows the numbers.
You'll have to show the tests.
Measure them yourself. (Hint: One company?s cores are 50% bigger than the other company?s on the same process.)
That's not really possible. But I am seeing conflicting numbers.
It's possible. But with AMD's record, they may not get it out on time anyway, and then they will be competing with upgraded chips from Intel as has happened the past two years.
I?m assuming you?re fine with Q3 2011 for Sandy Bridge E? [/quote]
I'm concerned that AMD will again have the situation they had with Barcelona, and Phenom. You remember that fiasco?
Of course, no one is happy about Sandy Bridge being pushed back three months. It does give AMD some breathing time.
I did. It says 2 integer cores, each of them more powerful than a K10 integer core. Hence my ?over 33% performance increase? comment on my last post.
Yes, I understand.
And why not ?up to 35%? for HyperThreading? Because according to the link you posted, that?s exactly what it is.
You can say that. but testing has shown that most of the time it comes pretty close, and sometimes has exceeded it. Again, you can go to Anandtech.
?because they will be offloading FP instructions to the GPU. [/qote]
Yes, we know that. It's what AMD hopes. But you also must have rad that they are taking a chance with this, because, so far, very few programs are taking advantage of the GPU. Even Adobe CS5 doesn't make much use of it, and of all the programs out there, it needs it really badly.
In addition, AMD can't count on knowing which GPU will be in any give computer. If it's a fast one that can e be used for this, then all may be well. But most computers manage with integrated graphics. It believe that's about 80% of all machines out there. They'll be getting no help from that GPU.
That means that machines that are being used for big spreadsheets (for example) that may need FP won't be getting any real FP punch at all.
They won?t have the same core counts at same prices or same TDPs. As I?ve shown, AMD will generally have higher core counts than Intel at equivalent prices/TDP. That?s been my main point here, and once I refuted your counterpoint you evade the situation entirely.
I never evade anything. I've already stated that I don't agree with that either. Intel can afford to be very aggressive and keep AMD's prices down where their profits are cut short. AMD competes, ironically, best on the HPC high end, and on the low end. Everywhere else, they can't compete. So if you are talking about the cheapest chips, then, maybe, assuming that their stuff comes out when it should, and it's as good as they say, then, for a while, they'll be competitive at that point, with those chips.
With Sandy Bridge, Intel?s not increasing core counts at all, except in two areas (high-end desktop and server), while AMD has slowly done so in all segments, especially servers. In servers AMD has double the cores of Intel, and that?ll continue next year. In desktops, if Thuban?s prices are any indication, Zambezi will be priced against 6-core, maybe 4-core Sandy Bridge. Llano brings quad-core down to even lower TDPs (as if 25 W Champlain won?t be enough).
AMD has had to because core for core, they can't compete.
So what ends up from there is that AMD can make a CPU comparable to (sometimes worse, sometimes better, depending on the segment) Intel?s in multithreaded situations by adding more cores, even though Intel spent all the work designing one or two new microarchitectures (Nehalem and Sandy Bridge). That?s AMD?s small-core strategy, whether accidental (because they didn?t have the resources to make a new microarchitecture) or deliberate. More cores for less money and lower TDP.
They have no choice. They're getting clobbered core for core. They need more cores just to keep up, or make a small advance over Intel. But you keep forgetting that Intel drops the process several times over the lifetime of these chips, and AMD is forced to follow.
Intel has and will have single-threaded advantage, AMD will have multithreaded advantage (they already do in a number of cases), and AMD will have the GPU advantage. AMD took a big hit in 2006-2009, but from 2010 the tables are slowly turning.
AMD won?t have the fastest CPU in notebooks and desktops, even with Bulldozer. But for Apple, that?s not an issue because they aren?t using the fastest notebook and desktop CPUs anyway.
Doesn?t matter because now they have the advantage. Or how about this? Why didn?t Intel continue using their superior dual-die approach? 12-core Westmere would beat 12-core Magny-Cours, no questions there.
How many of those apps use 100% of a core? I?d like more cores for video encoding and other intensive tasks thank you very much.
Plus with only 2 or 4 task threads the HyperThreading you mentioned won?t be useful on a number of Intel CPUs.
That?s why Apple hasn?t used AMD CPUs yet. I am looking at the future. Many people who criticize AMD?s chips are looking at only the past or present. When software does catch up (in addition to the software that already use multiple cores and GPUs), AMD will be in a very good position. What about Intel?
AMD has done well in HPC. It's an area that Intel has ignored for several years. Look at the power ratings in various Anandtech articles rating Intel and AMD. Intel is ahead virtually all the time when comparing comparable chis, and even some that are not so comparable.
I?ve read them. Magny-Cours is close to Westmere in performance. And you?re missing price again. Magny-Cours is cheaper.
30 W quad-core with 480 shader GPU.
Go read the article I linked to. Over 33% more integer performance for 33% more cores. Bulldozer also increases FP performance so one module would be faster than two K10 cores. Oh, and for about the die area of one K10 core.
That's all rehashing earlier stuff. Look, I hope that AMD stays around, and is competitive, but right now, they aren't. With Barcelona, they promised the world, 40% better performance per MHz, and what did they deliver 6 months late? Chips that not only were at lower speeds, but were 20% SLOWER per MHz!. This carried on to the Phnom as well, though not quite as badly.
This is what I'm concerned about. They will have to show that they won't repeat that. When Intel backs off a few months, the industry doesn't get concerned. It's usually because they'r looking back at AMD and don't see much competition, so they figure they can make some more money on current designs. When AMD is late, everyone is thinking that they are having major problems.
Let's see how AMD does and if they show any signs of closing the gap with Intel chips in performance.
http://www.anandtech.com/show/3674/a...1055t-reviewed
Let's see how AMD does and if they show any signs of closing the gap with Intel chips in performance.
I'm not seeing anything with the Phenom II that would make Apple want to switch. They really need a mobile contender, which needs to be more than just a good price for the performance but good power usage for the performance.
I'm not seeing anything with the Phenom II that would make Apple want to switch. They really need a mobile contender, which needs to be more than just a good price for the performance but good power usage for the performance.
They wouldn't want phenom II. They'd want a system on chip solution using fusion and AMD's far superior graphics technology.
You aren't seriously comparing the 13" MBP to a netbook?
No not exactly what I'm saying here is that it has become clear that people are willing to give up certain technologies to get an extremely mobile PC. An Optical drive being one thing people find that they don't need.
It definitely be a different kind of machine. I can agree with that.
Think of it as a device that corrects the mistakes made on AIR! In fact that is a very good way to look at it. It would be a laptop where what can be deleted is, but the stuff that many of us believe is important is still there. The machine would stress good performance and battery life.
Dave
I'd buy one in a second. 3D rendering is my thing.
http://www.anandtech.com/show/3674/a...55t-reviewed/3
http://www.anandtech.com/show/3674/a...1055t-reviewed
Let's see how AMD does and if they show any signs of closing the gap with Intel chips in performance.
Here is another but more objective review.
http://benchmarkreviews.com/index.ph...=508&Itemid=63
Given the new AMD six core chips are actually beating i7s at some workloads, and cost less than one third of Intel's prices, it may be the high end we see AMDs in. OS X is all about parallel computing, from threading, Grand Central, OpenCL, to distributed compilations. If apple can knock $700 (6 core) or $1400 (12 cores) off the price of the Mac Pro, wouldn't they?
I'd buy one in a second. 3D rendering is my thing.
http://www.anandtech.com/show/3674/a...55t-reviewed/3
Exactly. With the money i'm saving over an Intel Core i7 I could afford to put in a faster GPU or SSD.
Here is another but more objective review.
http://benchmarkreviews.com/index.ph...=508&Itemid=63
What an outstanding value. I think Apple would be crazy not to start delivering some AMD solutions next year with Llano. I like Intel as much as anyone but truth be told a lot of my work leads to the computer sitting here waiting on me. I think AMD certainly delivers a lot of processing power per buck. I'm content to let the "I gotta have the fastest computer available" people spend a premium on Intel based Macs. I'm more of a system balance kind of guy. Which is why I like SSD and will like faster GPU when they represent more than just increased fps in games.
Given the new AMD six core chips are actually beating i7s at some workloads, and cost less than one third of Intel's prices, it may be the high end we see AMDs in. OS X is all about parallel computing, from threading, Grand Central, OpenCL, to distributed compilations. If apple can knock $700 (6 core) or $1400 (12 cores) off the price of the Mac Pro, wouldn't they?
Yeah, I saw a review of the new 6-core AMD this morning and the i7-980 was 30% faster in some tests and I was ready to have the usual dismissive reaction but the price difference hit home. The chips were 30% slower only in the worst case but are under 1/3 of the price so as you say, Apple could build a 12-core Mac Pro $1400 cheaper than they could with Intel's chips (depending on which Intel ones they chose).
I'd much rather have an $1800 6-core AMD Mac Pro than an Intel one at $2500 when you're only going to see 30% improvement doing the long rendering stuff, which you're going to be waiting for anyway.
Plus, the 12-core AMD machine would be cheaper than a 6-core Intel so in terms of performance-per-dollar, the AMD offering will win in most cases.
I wonder if Apple will make the jump at this refresh. I expected the Mac Pro update to be out by now.
I really wish they'd make a smaller one. 12-core, 2/3rds the size and maybe 12kg instead of 18kg.
You aren't seriously comparing the 13" MBP to a netbook?
No, he's comparing them to the 13" CULV powered machines like the Thinkpad Edge, HP DM3, and others that use 1.2-1.6Ghz Intel and AMD CPUs and deliver performance halfway between a netbook and a full sized notebook like a Macbook Pro. What he neglects to account for is that the current Macbook Pro delivers better battery life than any of them while being around the same weight with significantly more power. Again, this is an example of the machine a user wants existing and being readily available from anyone but Apple.