eGPU + 9550

Does anyone have any current recommendations / experience for an eGPU box + 9550?

I know about the T3 two lane limitation, but am thinking about using the box as a docking station to my desktop monitors so should not be an issue.

The following has caught my eye: https://www.newegg.com/Product/Product.aspx?Item=N82E16814932011

 

I recently bought an egpu setup for my 9560. Those Gigabyte units should do the job fine as long as you're ok with keeping that whole setup in the long run rather than upgrading the card later. I went the Akitio Node ($246) route with an Evga 1070 Ti card ($499). I was lucky enough to pickup a card before the whole GPU mining inflation happened. End result is great even with the limitation (which i think was over exaggerated). If you want to game on the internal lcd just make sure you disable your onboard nvidia, reboot and you should be good to go. Using external monitors *should* be plug and play. I stress should because you know it never ends up 100% when you expect it lol

For gaming here are my quick FPS comparison with Resident Evil 7:
1050 4K Max Laptop Display - 8-10 fps
1070 Ti eGPU 4K Max Laptop Display - 39-42 fps
1070 Ti eGPU 4K Max External TV - 49-52 fps

 

It's usable but I'd say anything above a 960/1050 Ti or 1060/970/980 will get extremely bottlenecked to the point where it's not worth the investment

 

This site has a lot of useful posts on egpu including:

https://egpu.io/external-gpu-buyers-guide-2018/

https://egpu.io/forums/pc-setup/xps-15-9550-not-able-to-reach-40gbps-over-tb3/

 

Is that true with an external monitor plugged directly into the eGPU (no round trip back to the internal screen)? I am thinking about an external monitor / VR setup.

 

Yes, PCIe 3.0 x2 is extremely slow by today's standards seeing as modern high-end GPUs are fully saturating PCIe 3.0 x8 and are actually seeing a performance improvement, even if minimal, by going from x8 to x16

 

I can understand that would be an issue when the video output is sent back to the laptop for display on the internal monitor, but when its being sent direct from the eGPU to an external monitor then I would still expect a big jump between a 1060 and a 1080 (aren't the graphic primitives being sent from the lappy to the card the same?)

 

The problem is not the GPU signal - the problem is the speed at which the CPU can issue draw calls to the GPU. Using an external monitor reduces the performance hit by a bit but it's still pretty massive.

 

Indeed, but what I don't understand is why the draw calls to a 1060 would be any different to the draw calls for a 1080 (but the 1080 will render them better).

But yes 4k@60hz is going to work better than 1440@165hz but that is independent to the card in the eGPU?

 

You're missing the point - the GPU performance is less relevant than what the connection can allow in terms of throughput. Meaning, the 1060 will be more saturated than a 1070 or 1080 - sure, the 1080 will perform better but you'd still get better performance from a 1070 laptop than that 1080.

 

If the point your making is that a 10x0 will not work as well in a eGPU setup compared to an internal graphics card then yes I fully understand that.

But I was under the impression that the % hit is consistent (when attached to an external monitor) irrespective of the card used.

IOW the % improvement of a 1070 or 1080 over the 1060 in a eGPU setup would match the % improvement you would see in a desktop setup?

 

The performance hit is not linear - more powerful GPUs require more bandwidth to reach full potential. Check GN's Titan V tests - the 1080 doesn't care if the PCIe is 8x or 16x but TItan V actually sees an improvement on 16x.