Desktop Builds for Faster Render Times for H.265

Hello everyone!

I am currently researching the most feasible way to deliver H.265 or Uncompressed 10-12bit with the fastest render benchmarks under Mac OSX.

With that being said, I have been researching into Hackintosh's yet the components that are for the current operating system (12.2.1) aren't that powerful, and Nvidia is not partnering up for coding compatibility GPU's with Apple lately including certain ATI Processors so that limits the components list naturally.

What are your benchmarks with your rigs? For example, 8K RAW 12bit H.265 output delivery times and any other high quality output parameters.

With our current specs, a reel of 45 videos is taking 27 hours+, not practical for our business.
We also have a budget since the 2022 Mac Pro with the Dual Radeon x6700 64gb and the Intel 2.43Ghz Xeon 24 core is 25,000$ so were trying to steer away from that price point in general.

Thank you!

Hello everyone!

I am currently researching the most feasible way to deliver H.265 or Uncompressed 10-12bit with the fastest render benchmarks under Mac OSX.

With that being said, I have been researching into Hackintosh's yet the components that are for the current operating system (12.2.1) aren't that powerful, and Nvidia is not partnering up for coding compatibility GPU's with Apple lately including certain ATI Processors so that limits the components list naturally.

What are your benchmarks with your rigs? For example, 8K RAW 12bit H.265 output delivery times and any other high quality output parameters.

With our current specs, a reel of 45 videos is taking 27 hours+, not practical for our business.
We also have a budget since the 2022 Mac Pro with the Dual Radeon x6700 64gb and the Intel 2.43Ghz Xeon 24 core is 25,000$ so were trying to steer away from that price point in general.

Thank you!

8K RAW of what? ProRes RAW? BM RAW? RED raw?

You generally get better H265 encoding/decoding performance with NVidia cards.

If you want top-end performance, you are looking at multiple GPU solutions. With current PSUs, you are limited to 2 (maybe 3) 3090ti unless you use a PCIe extender box.

At a certain point you may need to consider the performance of your storage. You need to make sure you are feeding the system with data at the proper rate.

The right system really depends on your source media and delivery formats. You'll get more informed feedback if you briefly describe your workflow.

Here's some generic thoughts for a new system.

Current generation- it depends. AMD's 5950X is actually better in some video apps than Intel's 129000k.

However, the next generation of AMD processors are expected in September. Intel's are expected in October. Will have to wait to see who comes out on top.

As for graphics cards, there's pressure from board manufacturers to get Nvidia to lower prices and delay the 4000 series until next year because they are overstocked, thanks to Nvidia raising their MSRP's during the GPU shortage. They cut prices on the 3090 in response with more to follow.

And keep in mind, you'll need a HUGE rated power supply, especially for an Intel/Nvidia rig. You'll need a new power supply built to the ATX 3 standard and a MB that supports it. Especially with the 4000 series cards and 13th gen CPUs. You'll also want one of the cases designed to maximize airflow.

That aside, no matter which CPU/GPU you go with, you'll want a workstation level motherboard with the top of the line chipset from either manufacturer. If you want to run a couple of GPUs, you need this level of motherboard so both cards will have more PCIe lanes in each slot under PCIe 5.You'll also have more lanes for your drives. Gamer MBs won't do this.

You'll want not only screaming fast DDR5 memory, but also the fastest NVme / SSD drive you can get.

Memory will also be faster if it's Quad channel.

These are the generic things to keep in mind. Specific model choices are going to depend on how things shake out in the next couple of months. Waiting to see if the next gen from Nvidia or AMD will support A1 encoding.

fastest render benchmarks under Mac OSX.

No chance of moving to Windows, cause this screams using a 3000 series RTX card. (Looks damn good, too.)

HDR Deliver.png (67.9 KiB) Viewed 646 times

Yeah, even with an RTX3060 and i7 8700 my system exports UHD footage at around 60-100FPS depending on the complexity of grading. More GPU-intensive operations such as temporal noise reduction or optical flow throw a spanner in the works but the NVENC side itself is pretty damned fast.

If you want to stay in MacOS, you should go M1 Max. I'm not sure Ultra will give you much more at the current point of development. I can give you some rendering times on my laptop in two or three days.

Thank you everyone for responding.

To give a little more of the type of jobs I’m doing I’ll list a process down. I’ll list source media as well.

Source media: R3D, BRAW, SRW, ARRIRAW, FUJIRAW, .MOV (gearing up towards Kinefinity RAW, CRW)
All these source media are shot on the HIGHEST bitrate, resolution and always usually at 17:9)

- Importing 200+ clips into a timeline
- having 6+ nodes (Primary, Secondary, Hue/Sat, Qualifier nodes, and more) including Noise Reduction where Temporal is above 20% on the slider, max 100%
- Watermark burn in at (4K-8K) resolution
- Uncompressed version at 10-12Bit
- version for 8k UHD 17:9 h.265 at target bit rate at 10,000kbps+
- wrappings for Brightcove, YouTube, custom engineered video streaming service video
- to have these 100-300+ clip reels delivered under several hours

Essentially I need something for Mac OSX since we have created a custom python script for Mac OSX where we cannot transition to Windows. So we need a system that will be easy to build (not establish custom Linux/python logic to have components function with OSX); cheaper than the full decked out 2022 Mac Pro @ 35,000$, and will be functionally stable to run this program 24/7 without hiccups.

Thank you!

Hello everyone!

I am currently researching the most feasible way to deliver H.265 or Uncompressed 10-12bit with the fastest render benchmarks under Mac OSX.

With that being said, I have been researching into Hackintosh's yet the components that are for the current operating system (12.2.1) aren't that powerful, and Nvidia is not partnering up for coding compatibility GPU's with Apple lately including certain ATI Processors so that limits the components list naturally.

What are your benchmarks with your rigs? For example, 8K RAW 12bit H.265 output delivery times and any other high quality output parameters.

With our current specs, a reel of 45 videos is taking 27 hours+, not practical for our business.
We also have a budget since the 2022 Mac Pro with the Dual Radeon x6700 64gb and the Intel 2.43Ghz Xeon 24 core is 25,000$ so were trying to steer away from that price point in general.

Thank you!

8K RAW of what? ProRes RAW? BM RAW? RED raw?

You generally get better H265 encoding/decoding performance with NVidia cards.

If you want top-end performance, you are looking at multiple GPU solutions. With current PSUs, you are limited to 2 (maybe 3) 3090ti unless you use a PCIe extender box.

At a certain point you may need to consider the performance of your storage. You need to make sure you are feeding the system with data at the proper rate.

I would love to get the Ryder threadripper 24core; Nvidia A5000/5500, or the ATI Radeon Pro W6800 series; possibly SLI/Crossfire.

Obviously get fastest SSD, and 120GB of ram which seems not to have a compatibility issue with Hackintosh.

Motherboard is limited for current OSX build and those graphics cards don’t have community support yet from my research as informed me

Here's some generic thoughts for a new system.

Current generation- it depends. AMD's 5950X is actually better in some video apps than Intel's 129000k.

However, the next generation of AMD processors are expected in September. Intel's are expected in October. Will have to wait to see who comes out on top.

As for graphics cards, there's pressure from board manufacturers to get Nvidia to lower prices and delay the 4000 series until next year because they are overstocked, thanks to Nvidia raising their MSRP's during the GPU shortage. They cut prices on the 3090 in response with more to follow.

And keep in mind, you'll need a HUGE rated power supply, especially for an Intel/Nvidia rig. You'll need a new power supply built to the ATX 3 standard and a MB that supports it. Especially with the 4000 series cards and 13th gen CPUs. You'll also want one of the cases designed to maximize airflow.

That aside, no matter which CPU/GPU you go with, you'll want a workstation level motherboard with the top of the line chipset from either manufacturer. If you want to run a couple of GPUs, you need this level of motherboard so both cards will have more PCIe lanes in each slot under PCIe 5.You'll also have more lanes for your drives. Gamer MBs won't do this.

You'll want not only screaming fast DDR5 memory, but also the fastest NVme / SSD drive you can get.

Memory will also be faster if it's Quad channel.

These are the generic things to keep in mind. Specific model choices are going to depend on how things shake out in the next couple of months. Waiting to see if the next gen from Nvidia or AMD will support A1 encoding.

All concerns with PSU is not the problem. My problem is the specific parts wanted aren’t recognized on current Monterey build of OSX. Community support for Hackintosh is limited for the MB, GPU, and CPU.

If anyone knows work arounds; or a build of an effective Hackintosh with those components without the hassle of inventing custom OSX drivers for compatibility stability.

I'm afraid you are heading in the wrong direction trying to go Hackintosh these days. It's difficult enough to find the right drivers for a strong GPU under Windows, they are one of the most frequent issues around here. With no support from Apple or Nvidia for some of the strongest GPUs out there, you are asking for a lot of trouble. A Hackintosh is a dead end street by now, with Apple leaving Intel behind (pun intended). You'll sooner or later run into more and more incompatibilities.

I'm not a programmer, but IMHO it shouldn't be too difficult to port that Python script to Windows. So, my first recommendation would be doing that and building a beefy Windows machine. There is no sign yet when Apple will offer a new MacPro (and what it'll be, after all).

If that is absolutely out of the question, why not go Mac Studio? Heck, you may even want to get two machines, one base model with some RAM added for editing and grading with proxies and a second beefy one with all the bells and whistles to render along. BM did a pretty good job porting DR to Apple silicon and Apple's hardware support for H.265 encoding is second to none.

I would love to get the Ryder threadripper 24core; Nvidia A5000/5500, or the ATI Radeon Pro W6800 series; possibly SLI/Crossfire.

Obviously get fastest SSD, and 120GB of ram which seems not to have a compatibility issue with Hackintosh.

Motherboard is limited for current OSX build and those graphics cards don’t have community support yet from my research as informed me

There is no SLI support on the A5000 series. Actually, there is no SLI support on any recent NVidia GPUs.
Some GPUs support NVLink. Similar but not the same.

Im guessing there is not really a path to take via Hackintosh and getting some epic components?

Does anyone have benchmarks on any of those codecs and parameters I listed?

I would just like to get a grasp on your rigs and see if even the Mac Studio fully decked would be plenty for the load I am trying to put it through.

So, let's get over the details on my hardware:
Source media playing smoothly at 25 fps in a 4K timeline:
R3D, BRAW, SRW, ARRIRAW, Canon RAW: check

FUJIRAW: sample please
Kinefinity RAW: too
.MOV: this is just a container, what's in it?

All these source media are shot on the HIGHEST bitrate, resolution and always usually at 17:9)
For bitrate mainly storage and throughput matter, what will you use? I can play 6.5K ArriRAW at 30 megabyte per frame smoothly from a Thunderbolt SSD.
- Importing 200+ clips into a timeline: check
- having 6+ nodes (Primary, Secondary, Hue/Sat, Qualifier nodes, and more) including Noise Reduction where Temporal is above 20% on the slider, max 100%
What matters here is mainly temporal radius, not percentage. A high radius won't be RT on any hardware.
- Watermark burn in at (4K-8K) resolution: check
- Uncompressed version at 10-12Bit: Output? Again, data throughput and storage. But who is doing this today? It's a waste of space. Even EXR can be compressed.
- version for 8k UHD 17:9 h.265 at target bit rate at 10,000kbps+
Encoding 1 minute in a tad above 2 minutes of time for 8K H.265 10 bit 4:2:2 at 80 mbps
- wrappings for Brightcove, YouTube, custom engineered video streaming service video
Not tested
- to have these 100-300+ clip reels delivered under several hours
Do you mean the whole process or rendering? Only you know how fast you are editing and grading. For encoding see above.

All of this applies to my humble laptop (see sig). A Mac Studio Max would be roughly twice as fast. I doubt that an Ultra is four times faster (typical diminishing returns for Apple's top model), but it'll offer you more TB connections for your massive storage needs.

Feel free to ask.