- MacBook Pro (Retina, 15-inch, 2018) - MacBookPro 15,1
- CPU: 2.6 GHz Intel Core i7 (i7-8850H)
- GPUs: Intel UHD Graphics 630 + AMD Radeon Pro 560X
- Memory: 32GB
- HD: 1TB SSD
- macOS 10.14.6 Mojave
- Disconnect eGPU
- Start the MacBook Pro with Windows 10
- Hot-plugging eGPU
- Log into the Windows 10 – no errors and everything works fine
I was skeptical about the setup because I couldn't find any real case. So I brought a used Razer Core X to reduce my risk of possible financial loss. Everything seems really smooth so far. When training the Convolutional Neural Network model, the fans of the Titan RTX were running in their full capacity. However, surprisingly, they are not as aloud as I expected 🙂
Images of the setup
@airmc2018 What a beautiful setup! Thank you for sharing.
Thank you for your comments.
I have been benefited a lot from this website before I built up my eGPU. Hopefully, my brief guide can help clear up some doubts of building a portable ML with Titan RTX
I would love to share more if I can get the Titan RTX to work on the macOS.
Hello! This is exactly the setup I was looking for. Do you happen to have benchmarks to compare against a Titan RTX being used in a different setup (i.e., a more traditional setup such as a Windows desktop or workstation)? Many thanks.
I did not install the Titan RTX to other setups. However, I have a Lenovo P410 workstation with Xeon 3.5Ghz CPU, 64GB RAM, 512GB SSD, and Quadra M4000 8GB. When I was training the convolutional neural network on both machines, the MacBook Pro setup is 4X faster than the workstation.
Hopefully, such info will be helpful to you.
How well does that enclosure do in terms of cooling? Also, do you know if it would be possible to add a second Titan RTX in the same enclosure in the future and it it would be possible to use the link bridge if you did that?
I have a much older MacBook Pro, but was thinking about getting a new Linux laptop and an eGPU to do some more work with DL/NLP applications.
The cooling is great in term of running intensive CNN training. You will feel warm/hot air blowing out of the enclosure as the fan speed gets higher.
For the second Titan RTX, I don't think you will be able to add to the enclosure. This is primarily limited by the rating of the power supply as well as space inside.
Hi @airmc2018 I am looking for an enclosure for the TITAN RTX. I saw in the specs, it is required a power supply of at least 650W and the Razer Core X only provides 500W to the GPU. Didn't you have an issue with these specs?
Hi @jorge_alejandro_pena_munoz, I believe that 650W rating is for the entire system including CPU, motherboard, ram, HD, and etc, because those requires additional power.
The actual maximum power (Thermal Design Power) drew by Titan RTX is about 280W and the Razer Core X is rated 500W that is enough for the card. So far I don't have any problems for training deep neural networks using the enclosure.
@airmc2018, Hi, After reading this post, I also ordering eGPU enclosure and Titan RTX to try. I do have same MacbookPro model with bootcamp Windows setup. Does it requires additional software setup to install tensorflow-gpu or is it standard conda install tensorflow-gpu command ?