System specs:
- Notebook Model: Razer Blade 15 Base Model (Early 2020) - RZ09-0328
- OS: Windows 10 (10.0.19042 Build 19042)
- CPU: i7-10750H
- iGPU: Intel UHD Graphics
- dGPU: RTX 2060 Max-Q
- Internal Display: 15" 1080p 144Hz
- External Monitors: 2 x Dell S2719DGF (27" 1440p, G-SYNC Compatible)
eGPU hardware:
- eGPU: RTX 3060 Ti (Gigabyte GeForce RTX 3060 Ti Gaming OC Pro)
- eGPU Enclosure: Razer Core X
- TB3 Cables: .5m passive TB3 cable (included with Razer Core X), Belkin 2M active TB3 cable
Hardware pictures:
Installation steps:
Thunderbolt 3 eGPUs have been plug-and-play on Windows 10 for a while. First, I downloaded all Windows Updates. Then I deleted the existing NVIDIA dGPU driver using DDU v18.0.3.4 and rebooted my computer with the eGPU plugged in. Finally, I installed the latest nvidia drivers for the RTX 3060 Ti. The single driver installation installed drivers for both my dGPU and eGPU.
Benchmarks:
CUDA-Z H2D: 22xx.xx MiB/s (~18.4 - 18.8 Gbps)
I ran Fire Strike, Fire Strike Extreme, and Time Spy 3DMark benchmarks on the eGPU connected to the two external monitors and on the dGPU connected to the internal display.
I wanted some kind of comparison to desktop performance, but I didn't have the means to run such benchmarks myself, so I looked up results for benchmarks on an RTX 3060 Ti paired with a Ryzen 5 3600 in a desktop. According to UserBenchmark, Ryzen 5 3600 is a desktop processor comparable to the i7-10750H mobile processor; they have comparable benchmark performance, they're both 6 core processors, and although the Ryzen 5 3600 has a higher base clock speed, both it and the i7-10750H run a similar average turbo clock speed. Also, for each RTX 3060 Ti + Ryzen 5 3600 benchmark, I chose the single benchmark whose 3DMark score was closest to the average score for that GPU & CPU combo.
3DMark Benchmark | GPU + CPU | ||
RTX 3060 Ti (eGPU) + i7-10750H | RTX 2060 Max-Q (dGPU) + i7-10750H | RTX 3060 Ti + Ryzen 5 3600 | |
Fire Strike (1080p, D11) | 3DMark Score: 17016 | 3DMark Score: 14779 | 3DMark Score: 23103 |
Graphics Score: 19463 | Graphics Score: 16759 | Graphics Score: 28900 | |
Physics Score: 17004 | Physics Score: 16785 | Physics Score: 19871 | |
Fire Strike Extreme (1440p, D11) | 3DMark Score: 10918 | 3DMark Score: 7541 | 3DMark Score: 13367 |
Graphics Score: 11404 | Graphics Score: 7746 | Graphics Score: 14247 | |
Physics Score: 17308 | Physics Score: 17610 | Physics Score: 19838 | |
Time Spy (1440p, D12) | 3DMark Score: 8967 | N/A | 3DMark Score: 10833 |
Graphics Score: 9775 | Graphics Score: 11828 | ||
CPU Score: 6019 | CPU Score: 7337 |
From the above synthetic benchmarks alone, the RTX 3060 Ti as an eGPU outperforms the RTX 2060 Max-Q dGPU by ~15% and ~45% on 1080p and 1440p gaming, respectively. Similarly, the RTX 3060 Ti loses about ~26% and ~18% performance on 1080p and 1400p gaming, respectively, as an eGPU over Thunderbolt 3.
Note: Outside of the 3DMark benchmarks, I never experienced any kind of performance issue in any game or application while having the iGPU, dGPU, and eGPU all simultaneously enabled in Device Manager. However, when I initially attempted to run the eGPU benchmarks, I experienced frequent, persistent stuttering in all of them, resulting in scores that were roughly half of those posted. I had to disable the dGPU to resolve the stuttering issue and achieve the posted eGPU scores. Disabling the iGPU alone, or both the iGPU and dGPU, achieves similar scores to those above.
Comments:
Performance in Destiny 2: Obviously, synthetic benchmarks don't always translate directly to real-world performance. The most prominent example of this that I've experienced with this setup is my performance in certain areas of Destiny 2, a shared-world first-person looter shooter which I play at 1440p max settings using the eGPU. In the Tower, a Destiny 2 "social area" populated with other players and NPCs, my FPS can fluctuate from ~30 FPS to >90 FPS depending on how many players are in my instance, or even within the same instance standing in the same position depending on where I look. Such drastic FPS dips within the game are an anomaly though, as my FPS mostly stays within the range of mid-50's to low-70's in PvP and within the range of mid-50's to >80 in PvE environments, whether or not I'm around few or many other players in either. As there have been complaints of low FPS in certain environments in Destiny 2 since the Beyond Light expansion, I'm not sure what of this to attribute to the game itself vs. TB3 bandwidth, latency associated with an external TB3 controller and routing through PCH, etc. However, the dGPU doesn't have such drastic FPS dips in these similar problem areas.
PCH Cooling: After reading about @itsage & @nando4's discovery about lack of PCH cooling (and the associated performance degradation) in certain 8th generation H-CPU hosts, I wanted to make sure that my 10th generation H-CPU model wasn't affected. Certain Razer Blade 15 Base models have been affected in the past. After opening mine, I'm happy to report that the early 2020 model (RZ09-0328) uses a thermal pad to dissipate heat from the PCH to the aluminum chassis.
|
|
@erwin_holzhauser, Awesome build and thank you so much for sharing desktop comparisons!
@erwin_holzhauser, Thank you for sharing this excellent build! Nice to see there's cooling pad on the PCH chip. Since you have a spare M.2 NVMe slot, you may want to explore M.2 eGPU adapter such as the ADT-Link R43SG. It's not a slick solution like Thunderbolt 3 but will have better performance.
LinkedIn | Youtube | Instagram
• external graphics card builds
• best laptops for external GPU
• eGPU enclosure buyer's guide
2021 Gigabyte Z590i Vision D [11th,6C] + RTX 3060 Ti @ 32Gbps-TB4 (Gigabyte Gaming Box) + Win11 [build link]
@itsage, I'm definitely planning to do that at some point. I almost wrote about it in the OP, but I haven't fully done my research on connecting a GPU via M.2, so I didn't want to write anything that was incorrect.
I also have a mid-2018 15" MacBook Pro with a i7-8850H and a GTX 1070. When I have time, I'll probably do some benchmarking on the RTX 3060 Ti with the MacBook to get some data on the performance increase from a more direct TB3-to-CPU connection.
I actually got the Blade 15 Base to replace said MacBook. I'm a software developer, and I didn't like having to constantly switch between macOS and Windows to work and play. Doing some research on Windows Subsystem for Linux was enough to entice me to fully make the jump into Windows for work and play.
@erwin_holzhauser, Hi Erwin, thanks for you post! I'm very interested in this exact setup, possibly the 17". Can you game with the laptop closed?
To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.
.@erwin_holzhauser, wow. It looks like you wasted a fairly substantial amount of money for very little performance gain. You actually even scored lower on one benchmark with the eGPU than your onboard dGPU. Would have been better off not buying a crappy, overpriced laptop in the first place and instead getting something with a 2080S from the factory. Likely would have outperformed your next-Gen eGPU.
To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.
.