2016 nMP

[Serban]

Wauw, 1 game Will it be released 1 year after the big release on console / pc ?

just teasing you.. games are simple not on this platform. The 99% of all the great games is not here.. and we know that it will be so for a long time

hmm some of you may think all blizzard games are great, like Wow starcraft diablo and all are for Osx too
but yes games are working better with windows

 

[AidenShaw]

What Apple needs is something that takes the cooling requests from each of the three sources and applies that to the one fan. Suboptimally can fake that with another set of sensors that are centrally link than generate the indirect feedback but.

Wouldn't it be innovative to put a sensor and small PWM fan on each major heat source, and let the PWM case fan run as necessary to keep the ambient temperature inside the case at a comfortable level?

 

[filmak]

The difference between a 450W or 550W PSU isn't that big.

Isn't that big but many times these small details make all the difference.
Performance never comes for free... needs more resources.
or at least with the extra watts the PSU could be less stressed and cooler.

 

[h9826790]

But keep in mind that heat transfer from the cooler to the air works better with turbulent flow that with laminar flow. While the fans should be placed reasonably to generate the air flow, too much aerodynamic optimization could lead to suboptimal heat transfer.

I doubt about this. AFAIK, the best airflow is laminar attached airflow. However, the world is not ideal, laminar flow usually seperate earlier than optimum. Therefore, the next best option is turbulent attached airflow.

And this is why there are some small turbulent generator (or small cut) on the fan blade, which is actually for airflow optimisation (keep the airflow attached, but the downside is that the flow will become turbulent but not luminar).

This should be correct about the fan efficiency. I am not 100% sure if this is also true for heat exchange. But turbulence means less contact between the heatsink and air particles. Hard to believe that can improve heat exchange.

 

[lowendlinux]

The problem with the thermal core is the heat of one part directly effects all the other parts

 

[linuxcooldude]

The problem with the thermal core is the heat of one part directly effects all the other parts

As was mentioned, its not as easy as that. It depends on the materials used in heat sink, the length, size and shape, the type of cooling fan, whats being cooled, Not many forum users will know or understand the science behind it.

 

[SeaCaptainBiscuit]

The whole design is relatively rack hostile so why would rack measurement units be a completely non negotiable design constraint???? The previous Mac Pro was 8.1 between 4U ( 7" ) and 5U ( 8.75" ) and the world didn't end. Given you'll need a cradle to hold the round object to fit in a square hole the 0.4" seems a bit of a tight squeeze for a strict 4U fitting.

Think of it as a "monkey's paw" type of thing: people keep asking you for something easier to rack-mount, and no bigger than 4U, you give them a computer they can put 2 of in a 4U rack space...why are they still complaining?

Yeah, sure, it's also some goofy round thing, but that's the problem with getting your wishes from monkey paws, no?

Heavier and slightly more expensive components? I don't really see the 'show stopper' issue. Perhaps a put more money in Apple's pocket issue, but technical one? Yeah it is quite opaque as to what these might be.

One of the issues is how the thermal design and the use of the central switchboard in the base both benefit from having the CPU-and-GPUs as close to the bottom as you can reasonably get them: moving the components "up" adds to the wiring length -- and if you move CPU/GPU up uniformly, you can be adding ~ 2X that to the wiring length -- and similarly for thermal reasons having them placed lower is better than having them placed higher.

So it's not that you couldn't make it an inch or two taller, there's just not much benefit to it and there's some minor drawbacks. Going wider is pretty logical but the 4U thing is what it is.

EG: say you make the thing an inch taller. The optimal place for CPU/GPU/etc. is still going to be about where they are vertically, so if you leave them in the optimal spot you've added dead space at the top for no real benefit. If you move them up half an inch or so you add about an inch to some of the data paths -- not a big deal, but why do it if you don't have to? -- and don't appreciably change the thermals. If you try to use the extra space to cram in a CPU2 next to CPU1, you overload the power supply, the thermals, make the wiring that much longer, and still only have 4 ram slots (so 2 per CPU...yeah...).

Not much benefit here unless the inch is enough to actually allow a larger power supply and also provide enough thermal improvement to accommodate the extra power draw (without materially increasing fan noise, etc.).

Rather than adding an inch, the next logical height to consider is having the central area be "2 RAM sticks tall", but that's a significant redesign and would likely need to be wider just for practical reasons, at which point you may not need to go that tall anyways.

 

[Stacc]

The most important measures of a heatsink are the fin area (i.e. how big is the heat sink) and the air flow produced by the fan that is blowing over it. Things like laminar vs turbulent matter but are relatively minor concerns.

A traditional PC tends to have many smaller fans. Modern CPUs and GPUs tend to have moderately sized heat sinks and moderate sized fans attached to them. The heatsinks on GPUs are especially constrained given the large power output ~ 250 W on high end cards and the limitations in size and airflow since it must fit in a PCIe slot. Then you must add in a couple of moderately sized case fans to insure proper airflow throughout the large case. So counting up the fans you would need in a modern workstation we would need 1 for the CPU, 1 blower fan on the GPU (which tend to be noisy) and 2-4 case fans. That gives us 4-6 fans and the system will be as loud as the hottest part with the smallest fan.

The problem with this approach is that the air moved by a fan increases as the fan size increases. If you want a quiet system and still maintain cooling performance you want big, slow fans. The problem with the traditional PC tower is that often you are limited to 80 mm or 120 mm fans at best on most components. Thus they have to spin faster and make more noise. This is the same reason why a macbook pro's tiny fan has to spin at 6000 RPM to cool a 45 W processor while the mac pro can dissipate 450 W at 1000 RPM.

This is where the mac pro comes in. If your goal is to design a very quiet system you want the biggest fan(s) possible. The compromise is that they left behind the standard PC form factor and commodity hardware. What it got them is a big fan that adjusts its speed to the hottest component and a huge heatsink. Now instead of having your smallest fans be the weak link in your cooling system you can have your one big fan cooling everything. This also benefits from the fact that the use cases where the CPU and both GPUs are running full tilt is going to be fairly limited. So if you are running a mostly GPU load then those GPUs can take advantage of that big heatsink and the fan can easily cope with the load.

Think of it as a "monkey's paw" type of thing: people keep asking you for something easier to rack-mount, and no bigger than 4U, you give them a computer they can put 2 of in a 4U rack space...why are they still complaining?

Yeah, sure, it's also some goofy round thing, but that's the problem with getting your wishes from monkey paws, no?

What is the use case where it would make sense to use a mac pro as a headless server and not just buy a linux server blade? I have to imagine this use of the mac pro is so limited that it makes no sense for Apple to build a machine that caters to this niche.

 

[SeaCaptainBiscuit]

What is the use case where it would make sense to use a mac pro as a headless server and not just buy a linux server blade? I have to imagine this use of the mac pro is so limited that it makes no sense for Apple to build a machine that caters to this niche.

I meant exactly what I said: suppose that when you did happen to talk with your customers, the most-common feedback about the case was "it's a great case, but it'd be nice if it was easier to rackmount", and 4U kept coming up as an upper-bound...

I didn't say there's a common or practical use for rack-mounted nMPs.

 

[lowendlinux]

As was mentioned, its not as easy as that. It depends on the materials used in heat sink, the length, size and shape, the type of cooling fan, whats being cooled, Not many forum users will know or understand the science behind it.

Actually it is easy as that

 

[Stacc]

I meant exactly what I said: suppose that when you did happen to talk with your customers, the most-common feedback about the case was "it's a great case, but it'd be nice if it was easier to rackmount", and 4U kept coming up as an upper-bound...

I didn't say there's a common or practical use for rack-mounted nMPs.

Is this a hypothetical or are people requesting this to you? I'm just curious what they are doing with the mac pros that they would want to rack mount it.

 

[fuchsdh]

Is this a hypothetical or are people requesting this to you? I'm just curious what they are doing with the mac pros that they would want to rack mount it.

Don't ask me, but people do it, otherwise Sonnet wouldn't be selling these: http://www.sonnettech.com/product/rackmacpro.html

 

[wallysb01]

Don't ask me, but people do it, otherwise Sonnet wouldn't be selling these: http://www.sonnettech.com/product/rackmacpro.html

Errr, that's fallacious. Tons of things are made to be sold, but how many are actually sold....

 

[beaker7]

Please pass whatever you're smoking if you think Apple designed the trash can to be rack friendly. It's easily the most rack hostile design of the modern computing era save for maybe the Cube.

 

[developer13245]

Thermodynamics is not something you learn reading popular mechanics.

1st your assumption doesn't considers there is an active component (the fan) blowing air that takes away heat immediately.

2nd in case there is no fan (passive heatsink) the temperature at adjacent components will not be the same as the cpu but the mean thermal along the heatsink area, explained in a way you can understand, if you heat a pan with am cpu (a big oan) and this cpu delivers 135W does all the pan have the same temperature as the cpu? No depending on the thermal conductivity of the pan's material the heat will dissipate higher close to the cpu than farther if it's conductivity is slow, and more symmetrical if thermal conductivity is high, and the pan actual temp depends on the area exposed to lower temperature material or vacuum (heat dissipation to the vacuum is the most difficult and represents the material's heat transfer limit).

So no matter if the cpu is 145W and the gpu are 110 as long the fan is active and the cooler TDP isn't exceed all components will cool down as required.

In case the thermal delta among components is a problem because requires the heatsink cooler to effectively cool down quickly, what they (Apple) need to do is to semi isolate sections of the thermal core this way delaying heat normalization along the core giving time to the air to cool down the critical components before the heat from other components reaches the area assigned to critical component.

The nMP thermal core is the most perfect solution, also better than liquid cooling (when no extreme temperatures are involved as on few cpu from AMD).

Apple to improve the thermal core either can improve the inner fins design to cool down quickly components with higher TDP also they can upgrade the fan (the cheapest) also switch from aluminum to copper in case the TDP delta is too extreme.

Your arguments don't have an engineering base, but seems influenced by those nMP haters.

The funniest part of your post:
"...what they (Apple) need to do is to semi isolate sections of the thermal core"

I'm sure they can figure out how to do that by reading a few issues of popular mechanics...

Your point is an outright contradiction and lacks coherent logic. It's a COMMON thermal core. By isolating sections it will no longer be COMMON, and thus will be a major redesign (regardless of whatever shape it morphs to).

Sorry, even popular mechanics adheres to basic science. You're "isolated common thermal core" idea is just fantasy, and not based in even basic science.

Any change to the thermal core will cascade into total redesign and increased retooling costs. Do you even have any experience with product development? How much do you think it would cost to retool the manufacturing line to make a 7,1 "trash can" that was 10mm different in diameter?

The trashcan is a flawed design. But even when it's past 1000 days since last refresh I'm sure the fanboys will still be fanboys.

 

[Mago]

The funniest part of your post:
"...what they (Apple) need to do is to semi isolate sections of the thermal core"

I'm sure they can figure out how to do that by reading a few issues of popular mechanics...

Your point is an outright contradiction and lacks coherent logic. It's a COMMON thermal core. By isolating sections it will no longer be COMMON, and thus will be a major redesign (regardless of whatever shape it morphs to).

Sorry, even popular mechanics adheres to basic science. You're "isolated common thermal core" idea is just fantasy, and not based in even basic science.

Any change to the thermal core will cascade into total redesign and increased retooling costs. Do you even have any experience with product development? How much do you think it would cost to retool the manufacturing line to make a 7,1 "trash can" that was 10mm different in diameter?

The trashcan is a flawed design. But even when it's past 1000 days since last refresh I'm sure the fanboys will still be fanboys.

Engineering isn't something you learn in forums my sir.

The thermal core it's an piece of extruded aluminum, you can design the profile you want and order to China (or wherever) and get in your hands in weeks.

When I wrote SEMI isolate I clear never write to take apart the thermal core into differ segregated heatsinks, if you knew a bit of mechanical engineering (thermodynamics) you should be aware on how you can thermodynamically SEMI isolate an area from a heatsink to others, I shouldn't do your homework but it's OK to educate the thread, to semi isolate some areas of a heatsink all what you need is to shrink the walls interconnecting this area to other areas from such heatsink, so all what you need is to put a few coefficient at Mathematica then use the results to rebuild your previously parametrized cad design for such heatsink, and email to the extruder company, when you receive the new heatsink since it's the same vertical profile (only it's internal horizontal profile need to be changed) all what you need is to replace the older in the production line no re-tooling and neither the employee need to know this is an revised HS,since externally is the same, no metrics change.

If you where a little aware of thermodynamics you'll knew that the heat conducted by an material is conducted quick as thicker is the area interconnecting the high and low temperature sections (more less similar to electricity as thicker the cable less resistance).

If you don't know a bit on thermodynamics keep your mouth away keep reading popular mechanics, if you want to talk on thermodynamics I suggest you at least read an good book about (besides being very proficient in calculus since it's mandatory to understand thermodynamics) .

 

[pat500000]

hmm some of you may think all blizzard games are great, like Wow starcraft diablo and all are for Osx too
but yes games are working better with windows

Some blizzard games are great.

 

[JesperA]

Any change to the thermal core will cascade into total redesign and increased retooling costs. Do you even have any experience with product development? How much do you think it would cost to retool the manufacturing line to make a 7,1 "trash can" that was 10mm different in diameter?

Yes, Apple (and or rather Foxconn) can retool their production line every time Apple want to launch a new product, they can retool the production line when they want to to make the iPhone 0.2mm thinner, they can retool the production line when they want to make the iPad 0.2mm thinner, they can retool the production line when they want to add 3 more speakers to the iPad, etc etc BUT for you it is such a hazel for Apple to retool the Mac Pro product line, they just created an entire new product line when changing from cMP to nMP but holy moly would it be such a pain in the ass to update the production line like they do several times each year for their other products. So they can retool their entire product line but in your world the cost is an issue for retooling Mac Pro product line.

Also, you do realize that YOU made a fictional change (10mm larger diameter) to the Mac Pro and then you criticized the result of your own imagination? Apple could improve the cooling of the thermal core without changing its external size at all which would not prompt a complete redesign (and retooling of the entire product line) of the Mac Pro that you seem so incredible worried about (im not)

 

[fuchsdh]

Errr, that's fallacious. Tons of things are made to be sold, but how many are actually sold....

Considering Sonnet's still selling it after two years or so, presumably it justifies its continued existence.

 

[developer13245]

Engineering isn't something you learn in forums my sir.

The thermal core it's an piece of extruded aluminum, you can design the profile you want and order to China (or wherever) and get in your hands in weeks.

When I wrote SEMI isolate I clear never write to take apart the thermal core into differ segregated heatsinks, if you knew a bit of mechanical engineering (thermodynamics) you should be aware on how you can thermodynamically SEMI isolate an area from a heatsink to others, I shouldn't do your homework but it's OK to educate the thread, to semi isolate some areas of a heatsink all what you need is to shrink the walls interconnecting this area to other areas from such heatsink, so all what you need is to put a few coefficient at Mathematica then use the results to rebuild your previously parametrized cad design for such heatsink, and email to the extruder company, when you receive the new heatsink since it's the same vertical profile (only it's internal horizontal profile need to be changed) all what you need is to replace the older in the production line no re-tooling and neither the employee need to know this is an revised HS,since externally is the same, no metrics change.

If you where a little aware of thermodynamics you'll knew that the heat conducted by an material is conducted quick as thicker is the area interconnecting the high and low temperature sections (more less similar to electricity as thicker the cable less resistance).

If you don't know a bit on thermodynamics keep your mouth away keep reading popular mechanics, if you want to talk on thermodynamics I suggest you at least read an good book about (besides being very proficient in calculus since it's mandatory to understand thermodynamics) .

You continue to imply a lack of knowledge on my part, but yet you fail to answer a simple question: Do you have any experience in actual tech product development? Have you ever actually worked on a real world product that uses electricity and is sold in stores to the public?

Yes, Apple (and or rather Foxconn) can retool their production line every time Apple want to launch a new product, they can retool the production line when they want to to make the iPhone 0.2mm thinner, they can retool the production line when they want to make the iPad 0.2mm thinner, they can retool the production line when they want to add 3 more speakers to the iPad, etc etc BUT for you it is such a hazel for Apple to retool the Mac Pro product line, they just created an entire new product line when changing from cMP to nMP but holy moly would it be such a pain in the ass to update the production line like they do several times each year for their other products. So they can retool their entire product line but in your world the cost is an issue for retooling Mac Pro product line.

Also, you do realize that YOU made a fictional change (10mm larger diameter) to the Mac Pro and then you criticized the result of your own imagination? Apple could improve the cooling of the thermal core without changing its external size at all which would not prompt a complete redesign (and retooling of the entire product line) of the Mac Pro that you seem so incredible worried about (im not)

Umm... 836 days (and counting) does indeed equal "a hazel".

So in summary: Really? If it's so easy to just throw some new numbers into a CAD app and email the output to a supplier (wherever = USA BTW, not China) to get a spec bump refresh or new tech (GPU, etc) in the same form factor, why hasn't it happened in 836 days (and counting)?

 

[Mago]

You continue to imply a lack of knowledge on my part, but yet you fail to answer a simple question: Do you have any experience in actual tech product development? Have you ever actually worked on a real world product that uses electricity and is sold in stores to the public?



Umm... 836 days (and counting) does indeed equal "a hazel".

So in summary: Really? If it's so easy to just throw some new numbers into a CAD app and email the output to a supplier (wherever = USA BTW, not China) to get a spec bump refresh or new tech (GPU, etc) in the same form factor, why hasn't it happened in 836 days (and counting)?

I could write here "I'm a former space shuttle lead engineer" it's meaning less if I write bull &hit about whatever I want.

Since you are here in evidence, you only option it's to try disqualify

P.D. I don't imply what you loudly testifies.

P.D. 2 if you check my post history you'll find I've plenty experience in hardware development and what kind of hardware I used to design in my previous occupation before becoming a independent entrepreneur.

 

[Zarniwoop]

On Anandtechs Broadwell-EP review there's interesting mention:
"So far we can conclude that if you were to upgrade from a Xeon E5-2xxx v1 to a similar v4 model, your single threaded integer code will not run faster without recompiling and optimizing."

And same continued with multicore tests.. they had to be optimized for the processor. Summa summarum.. during past four years, perf/dollar has not changed. All software has to be recompiled for E5 v4 in order to benefit from it at all. So there... if you look to update your system but going to run old software.. no advantage.

All the noise for Apple not upgrading their HW... when it comes to CPU.. its been pretty pointless. And unless Apple recompiles their OS for E5 v4, and app makers too, there wont be much benefit to update the CPU.

More cores for higher price. That's Broadwell-EP. CPU market really needs competition.

 

[zephonic]

That isn't surprising at all and has been mentioned before. With all the clamoring for updates, there wasn't any substantial gain from v2 to v3, but that seemed to escape the notice of those complaining the loudest. So v4 seems again like a marginal update.

 

[JesperA]

Umm... 836 days (and counting) does indeed equal "a hazel".

So in summary: Really? If it's so easy to just throw some new numbers into a CAD app and email the output to a supplier (wherever = USA BTW, not China) to get a spec bump refresh or new tech (GPU, etc) in the same form factor, why hasn't it happened in 836 days (and counting)?

Did you see and understand what we where talking about? We where talking about retooling the product line, Apple:s decision to not update the Mac Pro in "836" days or whatever it is, has nothing to do with Apple:s willingness or unwillingness to update their production line or retooling of it. The decision to not update the Mac Pro have other reasons than that. Hint: look at "Zarniwoop"s post to see one possible reason.

Also, Apple would not need to change anything to do a "spec bump", the newer Xeon:s have the same TDP, the upcoming GPU:s from AMD will also have the same (or lower) TDP so why are you and that other guy so extremely insistent on the need for Apple to change the Mac Pro to do a simple spec bump? The other guys 10mm bigger diameter of the Mac Pro is just fictional, based on no facts whatsoever.

The only reason Apple would have to redesign the thermal core (but still keep its current dimension) is if they want to throw in other components that have a higher TDP than what is currently on offer, but yeah, the only update to the Mac Pro would be a higher wattage PSU and improved cooling in the thermal core but still, in that case, the dimensions and general design of the nMP would not need to change.

 

[wallysb01]

Considering Sonnet's still selling it after two years or so, presumably it justifies its continued existence.

Or its the same inventory that's been sitting on shelves for 2 years ... ... ...
[doublepost=1459730936][/doublepost]

On Anandtechs Broadwell-EP review there's interesting mention:
"So far we can conclude that if you were to upgrade from a Xeon E5-2xxx v1 to a similar v4 model, your single threaded integer code will not run faster without recompiling and optimizing."

And same continued with multicore tests.. they had to be optimized for the processor. Summa summarum.. during past four years, perf/dollar has not changed. All software has to be recompiled for E5 v4 in order to benefit from it at all. So there... if you look to update your system but going to run old software.. no advantage.

All the noise for Apple not upgrading their HW... when it comes to CPU.. its been pretty pointless. And unless Apple recompiles their OS for E5 v4, and app makers too, there wont be much benefit to update the CPU.

More cores for higher price. That's Broadwell-EP. CPU market really needs competition.

More competition isn't really going to help if what's being done is actually just really freaken hard and even achieving these small improvements is tremendously costly.

And from that Anandtech review, we can see going from the 2012, ~$2500 processor (2697 v2) to the 2016 ~$2500 processor (2695 v4), you get about a 50% increase in performance once you recompiled to take advantage of the instruction set.

This shouldn't just be brushed under the rug as if recompiling is some insurmountable task or that the older equipment would benefit just as much for a software update. As time moves on more and more apps will make this switch. It wasn't long ago that the same things were being said about adding multiple cores to a processor and it was up to the software to utilize this change in the technology to see the performance gains. For the most part people have accepted multicore processing now (when possible). People will make this change too.

Oh and that "More core for higher price" is BS. The 2630 v2 is a 6 core and listed at $616. The 2630 v4 is a 10 core and listed at $667. The 2697 v2 (mentioned above) is a 12 core, the 2695 v4 is an 18 core, both around that $2600 mark. We're getting 4-6 more cores for the same price compared to v2.