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.
No more nvidia chips in Apple's computer line
- Thread starter netkas
- Start date
- Sort by reaction score
Im hoping the new nMP has a fury.. Would love a fury x for its openCL strength in my cMP
And with octane 3 have openCL as compute would be interesting to see them stack against the 980ti's in cuda
seriously man, you're running out of ideas here.. i'll keep going just to see what you'll say next but..
the 2014 mini uses non-standard torx security screws.. the teardown_ers had to redneck-prototype a driver to work on those screws because they aren't standard.. in fact, they said they've never seen them before.
their complaint wasn't because 'SECURITY TORX'.. it was because of the size of the 'SECURITY TORX'
are you sure ifixit wasn't referring to the following in their bad macmini reviews:
• ram is soldered to the logic board
• cpu is soldered to the logic board
?
also lets not forget something else.. when i first mentioned the screws a few years ago, it wasn't "hey checkitout.. torx screws yay!"
more like -- "hey, check it out.. there are screws" ...ie- "hey- nothing is soldered or welded or glued"
you're the one that brought TORX into the mix.. funny thing too, go watch those old threads and you'll see yourself saying the same crap then as you are right now except it was about the normal torx screws.
you- nov10_2013 said:
Last edited:
say whatever lame thing you think you can get away with saying around here but so what.. it's only reflecting onto yourself.
plugging in a thunderbolt cable is far more user friendly than installing a pci card on cmp.
i don't even know how you would begin to argue otherwise and conclude 'plugging in the usb-c was harder than putting in the card'.. but please, let's hear it.
you're annoyed with me.. that fine. it's likely you're just annoyed with my personality and you're letting that skew your judgement about everything i say.. (such as now thinking cmppci is easier than thunderbolt).
why don't you just say what your real beef is instead of calling me an idiot over something that's so clearly obvious to be considered factual?
plugging in a thunderbolt cable is far more user friendly than installing a pci card on cmp.
i don't even know how you would begin to argue otherwise and conclude 'plugging in the usb-c was harder than putting in the card'.. but please, let's hear it.
you're annoyed with me.. that fine. it's likely you're just annoyed with my personality and you're letting that skew your judgement about everything i say.. (such as now thinking cmppci is easier than thunderbolt).
why don't you just say what your real beef is instead of calling me an idiot over something that's so clearly obvious to be considered factual?
So... a regular Torx would've been more user friendly, then?
no.. they're the same thing.. they are used in an identical manner and if you didn't see the head, you couldn't tell the difference between the two.
it's nearly impossible to make a mistake with a torx.. they don't slip out.
imagine driving a car that you simply could not crash.. that could be considered the most user-friendly car.
these are user friendly screws.
do you make the connection?
....
seriously.. go find a torx screw and bit tomorrow.. do some screwing.. get slotted and phillips too..
you'll see what i mean immediately.
like, you really will see what i mean and be like "nice, those things are sweet"
or just ask @MacVidCards (since he unscrewed the torx for the pictures above).. ask him
"hey mac, how many screw heads have you stripped when installing/swapping hard drives on cmp?"
then ask
"how many torx heads have you stripped?"
Last edited:
If they wanted it to be user replaceable there is no need for security torx. Normal torx would do. Simple as that.
But I don't even know why we're having this conversation when we all agreed already that the GPUs weren't user replaceable because they're bonded to the thermal core with thermal paste.
But I don't even know why we're having this conversation when we all agreed already that the GPUs weren't user replaceable because they're bonded to the thermal core with thermal paste.
Will discuss in point form for you i think it will be easier for you.
- This is my last reply to you, Ever.
- The thread is about GPU's - You seem to take EVERY thread off topic.
- Saying thunderbolt is easier well for a HDD or SSD yes, a raid system sure although pricier.
- How bout UPGRADING to a Titan X? Fury? - Possible through thunderbolt but the opposite of user friendly and unstable.
- You are a frustrating person you have no compromise in your arguments, You enjoy the attention that your getting from saying silly things so yes i don't like your personality at all.
- Your not stating facts, They are opinions, and yours at that.
- " the nmp is the much better example for easy access to pci." that comment from you will always make you look a little silly
- Goodbye
So I see we are all in agreement on this point: Apple doesn't want most people poking around in the nMP.
Contrast with the cMP which needs only your bare hands to work with the PCI slots.
Contrast with the cMP which needs only your bare hands to work with the PCI slots.
I'll tell you a funny story. I've been working on eGPU on the nMP. I decided to remove one of the GPUs to see what happens with just one connected. Imagine my surprise to discover those security screws. Fortunately, I had a set of Torx Security screw bits. I had purchased them for upgrading a couple 2014 Minis.
So, just a couple sizes larger. Same screws, same drivers. There for the same reason, to keep the owner of the machine OUT of the machine.
Got it. I'll let him slather the place with BS for awhile.
Last edited:
please mac, tell me you agree with the following:
apple did not put an access latch on the nmp in order to keep the owner OUT of the machine.
if we can not agree on that point, which remember-- was the very original and main point i ever said about any of this prior to years of arguing secondary and thirdary and fourthary points of the initial statement..
if we can't agree that apple has not intended to keep owners OUT of the machine based on the presence of an access latch, then there's no other subsequent points to be made.
i already live on an island.. i'll stay put and mvc can come here so we can kiss.. i'll youtube it for you.
Did it boot?? I always wondered if you could mix a D300 and D700 for example.
This thread is definetly going somewhere... Not sure where anymore...
It's going from California to Manhattan so two gents can box then kiss and be put on youtube via an iMac
From the hardware point of view:
OpenCL kernels are by default parallel operations on vector and/or matrix data (video frames are matrix data). Just because it's just one kernel, it doesn't mean it has to do serial operations. Most video processing can be done independently for each pixel. If you had 2073600 pipes in your hardware, you could do the processing of one full hd frame in parallel at one swoop.
TLDR; yes, the OpenCL drivers profit massively from the hardware parallelism of the AMD chips.
From the software point of view:
And the biggest complaint about OpenCL is the bad documentation. But that's not the fault of AMD or their chipsets. It's mainly Apple's fault. Which means, it hits Intel, Nvidia and AMD alike.
Last edited:
Front end - low overhead and exposure of the GPU for the application - that is for the AMD hardware.
The problem is that pipeline is streamlined in-order execution. Looks like designed for Intel and Nvidia GPUs.
Your statement is correct. However, there is almost no upgrade parts coming with thunderbolt connector. Most of the TB parts are the EXTRA parts.
HDD? SATA
SSD? SATA, PCIe, M.2...
GPU? PCIe
Audio / Video cards? PCIe
You can't just plug in the cable then finish the upgrade, that "plug in the cable" is an extra step on most of the upgrade (if not every single one of them), you need to have an extra PCIe box, may be an extra PSU... And you can't skip the "installing PCIe card" steps.
There is no way to compare a single step to the whole upgrade. You must compare the whole processes on both side.
And TBO, I doubt if there is any box can offer a PCIe solution as good as the cMP does. Zero tool require, well secure inside the machine, full PCIe2.0x16 bandwidth...
Right, because massive parallelism is a feature that only AMD provides in their GPUs? My point is that CL kernels for image processing usually take the form of one kernel that is run on the entire image, not several independent kernels that need to be run at the same time, which is what AMD's asynchronous compute allows. If you are using one CL kernel, then you throw it at the entire GPU and let it crunch the numbers for a while, and both AMD and NVIDIA do very well with a workload like that.
I give AMD two years as an independent company.
http://www.anandtech.com/show/9722/...computing-and-graphics-sales-hurt-bottom-line
Holy crap. You are just trolling at this point, you cannot possibly believe this. Maybe you can tell us if we are at war with Eurasia or Oceania today.
The front end is designed the way it is to eliminate overhead from the driver/ABI stack on the CPU. It's not for anyone's architecture.
I'd guess in order execution has nothing to do with Nvidia. Metal for OS X is nearly a straight port of what they're running on mobile, which means it's still designed to target the A series GPUs that Apple has on iOS devices. They didn't change anything significant when coming to OS X. They didn't design anything at all specifically for discrete cards besides some rearranging of resource management on the front end.