RTX 3080 + DELL XPS 7590 + Razer Core X : Is there any way I can significantly increase GPU performance?
-> GTX 1650
-> 32GB RAM
-> 1TB SSD
I recently purchased the RTX 3080 hoping to get decent performance in an EGPU. I understand it is typical for performance to vary dependant on benchmark or application or game, but I feel as though the loss I am obtaining is quite significant.
In Passmark's G3D benchmark my card received: 13800
Typical RTX 3080 in Passmark G3D score: 23000
User benchmark my card was scored in the bottom 2% of all RTX 3080 cards.
And in games like Metro Exodus, in 4K, Ultra settings, DX12, I am getting a 58% loss in performance than a typical RTX 3080.
In Far Cry 5, 4K, Ultra, I am getting a 36% loss.
In Skyrim, I can't get more than 40fps at 1080p.
Please let me know if there is anything I can do to increase this performance.
Ouch, I would expect a 15% loss over Desktop but certainly not more than half?
This is connected going directly into the laptop with the cable provided from the Razer Core X unit itself as well?
@natmar99, Do you know if your particular XPS 7590 has a thermal pad on the PCH chip? I tested an older XPS 15 and the lack of cooling on this PCH (which TB3 connection goes through) really hinders eGPU performance [build link].
@itsage Based on the pictures from the link you provided, it looks like there is a thermal pad on the PCH chip. I have attached a photo of my XPS internals for confirmation.
At this point my only suggestion would be to try a different TB cable and as wonky as that sounds, I've learned that all cables are not created equal. With my external NVMe and using a different cable, my data transfer speeds doubled. I couldn't believe I was being capped by a shoddy cable so while that may not be the case with your use here, it is something to consider.
I had never run passmark before, I just downloaded Passmark PerformanceTest 10 and ran the suite of 3D tests and got 5969 on my 2070 + i7 8565U on the laptop's internal monitor (I don't have an external monitor). It seems about right IMO considering the differences in our systems that you got ~130% increase.
I've noticed running some games and benchmarks that some things are less affected than others by being on TB3. I don't know if some benchmarks and titles do more communication over the PCIe bus than others. I tried using tools to monitor bus usage of different benchmarks and titles and couldn't come up with any reasonable smoking gun. It could be that some software titles are just more susceptible to the raw latency than others and the actual bus bandwidth usage doesn't show anything significant, or the tools have a hard time accurately measuring it, probably the latter.