Again, that's where we disagree. A gpu is a part they buy off the shelf. The two different daughtercards are different proprietary parts that they need to design themselves.
You just cherry picking what you want to focus on. Frankly you missed the part were one of those was a 1GB and the other a 2GB. The number of VRAM chips is likely different hand yet that doesn't prohibiatively drive up costs.
fixed upfront design R&D costs of moderate size amortized over a large enough base of units sold don't drive product differences. Amortized costs of $20 or less don't do separate products in the Mac Pro price range ($2000+). It is 1% or less than the cost of the system.
As would be the case if Apple used i7 on the low end and dual xeon on the high, yet in that case you'd insist it's suddenly two different products.
You are changing component that will be bought in quantity. Xeon generally use the same RAM. So the higher volumes in combining the single and dual offerings. Some with IOHUB... higher overall volumes if combine.
Same with amortizing design costs for the dual. If substantially overlap with the single version than can amortize R&D costs over a much larger base.
All that using a i7 and forking single dual design does is drive component and amortized design costs for the dual processor product up. That's it. There is zero cost saving for the user on the single model. The CPU packages largely cost the same (for the same socket and performance) and the marginal RAM difference won't get reflected in Apple's pricing. The margin is higher for Apple but there is no net increase in value for the customer. Nope, zip, nada. It makes for a weaker separated upper end product. ( market forces more prone to kill it off) and a juiced for no good reason lower end product ( likely higher long term costs for Apple since more tweakers with misguided overclocking leading to more long term support costs. )
So you're saying that the design needs to be different for single and dual xeon configurations. That's exactly my point - it's not like they're selling one design and just leaving off the second CPU.
You are grossly exaggerating the differences. The primary inputs and outputs to the daughtercard can be designed to exactly the same specs. That means no design changes between the two.
The CPU + DIMM subcomponents can be either be copy pasted to make two or even less expensively tweaked cut and deleted to make one. If design the dual package daughtercard first then the single card is largely a set of removals from the first. Characterizing that as some costly design change in an era of electronic design tools is humorous. No, not really. zero cost difference no. Large significant cost difference also no.
It isn't even the whole daughter card that is changing. Being very generous it might be 50% of the card. It is likely less than that if start with the dual and mutate that into a single. Removal of the 2nd CPU socket and DIMMs actually makes layout easier. Just center the socket into the "hole" created with the deletion.
Your commentary is indicative that your mindset is warped with by perhaps a methodology of dual the single first and then some substantial juggling to fit two. Removing parts from a design isn't that hard.
I assume you mean apple is not?
sorry about that. Yes Apple will not be selling empty CPU socket product offerings. Therefore. also not DIMM sockets that won't work at all without another CPU package either. No extra socket => no extra DIMMS.
And would that be a bad goal?
A single CPU machine is by definition going to be less capable than a dual.
Not really. Depends upon software being utilized. It is increasingly common that two GPUs would bring more computational impact to computational analysis than two CPUs. Depends upon the software.
The other issue is that higher core count typically leads to slower clock speed. Single CPU package offerings will generally clock higher. There are folks with legacy software stacks that need faster clocks to go faster.
The biggest problem with the low end MP is that it's way overpriced for the performance it gives,
That is largely driven by very bad Intel offerings. The low end is still stuck on Nehalem 3500 offerings that are several generations back.
so I don't see what would be wrong with an update that prices it appropriately.
There are several Xeon E5 1600s that do just exactly that and have much higher component synergy with dual Xeon E5 2600 offerings.
Apple probably should find a way to move the base MP price back close to the $2000 boundary between the iMac and Mac Pro. Something like $2,199 or ($2,099) instead of $2,500 starting point would boost Mac Pro sales without unnecessarily fratricide on too much of the upper end iMac BTO configs. There is too large a gap between the base price offerings.
Since folks seem to be wishing to inflict Thunderbolt on the Mac Pro a embedded GPU only BTO offering at $2,199 would work well . Folk who want their own GPU card could also boost the 3rd party card market also. (minimize to 4GB RAM and minimal 500GB HDD since folks are largely going to toss those those two components also.)
But driving into the sub $2000 range? No. that really isn't a Mac Pro anymore. That is basically a product to shrink iMac sales which makes no sense.
But how much of that is due to Apple not offering a real update in years?
HP , Dell, etc didn't offer any substantive updates in 2011 either. Intel didn't offer anything new.
The overall workstation market has been highly suppressed over most of this "Great Recsssion".
" ... The results from the workstation market's first quarter were disappointing, as it marked the first two consecutive quarters of negative growth since 2008. And if industry observers needed more evidence of the market's listlessness, they need look no farther than the second quarter's numbers ... "
http://jonpeddie.com/press-releases...-sluggish-in-q212-reports-jon-peddie-researc/
" ... The workstation market proceeded ahead in the fourth quarter of 2011, completing a long climb back from the 2009 depths of the global economic recession that had slashed its shipments by over 40 percent. But that fourth quarter advance was more a case of "steady as she goes" than "full speed ahead", reports Jon Peddie Research senior analyst Alex Herrera. ..."
http://jonpeddie.com/press-releases/details/steady-as-she-goes-for-the-workstation-market-in-q411/
So yeah.... this isn't an Apple only problem. Characterizing it as such doesn't have much if any quantitative evidence to back it up.
Most of the talk is "Oh I'm gonna when apple finally updates it" - until Apple releases a MP with hardware that's not behind the times, they're not going to sell many of them.
Frankly, this is largely the same bluster that accompanied the XServe. If only XServe had dual power supplies you're really going to see sales spike then. Apple has the numbers of units sold for both Mac Pro and overall sales of competitive offerings.
Expensive capital equipment gets refreshed on a longer lifecycle period than less expensive stuff does.
And for the most part it mainly IS just Apple, who else is offering a product line that for example doesn't have sata III, among other things that are way out of date on the MP?
And how many of those have anywhere near Apple's cash flow from operations from their PC line ups as Apple has from their Mac line up ?
Dell is going private. HP is floating around the idea of splitting off PCs again.
Markets that are going backwards generally get low priority assignments at Apple if not complete exits. If workstations were growing at a 15% year-over-year clip Apple would be all over them like stink on *****. It isn't so they are not.
There is a chance with an increasingly larger base of software that can leverage more GPGPU resources that may change with an influx of new users (formerly on clusters but not on individual workstations) to offset the relatively slow capital upgrade lifecycle. If it doesn't materialize this will probably be Apple's last stab at it.