If 8 GB unified is 2x better than regular RAM in performance, which should i get for occasional Final Cut and Logic work, 16 GB unified?
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.
Is 8 GB unified RAM equivalent or better than 16 GB regular RAM as used with Intel?
- Thread starter Luba
- Start date
- Sort by reaction score
16GB. RAM is RAM. If you needed 16GB before you’ll still need 16GB now. M1 isn’t a magical RAM doubler. I can’t believe this nonsense is still being spread around.
Since nVME SSD are faster, the need for RAM may become less, as data swap on 3GB/s NVME disks are not that miserable, comparing with 500MB/s SSD, or 140MB/s of HDD.
Intel machines had NVME as well.
Swap is many times slower than RAM.
There's no substitute for RAM if you need it.
NVME may be 3 GB/sec. RAM is around 10-12x that (around 35+ GB/sec for DDR4-3733 as found in my 2020 MBA) and has far, far better latency.
Would you think a CPU running at 100-380mhz would be reasonable in 2020 (when 1-3.8 ghz is typical for macbooks)? I think not...
NVME is better than SATA, but its still not a RAM replacement.
because UNIFIED!
Seriously though apple has been talking up the unified RAM thing a lot, but the intel machines had unified memory as well with the integrated graphics.
there's pros and cons to it - with so many different but related purpose-built processing units in the M1 it kinda makes sense (so each "engine" in the M1 doesn't need copies of data moving around), but the flip side is that everything is sharing the same memory and LPDDR4 isn't particularly fast.
M1 has big processor caches to help somewhat with the slow RAM but at the end of the day it is a cost compromise. "unified" does help somewhat with removing the need to copy data around everywhere to different RAM silos, but its still only 16 GB tops. Shared for everything. Run something that needs 8GB of video ram and guess what... that's coming out of your unified memory bucket - instead of say, a discrete GPU's dedicated VRAM, which would leave your processor ram mostly untouched by that.
It will be interesting to see if Apple maybe use HBM for the Mac Pro M1-derived packages.
I didn't mean performance as in Ghz CPU speed. Actually, I don't know what word to use. I just know that some types of memory are closer or more easily accessible by the CPU and therefore makes the computer run faster. "Unified" makes it sound like the RAM is part of the CPU ("sitting on top of it"?) and implies to me that "regular" RAM is a bit farther away, like "next door".
M1 Mac mini review - 2 weeks later
After using the 16GB M1 Mac mini for two weeks, I have lots of thoughts. Check out my M1 Mac mini review to see if you should buy one.
markellisreviews.com
This may be worth considering even if it does not address the OP's exact question.
What you looking for in video editing is capacity, not the memory speed itself, though.
For the work you are wanting to do 16GB is the minimum, but if you do a lot with 4K/8K video, you may want to wait for the "pro" level Mac in the future.
For photo editing work my 8GB M1 is more than enough to do the job and then some.
Can't compare the Unified Memory against the DRAM module Intel Macs - they are definitely not the same.
I ditched my Intel Macs with double or quadruple the RAM as they can't keep up.
8GB on an M1 runs much faster without app opening delays, spinning beach balls, and fan noise.
Oh no, not another RAM thread.
Just get 16GB. 8GB will never be enough in the long term as an iMac should last easily 5-7+ years. You will never be able to upgrade from 8GB and there's no way memory footprints for apps and OS's are going to shrink any time soon.
Just get 16GB. 8GB will never be enough in the long term as an iMac should last easily 5-7+ years. You will never be able to upgrade from 8GB and there's no way memory footprints for apps and OS's are going to shrink any time soon.
While the RAM is closer to the CPU (which may help with latency and speed due to shorter signal paths), the "unified" refers to that the CPU and GPU (and other processors) don't portion off parts of the available memory to themselves (eg., if you have 8 GB on an Intel machine, the GPU might take 1.5 GB and the CPU has 6.5 GB) but instead both the GPU and the CPU can operate on all data in the RAM. This is helpful in some cases, but doesn't really act like "more" RAM in general.
1. RAM isn't user upgradeable.
2. Most programs require more RAM not less, over time.
While the SoC has made it easier to access the RAM on board, it hasn't increased the RAM on board. As such, when the time comes that you need the extra RAM, efficiency will go out the window with the bath water. In the end, programmers are great at writing bloated code, and since RAM is plentiful, they can get away with it.
I can assure you, the non-entry level computers running ARM in the future will offer you far more RAM capacity... for a reason. These early models are technologically crippled on purpose.
It's up to you, skimp now and replace the thing in less than 2 years or bite the bullet and give yourself another year or two.
2. Most programs require more RAM not less, over time.
While the SoC has made it easier to access the RAM on board, it hasn't increased the RAM on board. As such, when the time comes that you need the extra RAM, efficiency will go out the window with the bath water. In the end, programmers are great at writing bloated code, and since RAM is plentiful, they can get away with it.
I can assure you, the non-entry level computers running ARM in the future will offer you far more RAM capacity... for a reason. These early models are technologically crippled on purpose.
It's up to you, skimp now and replace the thing in less than 2 years or bite the bullet and give yourself another year or two.
People do forget one thing - Intel GPUs need to have a copy of the memory in order for the GPU to work properly. M1's GPU has access to ANY and ALL of the 16GB of RAM. No copying necessary.
There are workarounds on this for Intel GPUs, but it is a programming change.
This doesn't mean 16 GB is the same as 8GB, but it does make some impact.
Also, releasing NSObject references is much faster on M1 - so that can help speed things up a bit too.
There are workarounds on this for Intel GPUs, but it is a programming change.
This doesn't mean 16 GB is the same as 8GB, but it does make some impact.
Also, releasing NSObject references is much faster on M1 - so that can help speed things up a bit too.