2020 upgrade- 5700xt vs lower spec dGPU plus 5700xt eGPU

[thermalthrottle]

Hi folks, I'm looking at upgrading to a new 27inch iMac and wondering if it's worth upgrading to the 5700xt 16gb dGPU over the base model 5500 8gb model or if I'm better off saving the money and put it towards a 5700 xt eGPU? I already have an eGPU enclosure and the 5700 xt cards are going for around £350 so I would be saving some money over upgrading outright to the 5700xt dGPU option at £500. My priority however is on performance rather then cost, otherwise I could have considered the mid spec option with the 5300 4gb card plus the eGPU but having used a 16in MacBook Pro with the same card, I can tell right away that 5300 GPU would be too sluggish for my needs.

The main issues would be thunderbolt 3 bottlenecking and how that would affect the performance of the 5700 XT eGPU although Geekbench metal scores seem to show that card averaging in the high 60k mark whereas early benchmarks of the dGPU 5700 XT on the iMac is averaging 12k less on the metal benchmarks. The other issue is if I would be better off having 16gb vram on one card or split over two 8gb cards. Some of the apps I use can leverage multiple GPUs although whether that is an improvement over the one card remains to be seen. If anyone who already has a new iMac with the 5700xt 16gb could you please download and run NeatBench and report the results, thanks in advance.

One final point for the eGPU solution is added flexibility, being able to use it with my 15 inch 2013 MacBook Pro which although showing it's age, is still a capable machine when paired with the right eGPU. I'm mainly using this for video editing and use FCP and starting to use Davinci Resolve, both with heavy use of plug-ins such as Neatvideo.

 

[sco]

Sounds like the 5700XT steals power from the CPU. Depending on your scenario you may be much better with an external 5700XT to free up power for the CPU:

 

[0279317]

Hi folks, I'm looking at upgrading to a new 27inch iMac and wondering if it's worth upgrading to the 5700xt 16gb dGPU over the base model 5500 8gb model or if I'm better off saving the money and put it towards a 5700 xt eGPU? I already have an eGPU enclosure and the 5700 xt cards are going for around £350 so I would be saving some money over upgrading outright to the 5700xt dGPU option at £500. My priority however is on performance rather then cost, otherwise I could have considered the mid spec option with the 5300 4gb card plus the eGPU but having used a 16in MacBook Pro with the same card, I can tell right away that 5300 GPU would be too sluggish for my needs.

The main issues would be thunderbolt 3 bottlenecking and how that would affect the performance of the 5700 XT eGPU although Geekbench metal scores seem to show that card averaging in the high 60k mark whereas early benchmarks of the dGPU 5700 XT on the iMac is averaging 12k less on the metal benchmarks. The other issue is if I would be better off having 16gb vram on one card or split over two 8gb cards. Some of the apps I use can leverage multiple GPUs although whether that is an improvement over the one card remains to be seen. If anyone who already has a new iMac with the 5700xt 16gb could you please download and run NeatBench and report the results, thanks in advance.

One final point for the eGPU solution is added flexibility, being able to use it with my 15 inch 2013 MacBook Pro which although showing it's age, is still a capable machine when paired with the right eGPU. I'm mainly using this for video editing and use FCP and starting to use Davinci Resolve, both with heavy use of plug-ins such as Neatvideo.

What did you end up doing? Are you happy with the results?

 

[thermalthrottle]

I borrowed an 8 core iMac 2020 with the 5500xt 8gb and paired it with a 5700 xt 8gb egpu via thunderbolt 3. The egpu is supposed to be more powerful then the dedicated card in the iMac yet overall performance was pretty much about the same when using handbrake or rendering in FCP/davinci resolve. I chalked this down to the limitation of thunderbolt 3 but also noticed that some of the processes weren't really maxing out the graphic cards regardless of which one was used which suggests the programs aren't fully optimised for both or have some sort of universal optimised code which maximises older slower hardware, but bottlenecks newer hardware.

I'm going to be using a 10 core with the 5700xt 16gb next month and will run some tests to see if hardware makes much difference or if the software is bottlenecking the processes.