This is not a "Pro" machine.

[GroundLoop]

I hope you're not wearing white.

Kool-Aid stains are holy HECK to get out and I see a drenching in your future.

I am unable to recall a time where computers went BACKWARDS in terms of connectivity speeds. Having TB2 replace PCIE is like having FW800 replaced by FW400.

I guess I need to get dressed like I am in the front row of a Gallagher show...

(wow I am old)

GL

 

[flat five]

Can't argue the technical point

i've tried but nobody wants to talk about it in human terms.. it's always (n-7)bbps/44 * whgvsfk = i am right

numbers aside, what exactly are you arguing for? according to this thread (well, this forum even), there's something super important you're arguing about and i fail to see it. maybe you can lay out a scenario where this speed you're arguing about is highly beneficial to the user.

 

[throAU]

If they used standard PCB and component yes. But we are talking about a limited run new form factor component for the GPU and CPU board...

The Mac Pro doesn't produce enough sales to justify a big production of those parts.

Remains to be seen.

Maybe Apple and AMD have a deal across the product line -

AMD gives apple cheap mac pro GPUs
Apple promises AMD business of x million GPUs across their product line over x years


There's every possibility apple may be getting the dual GPUs at significant discount to others due to something similar to the above.

AMD are in a pretty shaking position at the moment, if they could get guaranteed future earnings like that I'm pretty sure they'd be keen to sign up.

 

[aggri1]

I am unable to recall a time where computers went BACKWARDS in terms of connectivity speeds.

How did IDE compare with then-current SCSI, when Macs switched? (Yeah I realise that there were also other connectivity options...).

 

[mcnallym]

Well said.

To compare, we need to try to guess at what Apple's cost is for the current Big Old Aluminum case is a challenge...on the one hand, its a really nice case. On another, it has also been around since 2003 (virtually unchanged). We do know that one can buy a generic PC case for $100 retail, so after some SWAG massaging, its probably safe to say that Apple's wholesale cost is sure to be under $100.

From there, we would need to figure out the incremental cost of that $100 case being bigger or smaller ... quite frankly, it really isn't worth all that much of an effort, because we can assume it is huge ... 30% ... and that number means that it equal to the cost of just the Thunderbolt cable in the above inventory list and our cost comparison is over.



Extruded deep draw aluminum, which is then treated somehow to make it glossy black.



Agreed: Apple is taking a significant departure from a lot of well established and cost-amortized designs & interfaces from the WinTel PC side of the world, which means that Apple gets stuck paying 100% of their lifecycle fixed costs...or more accurately, Apple pays them upfront, and then sets the price so as to pass 100% of these higher expenses along to the customers to pay.

Golly gee thanks, Apple!



-hh

Quote:
Originally Posted by mcnallym View Post
Agreed: Apple is taking a significant departure from a lot of well established and cost-amortized designs & interfaces from the WinTel PC side of the world, which means that Apple gets stuck paying 100% of their lifecycle fixed costs...or more accurately, Apple pays them upfront, and then sets the price so as to pass 100% of these higher expenses along to the customers to pay.

You forgot to put "FIXED COST" in big bold italic letters and link it to the wikipedia article

Of course I got "yaddayadda'd"' when I point out basic micro-economic concepts. Logic is no match for unchecked and unfounded conjecture. Apparently my old Mac Pro case is made of solid gold and the new one will be made out of fairy dust.

I was wrong about the plastic in the new case, but that just makes the argument even more ridiculous.

I think you might have gotten the quoting slightly out there as the part you are quoting me on was actually by -hh. I can't think in language terms like that quotation.

I never mind being corrected, (my name isn't Mr A Right), as long as it is something I have actually said.

 

[throAU]

How did IDE compare with then-current SCSI, when Macs switched? (Yeah I realise that there were also other connectivity options...).

SCSI had many, many different formats - regular, fast, wide, fast-wide, ultra-wide, etc.

So it depends. Then-current (at the time) ATA would quite likely have been faster than the SCSI standard it was used to replace.

Also, the big benefits from SCSI came from being able to connect many drives, command queuing, etc - and a lot of those benefits are marginal on a single user machine, for a lot more cost.

 

[tuxon86]

Remains to be seen.

Maybe Apple and AMD have a deal across the product line -

AMD gives apple cheap mac pro GPUs
Apple promises AMD business of x million GPUs across their product line over x years


There's every possibility apple may be getting the dual GPUs at significant discount to others due to something similar to the above.

AMD are in a pretty shaking position at the moment, if they could get guaranteed future earnings like that I'm pretty sure they'd be keen to sign up.

Not in the real world...

1- the w9000 is the top of the line AMD gpu. No one give discount on the top end model. Beside, Apple market for worksation is ridiculously small and is shrinking even more as we speak.
2- The nMP won't sell in the millions. This is a workstation not a simple desktop computer and it will be priced at such. It used workstation parts, like a Xeon CPU and ECC memory. Those are "niche" low production parts hence the markup on them.
3- Apple never ever undercut their hardware.
4- While Apple is filthy rich, their overal computer division is lagging way behind the other manufacturer like DELL, HP or Lenovo. If they can't get the parts cheap while selling way more workstation than Apple does and will then why should Intel give them a pass?

 

[freejazz-man]

No, but yo could if you needed to at no extra cost except for the drives themselves and a couple of $2.00 sata cables. That's his point.

I'm aware that's his point - my point is that "I need it to be in the case" is rather weak reasoning

 

[thekev]

Remains to be seen.

Maybe Apple and AMD have a deal across the product line -

AMD gives apple cheap mac pro GPUs
Apple promises AMD business of x million GPUs across their product line over x years

I think you guys are examining this from too narrow a view. If our assumptions are that it must be examined from where the line was a year ago and at that same pricing, I think 2 x W9000 would be unrealistic. It would be the most expensive hardware they've used to date. The firepro thing could be more branding than anything here. As for millions, even with 2 gpus per machine it is not feasible.

Not in the real world...

1- the w9000 is the top of the line AMD gpu. No one give discount on the top end model. Beside, Apple market for worksation is ridiculously small and is shrinking even more as we speak.
2- The nMP won't sell in the millions. This is a workstation not a simple desktop computer and it will be priced at such. It used workstation parts, like a Xeon CPU and ECC memory. Those are "niche" low production parts hence the markup on them.
3- Apple never ever undercut their hardware.
4- While Apple is filthy rich, their overal computer division is lagging way behind the other manufacturer like DELL, HP or Lenovo. If they can't get the parts cheap while selling way more workstation than Apple does and will then why should Intel give them a pass?

Again AMD doesn't have much to lose by lending the firepro name to indicate some amount of opengl optimization. I do not think Apple would go for $6000 of gpu power. It seems completely ridiculous. On that note the cheaper mac pros in their current line don't have much in the way of expensive parts. Even when they debuted they were overpriced from day one. It's not a matter of whether a market exists, just that they have poorly aligned value. People would buy due to a requirement of OSX and that the lower product lines didn't present an adequate solution.

 

[tuxon86]

I think you guys are examining this from too narrow a view. If our assumptions are that it must be examined from where the line was a year ago and at that same pricing, I think 2 x W9000 would be unrealistic. It would be the most expensive hardware they've used to date. The firepro thing could be more branding than anything here. As for millions, even with 2 gpus per machine it is not feasible.




Again AMD doesn't have much to lose by lending the firepro name to indicate some amount of opengl optimization. I do not think Apple would go for $6000 of gpu power. It seems completely ridiculous. On that note the cheaper mac pros in their current line don't have much in the way of expensive parts. Even when they debuted they were overpriced from day one. It's not a matter of whether a market exists, just that they have poorly aligned value. People would buy due to a requirement of OSX and that the lower product lines didn't present an adequate solution.

What they have presented in the nmp correspond specs and vram wise to the w9000. Those retails at $3.2k each.

 

[deconstruct60]

Not in the real world...

1- the w9000 is the top of the line AMD gpu. No one give discount on the top end model.

But it actually isn't a W9000 that Apple will be selling. An equivalent performance/computational features but not the same.

2- The nMP won't sell in the millions.

Not particularly material. The old Mac Pro didn't sell in the millions. The sky is blue.... so what.
There is no need to get into the millions range to get discounts and better GPU functionality pricing.
FirePro cards are high priced because of the sky high ( higher than Apple's standard mark up) profit margins. Neutralize that somewhat and the price will be more affordable. It has nothing to do with the cost of the components or volume level buying of component discounts.

3- Apple never ever undercut their hardware.

The W9000 is an AMD card. Apple's card is an Apple card. Undercutting a card not made by them ( or technically contract made by them) is not undercutting their own hardware.

The FirePro series is the only one where AMD made all of the cards itself. The mainstream stuff is *NOT* made by AMD (contract built). From all appearences ( board design being very Apple like ) this nMP board is an Apple board, not an AMD one. All Apple needs from AMD is the actual GPU package and the certification that this is a "FirePro" or "FirePro compatible" GPU subsystem. The rest of the components on the card Apple can buy itself and put Apple's, not AMD's , mark-up on.

4- While Apple is filthy rich, their overal computer division is lagging way behind the other manufacturer like DELL, HP or Lenovo.

Lagging way behind in profits? Profit margin? What metric are you talking about that is actually relevant to Mac Pro and associated GPU card pricing ?

I think you guys are examining this from too narrow a view. If our assumptions are that it must be examined from where the line was a year ago and at that same pricing, I think 2 x W9000 would be unrealistic.

Technically it isn't not a W9000 card. It is an equivalent of a W9000 card.
The flawed assumption is that an equivalent W9000 card has to be priced exactly the same. It doesn't. Even more so when they are not physically interchangeable. A "cheaper" W9000 equivalent card does little to drop off the sales of a standard PCI-e card format W9000. There is a close to perfect market segmentation because Apple's card won't work in other machines, so it doesn't stop the sales of W9000 into those other machines.

A user would have to buy the whole Mac Pro system instead of the other system is the only way get some movement to this cheaper W9000 equivalent. Given the folks buying the "box with slots" are likely bigger "box with slots" lovers that movement probably isn't going to be very significant.

Apple can put in a much more affordable W9000 equivalent card by just assigning "normal" Apple margins to the card.

 

[thekev]

A user would have to buy the whole Mac Pro system instead of the other system is the only way get some movement to this cheaper W9000 equivalent. Given the folks buying the "box with slots" are likely bigger "box with slots" lovers that movement probably isn't going to be very significant.

Apple can put in a much more affordable W9000 equivalent card by just assigning "normal" Apple margins to the card.

The firepro cards also aren't that critical to AMD the way Quadros are to NVidia. NVidia really owns most of that market on the PC side.

What they have presented in the nmp correspond specs and vram wise to the w9000. Those retails at $3.2k each.

Those have been on the market since 2012 as higher end gpus run for long life cycles. To expect both the price to remain static and for Apple to implement $6000 worth of gpus given their past behavior seems highly unlikely. If the 12 core cpus follow the current path, those will be upwards of $1500. I've looked at the breakdown with the current model. They aren't going from $2900 worth of cpus (at launch) to $1500-2000 for cpu + $6000+ for gpus. In my experience trying to shoehorn existing parts at launch pricing never works out.

 

[tuxon86]

The firepro cards also aren't that critical to AMD the way Quadros are to NVidia. NVidia really owns most of that market on the PC side.



Those have been on the market since 2012 as higher end gpus run for long life cycles. To expect both the price to remain static and for Apple to implement $6000 worth of gpus given their past behavior seems highly unlikely. If the 12 core cpus follow the current path, those will be upwards of $1500. I've looked at the breakdown with the current model. They aren't going from $2900 worth of cpus (at launch) to $1500-2000 for cpu + $6000+ for gpus. In my experience trying to shoehorn existing parts at launch pricing never works out.

Depend on which market they are targetting. For media/3d market where most of the processing can be delegated to the GPU then yes, having a lower $$$ and spec CPU + $6k of GPU is the thing to do.

It's a question of supply and demand. AMD isn't going to start and retool a totaly different production line to accomodate apple new design without making apple pay for it, which in turn will be passed on to the consummer. Apple isn't microsoft or sony, they don't sell at a loss. The nMP doesn't use standard pcie pcb and they'll probably be the only company to use it, making it quite a risk for AMD.

You have to keep in mind that both the overall workstation and particulary the Apple workstation business is shrinking. There are also new player in this field like Boxx that target the media/3d market succesfully. HP and DELL are also the two major player and they use standard PCIe pcb based cards.

The economic turmoil that the world presently go through also mean less corporate capitals to spend in niche market IT gear. Meaning that the industry and research is looking at lower price workhorse instead of trendy design.

 

[deconstruct60]

The firepro cards also aren't that critical to AMD the way Quadros are to NVidia. NVidia really owns most of that market on the PC side.

You are kidding right? It is important to both vendors as discrete card vendors. ( I suppose if mean AMD in total (GPU + non GPU business) also then yes. But the non GPU parts aren't going to subsidize the discrete GPU business. The non discrete GPU parts of AMD would retire the cards just as much as enable them.).

http://www.anandtech.com/show/6137/the-amd-firepro-w9000-w8000-review-part-1/3

The lower end discrete graphics card market is largely a race to the bottom at this point. These Pro market cards are largely keeping the heavy R&D investments going.

Long term if the discrete card portion of AMD wants to survive they have to get back getting a decent share of the Pro market. It isn't a question of outselling Nvidia any more than it is a question of Apple outselling HP, Dell, or Lenovo.

It is more likely that Nvidia scoffed at trying this idea. They aren't in any of the upcoming game consoles either. AMD is more flexibility because they are behind and not addicted to a single extremely high mark-up solution.

To expect both the price to remain static and for Apple to implement $6000 worth of gpus given their past behavior seems highly unlikely.

The whole point is that there never was $6000 worth of components there in the first place. The age , in year one versus year two , isn't particularly material. I'm certain Apple never even offered to pay anywhere near $6000 for this stuff. It is probably a number higher than some of the mainstream card vendors AMD deal with, but Apple would have stripped out extremely high mark ups. Apple doesn't roll that way.

If the 12 core cpus follow the current path, those will be upwards of $1500. I've looked at the breakdown with the current model.

And? The top end part now is $1440

http://ark.intel.com/products/47919/Intel-Xeon-Processor-X5667-12M-Cache-3_06-GHz-6_40-GTs-Intel-QPI

Two is $2880 ... if the new E5 v2 12 core part is $1800-2000 that is a $800-1000 cheaper. It is still going to be in same price range (over $6K price range).

They aren't going from $2900 worth of cpus (at launch) to $1500-2000 for cpu + $6000+ for gpus.

Why not if the price goes up?

I suspect suspect there will be a

12 core + W7000 equivalent + W9000 equivalent in the around the same $6+K range the old top end Mac Pro priced out at.

Dual W9000 equivalent could easily just be priced higher. There is no likely rule that all the BTO configuration have to come in at the same price levels are the old ones.

However, I think you are still missing the boat. Even at $2000 a pop a W9000 is still a high enough mark up for both Apple and AMD to make some profits. I suspect that Apple's price for there isn't going to be anywhere near the current $3K street price for W9000's. When throw in the cheaper CPU cost and swap the old GPU cost and don't pair the W9000, the price is going to be in a very similar ballpark as the top end for Mac Pro's now.

I doubt either the W9000 equivalent or the 12 core CPU package are part of any of the standard configurations. So really primarily in BTO pricing land which can just go higher than the pricing on current standard configurations.

In my experience trying to shoehorn existing parts at launch pricing never works out.

Again Apple's new card isn't an already existing part. So that isn't even close to what I'm outlining.

 

[GroundLoop]

And? The top end part now is $1440

http://ark.intel.com/products/47919/Intel-Xeon-Processor-X5667-12M-Cache-3_06-GHz-6_40-GTs-Intel-QPI

Two is $2880 ... if the new E5 v2 12 core part is $1800-2000 that is a $800-1000 cheaper. It is still going to be in same price range (over $6K price range).

I don't believe that this is an accurate comparison. A better proxy for the upcoming 12 core CPUs would be the following:

http://ark.intel.com/products/64596...E5-2690-20M-Cache-2_90-GHz-8_00-GTs-Intel-QPI

This is the current top end 8-core single CPU (price: $2057). I believe that the 12-core processor that we have seen in benchmarks (Xeon E5-2697 v2) will likely carry about a true 15-20% premium on top the current E5-2690. The real issue is that Apple significantly overcharges for CPU upgrades in their BTO options. Example: Apple charges $2,400 to upgrade two E5645 to X5675

E5645:
http://ark.intel.com/products/48768/Intel-Xeon-Processor-E5645-12M-Cache-2_40-GHz-5_86-GTs-Intel-QPI

X5675:
http://ark.intel.com/products/52577/Intel-Xeon-Processor-X5675-12M-Cache-3_06-GHz-6_40-GTs-Intel-QPI

The "true" price difference between those CPUs is $1440 - $551 = $889. Therefore, Apple is charging a 35% premium for the upgrade ($2400/$1778). This premium is consistent across all of Apple's Mac Pro CPU upgrades. So, using the best case scenario (assuming pricing parity between the new 12 core with the old 8-core), we are looking at at least $2057 * 1.35 = $2,777. But, I believe that there will be around a 15% premium of the E5-2697 v2 over the E5-2690. That would bring the price of the E5-2697 v2 to $2,777 * 1.15 = $3,194.

Note: I used these part numbers as they are known to be used in the new Mac Pro. The entry level 12 core processors will be less expensive.

GL

 

[MacMilligan]

It is pretty obvious that the cards aren't going to cost Apple anywhere near $2000-$3000 retail each. Apple will buy the cards at a decent wholesale price, pay a bit of licensing to use the Fire Pro name, and slap the cards into the Mac Pro with tweaked drivers. Apple's cost will most like be in the hundreds per card, not thousands.

These cards will not compete with AMD's real Fire Pro lineup because they can't be used in anything other than a Mac Pro. They have their own form factor!

 

[Zemzil]

Technically it isn't not a W9000 card. It is an equivalent of a W9000 card.

What is costly on FirePro Wxxxx card is drivers certification. Beside some minor or specific differences (ECC Ram, SDI output), GPU and hardware are almost identical as Radeon 79xxx series.

The price is for having drivers that guarantee true mathematical results on display and deep support/reliability for CAO/3D architecture/design software.

 

[barmann]

Here is a PCIe SSD that was released three years ago...

http://www.techpowerup.com/134638/lsi-announces-warpdrive-slp-300-enterprise-pci-e-ssd.html

It has roughly the same performance as the "standard" SSD to be included in the new Mac Pro...at the time it cost over $11,500

So, in 2-3 years from now, well within the Mac Pro's expected life, there will be SSDs with this type of performance in the $500 range. Yet the new Mac Pro will not be able to support it.

My opinion is that bottlenecking the nMP's I/O to TB2 speeds for the entire life of the workstation is a bad decision. Technology moves too fast to artificially limit yourself.

That's so out of my depth .

First CPU performance is to be replaced by GPU performance, now drive speed is supposed to take over .

Can't wait for the boot time comparisons; will buy it from the same place that's selling the regular 1TB SSDs for 200 bucks . The ones predicted 2-3 years ago .

 

[subsonix]

That's so out of my depth .

First CPU performance is to be replaced by GPU performance, now drive speed is supposed to take over .

Can't wait for the boot time comparisons; will buy it from the same place that's selling the regular 1TB SSDs for 200 bucks . The ones predicted 2-3 years ago .

Memory hierarchies on PC hardware is organized from L1-L3 cache to RAM to disk. As you get further away from the CPU the access times gets worse by several orders of magnitude, this is the Von Neumann bottleneck. Faster storage is very much needed as almost nothing has happened there for decades , especially compared to the development on the CPU front in the same time.

 

[deconstruct60]

I don't believe that this is an accurate comparison. A better proxy for the upcoming 12 core CPUs would be the following:

http://ark.intel.com/products/64596...E5-2690-20M-Cache-2_90-GHz-8_00-GTs-Intel-QPI


This is the current top end 8-core single CPU (price: $2057).

believe what you want. The issue is far more so about how much Apple will be willing to spend on a major CPU component. It is far more grounded to select a package they have actually bought in large number rather than something they never did.

There is a cut price off for high end Intel Xeon chips though that Apple isn't likely to cross; at least for standard configurations. If there is a $1800 , $2000 , and $2200 12 core model, Apple is far more likely to pick the $1800 one, no matter what its base clock is.

If Apple had done a classic Mac Pro configuration they likely would have taken the either the $1440 E5 2665 (and staid in sub 2 * 130 TDP zone) or E5 2667 at $1552. It is highly unlikely they would have used the most expensive E5 available. That isn't the point of the Mac Pro. Never was. Being the most expensive machine they can trott out the door is not a primary objective.

The real issue is that Apple significantly overcharges for CPU upgrades in their BTO options. Example: Apple charges $2,400 to upgrade two E5645 to X5675

If you think the mark-up that Apple charges ( around 30% ) is high then the mark-up that AMD throws on the FirePro GPUs and cards is in the stratosphere. That is the whole point. If Apple just retreats the FirePro equivalent cards back to just "normal" Apple margin markup then those FirePro equivalent cards will be significantly cheaper.

Similarly for the 12 core E5. If Intel has a super high market up, Apple will just put theirs on top, but not going to select the one with the highest Intel mark up. It is a BTO option so those can spiral up pretty far, but even Apple knows there is a diminishing market the higher the prices go.

But, I believe that there will be around a 15% premium of the E5-2697 v2 over the E5-2690. That would bring the price of the E5-2697 v2 to $2,777 * 1.15 = $3,194.

Far more likely that Apple uses the E5 2695 v2. 12 cores caps out at 2.4GHz and only 115W ( rather than 2.9 and 130W ).

You'll have 12 cores just not at the high clock rate. It is all dependent upon Intel's mark-up. Most likely though Intel is going to price the 2697 v2 out of the zone that Apple will pay for.

The 2697 v2 isn't comparable to the E5 2690. The E5 2690 v2 is comparable to the E5 2690 (v1). The 2697 v2 is likely about not 15% more but $150-200 more than the 2690 v2. Intel prices are mainly going to go between the v2 models. Not across the old model. Between generation you want to match up equivalents. Those typically don't have 10+% swings in price

----------

What is costly on FirePro Wxxxx card is drivers certification. Beside some minor or specific differences (ECC Ram, SDI output), GPU and hardware are almost identical as Radeon 79xxx series.

The drivers on Mac OS X likely are not going to be segmented. There are apps now certified on the Mainstream cards. I doubt there is going to be much of a cost increase to that of the past. The Mac GPUs won't be the same as the mainstream cards ( as in the past ) but no where near the Pro card mark up.

It isn't just the drivers but some of the applications to certify that drive the costs up. More than a few of those aren't on OS X anyway so that "Drama" isn't going to necessarily drive up costs.

If Apple is primarily changing the physical layout and packaging there is really no new huge certification process to go through. Apple would have to pay a price for the base reference design ( high but recoverable ) but not particularly necessary to reinvent the wheel. It isn't like there is big competition for other replacement cards so new some new gimmick over the standard W9000 card to differentiate. They don't compete so no tweaks/quirks/gimmicks necessary.

The price is for having drivers that guarantee true mathematical results on display and deep support/reliability for CAO/3D architecture/design software.

And yet some of these same packages run on current Mac Pros. Having supported/reliable/stable drivers would be a good thing across all Macs. If OS X doesn't go chasing after gamer/ high frame rate corner case driver instabilities then there is no big gap between mainstream graphics drivers and the "pro" ones.

 

[goMac]

The drivers on Mac OS X likely are not going to be segmented. There are apps now certified on the Mainstream cards. I doubt there is going to be much of a cost increase to that of the past. The Mac GPUs won't be the same as the mainstream cards ( as in the past ) but no where near the Pro card mark up.

This. I think the Mac Pro's GPUs are highly likely to be 7970s, with 7970 prices, and just a different sticker. Maybe ECC memory (which isn't going to be expensive.)

 

[phrehdd]

Memory hierarchies on PC hardware is organized from L1-L3 cache to RAM to disk. As you get further away from the CPU the access times gets worse by several orders of magnitude, this is the Von Neumann bottleneck. Faster storage is very much needed as almost nothing has happened there for decades , especially compared to the development on the CPU front in the same time.

This is really an excellent point you make. Early computers had many bottlenecks and the usual mentality was to sell people on the notion that a more powerful cpu was the answer. RAM, L1,L2,L3 cache and faster bus along with matching storage/drives was by far a better solution than just running for the fastest cpu on the market. The other frailty of exploiting hardware are the operating systems and software. Of the latter, efficient code has gone by the wayside which furthers the issue of not getting the best bang out of the hardware.

More cores isn't always better but code that exploits distributed cores should be more a priority than a cpu that "more" cores. I remember back when certain networking software would wait until one cpu was maxed before it engaged another cpu on dual cpu boards. This was terribly inefficient. They later made strides to correct this problem. We see much the same these days with OS's and software simply not designed properly at any level to go beyond 30-50 percent of the power of these multi-core cpu's. We can thank indirectly IBM and Microsoft for leading the way to this ugly way of putting a cpu in a bog.

My past exploits - DOS, OS2, MS Windows, Unix, Linux, OSX T - ML. Light work with C, Assembly, Fortran, REXX, BasicA.

 

[GroundLoop]

Far more likely that Apple uses the E5 2695 v2. 12 cores caps out at 2.4GHz and only 115W ( rather than 2.9 and 130W ).

Just going on what we have seen from the leaked benchmarks. Apple still has the choice not to offer this CPU option, but it is the only piece of information about CPU options in the new Mac Pro in the public domain.

https://forums.macrumors.com/threads/1599368/

The 2697 v2 isn't comparable to the E5 2690. The E5 2690 v2 is comparable to the E5 2690 (v1). The 2697 v2 is likely about not 15% more but $150-200 more than the 2690 v2. Intel prices are mainly going to go between the v2 models. Not across the old model. Between generation you want to match up equivalents. Those typically don't have 10+% swings in price

So you are saying that the 2697 v2 outclasses the 2690 v2? I guess that would further justify the 15% premium that I discussed earlier.

GL

 

[Zemzil]

The drivers on Mac OS X likely are not going to be segmented. There are apps now certified on the Mainstream cards. I doubt there is going to be much of a cost increase to that of the past. The Mac GPUs won't be the same as the mainstream cards ( as in the past ) but no where near the Pro card mark up.

It isn't just the drivers but some of the applications to certify that drive the costs up. More than a few of those aren't on OS X anyway so that "Drama" isn't going to necessarily drive up costs.

If Apple is primarily changing the physical layout and packaging there is really no new huge certification process to go through. Apple would have to pay a price for the base reference design ( high but recoverable ) but not particularly necessary to reinvent the wheel. It isn't like there is big competition for other replacement cards so new some new gimmick over the standard W9000 card to differentiate. They don't compete so no tweaks/quirks/gimmicks necessary.

Yes it's not revealing to compare new MacPro GPU to retail FirePro ATI cards (technical pricing charts).

But ATI is a good choice, because it has strong reputation in GPGPU computing. CUDA is well supported by pro(sumer) software but ATI is a solide reference for science, simulation or brute calculation, as OpenCL is. And it does the minimal of what you could except for such hardware in other pro software that Nvidia had favor.

 

[deconstruct60]

Just going on what we have seen from the leaked benchmarks. Apple still has the choice not to offer this CPU option, but it is the only piece of information about CPU options in the new Mac Pro in the public domain.

There may be a super high end BTO option Apple goes with this time just to cover a subset of the old dual performance space. I doubt that will be a standard config. ( at least not without GPUs stripped way down from W9000 class ).

Going that high on price would be a change from their past practices. I doubt it is actually going to be successful offering because the Intel+Apple mark up at the extreme starts to eat significantly into $/performance. At least as many folks are going to balk at the pricing as buy the configuration.
There were both design and marketing problems with those options before that don't really go away with the new Mac Pro.

So you are saying that the 2697 v2 outclasses the 2690 v2? I guess that would further justify the 15% premium that I discussed earlier. [/quoet]

No, I'm saying if you want to project 2013 v2 prices then far cleaner and likely far more accurate to project up the prices for the 2690 -> 2690 v2 and then map to new "levels" to fill in the new v2 line up. The high end v2 prices will be relative to the low and mid range v2 prices. Not cross generations like you were doing.

For example if go to the v1 pricing

http://www.cpu-world.com/news_2012/2012030701_Intel_rolls_out_Xeon_E5-1600_and_E5-2600_CPUs.html

2670 -> 2680 +11%
2680 -> 2687 +9%
2687 -> 2690 +9%

The prices are already in "loopy high" land they aren't pushing them up at 15% increments at the top.