The R270 has one big advantage many oversee for the nMP: It has only 130W max in the default config so throttled it would fit nice into the 450W of the box.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
New Mac Pro GPU Overkill
- Thread starter kirkbross
- Start date
- Sort by reaction score
What part of "they changed the ***** cards to a proprietary design" aren't you getting? You keep referencing how Apple has done business with previous Mac Pros but ignoring that this is an entirely new design. You simply cannot assume everything will remain the same. Because it's not the same.
And no, they don't look easy to replace. You're still trotting out that opinion? Oh, right. There are screws there for access. I'm done. It's the same talking points coming from you over and over. We completely disagree. You are overly optimistic about the GPU upgrade path, whereas I feel I'm closer to reality. We'll find out soon enough, but I'm confident that you're flat out wrong.
most of the stuff i'm saying are attempts at supporting arguments but they're not even the real point - the design point- that actually has me interested to begin with..
i don't necessarily have the energy to keep going through it at every angle.. especially when considering the mp release is imminent..
the bottom line point i'm trying to make is because of the latch then the screws.
if those screws aren't meant to be tinkered with by the owner then apple has incorporated a huge design flaw.. and it's such an obvious mistake that i can't imagine in a billion years how at least one person on the team couldn't miss it.. i mean, i see it and these guys are better designers than me..
in story form:
you get your shiny new mac pro with a 256g drive because you still need to buy the rest of the family their christmas gifts.. around april, you spring for the 1TB upgrade which apple is so kindly allowing you to buy and change yourself.. off to homedepot you go to get the required t15 (or whatever) screwdriver.. since this is your official nMP servicing tool, you spring for the fancy ten dollar version with magnetic head and maybe some titanium in the mix.. so you drive home all proud and hyped on your new computer tool..
get to your desk and unplug the power source and peripherals (since you have to), push the button(?) and remove the shell.. get out your trusty nmp servicing tool and unscrew that one screw.. just ignore all the other screws which are an inch away and would perfectly accept your nmp servicing tool.. if you touch any other screw except the one in the middle then you're going to void your warranty..
[/story]
i mean, even using phillips head for that one screw is way way better than this set up (though i'd still be claiming they made a mistake by having ANY accessible screw which isn't meant to be user serviceable).. i can think of a thousand different ways which would make more sense than this wacky setup (one of which is for you to still use your nmp servicing tool for that one screw except the gpus are stapled to the core with an office stapler)
those screws are meant for you, the owner, to stick a screwdriver in..
This is one of the most facetious arguments out there against the nMP. People keep forgetting this is a workstation class computer; sure, some prosumers and regular consumers with more money than sense will might buy these. But if you are actually getting one for work, more than likely it'll be replaced well less than five years time. In enterprise/professional fields, high-end machines get obseleted and replaced far more quickly than for regular consumer-class machines. Why else do you think there's a gold mine of enterprise grade components for sale so cheaply all the time?
In my particular line of work (bioinformatics research), I can tell you that when it comes to servers and workstations being used for heavy crunching, they won't last three years. If you're a consumer buying a computer you don't need, then sure, you might be bummed the GPUs are sub-par in a few years. If you're an actual professional user, you won't care because you'll already have a new machine for your work by then.
Last edited:
Well, for one thing, we have no insight into how the GPU silicon on the other side of that daughterboard mates with the internal, centralized heat sink. It's possible that all that you'll need to do is clean off the old stuff, apply a dab of good heat sink compound, and you're good to go. That would be in the realm of most end users (but still significantly more difficult than swapping a PCIe card in the old Mac Pro). Your theory relies on this being the case.
However, it's possible that Apple have gone with a more esoteric or delicate heat transfer mechanism in order to achieve their aggressive cooling goals with the new design. We just don't know how tricky that process will be, or how precise the installation needs to be in order to maintain adequate cooling. The tolerances for heat sink compound may be beyond what can reasonably be expected an end user to achieve. Too much can be just as big a problem as not enough. What you don't want to see is people who face perpetual cooling issues or fried silicon because they botched the replacement. Will that be the case? I don't know -- and neither do you. Ergo,it's hard to predict.
Until these devices are available for teardown, it is very hard to predict what the process will involve.
Because this is Apple and not a piece meal HW company. You get what they think you need and offer you. I would want a single super high end Nvidia option. Next iteration can have all the AMD love once SW (That's you Adobe) actually codes appropriately for it. It appears to be bad timing but I bet Apple has a closer ear to the ground than I.
LOL
Apple sells roughly 400,000 iPhones per day, and almost everyone buys at least one extra cable or charger.
It is estimated that Apple sells around 300,000 Mac Pros per year, which works out to about 822 per day. If we assume what, maybe 10% of them will wish to upgrade the proprietary GPUs within a standard 3 year lifetime, you'd eventually level off at 30,000 potential customers per year.
So the customer base for lighting connectors is literally almost 5000 times larger than its likely to be for a proprietary PCI-E GPU connector. And that gap is only going to widen.
So yeah, not an apt comparison. Not even a little.
The only reason the existing Mac Pro's official GPU upgrade options were merely sad instead of nonexistent is because the R&D in converting a PC board was minimal.
consider they've done the same thing with the ssd and how much more pricey one of those are compared to a lightning cable.. then things start to come more in tune with the lightning connector comparison..
apple is going to make a killing off these proprietary parts.. the way they have it set up is that they're in the position to make similar money regardless of who makes what.. if apple wants to sell a new accessory, they have to design it, manufacture it, ship it, advertise it, etcetc.. once all that is paid for they're left with X amount of profit.. licensing the connector puts all/most of the r&d + logistics on the other company but apple still collects their X amount of dollars on the part.. in many ways, it's better/easier for apple to sub out the connectors to other corporations-- ultimately, they own the most valuable piece of the whole equation and that being the trademark plus the ability to control the 'competitors' prices..
the ONLY reason so many 3rd party companies play the apple game is because they stand to make money on the parts.. in the past, they were able to make a lot more money because they didn't have to share profits with apple.. apple didn't like that so they've put a squeeze on it but not so tight in order to push other manufacturers out of the game.
i get where they're coming from.. i mean, plugs etc for iphones would have absolute zero chance to be profitable if it werent for apple making it in the first place..
just take a guess at how much profit you think nvidia or seagate or owc etc have made buy supplying upgrade parts for macs.. billion$? yes.
apple wants some of those billions for themselves.. that's a huge reason for them to own the technology which is being used in their computers..
they're not buying the technology simply to make it unavailable to end users in order to force them to buy more computers.. i don't see how it's possible for them to make more billion$ going that route instead of allowing parts/replacement/upgrades to be sold individually.
----------
another thing to note is that apple will assist their 'competitors' in r&d.. it's not like 3rd party manufacturers have to start with a blank slate when making parts..
companies like owc and belkin and amd (personal computing division) etc... they're not competitors to apple.. probably makes more sense to view them as partners to apple.. except apple is still the boss in the relationship
Last edited:
my personal guess is that there's no thermal paste involved at all..
in the apple renders, we can see the copper(?) square where the processor touches but what's curious is the secondary plate in between the gpu and the thermal core.. i'm going out on a limb and assuming that plate is made of a material which absorbs/disperses heat more rapidly than past heat sink material.. this plate is possibly doing a more effective job than applying paste directly in between the main core and the processor.. (?)
another guess is that the mounting bracket is designed to allow a typical person with a screwdriver to achieve proper tension/force on a consistent basis.. a similar idea can be seen on the cpu bracket.
in the case of the gpu bracket, i'm willing to bet it's spring loaded (for lack of better words) in that the central ring will sit lower than the outer four flanges if sitting by itself on a flat surface.. like so:
(for clarity, exaggerated side view..)
when tightening, there will be a bit of play in there with where exactly the final turn must stop..
really, the only practical reason i can see for using this type of bracket is to assist the end user with non-specialized tool to achieve proper fit and pressure.. if it were meant to be serviced only by specialists, i'd imagine it would be easier to achieve optimum mounting via computer controlled torque drivers.. or solder or glue etc.
Last edited:
That's quite a limb! I don't think you can assert that Apple have discovered or invented a new kind of metal that nobody has ever seen before and which will make heatsink compound unnecessary while still claiming:
Because to me, what you just said sounds like little more than wishful nonsense as you contort further and further to try to justify your beliefs.
well it is different than anything we've (i've) seen so far in computer cooling.
I highly doubt apple invented it themselves in their own labs or whatever. possibly some sort of ceramics in the mix? maybe nasa invented it 30 years ago.. I obviously don't know and I don't claim to know.
what I do know is that expecting users to apply gooey pastes to the innards of their electronic devices is pretty dumb.
ahh.. the old mix and match separate contexts game
Yes, which is one of the reasons that many think that the GPUs in the nMP will not be user-swappable. That's certainly the least contorted conclusion.
Which guess would Occam endorse? That Apple will be using some secret, NASA technology that no other computer manufacturer is aware of? Or that Apple don't intend for end users to swap the CPU or GPU in the new Mac Pro?
(For the record, ceramics are used in industry because they are exceedingly poor conductors of heat, which is why the space shuttle heat shields are made of ceramic. That's the opposite of what's needed in this application.)
Last edited:
hey wait.. as far as i can tell, i'm the one that brought up the argument earlier this morning for the first time on this forum.. to me, it's one of the better arguments against nmp upgradability, but again it was me arguing against myself..
when phrased that way, occam leans towards no cpu/gpu upgrades via users.
when considering apple is claiming -quote- 'ingenious' solutions have been found for heat transfer/dissipation as well as knowing damn well they can afford to hire nasa engineers and/or other true experts in the field.. well, occam starts getting a bit foggy.
ha. right
Then there would be a pad. This was talked about before.
No. Metel to metel contacts arent that good (relative to thermal paste) at conducting heat because there are inevitably pockets of air trapped between them, creating an insulator effect to some degree.
Youd need a to have torque screwdriver and even then putting CPUs into a socket with 4 screws would require even tightening. Not saying its impossible, or even necessarily hard for apt tinkerer, but its certainly less easy that the pressure levers on any other LGA mobo.
Normal LGA sockets require NO tools at all. And to ensure no over tightening there is just no way to avoid a torque screwdriver unless basically finger tight is good enough. But Id still be surprised if screws where used at all. You just dont see screws going into computer boards like that and if there are screws, that makes it harder, not easier.
As far as the CPU and GPUs go in the new Mac Pro, it's possible that Apple is using lidless (naked) chips without integrated heat spreaders to improve heat transfer between the silicon and HSF. This is often how GPUs are implemented and Apple went this route with the dual CPUs in the 4,1/5,1 MP (which is very unconventional for CPUs). If that's the case, it's an added challenge for upgrading, possibly making it impossible due to the lack of lidless alternatives and clearances involved using regular CPUs. On top of this, it can be risky... as there's the possibility of chipping a die if you don't carefully seat the HSF.
EDIT: Based on this render, it looks like the CPU has a heat spreader but the GPUs do not... And it also looks like the CPU is not using a standard LGA lever retention mechanism, which means replacing the CPU could be challenging without bending a pin.
EDIT: Based on this render, it looks like the CPU has a heat spreader but the GPUs do not... And it also looks like the CPU is not using a standard LGA lever retention mechanism, which means replacing the CPU could be challenging without bending a pin.
Last edited:
right.. with the cpu
https://forums.macrumors.com/posts/18413798/
dunno, it's beyond my knowledge.. if i were forced to guess, i'd imagine the heat is moving fast enough through the points of contact to make up for any residual air pockets.. or maybe the mystery material becomes elastic (at microscopic level) enough to fill voids as heat intensifies.. (which is actually another case for the spring loaded bracket idea)
sort of went through this before and was unable to go further because of lack of quality images of the cpu..
the part of the cpu bracket which has this tension control feature are the outer four screws.. the ones you screw in when connecting to the heat sink.
at that point, the cpu is already secure inside the socket via the other 4 screws (though i think those 4 screws are secondary locks.. i don't think you'll be sitting there trying to hold a cpu inside its socket while being required to drive 4 screws.. there's something else to it but the apple renders don't show it.. and there are no photographs of it.
you sort of do though.. the mp1 cpu is attached to the heat sink via 5 hex heads..
Define legit music?
My album... me on vocals, bass, keyboards and guitars... Paul McCartney's drummer and Alicia Keys' drummer on drums... Bryan Ferry's guitarist on electric guitars. Echo & The Bunnymen keyboardist and cellist on cello and string arrangements. Jesus Christ on tambourine... executive produced by the current president of music at Warner Brothers Pictures.
I also love the art and science of sound engineering. And I build PCs for gaming.
Holy *****, am I amazing or what?!!
There will always be thermal paste until you can fabricate metal surfaces with micron level smoothness on an assembly line level.
Your alternative hypothesis doesn't make sense because even if this magic thermal dissipation material exists, you would still need thermal paste at the junctions of these materials and the processors (unless of course, you had the micron level manufacturing ability, at which point, this entire academic exercise is moot).
----------
I'm not trying to be a jerk here, but your understanding of basic physics is pretty poor. The form factor may be unique and recently patented but the cooling technology itself is neither novel nor particularly interesting. These suggestions excite the imagination but aren't even in the ballpark of plausible.
As for expecting users to apply thermal paste; you're right, Apple doesn't expect that because Apple doesn't expect end users, even the professional market segment, to do major assembly or disassembly ever. No OEM expects this.
There's a misconception that just because a computer is "pro-grade" or workstation-class, the end user is expected to tinker with it or have it last longer. Enthusiasts are not the same as professionals. Truthfully from personal experience, I contend that these users are generally *less* likely to do these types of major surgery on their computers because they will 99% of the time be work computers. Have a problem? Have the IT crew work on it. Or send it in for warranty. That's why they cost so much; you're paying for the components but really, you're paying for support.
Last edited:
No, you're wrong. The engineers at Apple have actually discovered this new magical element called Macpronium. They're probably going to win the next Nobel prize for chemistry.
Yes. Yes it is.
I stand corrected. This sounds like trufax to me. I should file a patent on this so I can sue Apple later. Anyone else want in on this venture?
The pictures you posted don't support that; even the best physical contact is going to be imperfect, this is pretty much why thermal paste is used to begin with, as any chance of imperfections means a reduction in heat exchange, the paste just helps to reduce that.
It wouldn't require much for the assembly line machine to just apply a quick blob of thermal paste and smooth it out, ready for someone to just screw the card into place. If Apple does plan to provide replacement cards then they could just as easily leave that up to end users, e.g - with a protective film. Just clean away any existing thermal compound from the previous card(s), remove the protective film and screw into place.
But it's all conjecture; my personal guess is that there is thermal paste involved, as it makes less sense to not use it. We just don't know, so there's not really any point theorising so much from pictures that don't show nearly enough on which to base a theory.
why are you trying to ridicule me about this?
do you see the piece of material in the pictures or not?
I'm saying "oh look, there's a new piece of metal in between the processor and heat sink"-- and I'm, at this point, interpreting your response as "no, there's not.. you're an idiot for saying there is." ..again, do you see the plate in the pictures or not?
(and I doubt it' called "macpronium".. more likely it's called "graphite")