RTX 3070 + Razer Core X (Thunderbolt 3) = Bottleneck ?
The comments :
- Judging from how the eGPU performance offset due to Thunderbolt 3 bandwidth, more powerful GPU means more power unable to utilized by Thunderbolt.
- I agree with this statement. The cost-performance ratio is really bad for highend cards. Stick with higher midclass. It does not make sense to pay double to triple the price just to get 10% more fps.
- yea,it's crazy with generation how bad the thunderbolt is bottlenecking the whole system. I wanted to take a 3090 for myself (with a Core V1) but clearly not anymore. Wondering if it's possible to do some compression tho.
I would like to have your opinion.
@bloodstainer, When I was in high school, Fast & Furious came out and all the kids my age spent every dollar we had on shitty mods to make our cars look faster. When I got a full-time job I was able to afford more appropriate performance mods. I soon realized the car was faster but it's still the exact same car to everyone else. Ultimately you get to decide what is worth doing and others' opinions don't matter. The chase for better performance doesn't have an end. You decide when and where to stop.
I'm more grown up now and approach car mods with a better plan. The starting platform has to be strong enough to accommodate higher performing components I would eventually add to it. All that to say eGPU performance has a lot to do with the computer you attach the external components. If it's a 2017 iMac 4K in your signature and you plan on using an external 4K monitor with the RTX 3070, the gain will be substantial over the RTX 2060.
@itsage, thanks for your answer.
In fact the comments I quoted talked about a bottleneck about the "Thunderbolt 3" bandwidth with RTX3070. I agree the performance depends on the base materiel but that was not my concern (I plan to upgrade my Mac too).
So the people that I quoted say that the bandwidth of TB3 is a bottleneck for the news RTX3XX. That's why I would like to have the opinion of technical experts of this forum (like you).
(Sorry for my English, I'm French).
Why not look at some actual RTX 3070 eGPU performance metrics?
Unigine Superposition at 4K DirectX
My old RX 5700 XT eGPU: 47 FPS average
Increase in performance with the RTX 3070: 79%
Cool. There’s also another good 3070 eGPU build with gaming benchmarks here:
Another comp, this time with my current Vega 56 eGPU:
Shadow of the Tomb Raider at 4K (without DLSS or Raytracing)
My AMD Vega 56 eGPU: 32 FPS average
Increase in performance with the RTX 3070 : 62%
I think there is another thing you should consider. What kind of performance improvement are you trying to achieve?
If the main purpose is gaming, I think that the main factor is the target fps you are trying to reach.
Let's say you already reach between 30 and 50 fps in a given game with your current card and the gpu is the main limiting factor. You will probably see a pretty good improvement as others already stated.
But let's consider the case in which you are already reaching something like 100 fps. Then the improvement you will see will probably be much smaller, because at high fps the latency becomes a significant limiting factor.
I'm kinda new to this stuff, so if I said something wrong someone please correct me.