2019 15" Dell XPS 15 7590 (GTX1650) [9th,6C,H] + RTX 2070 @ 32Gbps-TB3 (Razer Core X Chroma) + Win10 [Emmanuel Barinia]
Acer Predator X34P (120hz, 3440x1440, IPS, G-Sync)
EVGA RTX 2070 XC ULTRA 8gb
Installation was rather plug and play compared to my previous RX 5700 XT. No need to reinstall older windows version or anything. Just plug, install drivers, and play.
- downloaded and installed latest Cuda drivers. I read here that normal driver could worsen battery life, so I have followed the instruction indicated here: https://egpu.io/forums/builds/2018-15-razer-blade-15-gtx-1080ti32gbps-tb3-asus-xg-station-pro-win10-1803-itsage/#post-63079
Now this is where it gets quite confusing for me. I know that the XPS 15 with dGPU is not ideal for an eGPU setup because of many many issues: Thunderbolt connected to the PCH (so even more performance loss), PCH temperature and potential throttling, poor CPU/dGPU cooling due to inadequate small chassis, Optimus which seems to be causing me another layer of problem... I mean, I chose this laptop with media consumption in mind, and the eGPU was an after thought for me. Sadly, not every Thunderbolt laptops are born equals.
Now, I think I have addressed most of these issues by doing the following:
- I changed the original thermal paste with liquid metal and undervolted my CPU to -150mV. This means that my CPU temperature never reaches more than 80 degrees on an 6 hours Aida64
- The 7590 (unlike the 9570 I had before) has a thermal pad (quite thick, about 3mm) already installed. HwInfo reported temps of 65 max. I have added another layer of 1.5 pads in order to make contact with the aluminium chassis, this has dropped PCH temperature further down to a maximum of 55, which I hope is enough not to thermal throttle.
Apologies for the links, I seems to be unable to post photos directly. Please note that all of the results below are with OC of 125mhz/600mhz applied to the GPU.
GPU Z report: https://imgur.com/0heRRDz
Superposition 1080p - 14223: https://imgur.com/hKupopw
Superposition 4k - 7105: https://imgur.com/oLWGO2J
Unigine Valley - 93fps: https://imgur.com/qsNln7b
Firsestrike - 14530: https://www.3dmark.com/fs/21662178
Timespy - 7915: https://www.3dmark.com/spy/10407625
Superposition 1080p - 10614: https://imgur.com/undefined
Superposition 4k - 6492: https://imgur.com/undefined
And... sadly this is as far as my internal display benchmarks could go. I seems to be unable to add Unigine valley to the graphics setting options in windows to enable it to be used with the eGPU, so it is rendered at 5fps. Same goes for 3dMark (the steam version).
As mentioned earlier, I had a RX 5700 XT prior to the RTX 2070, and even though my TimeSpy benchmark for both are quite similar, my Firestrike benchmark was much higher with the AMD card, at 17869 - https://www.3dmark.com/fs/21624601 Any idea what would be the reason for this?
All in all, I guess I'm a little bit underwhelmed by the performance. Granted, the eGPU setup allows me some flexibility and portability, but I guess the fact that the Thunderbolt is connected to the PCH makes the overall performance a little bit slower than I anticipated. I saw XPS 13 build with slower CPU getting better results, I assume simply because of the way the thunderbolt is connected directly to the CPU.
I will conduct more research on the internal display performance. A Superposition drop of 4000 points seems rather suspicious at 1080p, although the 4k benchmark is about where I would expect it compared to the external display. If anyone has any tip on how to improve the internal/external display performance, I'm all hear.
Those benchmark discrepancies aren’t too surprising.
TB3 performance loss (during loopback to the internal display) occurring at a much higher rate at 1080p than 4K is to be expected.
Nvidia RTX card performance can vary a lot on DX11 games. Sometimes performance is better, but a DX11 test (like Firestrike) being surprisingly low is also something that happen.
Hm. Thank you for making this thorough post. I love the XPS series (have had the last three) and was hoping to upgrade to the 7590/i7-9750/FHD. Would be using a Razer Core X (was thinking 1080ti/2070s or greater) to power an external 3k monitor (HP z34c) and/or my Valve Index (really only care about HL:Alyx when it drops), but it seems as though this is not a good option.
Given what you know now, what would you do different?
@pazza, I have an XPS 15 7590 and a WMR headset, and I'll be receiving a Aorus RTX 2070 gaming box which I bought mostly for HL:Alyx. I can report on general VR performance when I receive the eGPU, if you want. Keep in mind that, while not the most comfortable way of experiencing VR, XPS 15 7590 with its dGPU can definitely do some light VR gaming without an eGPU enclosure. I've been playing H3VR for a while on my laptop, and while not the most comfortable experience, Boneworks is also somewhat playable.
Hey, I reveived the eGPU today and ran a couple of VR games really quickly. While certainly playable, I really don't see a huge improvement over the GTX 1650 XPS 15 already has. (and by that, I mean there is certainly quite a bit of stuttering with Boneworks) I'll do a couple of tests with FPS counters, as soon as I figure out how to do that.
There's also a weird lack of information about eGPU enclosures being used with WMR headsets. I guess eGPU's are still quite niche. I might post a thread here if I can bring myself to do proper tests with this. It's a shame I don't have an external monitor for non-VR games.
Well, thats odd as hell to be honest, i cant see why 1650 would perform in similiar fashion with RTX2070 on tb3. Ive ordered a Aorus RTX2070 box so ill see whats up when it comes in. I do have a external monitor setup so we'll see. Thanks for the feedback!
So far I am very happy with the performance, but I did not run any benchmarks. In case you would like me to do that, let me know and I will gladly share the results.