Buying Razer Core X eGPU enclosure. Evaluating Radeon graphics cards

I have an iMac Pro 2017 with Radeon Pro Vega 16GB built-in graphics and 64GB RAM.
I’m trying to run a second HD monitor and process Video AI tasks. Crashing and burning everytime I push past 50% memory. I’ve done some research on the various options and besides video processing, the iMac works great. And I don’t need a new M2 laptop, so eGPU seems like the most cost effective solution, given that Apple encourages it.

I don’t think I can go wrong with a Razer Core X enclosure. It’s the choice of video cards that has me asking for some expert advice. The only card option I have is Radeon, but they have a wide variety to choose from, making it a challenge to find the right mix. Given the enclosure costs the same price as a halfway decent GPU, I figured I wanted to put something in that really supports what I need, primarily video processing. If I’m not using Topaz Video AI, I’m editing in FCPX. In need some raw video processing power at an affordable price.

Tom’s Hardware has the a great summary Toms Hardware GPU Benchmarks 2023.

Here are some relative price/performance points from this.

~35fps using RX5808GB @$125 or $3.57/fps with 8GB VRAM
~80fps using RX6650XT @$400 or $5.00/fps with 8GB VRAM
~110fps using RX6750XT @$550 or $5.00/fps with 12GBVRAM
~120fps using RX6800XT @650 or $5.42/fps with 16GB VRAM
~140fps using RX6950XT @1000 or $7.14/fps with 16GB VRAM

That seems to suggest a sweet spot around the 6600 or 6700 family processors at $5/fps.

Looking the specs, it seems like most important variables are:
-Process Node
-GPU Cores
-Ray Tracing Cores
Tensor Cores

I would love to know anyone’s experience using an eGPU, and had at least one of these cards in it and what you think in terms of performance and cost.

Thanks in advance! I’m leaning towards the RX6750XT at this point, given current pricing.

Hold up… I just spotted the RX6950XT @699 on AMD.com.
That clocks in at 4.999/f/s and a reasonable buy forward at that price.

No matter what GPU you buy it won’t get faster because the CPU is limiting here.

Also on the net you can read that an egpu for FCPX does not help and that the internal gpus are faster.

And the benchmarks you look at in thw are all for gaming.

2 Likes

I have never heard of anyone using an eGPU on a mac. Do they make drivers for mac? My impression of macs in general is that you cannot put whatever hardware you want in them because of the lack of drivers… and Apple worrying about inconsistent user experience from unsanctioned hardware—therefore no drivers are ever made for most hardware that is not already in macs.

Apple Support-eGPu

Ah, so it’s limited to AMD GPUs and not the latest generation yet. Thank you.

Bingo. You’re right, the Tom’s test benchmarks were for rendering 1080p and 1440p video, which I thought would be a relatively good proxy for general GPU performance.
Where things get murky quickly is what Topaz uses for its rendering algorithms. You’ve got shaders, ray tracing, and tensorflow processors. What gives you the biggest bang for your buck in Video AI processing?

My iMac has a 3 GHz 10-core Intel Xeon processor. CPU performance has never been a problem. The problems come when I attach a second monitor and try to do anything else requiring GPU - transcoding, rendering, anything that adds a load to the GPU ends up with a “lights-out” system shutdown. When I reported this to Topaz, I was told to reduce max memory usage to 35% and start there, working my way up to see how far I could push it without crashing. That’s been a frustrating experience and requires I shut off my second monitor (useful for video editing) and run any rendering as a dedicated workflow which kinda sucks.

1 Like

That’s interesting about FCPX. I’ll have to dig into that. I suppose it would depend on whether FCPX’s dev team has optimized for eGPU or plan to in the future. Most people probably won’t remember this, but in the early days of Macintosh, post-Lisa, the Motorola processors just weren’t competitive with WinTel, so there was a company that made a bunch of add-in accelerator cards that boosted CPU/GPU performance by quite a bit. Can’t remember their name, but most Mac owners looking for peak power bought them. That market only lasted about eight years until Intel became the processor of choice for Apple (circa 2005?), but they made a huge difference. Rather ironic we’re seeing this play out again for Apple…

Hi! Glad to hear I’m not the only one experiencing crashes with Topaz using an eGPU. Did you find any solution?

I have an 27" iMac with 8GB Vega GPU but after reading about eGPU capabilities, I got an AMD RX6800 XT 16gb with a Chrome X enclosure (which is beautiful btw!).

Overall the computer graphics run faster and all is good, but with Topaz, my computer freezes immediately or sometimes it renders for 10-20 minutes and my computer just reboots.
I did lower the Memory usage 50, 20, 10% and while it held up a bit longer, eventually it just crashes (30 min is the max time for now).

Wondering if this is a Topaz issue that can’t handle the external cards or what. The only thing that I haven’t done so far is to “update the GPU drivers”, but I don’t have a Windows computer but is it necessary? as the drivers are related to the OS, not the card itself (unless there’s a firmware update?)

I’m starting to thinking if I should just return it all since Topaz is the main reason I got this, and maybe get those new M2 Pro computers…

Thanks!

Yes, the performance never really came through. I ended up purchasing Mac silicon with shared memory In the processing works just fine. Would certainly be my recommendation. eGPU is just an ugly solution.

1 Like

Thanks for the quick reply! Yeah, unfortunately I think I’ll have to do that too. Which Mac silicon did you get? I’ve been searching for comparisons so that at least whatever M chip I get is equivalent to the 6000 series xt gpu or something. I was looking into M2 Pro but those 19 - 30 core GPU numbers don’t tell me much.

Another solution is to keep the enclosure and get a Windows laptop with an Nvidia card, lol (Topaz is supposed to take advantage of it better than AMD cards).

thanks!

I do a lot of video and image editing So I ended up purchasing the 16 inch MacBook Pro M2 Max with 32GB of shared ram. I found the shared ram to work really well and give it big performance boost what Intel has. That said, I just bought a Dell workstation with an Nvidia card in it for anything I really need to get done quickly. It’s the Cuda architecture that’s perfect for topazlabs apps

1 Like

Oh, great!! I might wait until the Mac Studio M3 release. Otherwise, maybe a Windows workstation work too. Which Nvidia card did you get? I read that the 4060 is similar in speed of the 6800xt.
Thanks!

I ended up buying the Dell Precision 7680. It came with 64GB RAM, 13th gen i9 and two video processors, the Intel UHD for display and 4GB Nvidia RTX3500 Ada GPU, plus a 1TB internal drive. I’m using my Sandisk Extreme Pro 2TB USB-C drive for external storage. I’m using it primarily for 2D/3D mapping and modeling so I’m really pushing the system and I’ve been extremely happy with the performance.

Here are the specs. It’s expensive, but so are Macs…
(Precision 7680 Mobile Workstation | Dell USA)

1 Like

Thank you for the detailed description! Wow, that looks one powerful laptop! I’ll look into it as it wouldn’t hurt to have a Windows OS just in case. And it runs Topaz quite fast?
I returned my AMD card yesterday and I think I’ll return the enclosure as well, and just use that money towards a new laptop or a Mac Studio M3 later this year. Too much headache and space haha.
Thanks!!

Wow, that’s a hefty price tag that notebook carries.
Could you please provide a TVAI benchmark so that we can have a performance estimation vs. the comparably priced MacBook Pro 16" with M3 MAX?

Kind of a mixed bag at 1080p.

Prices as spec’d:
Apple Macbook M2Max: $4,800.
Dell Precision 7680: $5,200.

1080P Results.

Topaz Video AI v4.2.0

System Information

OS: Mac v14.0301

CPU: Apple M2 Max 64 GB

GPU: Apple M2 Max 48 GB

Processing Settings

device: 0 vram: 1 instances: 1

Input Resolution: 1920x1080

Benchmark Results

Artemis 1X: 12.03 fps 2X: 07.13 fps 4X: 02.47 fps

Iris 1X: 13.22 fps 2X: 03.98 fps 4X: 01.80 fps

Proteus 1X: 12.19 fps 2X: 07.89 fps 4X: 02.41 fps

Gaia 1X: 03.49 fps 2X: 02.41 fps 4X: 01.76 fps

Nyx 1X: 03.99 fps 2X: 04.22 fps

Nyx Fast 1X: 10.81 fps

4X Slowmo Apollo: 13.23 fps APFast: 51.88 fps Chronos: 03.75 fps CHFast: 06.15 fps

16X Slowmo Aion: 13.04 fps


Topaz Video AI v4.2.0

System Information

OS: Windows v11.23

CPU: 13th Gen Intel(R) Core™ i9-13950HX 63.692 GB

GPU: NVIDIA RTX 3500 Ada Generation Laptop GPU 11.749 GB

GPU: Intel(R) UHD Graphics 0.125 GB

Processing Settings

device: 0 vram: 1 instances: 1

Input Resolution: 1920x1080

Benchmark Results

Artemis 1X: 11.50 fps 2X: 07.30 fps 4X: 02.65 fps

Iris 1X: 12.88 fps 2X: 08.31 fps 4X: 02.22 fps

Proteus 1X: 10.88 fps 2X: 07.03 fps 4X: 02.35 fps

Gaia 1X: 03.87 fps 2X: 02.63 fps 4X: 01.74 fps

Nyx 1X: 04.93 fps 2X: 04.31 fps

Nyx Fast 1X: 09.13 fps

4X Slowmo Apollo: 16.55 fps APFast: 45.12 fps Chronos: 08.99 fps CHFast: 13.97 fps

16X Slowmo Aion: 28.66 fps

1 Like

Thanks. So they’re quite on a level (apart from the still abysmal Iris performance on Apple Silicon that Topaz somehow refuses to fix).

1 Like