Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Very clever design but not "innovative" given that integrated circuits in general and solid state drives specifically have been around a long long time. However, they were very resourceful in being able to combine those to form a "computer."

Sigh. :rolleyes:

"You keep using that word. I do not think it means what you think it means."

Well alas, perhaps your grasp of the language remains lackluster. I'll stand by my statement and you may remain with your own thoughts.
 
Power consumption

I am impressed by the whole package. Yet, outside of the performance bumps, I'm also eager to see what the reduction in power consumption will equal.

When I moved my 2009 dual processor out of my house and into an office, my electric bill went down by almost $40/month (I might add that my home office also got a lot chillier in the winter!) That big box put out some HEAT!

This nMP will be heading back into my home office. I'm gonna guess it wont need as much juice as the older model. People who leave these on to render overnight might also be able to consider that reduced electricity in the overall upgrade cost. Then again, with improved render and export times, we might just be able to put 'em to sleep when we go to sleep!

All around a good thing.
 
Actually you are more wrong than he is.

As others have already pointed out more eloquently, this is entirely wrong, and I wonder what your motivation is in posting it in the first place. It's complete nonsense.
Baloney! Take a look at those printed circuit boards, do you see an excess of chips, or a bunch of old technology? A designer can have a significant impact on system power usage simply by making decisions based on what to include and what to exclude.
Apple is achieving the lower power design through underclocking the graphics chips and binning.
It appears that the only thing under clocked is the D700 or top end GPUs. We don't even know if this is for power usage or reliability. These top end GPUs are pushed really hard and as a result have higher failure rates than Apple might find acceptable. We can't really say that the under clocking is only due to the power issue.
There is nothing more fancy going on than that. The real challenge in any system build is effectively venting the energy that will be released as heat.
Sure that is a challenge but reducing heat in the first place effectively reduces that challenge to something manageable.
What Apple *is* very good at is creating creative (note I say "creative," which is different than "effective," although they can and do often overlap) heat dissipation systems, and the Mac Pro is no exception.
I would agree with that, they seem to lead the industry in that respect. However removing heat is not all there is to heat management. I would say reducing heat production in the first place is a smart bit of engineering. You have to remember the CPU is only part of the heat equation on a motherboard. Simply addressing losses in the CPU chip power supply can have a big impact.
Also nonsense. No such complete system upgrade should be required.
I can not agree with you at all on this one. With today's hardware it is absolutely silly to consider CPU or GPU upgrades as something viable.
You're also making the false assumption that everyone's needs are bound by system-wide limitations, distributed evenly. The bulk of the Mac Pro's computational power is tied up in its graphics cards through OpenCL—it is not a particularly relevant computer if your needs are CPU intensive.
I can't agree with this either, many apps are computationally intensive and still bound to the CPU. That might be due to the code base not translating to OpenCL easily or it might be laziness on the part of the developer, but the fact remains the Mac Pro is still a machine where the CPU is important.

More importantly though is that in a year or two XEON will likely be on a different socket supporting a new RAM interface that will require a new motherboard anyways. GPUs are just as likely to go through major overhauls also. In the end you have to upgrade so much of a system to realize the benefit of new tech that it isn't worth it to consider GPU or CPU upgrades for the vast majority of users.

Think about it for a bit, the GPUs in these machines represent about 2/3rds of the cost of the whole machine. It depends upon the machine of course, but you will have a hard time convincing me of the value in putting that much money into an upgrade of an existing machine. This is especially the case if it has been around for more than two years.
The ability to upgrade the GPUs, then, is paramount to the useful lifespan of the machine, since GPU revisions and developments occur at a rapid pace.

Again baloney! This was certainly the case in the past, but we aren't seeing the massive bumps in GPU performance anymore every year. At best the GPU manufactures come out with new architectures every two years and even then the performance jumps are no where near as dramatic as in the past. So if it takes 2+ years to see a new GPU what has happened to the rest of the technology in the machine? By the time 3 years has passed I see no value at all in an upgrade to the GPUs that doesn't also address the CPU and at that point you might as well buy that new machine.

Even if all your apps are tightly GPU bound and you know that for a fact, I just don't see a huge value in a GPU upgrade these days. We just aren't seeing massive growth in GPU performance anymore.
 
I'll always think of the old Mac Pro case as the most beautiful tower computer ever, but the new circular design is probably the most innovative thing I've seen in the PC space since the PC came to be.

Simply genius.
 
With today's hardware it is absolutely silly to consider CPU or GPU upgrades as something viable.

If I swap a CPU and it actually gives me a major improvement in performance, how exactly is that "silly"?

And the fact is, it is possible to swap the CPU for a much faster one, and doing that does give a major improvement in performance.
 
If I swap a CPU and it actually gives me a major improvement in performance, how exactly is that "silly"?

And the fact is, it is possible to swap the CPU for a much faster one, and doing that does give a major improvement in performance.

Many people will consider this silly because in this scenario you do not pay Apple. Keep in mind that many members on this forum are Apple investors and their perspective is very different from yours.
 
You're on! I foresee us quaffing significant amounts of regional craft brews and co-authoring dirty limericks about Apple design engineers. :cool:

Oh my... what a thought (lol).

Once was an Apple designer from Nantucket, who made the new Mac Pro look like a bucket....

I can't do this! argh
 
]It appears that the only thing under clocked is the D700[/B] or top end GPUs. We don't even know if this is for power usage or reliability. These top end GPUs are pushed really hard and as a result have higher failure rates than Apple might find acceptable. We can't really say that the under clocking is only due to the power issue.

It also appears that GPUs are by far the main consumers of power in this machine (and many PCs in general). So obviously underclocking them is a big deal. And as just published AnandTech review shows D700 is twice slower than the modern AMD graphic cards for PCs. So there is no magic here. You can either have a powerful computer or a one with low energy consumption.
 
It also appears that GPUs are by far the main consumers of power in this machine (and many PCs in general). So obviously underclocking them is a big deal. And as just published AnandTech review shows D700 is twice slower than the modern AMD graphic cards for PCs. So there is no magic here. You can either have a powerful computer or a one with low energy consumption.

I didn't catch the TWICE slower. I though it was in the realm of 17% IIR?
 
Oh my... what a thought (lol).

Once was an Apple designer from Nantucket, who made the new Mac Pro look like a bucket....

I can't do this! argh

An Apple designer got plowed,
And crawled his way home much too loud.
His wife cried, "Just a minute,"

And that's as far as my lawyers advise I should proceed. ;)
 
I didn't catch the TWICE slower. I though it was in the realm of 17% IIR?

It's on the Gaming Performance page. For dual D700 with CrossFire enabled pretty much all comparisons there look like this one:

60704.png
 
As someone who's experience in fixing electronics (satellite receivers, security cameras) involved about 6 months in a little store that only had 3 employees in total, the capacitors seem like the most likely thing to go out in any electronics. I was continually amazed at how my boss managed to charge people $75 a pop for a job that involved replacing a single capacitor. Anything that improves the experience with that is inspiring (especially when so many company's use poor quality ones).
I've changed more than a few in my lifetime. High reliability capacitors are very expensive at least relative to standard capacitors. It is interesting that in some cases the electrical shock from the knock off I devices chargers has been blamed on cheap capacitors.

As for your boss charging $75 that is pretty damn cheap really. Your wages, operating costs and other factors like taxes means that companies like his need these sorts of jobs just to stay in business. On the jobs that take a lot more than 5 minutes it is far harder to make money and charge people for the repair.
For those that don't know, replacing a single capacitor is extremely easy, and takes less than 5 minutes even if you are inexperienced. This is something that you could probably teach to 90% of your average grocery store workers (provides they'd let you).

The replacement is easy, it isn't often easy to find the bad cap though. This especially the case if you are taking your first crack at a repair on something new.

As a side note years ago we had some DEC VT terminals (forgot which model) that would always fail in the same way. In talking with the field rep (yah back in those days you had contracts on everything) he would come in and replace a cap. That would be a nice service call for DEC and frankly an unhappy customer as one would expect a little engineering effort to fix a chronic problem. Between that and other issues it is no surprise that DEC died so quickly.

In any event what I was most concerned about is the idea that $75 dollars is a lot to charge somebody for a repair that takes 5 minutes. Depending upon the company and the overhead involved it could be the minimal he can charge and stay in business.
 
Its easier to access/work with if it can fit on your desk.

Also, on-site work becomes much more doable.

On one hand, my desktop real estate makes the size and form factor of this nMP perfect. On the other, I already have all of my work and data external so having a big-ass case full of internal storage bays would be a terrible waste, not counting my aversion to spending any time tinkering inside a computer case.

I don't give a rat's about upgradeability or any of that computer enthusiast stuff. I want a machine that comes out of the box ready to work hard for me, and when its life is done I'll retire and replace it. Just my personal philosophy and again I do not presume to speak for anyone else.

I thank you both for your replies. It seems that much of what I have read of the new MP is the wow factor over what it looks like, not what it does. You both have given examples of why the design is better for you, most others haven't so it was starting to look like a form over function attraction, IMO.
 
I believe there are two TB controllers.
I'm under the impression there are three TB controllers, one for two ports to the outside world.
But very possible about the PCIe lanes if there are 40 and each GPU probably takes up 16. That's a downside of only doing single cpu machines, and forcing dual GPU.
I wouldn't be surprised to see Intel address this in future sockets and chipsets. Frankly Intel could completely do away with SATA support in its server ships right now that might free up some logic and I/O pins.
Frankly for non graphics folks, substituting the second GPU for a card that has three more of those SSD slots would be absolutely killer. Or even having more channels of external TB2. I wonder if the next generation xeon will allow more lanes and if that could allow more expansion in that regard.
Even one more slot would be major. As it is, I don't see this as I possible in a future machine. Apple simply had to work with what they had at the moment.
I assume this machine is PCIe version 3, meaning just under a GB per second per lane?

Most likely. This becomes an interesting subject because you need enough lanes to properly support TB ports. As such I have much interest in finding out how they actually allocated PCI Express lanes.

----------

That was certainly the case in the past but I've seen nothing recently that makes me believe that is true today or will be true in the next generation of cards.

Nope. New GPUs can offer massive performance boosts over even a single generation. A card two generations removed from the one you have now can potentially give you a 100% performance increase running off the same bus, all without requiring you to spend another $8000 buying an entirely new machine to benefit from it.

I'm not saying the lack of upgradeable GPUs will break the new Pros. It's got enough of its own strengths to overcome this one weakness. But it would've been really, really nice if they were in there.
 
I thank you both for your replies. It seems that much of what I have read of the new MP is the wow factor over what it looks like, not what it does. You both have given examples of why the design is better for you, most others haven't so it was starting to look like a form over function attraction, IMO.

Exactly. Usually when it comes to designing a new workstation, one would start with the technical specs like performance, memory and storage capacity etc. As is often the case with Apple they do it differently. Apparently they start with size and then see how much they can cram into it. It's a very different task which will always produce a better size-to-performance ratio. And it works for some easily impressionable consumers. Will see how it goes with professionals for whom desk space usually is a much lower priority than performance.
 
Right -- they can select lower power parts, just like everyone else.
Yet not everyone selects lower power parts. This especially the case in server and workstation boards. Attitudes are changing here as more customers become focused on power usage but in general using low power parts have not been common in servers and workstations.
As to the other board components, they're not going to contribute greatly, in relative terms, compared with the 3 main power-suckers.
Well yes and no. Obviously the big chips take the greatest percentage of power, but little efficiencies add up. Save a few percent in a power supply, deletes some chips you don't need and use low power variants else where and you can save a significant amount of power.
As to the technical minded tear-down, Anand just posted his review. I'm in the middle of reading it. He usually has some interesting findings.

Yes! Hopefully he has gone deep enough to clear up some of the speculation here.
 
I thank you both for your replies. It seems that much of what I have read of the new MP is the wow factor over what it looks like, not what it does. You both have given examples of why the design is better for you, most others haven't so it was starting to look like a form over function attraction, IMO.

You're more than welcome. I really get annoyed by the hasty generalizations on these forums, by people who apparently feel as though they can speak for anyone but themselves--that goes for people both in favor of and against the nMP's design. It is what it is and if it's the right rig for you you love it, and if it isn't...then you just need to move along and find what works for you.
 
That was certainly the case in the past but I've seen nothing recently that makes me believe that is true today or will be true in the next generation of cards.

The performance gains per generation have slowed down quite a bit over the last few years, but there's still enough gain there that couple of years could make for some big differences.

The biggest thing to take into consideration isn't raw graphics performance, but it's GPGPU capabilities, since this is what the new Mac Pro is more or less focused on. With the former, you won't see much difference on a year to year basis, and isn't really all that important as far as the Pros usual workload is concerned. But the latter is still a relatively new thing, and has seen some huge performance increases as newer cards have come out to better take advantage of it.
 
most others haven't so it was starting to look like a form over function attraction, IMO.

function over form & form over function are both design fails (well, not necessarily fails but....)

it's the marriage of form and function which truly sets (modern) history's great designs apart from all others..
 

Very interesting article to say the least. I totally missed that the CPU chip had 8 extra PCI Express lanes. If Anandtech is right the TB ports should deliver excellent performance in most cases.

I just skimmed around to the good parts, that is the stuff I'm interested in. Might read the article in depth later. In general though it should be pretty clear that we won't get any extra SSD slots this year. However that could change if Intel ever gets USB 3 built into the chip set as Anand is indicating that Apple had to use a PCI Express to USB 3 bridge chip. So maybe with a little consolidation whenever Intel delivers the next Xeon chip set we might actually see a Mac Pro with two internal SSD ports.

In any event if anybody is really interested in the new Mac Pro the article is highly recommended. Notably he had to work extremely hard to get the machine to back off on clock rates due to thermal issues.

----------

Which says

Did you miss the part about the 8 lanes on the north bridge chip? That right there eliminates my confusion about where all the I/O goes. It is a very interesting design to say the least. Apple was very good at supplying plenty of bandwidth to the various chips in the platform.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.