[Lateralus]

Are we forgetting the DVD-RAM incident?

[Simon]

An octo-core 45nm Mac Pro with a 1600 MHz Seaburg bus and dual 16x PCIe lanes should be more than enough workstation for creative professionals. At its price point it will likely also look great next to the 'generic PC' competition.

[mduell]

Originally Posted by Simon 

An octo-core 45nm Mac Pro with a 1600 MHz Seaburg bus and dual 16x PCIe lanes should be more than enough workstation for creative professionals. At its price point it will likely also look great next to the 'generic PC' competition.

Going all octo (as I expect them to do) will really help differentiate the Mac Pro from the $600 2.4Ghz quad/2GB/160GB/nVidia graphics systems that Dell and friends are selling, but it won't really help Apple to compete with those generic PCs.

[Simon]

The Mac Pro is not intended to compete with the $600 Dell. It's a workstation and competes very well with other high-end boxes in that market. Of course we know that Apple would need to release a real HEM (beside just a mini and the iMac) to compete with the $600 Dell. Obviously they don't feel like it. We've been there. And as far as I can see not much has changed since.

[mduell]

Mea culpa, misunderstood 'generic PC.'

[goMac]

Originally Posted by mduell 

Going all octo (as I expect them to do) will really help differentiate the Mac Pro from the $600 2.4Ghz quad/2GB/160GB/nVidia graphics systems that Dell and friends are selling, but it won't really help Apple to compete with those generic PCs.

I've actually been trying to figure this out. I'm waiting for the new Mac Pros, and I do coding for a living. 4 cores sounds nice. 8 cores? Honestly, I don't know what I'd do with 8 cores. Seems a bit overkill. I don't think for graphics pros PS even uses 8 cores. I know XCode could compile 8 classes at once on an octocore... and I do a lot of coding of OpenGL/graphics applications presently, but even then my own code isn't really optimized for 8 cores, not past any auto-magic acceleration Apple gives me.

Mostly I just want some high ghz Cores, the 1600 mhz system bus, and the large amount of L2 cache, along with a new GPU. I can't say I'm quite as excited for octo cores across the line.

[Simon]

The idea is that once CPU-limited codes are properly parallelized and scale efficiently to four cores, they should also scale well to eight. I think in this respect hardware is ahead of software development right now. I can't understand why people have to wait for minutes for some PS filters to finish while not all of their eight cores are at >80% load (and no, that's not because they are RAM starved). Once pro apps overcome such issues, eight cores will clearly make more sense than four. Especially with the latest high-end CPUs where you don't have to trade clock frequency for number of cores.

[SierraDragon]

When buying a new box it should primarily be about expected future performance rather than simply about how apps run today.

The benefit of 8 cores, or of other hardware inprovements that require app or OS mods to be fully beneficial, is for the future. E.g if multiple cores are available cost-effective from the hardware side, can we expect OS/app versions during the life of the box to take advantage of those multiple cores? IMO absolutely yes. Heavy apps designers who fail to scale their apps to take advantage of increasing core multiples would lose market advantage. Heavy apps - and the Mac OS - will be scaling up.

-Allen Wicks

[goMac]

I'm sure in the future things will scale up, but it's a long ways off. A large majority of software isn't even really optimized for dual cores. The 8 cores is nice for giving each app it's own space to work with, but really, I think 4 cores is a pretty good number. 8 cores is still a bit much.

[Simon]

No matter what our opinion on the matter is, it's very likely Apple will switch to dual quad-cores across the line with the next MP release. And every Mac sold today comes with at least two cores. At least in Mac land there is definitely no reason for developers not to at least think about parallelizing their code.

[~bash $]

Originally Posted by Simon 

No matter what our opinion on the matter is, it's very likely Apple will switch to dual quad-cores across the line with the next MP release. And every Mac sold today comes with at least two cores. At least in Mac land there is definitely no reason for developers not to at least think about parallelizing their code.

Which, just to emphasize, is only parallelizable insofar as it makes sense to do so. If you have a loop where the nth iteration depends on the result from n-1, there isn't a whole lot that you can do to speed that up by throwing more cores at it.

It will take very good software development to exploit all the little ways that multiple cores are useful. Encoding in Max v. iTunes is a nice example of how one can distribute a task over multiple cores even when the individual task is not running in a truly parallel manner.

My own code is able to take advantage of new multithreaded routines in MATLAB, so the speed improvement for code that takes this into consideration is often incredible. But I'm finding that I could write the code even better to get even more out of it. (A figure from an Apple developer tech talk on this is coming to mind regarding optimization.)

Anyway, this is all exciting in the sense that, to a first approximation, it looks like the multithreaded "myth" is more substantial than the megahertz myth. Is it my feeling only or is it supported by the facts that the actual clock speeds haven't been progressing quite as quickly as they "should" be?

[goMac]

Originally Posted by ~bash $ 

Which, just to emphasize, is only parallelizable insofar as it makes sense to do so. If you have a loop where the nth iteration depends on the result from n-1, there isn't a whole lot that you can do to speed that up by throwing more cores at it.

Which is my point. True, there are tasks that can be speed up by multiple cores. But you're still going to hit a jam when you reach the single GPU, and you're still bound by disk speeds when you hit the hard drive. I think Apple may be adding Cores too fast while ignoring the market for dual core and quad core destops. Having an 8 core Mac Pro leaves and uncomfortable void in Apple's product line for people who aren't studying the human genome.

[Simon]

I understand where you're going and I agree with you that there's a void between the mini and the MP, but I do not think this is so much a MP problem than one of the missing HEM. The MP at ~$3k needs to go Harpertown. It just doesn't make sense to use (expensive) dual Woodcrests instead just because "there's some codes that don't scale well". OTOH I absolutely agree that Apple should consider a 'normal' desktop Mac (the HEM) and for that Kentsfield/Yorkfield are more than sufficient.

[goMac]

Originally Posted by Simon 

The MP at ~$3k needs to go Harpertown. It just doesn't make sense to use (expensive) dual Woodcrests instead just because "there's some codes that don't scale well".

Who said anything about Woodcrest? What is wrong with offering cheaper Wolfdale Mac Pros?

[Simon]

Originally Posted by goMac 

Who said anything about Woodcrest? What is wrong with offering cheaper Wolfdale Mac Pros?

Woodcrest is what the MP has now. If you're going dual CPU Kentsfield/Wolfdale are not an option. I think your question is actually why is Apple using dual duals rather than single quads.

I believe the answer lies in the chipsets. The chipset Apple's using with the Xeons (5000X chipset) offers more PCIe lanes and has a higher bandwidth FSB than the Glenwood/Bearlake chipsets used with the exiting Core 2 Quads. In addition Woodcrest's TDPs were substantially lower than comparable Kentsfields at the time. Obviously you then have twice the dissipation to take care of with dual Woodcrests, but that's still favorable because the problem with heat dissipation form modern CPUs is rather power flow density rather than absolute power flow. Cooling two separate dies at 65 W could be a lot easier than cooling one die at 125 W (keyword: quiet workstation). In addition, Seaburg will offer up to 128 GB memory support wich Apple also might be interested in.

In a broader sense I think dual Xeons are the right choice for the MP which is a workstation and not just some 'desktop' computer. The actual problem IMHO is the missing HEM (where Kentsfield/Woodcrest would be absolutely appropriate), not the use of dual Xeons. Just my 2¢.

[goMac]

Originally Posted by Simon 

I think your question is actually why is Apple using dual duals rather than single quads.

My question is why there will be such a broad chasm between the prosumer iMac and the octo core Mac Pro. They might offer the low end as a 4 core, but really they should offer a 4 core Wolfdale 3.16.

I don't mind them using dual CPU's. I just think there are a lot of professionals out there who would like a dual or a quad with a nice GPU, would like the high ghz for traidtional apps, and don't really need the octo cores.

[Simon]

Well the discussion about the missing quad-core HEM is about as old as the MP itself. Apple has simply chosen not to use Conroe/Kentsfield in the past. I think on this board we have come up with many scenarios where a quad-core HEM would be of great benefit. It's obvious not everybody who wants two bays, an exchangeable GPU and some decent CPU power needs a MP — and of course even less are willing to pay over $2k for such a machine.

I still believe though that Apple will maintain the MP as a high-end workstation (and actually even more so if they go octo across the line) and won't use the new Intel processors to launch less expensive "Yikes" MPs to close the gap to the mini or iMac. But there is also some good news for those people interested in a less expensive HEM. Once Apple releases new octo MPs the previous MPs should drop considerably in price. And there will also be refurbs (as there are now) going at significantly lower price points. A MP for $1799 isn't out of the question if you're willing to take dual Woodcrests and an X1900 XT. And that I consider a very good deal actually. The downside is the need for expensive FB-DIMMs and the large case size. It's no real replacement for a dedicated HEM, but at least there will be alternatives.

[MarkLT1]

As a scientific researcher who writes my own simulations, I say, the more cores the better.

Away from simulation, for thinks like image editing and video editing, where things are done on a pixel-by-pixel or frame-by-frame basis, and each calculation is independent, or loosely correlated to surrounding calculations, throwing more cores at the problem is of great interest.

And goMac- as far as there being issues with multiple CPUs and a single hard drive/GPU, this is rarely an issue at the desktop level. Typically, the problem is loaded into memory, you use your multiple procs to tackle said problem, then you output the data to the drive or GPU. The amount of I/O to/from the hard drive or display is the same, since you are using a shared memory bus, its just you have more workers to work on that memory space. I am generalizing with that statement, but for most processing applications a typical power user would see, it is not an issue.

[AngelaBaby]

Originally Posted by MarkLT1 

As a scientific researcher who writes my own simulations, I say, the more cores the better.

Away from simulation, for thinks like image editing and video editing, where things are done on a pixel-by-pixel or frame-by-frame basis, and each calculation is independent, or loosely correlated to surrounding calculations, throwing more cores at the problem is of great interest.

And goMac- as far as there being issues with multiple CPUs and a single hard drive/GPU, this is rarely an issue at the desktop level. Typically, the problem is loaded into memory, you use your multiple procs to tackle said problem, then you output the data to the drive or GPU. The amount of I/O to/from the hard drive or display is the same, since you are using a shared memory bus, its just you have more workers to work on that memory space. I am generalizing with that statement, but for most processing applications a typical power user would see, it is not an issue.

Does throwing more core's drastically improve rendering time, or just shave off a few seconds. Cause frankly, unless you're running a rendering farm I don't see the point in having more cores.

[SierraDragon]

IMO those of us dealing with batches of digital image files (i.e. all DSLR and video photogs and their editors, which is quite a large and rapidly growing group) pretty much are running rendering farms.

Existing Mac Pros are more than adequate for mainstream desktop apps like surfing and office apps. If it were not for perception marketing issues a couple of the existing MPs couId simply be kept unchanged as the basic low end desktop towers. It is heavy apps that actually need the performance improvements of a new Mac Pro, and most of those apps can and should be written to scale to multiple cores.

-Allen Wicks

[goMac]

Originally Posted by MarkLT1 

And goMac- as far as there being issues with multiple CPUs and a single hard drive/GPU, this is rarely an issue at the desktop level. Typically, the problem is loaded into memory, you use your multiple procs to tackle said problem, then you output the data to the drive or GPU. The amount of I/O to/from the hard drive or display is the same, since you are using a shared memory bus, its just you have more workers to work on that memory space. I am generalizing with that statement, but for most processing applications a typical power user would see, it is not an issue.

Right, and as I pointed out, scientific usage is one area where multiple cores become more useful because most code is entirely CPU bound and can be done in parallel. That still doesn't mean I think it's a good idea for Apple to make all Mac Pro's octo core.

Originally Posted by Simon 

Well the discussion about the missing quad-core HEM is about as old as the MP itself. Apple has simply chosen not to use Conroe/Kentsfield in the past. I think on this board we have come up with many scenarios where a quad-core HEM would be of great benefit. It's obvious not everybody who wants two bays, an exchangeable GPU and some decent CPU power needs a MP — and of course even less are willing to pay over $2k for such a machine.

But again, some pros would like a quad core Mac.

Originally Posted by AngelaBaby 

Does throwing more core's drastically improve rendering time, or just shave off a few seconds. Cause frankly, unless you're running a rendering farm I don't see the point in having more cores.

Rendering is something that can be done in parallel. Usually you only get large performance increases when you add more cores when you're rendering in software though. So it's really only something that would be useful if you're a 3D artist, it's not as useful in general applications.

[MarkLT1]

goMac-

It seems to me that you are missing the point of the MacPRO. PROfessional level applications- i.e.- 3D rendering, image manipulation, video editing, scientific research, etc... Those who are purchasing the MP for the types of applications it is targeted for will be able to make use of "more cores the better" approaches (as long as software developers can keep up). I agree with you that Apple has a gaping hole between the MP and the iMac, but that doesn't change the fact that the MP will, in many (most?) cases be used in situations where 8 cores will be better than 4. The gaping hole (xMac, midtower mac, whatever you want to call it) issue you keep pointing to is a totally different issue than "Should Apple's top of the line machines keep moving towards more cores."

[MarkLT1]

Originally Posted by AngelaBaby 

Does throwing more core's drastically improve rendering time, or just shave off a few seconds. Cause frankly, unless you're running a rendering farm I don't see the point in having more cores.

Throwing more cores can very drastically improve rendering time. Each frame in say, a 3D animation, is independent. Thus, if the software is written correctly, and you have a 4 core machine, you can render 4 frames at the same time. If you have an 8 core machine, you can render 8 at a time (given enough system memory). This is referred to as an embarrassingly parallel problem, and scales VERY well to more CPUs (not quite linearly, but close)

[goMac]

Originally Posted by MarkLT1 

t seems to me that you are missing the point of the MacPRO. PROfessional level applications- i.e.- 3D rendering, image manipulation, video editing, scientific research, etc... Those who are purchasing the MP for the types of applications it is targeted for will be able to make use of "more cores the better" approaches (as long as software developers can keep up). I agree with you that Apple has a gaping hole between the MP and the iMac, but that doesn't change the fact that the MP will, in many (most?) cases be used in situations where 8 cores will be better than 4. The gaping hole (xMac, midtower mac, whatever you want to call it) issue you keep pointing to is a totally different issue than "Should Apple's top of the line machines keep moving towards more cores."

Not all professionals do image manipulation. I'm a programmer. I need lots of hard drive space, lots of RAM, a fast GPU, and fast 64 bit SSE4 processors. I don't need eight cores though. 4 cores would do me just fine.

[MarkLT1]

Originally Posted by goMac 

Not all professionals do image manipulation. I'm a programmer. I need lots of hard drive space, lots of RAM, a fast GPU, and fast 64 bit SSE4 processors. I don't need eight cores though. 4 cores would do me just fine.

No.. not all of them do. That being said, at this point, the majority of the group that Apple markets the MP to, DOES do image/video manipulation, and would make use of more cores. Its just the law of numbers- they are marketing to those who could use more procs, thus the machines have more procs..

[mduell]

Originally Posted by ~bash $ 

Is it my feeling only or is it supported by the facts that the actual clock speeds haven't been progressing quite as quickly as they "should" be?

Please, tell me, exactly how fast should clock speeds be progressing?
Intel and IBM would love to know.

[goMac]

Originally Posted by MarkLT1 

No.. not all of them do. That being said, at this point, the majority of the group that Apple markets the MP to, DOES do image/video manipulation, and would make use of more cores. Its just the law of numbers- they are marketing to those who could use more procs, thus the machines have more procs..

But even then, Adobe products use, at most, 2 cores. Only some 3D renderers and software that runs simulations will benefit from the multiple cores. For most Pros, 4 cores is more than adequate.

[zacharykrannert]

i dont get the whole idea of well i really dont need 8 cores argument. Its like saying I dont need more than 2gbs of ram instead of 4. Why turn it down. We all know that apple isnt going to raise the prices of the macs so be happy your getting 8 cores and maybe someday you will stop complaining about apple giving you more than you need. its hard to be spoiled with 4 extra cores just sitting around.

[goMac]

Originally Posted by zacharykrannert 

i dont get the whole idea of well i really dont need 8 cores argument. Its like saying I dont need more than 2gbs of ram instead of 4. Why turn it down. We all know that apple isnt going to raise the prices of the macs so be happy your getting 8 cores and maybe someday you will stop complaining about apple giving you more than you need. its hard to be spoiled with 4 extra cores just sitting around.

Because I could get 4 cores cheaper? I mean, Apple isn't giving you 8 cores for the price of 4 cores. You're citing the prices of the current Mac Pros, but everyone knows the prices as they stand today are way overinflated.

[Simon]

goMac, I think you are confusing your demands with those of Apple's model customer for the MP. Obviously you don't need an high-end octo MP. Nobody is disagreeing with you that it would be nice to have a quad-core HEM. And nobody is denying that some professionals would prefer that over an octo MP.

The conclusion to that though is that Apple needs to close the gap below the MP. By no means however does it mean that Apple shouldn't continue to crank up the MP to be at the high end of what you can buy in a 'desktop' machine. At ~$2.5k the MP has to be a serious workstation. Going Harpertown/Seaburg is exactly that.

But also, I don't quite see why you're so disappointed with the expected devlopment. As soon as the new MP comes out you will be able to buy a quad-core MP at a very nice price point. From what I understand, such a bopx comes pretty close to what you want. And it will also be a lot cheaper than a new MP. Isn't that a decent solution for people in your position (at least until Apple actually launches a serious HEM)?

[Simon]

Originally Posted by goMac 

You're citing the prices of the current Mac Pros, but everyone knows the prices as they stand today are way overinflated.

That BTW is not because of the number of cores. It's because the other specs (HDDs, RAM, GPUs, etc.) are already outdated and hence overpriced. The two X5130 in there are certainly not the main reason the machine appears expensive now (when it ame out it was actually a quite a deal btw!).

[goMac]

Originally Posted by Simon 

But also, I don't quite see why you're so disappointed with the expected devlopment. As soon as the new MP comes out you will be able to buy a quad-core MP at a very nice price point. From what I understand, such a bopx comes pretty close to what you want. And it will also be a lot cheaper than a new MP. Isn't that a decent solution for people in your position (at least until Apple actually launches a serious HEM)?

Again, I want the new dual core Xeons with SSE4 and the higher L2 cache and bus speeds. Intel is coming out with the same new Xeons in dual core, and I think Apple should offer them as an option.

[MarkLT1]

Originally Posted by goMac 

Again, I want the new dual core Xeons with SSE4 and the higher L2 cache and bus speeds.

Sing it with me now...

I saw her today at a reception
A glass of wine in her hand
I knew she would meet her connection
At her feet was her footloose man ....

And you know the rest..

[mduell]

Originally Posted by goMac 

Again, I want the new dual core Xeons with SSE4 and the higher L2 cache and bus speeds. Intel is coming out with the same new Xeons in dual core, and I think Apple should offer them as an option.

E5450 3.00 GHz 80W 12MB $851
E5260 3.33 GHz 65W 6MB $851

I doubt many people will find trading half the cores for 11% on clockrate to be an attractive option.

[dantewaters]

Originally Posted by mduell 

E5450 3.00 GHz 80W 12MB $851
E5260 3.33 GHz 65W 6MB $851

I doubt many people will find trading half the cores for 11% on clockrate to be an attractive option.

Those 2 configs are the 4core and octo core?
And their going for the same price??

[Simon]

Originally Posted by dantewaters 

Those 2 configs are the 4core and octo core?
And their going for the same price??

Yes. The E5450 is the quad-core Harpertown and the E5260 is the dual-core variant. And yes, Intel is planning on selling them at the same price.

So going for dual dual-core Xeons doesn't make much sense if you want to get a lower BOM and hence a lower entry-level price. OTOH what Apple could do is release a non-Xeon single CPU MP with four cores, IOW Kentsfield (up to now) or Yorkfield (after Nov 11). That would actually decrease the BOM:

Core 2 Extreme QX9650 - 3.0 GHz / 12MB/ 1333 MHz - TDP 130 W - $999
Xeon E5450 - 3.0 GHz / 12MB / 1333 MHz - TDP 80 W - $851

IOW if Apple would use a single Yorkfield the quad-core CPU would cost $703 less than the two Harpertowns in the octo MP.

Of course in reality it's not quite that simple because Apple would have to design a totally new system: Yorkfields require a different chipset than the Xeons and actually it's a chipset Apple has previously never used in a Mac (the iMac and mini use mobile chipsets). That means a lot of design work, new components, less commonality and hence increased component costs, etc. Apart from that there are also other issues: Going with a single Yorkfield on its Bearlake chipset rather than dual Xeons with Seaburg means no dual 16x PCIe lanes, no 128 GB RAM support, less FSB bandwidth, no FB-DIMMs, etc.

Personally, I wouldn't mind seeing a new desktop Mac model positioned between the mini and the MP with a Yorkfield, some PCIe slots, disk bays, etc. (IOW the HEM). But I think this needs to be a new model, not a MP variant. And obviously Apple seems reluctant to do it at all. IMHO they're presently trying to play it safe. By positioning the MP as an expensive workstation, the iMac as a stylish and silent AIO, and the mini as a ultra-compact entry-level desktop they can stay out of the most dangerous part of the market. Once they start selling Wolfdale or Yorkfield HEMs they will immediately be compared with the $600 Dell tower. I am well aware that that's where the majority of the desktop computer market is, but honestly, (and keeping in mind Apple's history) who here actually believes Apple wants to try and compete with that $600 Dell?

[ninahagen]

Originally Posted by goMac 

Again, I want the new dual core Xeons with SSE4 and the higher L2 cache and bus speeds. Intel is coming out with the same new Xeons in dual core, and I think Apple should offer them as an option.

As Simon said, if Mac gave you your wish, you would be able to save $703 minus the cost of development amortized over the expected number of sales. Since the number of sales would be a small segment of the market, the cost to each unit would be significant. To be sporting, let's say Mac did all the development and it came to $403 per unit (conservative since the economies of scale are shot), making your savings $300, all of which would be lost when you go to resell it in an 8 or 16-core world a few years from now. Are you saying that you would not spend an extra $300 to have 4 more cores, higher capacity PCI lanes, better RAM support, higher bandwidth and increased resale value? Without having any idea what features Leopard will have to optimize those 8-cores? If so, you are in a sub-1% minority, and must have some masochistic streak.

[mduell]

Between the chipset, logic board, and memory there's a lot more than $703 worth of component savings to be had. Development is never cheap, but a the logic board could probably be adapted from one of Intel's many existing options.

Contrary to ninahagen's assertion, I think the market for such a system would be quite large compared to the market for an 8-core/FB-DIMM system.

[SierraDragon]

Simon, nice analysis.

Remember the IIfx, introduced as "wicked fast?" For midrange apps like office-type apps, existing MPs are already wicked fast. Perhaps Apple lets the used Mac Pro marketplace go after the midrange; a strong used market giving high end folks incentive to shorten life cycles. However - arguing against myself - a better used market would not cause me to shorten life cycles, which I maintain as fairly long (4+ years) with the Mac boxes I manage.

Also I do agree with Mark about the atrractiveness of the midrange market. My guess is Apple can build a $1000 box as well as anyone using stock componentry. But Apple has OS 10.5 Leopard + Win as an option, while the Dells of the world are stuck with only Microsoft's security-nightmare OS (Linux still being a geeky specialty item). That, especially with the iPod/iPhone halo, means Apple could get away with an extra $100 of pure profit retail per box above Dell for similar hardware.

-Allen Wicks

[mduell]

Originally Posted by SierraDragon 

Apple could get away with an extra $100 of pure profit retail per box above Dell for similar hardware.

I think the markup would be closer to 100% than $100.

Dell often has a 2.4Ghz quad/2GB/80GB/SuperDrive/8300GS 128MB desktop for $609. With the expansion the Dimension 9200 has, that would be closer to a $1199 box from Apple than a $699 box.

But as much as I'd like it, I really don't think Apple will get into that market in the next 3 years.

[ninahagen]

Originally Posted by mduell 

Contrary to ninahagen's assertion, I think the market for such a system would be quite large compared to the market for an 8-core/FB-DIMM system.

Mark, would you buy one? (not rhetorical, I really am asking)
I wouldn't.
I am not saying there is no market. However, I bet very few Mac Pro users would step back to such a machine.

[Veltliner]

Originally Posted by MarkLT1 

Throwing more cores can very drastically improve rendering time. Each frame in say, a 3D animation, is independent. Thus, if the software is written correctly, and you have a 4 core machine, you can render 4 frames at the same time. If you have an 8 core machine, you can render 8 at a time (given enough system memory). This is referred to as an embarrassingly parallel problem, and scales VERY well to more CPUs (not quite linearly, but close)

How real is such software taking advantage of 8 cores? Does it exist now?

[Veltliner]

Originally Posted by goMac 

But even then, Adobe products use, at most, 2 cores. Only some 3D renderers and software that runs simulations will benefit from the multiple cores. For most Pros, 4 cores is more than adequate.

I heard that, too, and repeatedly.

Doesn't this look like hardware development is way ahead of software development?

[SierraDragon]

Originally Posted by Veltliner 

How real is such software taking advantage of 8 cores? Does it exist now?

I would bet that many major apps will be taking good advantage of 8 cores during 2008; certainly one or more of the ones I use. That would be less than 25% of the time into the life cycle of new Mac Pros. Who cares if it exists now?

-Allen Wicks

[SierraDragon]

Originally Posted by Veltliner 

Doesn't this look like hardware development is way ahead of software development?

Yes, hardware development typically is ahead of software development. That means that all one has to do is apply a little intellect to Intel's roadmap to see where the OS/apps world will be in the proximate future: 8 cores.

-Allen Wicks

[Veltliner]

Originally Posted by SierraDragon 

I would bet that many major apps will be taking good advantage of 8 cores during 2008; certainly one or more of the ones I use. That would be less than 25% of the time into the life cycle of new Mac Pros. Who cares if it exists now?

-Allen Wicks

Would you bet on your Mac Pro's CPU?

With "now" I obviously did not mean "right now". Was easy to see, wasn't it?

2008, a few months from now, can be practically counted in the realm of "now".

[Veltliner]

Originally Posted by SierraDragon 

Yes, hardware development typically is ahead of software development. That means that all one has to do is apply a little intellect to Intel's roadmap to see where the OS/apps world will be in the proximate future: 8 cores.

-Allen Wicks

Then why not apply a little bit more than just a little intellect and look at the roadmap of the software developers?

Now that would be interesting.

To see if they are ready.

And in what time frame - what challenges they are facing.

[awcopus]

Fascinating conversation, gentlemen. Information overload.

Still have the Mac Pro Waiting Blues, of course. But thanks to your speculations I feel better informed about the computer I may be able to preorder within the next 60 days for delivery in Q1 2008.

Is it just one of those crazy coincidences that turning the growth chart of iPod sales upside down accurately depicts the sales of Mac Pros?

[Simon]

Indeed an interesting discussion here.

For the sake of the argument let's assume Apple goes all octo on the MP after Nov 11 and let's also assume they would actually take the plunge and build a new Mac around Yorkfield.

I'm taking it that since nobody replied to this

Originally Posted by Simon 

Once they start selling Wolfdale or Yorkfield HEMs they will immediately be compared with the $600 Dell tower. I am well aware that that's where the majority of the desktop computer market is, but honestly, (and keeping in mind Apple's history) who here actually believes Apple wants to try and compete with that $600 Dell?

I'll assume nobody here believes Apple wants to directly compete with the $600 Dell. But if we put together a Yorkfield HEM, what's a realistic price point?

• Core 2 Extreme QX9650, (3.0 GHz, 12 MB L2, 1333 MHz FSB) ~$900
• 2x1 GB DIMM ~$50
• 320 GB SATA HDD ~$50
• 8x DL SD ~$25
• Radeon HD 2600 XT (256 MB) ~$100
• Bearlake chipset ~$100
• case, other material, documentation, etc. ~$75

That makes a BOM of ~$1300. Add in 40% margin. That puts as at ~$1800.

Two conclusions:
• as expected in terms of price it can't compete with the $600 Dell — but more importantly can it compete with similarly configured desktops at all?
• it's price comes out comparable to the 24" iMac and well below the MP — lots of cannibalized sales there; is that good or bad (i.e. would the total margin [sold units x margin] be greater for this box than on the iMac or MP?

If we assume Apple isn't being run by idiots and if we also assume Steve's not being religious (which admittedly is a bold assumption) they must have come to the conclusion that the answer to both questions is no. I expect many people on this board will say yes twice though.

[Simon]

Another point. If they go for a cheaper Yorkfield that would bring down the BOM a lot.

For example if they'd go for the Core 2 Quad Q9450 (2.66 GHz, 12MB, 1333 MHz, Jan 08) for ~$300 they could reduce the overall price of the above Mac to ~$1000. That's one serious iMac killer! On the other it would also reduce their profit on every box sold by $120 (-23%).