Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

dGPU is dead with Apple Silicon, meaning that eGPUs are also dead?

Unregistered 4U said:
When you add the fact that Windows PC’s can be faster/cheaper across the board with the fact that folks don’t really care about faster (if they did, they would have went Windows a long time ago), you’re down to the point where folks in the market for a new Mac primarily just want to know how much faster it is than the current thing they’re using.
Click to expand...

Unless you're a niche user, for most of the things most people do all computers on the market today are fast enough pretty much.

The difference is how fast the user can tell the computer what to do. And for that, macOS is IMHO superior.

macOS gets out of my way and gives me tools to automate stuff easily.
 
  • Like
Reactions: Unregistered 4U
Michael Adams said:
and the ability for AMD & Nvidia to offer their wares still possible.
Click to expand...
Technically, Apple has described a processor setup where the CPU and GPU have access to the same memory without going over a slower external bus. Other things may be possible over PCIe, but, as of now, it seems like Apple’s expecting Silicon code to use Metal and their GPU rendering pipeline.

I was just looking into what Imagination Technologies has been doing for the past few years with Tile Based Delayed Rendering, and, while it looks like they don’t currently have any products slated to use their higher end tech (they license to other companies), if Apple’s anywhere in that general range, it could be that only high end gamers are going to be left out (and they’re pretty much left out already just ‘cause macOS).
 
GiantKiwi said:
Nope.

Macs are a niche market, eGPU users are a niche market within that niche market. GPU manufacturers have no reason to waste time providing driver support for such a minute market share.
Click to expand...
MacOS video drivers were Apple anyways. Hence why no Nvidia web drivers for newer OS's, Apple wouldn't sign any to work.
 
Maybe it will come back but I guess not with 3rd party GPUs. I assume Apple at some point (maybe with the AS Mac Pro) will make their own dedicated GPUs which might be usable with any Apple Silicon Mac as an eGPU. I was really sceptical about that at first, but seeing how fast the IGPU of the M1 is I am looking forward to see Apple's future offerings.
 
Erm... your question doesn't make any sense. This is an IGPU, why would you compare it to a dedicated GPU that can draw 50x more power?
Thing is that it is the fastest IGPU as of now appr. reaching performance of a GTX 1060 while fitting in a 15W TDP envelope. And if they are able to pull this off for an IGPU, they will be able to make high performance dGPUs for sure.
 
You are assuming that they will make GPUs similar to the current ~300 watt offerings. The M1 GPU can do a lot of the things that usually requires a good GPU, making the market even smaller. We are left with gaming, which Apple pretends doesnt exist, and niche GPU work, which leaves me to believe that really expensive afterburner-like cards will be the only thing offered.
 
toke lahti said:
Anybody think eGPU support will have "second coming"?
Click to expand...
If we ever get AMD dGPU support over PCIe in ARM Macs (and native drivers) we will probably have eGPU support, either officially or hopefully through hacks such as the purge-wrangler script (which still works on Big Sur).
Would Apple be able to replace AMD dGPUs in all its machines with Apple GPUs, integrated or not, and never adopt AMD (or nVidia) dGPUs in ARM Macs? Possibly but that would be a much bigger challenge than what they did with the M1 in my opinion.
 
CMMChris said:
Erm... your question doesn't make any sense. This is an IGPU, why would you compare it to a dedicated GPU that can draw 50x more power?
Thing is that it is the fastest IGPU as of now appr. reaching performance of a GTX 1060 while fitting in a 15W TDP envelope. And if they are able to pull this off for an IGPU, they will be able to make high performance dGPUs for sure.
Click to expand...
Erm... you don't seem to understand the problem. People need gpu power to do things that needs power. If you edit video or do 3d or cad or just play games, you need power. If Apple's iGPU is in all of its greatness 3x more efficient than anything else on the planet, that doesn't help people that are used to use 300W-gpus. Yes they can use Apple's 15W-solution to get power of 45W-traditional-solution, but that's it.
And it's a whole lot less than what they have or will buy next.
Apple is now destroying powerful gpu's from its ecosystem.
Maybe the answer is the classic one "Just buy a mac pro, over here"?
 
Last edited:
  • Like
Reactions: ETN3
CMMChris said:
I assume Apple at some point (maybe with the AS Mac Pro) will make their own dedicated GPUs which might be usable with any Apple Silicon Mac as an eGPU.
Click to expand...
Apple’s has indicated they’re not going to be doing anything that looks like today’s dGPU’s on a PCI bus. Their graphic solutions will all be internal but perform the way we expect externals to perform.
 
T'hain Esh Kelch said:
Any guess as to wether the M1 GPU uses regular RAM as shared VRAM, and thus one should always upgrade to 16GB RAM?
Click to expand...
Yes, it’s shared memory, so I would upgrade. Of course I am not privy to all of the details of how the magical pixies work in AS, but I would upgrade anyway considering that it is shared across the GPU, CPU and neural engine
 
  • Like
Reactions: T'hain Esh Kelch
T'hain Esh Kelch said:
I am sorry, but I do not know how that explains my question? :)
Click to expand...
Ah, sorry.

actually, not always, shared memory is better because the Intel GPU always allocates 1.5GB to 2GB (depending on the GPU model) of your RAM even when it’s out of use.

This means more free RAM in general use. So, 8GB is better on the M1 than on the Intel line.

For future proofing I’d totally going with the 16gb option. It’s also easier to sell it in the future.

Hope this answers the question.
 
  • Like
Reactions: T'hain Esh Kelch
PortoMavericks said:
Ah, sorry.

actually, not always, shared memory is better because the Intel GPU always allocates 1.5GB to 2GB (depending on the GPU model) of your RAM even when it’s out of use.

This means more free RAM in general use. So, 8GB is better on the M1 than on the Intel line.

For future proofing I’d totally going with the 16gb option. It’s also easier to sell it in the future.

Hope this answers the question.
Click to expand...
Exactly what I wanted to hear, thank you!
 
  • Like
Reactions: PortoMavericks
Depends on the use.
If somehow your gpu needs lots of memory, it used to have it.
If the iGPU needs 8GB of memory, like the dGPU used to have, then with shared memory, you're left with...
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back