In light of the tests performed by the Macworld (link below,) regarding geekbench Cinebench etc, which M chip gives best bang for the buck, specifically for VDMX?
It looks like the M3 holds its own, but I don’t know nearly enough about VDMX hardware loads to be confident.
Inviting educated discussion. Thoughts?
I can’t read that article (requires a subscription), and I have an M3 Max, but even the M1 Max Macbook Pro was a beast and could handle anything in VDMX I threw at it. I would think the difference for most machines would be with more memory and internal SSD vs CPU/GPU performance.
Here’s the meat of the article.
SPECIFICATIONS
The M4 Pro 16-inch MacBook Pro in this review has the same CPU, GPU, and RAM as the $2,899/£2,899 standard configuration, but the SSD is a 2TB upgrade ($600/£600) and it has the Nano-texture glass ($150/£150). Here are the specifications of the laptop in this review:
CPU: M4 Pro with 14 cores (10 performance cores, 4 efficiency cores), 16-core Neural Engine
GPU: 20 cores
Memory: 48GB unified memory (273GBps memory bandwidth)
Storage: 2TB SSD
Display: 16.2-inch Liquid Retina XDR; 3456-by-2234 native resolution at 254 pixels per inch; 1000 nits sustained XDR brightness, 1600 nits peak (HDR content only); 1000 nits SDR brightness; 1 billion colors; P3 color gamut; True Tone; Nano-texture glass
Ports: 3 Thunderbolt 4/USB-C; MagSafe 3; SDXC Card slot; HDMI; 3.5mm audio
Networking: Wi-Fi 6E (802.11ax); Bluetooth 5.3
Input devices: Magic Keyboard with Touch ID; Magic Trackpad
Weight: 4.7 pounds (2.14 kg)
Dimensions: 0.66×14.01×9.77 inches (1.68×35.57×24.81 cm)
Price (as tested): $3,649/£3.649
PERFORMANCE
The standard configuration of the M4 Pro has two more CPU and GPU cores than the M3 Pro it replaces. Apple added more unified memory in the standard configurations, too, which is a result of Apple bumping the starting level of all Macs with the based M4 chip to 16GB. The M4 Pro now starts at 24GB of RAM, up from 18GB in the M3 Pro.
Does Geekbench or Cinebench or Handbrake track and represent strongly correlated performance within VDMX? or is there a better metric? To my eyes it appears that for equivalent GPU cores and CPU doohickies the M3 is better. But doohickey is an accurate depiction of the depth of my computational nuance when it comes to performative parameter analysis. So I would appreciate a more insightful eye.
You asked what is the best “bang for the buck”, so what is the price difference for the models you are comparing? I think it is going to be difficult to exhaust the capabilities of either the M3 or the M4 with VDMX usage. Are you doing anything truly monsterous? Encoding multiple 4k streams? Running heavy TouchDesigner patches?
The M3 is probably fine. The M4 will possibly last a little longer into the future. I think we’re currently in a bit of a plateau for computer hardware, so you may wish to buy an M3, and factor in a quicker upgrade cycle. The reason I say this is that I think we will be seeing some architectural changes in the medium-term to run more AI workloads. So in 3-5 years you might want a new chip because of better capabilities beyond the ordinary advancement of capabilities. 
This is something I’m quite curious about. More on M4 Pro 24GB ram vs 48GB ram in VDMX for running say 4 layers of 4K content with alpha channels.
The work I mostly do with VDMX is corporate stuff, so basically layering videos on top of a background image, sometimes layering 4 videos on top of an image. All in 4K. Like for this kind of workflow, would there even be a difference in 24gb vs 48gb ram? GPU Cores in M4 Pro should be more than enough for what I do. Even if I do VJ stuff, it’s maximum of 3 layers in 4K with effects. Should 24GB be enough for that?
TouchDesigner .tox files can use up a lot of memory fast, but video playback is less of an issue. Internal SSD space is more important. It’s common to see people running a “slow system” only to find out they only have 3% of their SSD free, and the rest is always filled up. Best not to dip below 10%, or even 15% of capacity is a good idea.
Last I tested, which was with an M1 Max Macbook Pro and 32GB of memory using multiple external SSDs to get over limitations, VDMX was able to run 107 HD videos playing back simultaneously. I haven’t tested this again with the M3 Max I have now, but that was enough for me. For example, my 2015 Macbook Pro maxed out somewhere between 8-10 HD videos (1080p30). These were all in the HAP codec, but based on that test, one could assume 26x 4K (30p) streams. ~13 4K60fps (6x 8K 30fps). etc. And that was an older machine. The limit with video playback like that seems to have more to do with storage bandwidth vs CPU / GPU power.
When you start adding ISF, Vuo, TouchDesigner + multiple 8K streams with various FX, that’s where things start to differ (wildly), but the base rate performance for many of these newer machines is a pretty good leap beyond those of the older Intel machines (for most tasks).
One of these days, we’ll have to set up an actual VDMX benchmark!
