2019 13" HP Spectre X360 OLED [10th,4C,G] + RX 5700 XT @ 32Gbps-TB3 (Mantiz Saturn Pro) + Win10 2004 & Linux Pop!_OS 20.04 [itsage]
In my search for more Ice Lake ultrabook with Thunderbolt 3 ports, I stumbled upon this HP Spectre X360. It's a convertible with a 13-in touchscreen OLED display. I do not like 2-in-1 or touchscreen on a laptop, but the OLED screen was very appealing. This laptop has great build quality and I like the matte black and gold accent. There are two Thunderbolt 3 ports, one of which sits at an angle. The space in between the two TB3 ports contains a tiny indicator dot for USB-C/TB3 power state.
2019 13" HP Spectre X360 OLED – i7-1065G7/Iris Plus Graphics iGPU/16GB RAM/512GB SSD
Installation in Windows 10 2004 was straight forward. The system detected the Mantiz Saturn Pro + RX 5700 XT following a hot-plug. Rather than relying on Windows to install a much older set of graphics drivers, I manually download and install Radeon 20.7.2. It seems most Ice Lake laptops have Thunderbolt 3 security set the same way as Mac computers, no approval of Thunderbolt devices needed. The only issue I encountered is error 12 through hot-plug. The Radeon eGPU has to be cold-plugged in order to get around error 12.
Installation in Linux [Pop!_OS 20.04] was nearly identical. Thunderbolt detection happened immediate. There are AMD drivers but it didn't support the RX 5700 XT. I upgrade to the latest Oibaf PPA and the RX 5700 XT eGPU fired right up. It requires cold-plug for the eGPU to function as well. I have to say I enjoy using Pop!_OS a lot in the past couple years. Power management is great with a drop-down Menu item to set power and graphics mode. Even my bulky and power hungry Alienware 13R3 could get near 10 hour battery life with web use.
HP included a utility software called, HP Command Center. Of course I wiped the drive clean as soon as I opened the box and apparently there's no way to manually install this utility. I've seen screenshot that HP Command Center allows switching of different power modes. Lacking this choice I resorted to Max Performance through Windows Power settings.
Also interesting is when I boot into Windows with the eGPU connected, the boot screen shows an OMEN logo. By default (without eGPU), the boot screen shows an HP logo. Linux boots so fast I only saw a gray screen then it's ready for me to log in. 😀
|Loopback Mode||FHD Monitor||QHD Monitor||UHD Monitor|
There's overheating issues and over all strange behavior when I ran different benchmarks. For example, Superposition stuck at 40 FPS regardless of settings and resolutions. It's sad too many manufacturers focus on "the look and feel" but completely ignore how their products function. This ultrabook checks a lot of boxes on paper but it's a hot mess in real life.
@itsage, @burninglutz on page 71 on this thread https://egpu.io/forums/pc-setup/egpu-fail-after-update-to-windows-19-03/ just stated his experience after the latest 2004 update that the error 12 has been resolved for his mac. Can you confirm what he's saying to be true? Is coldplugging the new meta to get around E12? Because if that's the case, then maybe there should be a sticky that reviews this latest update and the current workaround for error 12 for official windows updates (and possibly remove the banner with the (!) thread?).
This would be the greatest news since build 1903.295 😯
@aurelius_pontius, There's something not right with the way HWiNFO64 read this system. You could see it identified the Thunderbolt 3 connection as PCI Express version 1.1. AIDA64 GPGPU benchmark confirmed otherwise. The eGPU was able to post full Thunderbolt 3 bandwidth of roughly 22Gbps which is through a 4 lane PCIe 3.0 connection.
@test1 Different versions of Win10 2004 and error 12 behavior are inconsistent between system + eGPU combos. I've been testing myself and keeping an eye on reports from members. It's not a sure fix yet imo. We're seeing error topics almost everyday. In this very build, I encountered error 12 through hot-plug on Win10 2004.
@itsage, All Thunderbolt 3 downstream bridges and devices (NHI, USB) report PCIe 1.0 speed (2.5 GT/s). This has always been true (whether Alpine Ridge, Titan Ridge, or even Ice Lake). Only the upstream bridge or real PCIe devices connected to one of the downstream bridges have real link rate and link width values. That's ok because those downstream devices are internal to the Thunderbolt controller. The only values that matter are the other PCIe devices connected to the Thunderbolt controller (other downstream or upstream).
In the following list, the first Thunderbolt 3 bridge is the upstream Thunderbolt bridge. It is connected to a PCIe slot (g3x8) so it must have real link rate (g3 = PCIe 3.0 = 8 GT/s) and link width (x4) values (g3x4). No thunderbolt device is shown in this example. For an example with many Thunderbolt devices connected, see the text file attachment at #417
Ice Lake is weird because it uses integrated Thunderbolt with no real upstream Thunderbolt bridge - so all the Thunderbolt stuff says 2.5 GT/s x4 just like the downstream bridges and devices of Alpine Ridge and Titan Ridge even though they all operate faster than that.
@itsage, which reminds me, did you ever get around to getting pcitree.sh to work on your Ice Lake MacBook Pro?
@joevt, Thank you for the insight as always. I unfortunately couldn't get it to run your pcitree.sh on the Ice Lake. I'm waiting for an i7 unit then give it another go.
@itsage, I added download and install instructions at