The ASUS XG Mobile is a device that can be used with either a desktop graphics card setup or an eGPU enclosure, and we put it to the test against a Thunderbolt enclosure setup featuring a powerful desktop graphics card. The results were not what I expected.
In terms of gaming performance at lower 1080p resolutions, the XG Mobile was able to achieve average frame rates that were significantly higher than those achieved by the Thunderbolt setup. On average over all 12 games tested, the XG Mobile was reaching a 42% higher average frame rate than the Thunderbolt setup. This made sense, as the higher bandwidth available to the XG Mobile mattered more at lower resolutions. However, some of the results were unexpected, such as Red Dead Redemption 2 and Watch Dogs Legion, which did surprisingly well on the XG Mobile.
As we stepped up to the higher 1440p resolution, the Thunderbolt enclosure was finally able to offer an improvement in average frame rate, with the XG Mobile being around 24% slower than the desktop graphics card setup. However, this number is primarily held up by the top 2 or 3 games, and when we consider other resolutions, such as 4K, the XG Mobile's performance began to fall behind.
The problem is that for GPU heavy tasks outside of gaming, the Thunderbolt enclosure can offer better performance than the desktop graphics card setup. In a 3DMark test, which is a stress test designed to measure the capabilities of a system's graphics processing unit (GPU), the Thunderbolt system was able to use substantially more power and achieve better results than the desktop GPU. This is because the Thunderbolt enclosure allows for more efficient power delivery to the CPU.
However, when it comes to actual games, the XG Mobile often outperforms the Thunderbolt setup. In a test of content creator workloads, such as SPECviewperf, DaVinci Resolve, Adobe Premiere, and Adobe Photoshop, the XG Mobile was able to consistently score higher than the desktop graphics card setup. This is particularly true in Adobe Premiere, where the XG Mobile offered a significant advantage.
Despite the desktop graphics card being more powerful on paper, I do not recommend using a Thunderbolt enclosure setup with a Z13 device. The issues that arise from this setup are too numerous to ignore, and even if you're serious about 4K gaming, you'll still be dealing with poor 1% lows. Even in these situations, the XG Mobile can often deliver better performance than the desktop graphics card setup.
I believe that ASUS is not doing this to sell more XG Mobiles, as the device is extremely difficult to get your hands on. However, it's possible that ASUS could fix the remaining issues with the Z13 and a Thunderbolt eGPU setup. I've spent months troubleshooting these problems with them, but I'm just kind of over it at this point.
A member of my Discord server, who bought the Z13 and tried to use it with an RTX 3090 in the Aorus Gaming Box, had similar issues. They couldn't get it to work at all and ended up returning it. This matches my own experience, as I've encountered a number of random problems with Thunderbolt.
In contrast, the XG Mobile worked flawlessly in my testing. Therefore, based on this test, I would definitely recommend using an XG Mobile over a Thunderbolt setup. ASUS has now released both AMD and Nvidia configurations for the XG Mobile, so you can compare the performance of each.
"WEBVTTKind: captionsLanguage: enThere’s absolutely no way I can recommenda Thunderbolt eGPU setup with the ASUS FlowZ13.I’ve had 3 major problems with it, thisis what you need to know!This ROG Flow Z13 gaming tablet has Nvidia’sRTX 3050 Ti inside, but it’s also got themost potential GPU expansion out of any deviceso far.Not only does the Z13 have Thunderbolt 4 support,so we can attach a desktop graphics card overThunderbolt, but ASUS also offers their XGMobile.Basically any laptop with a Type-C thunderbolt4 port can use the desktop GPU enclosure,while ASUS’s XG Mobile currently only workswith their devices that have this custom port.Right now this includes their Z13, X13 andX16.Now a Thunderbolt external graphics enclosurelike this one obviously gives you more upgradeabilitybecause you can install whatever desktop cardyou like.I’ve got Nvidia’s original RTX 3080 inthis one.The XG Mobile I’ve got uses a laptop RTX3080 GPU, so it’s less powerful comparedto the desktop 3080 in terms of the raw specs.The mobile GPU in the XG Mobile has less CUDAcores and a lower power limit compared tothe desktop version, but it’s also muchsmaller and more portable as a result, thoughat the expense of less upgradeability as you’dhave to replace the whole XG Mobile unit ifyou want to change the GPU.Interestingly the XG Mobile 3080 has moreVRAM than the desktop card, however the desktopcard is using faster GDDR6X memory.One of the biggest downsides of a ThunderbolteGPU setup is that it's limited to 4 lanesof PCIe Gen 3 over Thunderbolt, and dependingon the device you’re connecting it to, thatsignal might first need to go via a Thunderboltcontroller that’s external to the processor.Modern Intel architectures have that builtin to the CPU though, which does give a performanceimprovement.But still, this is the main reason why a desktopgraphics card performs worse in a ThunderbolteGPU enclosure compared to just running itin a desktop PC.There’s just more overhead with Thunderbolt.The XG Mobile on the other hand uses 8 lanesof PCIe Gen 3, it’s simply a more directconnection with more bandwidth.Alright now here are the 3 major problemsthat I had while testing a Thunderbolt eGPUwith the Z13.The first major problem was any time I connectedmy Mantiz Saturn Pro II eGPU enclosure overhere, the Z13 would blue screen within abouta minute.Then it would just sit there in a boot loopuntil I unplugged the enclosure.Now I was able to work around this by usingDDU to uninstall the Nvidia drivers then connectthe enclosure and then reinstall the Nvidiadrivers, but that’s not exactly somethingpractical you’d want to do every time youconnect your eGPU to Z13.Now fortunately ASUS were able to replicatethis a couple of months ago or so and theyhave since rolled out an update that fixesthe problem through the MyASUS software.So as far as I’m aware, that issue is fixed.But the second issue I had was equally annoying.When we connect an external GPU via Thunderboltto the Z13, the Windows operating system seesit as a hardware change.It’s got a new GPU.Now with Windows 11, which is installed onthe Z13 by default, Microsoft heavily encouragesif not forces you to use an online Microsoftaccount, and when it does this it enablesBitlocker encryption by default for your security.There’s just one small problem though.Whenever Bitlocker sees that hardware hasbeen changed, it freaks out and doesn’tlet you boot the machine.So basically after I had that blue screenboot loop issue before, I was presented witha Bitlocker screen.This means I had to use a separate devicewith Internet access to log in to the Microsoftaccount to retrieve the 48-digit Bitlockerkey.What a seamless process!Don’t have a separate device with Internetconnectivity?Then I guess you’re not booting up the Z13.My original workaround for this was to disableBitlocker, but again ASUS have said that they’vefixed this problem in an update.Connecting a new device via Thunderbolt shouldn’tregister as a hardware change that triggersBitlocker.Ok so that’s great, but I’m really startingto feel like I’m the BETA tester for thisproduct.Especially when the third issue still existsin mid-May 2022.The Z13 is charged with 100 watts over Type-C,and it comes with an ASUS branded 100 wattType-C charger.Now my eGPU enclosure can offer 100 wattsof Type-C charging over this single Thunderboltcable, so I should be able to plug this onecable in and fully power the Z13 right?Wrong.Apparently the Z13 can only get the full 100watt charge with that included ASUS Type-Ccharger.I’ve been told that the reason for thisis that the power gets sent directly to themotherboard, which is meant to be more efficient.Others apparently send the power first viathe battery which acts as a bit of a buffer.So with this design, if a third party chargerhas any sort of spike in power delivery, itcould potentially damage the device.So to prevent a third party charger possiblykilling the Z13, it is limited to 65 wattswith third party chargers, which includesthe Thunderbolt enclosure.So then the question becomes, is 65 wattsof power delivery enough to give you fullperformance with the Z13?The answer seems to be probably not.If I just run a CPU workload like Cinebenchon loop, CPU package power seems to limitto a 30 watt TDP in manual mode, while turbomode is completely grayed out, which happenswhen the charger isn’t connected.If I instead connect the 100 watt ASUS chargerto the bottom Type-C port and leave the eGPUconnected to the Thunderbolt port above, theCPU package power runs higher at 50 watts.Which is also the same limit if we don’thave the eGPU connected.So basically to get the full power limit witha Thunderbolt enclosure connected to the Thunderboltport, we also need to connect the ASUS poweradapter to the bottom Type-C port at the sametime.But Jarrod Technologies, maybe your eGPU justsucks?ASUS thought that was a possibility too, andthey said that the Z13 worked fine on twoseparate enclosures based on their testing.I did actually reach out to the Mantiz guyswho make this GPU enclosure to see what theythink, and as far as they were aware theycouldn’t understand why it would be a problemwith their enclosure.I spent more than a month waiting for ASUSto provide me with one of their eGPU enclosuresthat they tested on the Z13 to confirm itfor myself, but it just didn’t end up happeningfor whatever reason.Eventually I reached out to Razer and askedif I could borrow the Core X Chroma, as it’sbasically a top of the line Thunderbolt eGPUenclosure and a bit more of a name brand comparedto the Mantiz enclosure.Sorry buddy…Anyway the exact same thing happens with theRazer enclosure.The CPU TDP gets limited to 30 watts withonly the eGPU connected and the ASUS 100 wattcharger needs to be connected at the sametime for it to run up to 50 watts.Ok, so just run the Z13 with two Type-C cablesconnected, problem solved right?Unfortunately not.Although connecting both Type-C cables doesimprove CPU only performance, game performancewas still all over the place.Let’s see how 12 different games performat 1080p, 1440p and 4K resolutions.All game testing was done with Razer’s CoreX Chroma with both Type-C cables connectedto the Z13 unless otherwise specified.Let’s start out with Control, yeah it’san older game now, but it’s one of the onlygames out of the 12 where things actuallylook normal and how I’d expect, so it’sa good place to start.I’ve tested using the Z13’s screen at1080p, and then with an external screen connecteddirectly to the eGPU, which should performa little better because the signal doesn’tneed to get sent back from the eGPU to theZ13 and processed by the Z13’s GPU.Anyway at the lower 1080p resolution, thehigher bandwidth from the XG Mobile givesus higher frame rates.At 1440p and 4K though, the Thunderbolt eGPUsetup is able to reach higher average FPS,and this is because the desktop 3080 GPU ismore powerful and it can get to work.That said, the dips in performance shown bythe 1% lows are extremely low with the Thunderboltsetup, possibly due to the lower bandwidth,or maybe the CPU power limit shenanigans.Alright now here’s Red Dead Redemption 2,and excuse me, but what the heck is goingon here?There’s clearly some sort of bottleneckfrom the Thunderbolt eGPU setup, because it’sonly reaching a little over 30 FPS regardlessof the resolution tested.I tested this game on both the Mantiz SaturnPro II and Razer Core X Chroma enclosuresand got the same results, so this is not aproblem with a specific Thunderbolt enclosure.For some reason the desktop graphics cardjust isn’t running properly, it’s runningat like 50% utilization with low power levels,but I can’t work out why.I get the same results whether I run the testwith one or both Type-C cables connected.I also tried a fresh Windows reset, but thisdidn’t change anything either.Just for comparison, last year the X13 withthe same desktop GPU was reaching 61 FPS inthis test at 4K, so yeah, there’s an unknownproblem here.Watch Dogs Legion was running like garbageon the Thunderbolt eGPU too.I mean, yeah sure if we lowered the settingsdown a bit or used DLSS it would perform abit better, but come on, the XG Mobile hasa massive lead at all resolutions.Even the 1% lows from the XG Mobile were farahead of the average FPS coming out of theThunderbolt setup at 1080p and 1440p, an embarrassingresult for Thunderbolt.The numbers coming out of Shadow of the TombRaider on the other hand look good.The desktop GPU is able to take the lead athigher resolutions as expected, but this gamedemonstrates yet another problem I had withthe Thunderbolt enclosure.It wasn’t even able to load the game upon the Z13’s screen, it just kept crashing,and that’s why there’s no data for it.Granted I think most people would probablyuse an external screen on a Thunderbolt setup,but regardless, not a good look.The XG Mobile is far more portable so it’sa better option if you’re taking it withyou to boost gaming performance on the gowith the Z13’s screen.The Witcher 3 is another older game, but isanother where the game crashed straight awaywhen loading it up on the Thunderbolt setupon the Z13’s screen, so at least 2 differentgames that are unplayable unless you’reusing an external monitor.Cyberpunk 2077 looks alright in terms of averageFPS, but the 1% lows coming out of the Thunderboltsetup are beyond a joke.The dips in performance and stuttering weremore obvious compared to the XG Mobile, despiteit having a lower powered laptop GPU.Sure the desktop GPU was giving a higher averageFPS at 1440p and 4K resolutions again, butit just wouldn’t be a great experience withthose dips.Especially considering how much money a setuplike this would costThis wasn’t the case for all games though,some like Dying Light 2 seemed to be fine.The 1% lows weren’t a problem here.The Thunderbolt setup was ahead in all regardsat the higher 1440p and 4K resolutions, whichis what I expected from all games going intothe testing.Assassin’s Creed Valhalla was almost alwaysdoing better on the XG Mobile.The Thunderbolt setup was only a couple ofFPS ahead at the highest 4K resolution, andit was only ahead for the average frame rate.The dips in performance, measured by the 1%lows, were lower on the Thunderbolt setup,meaning the XG Mobile is more consistent withless stuttering.Metro Exodus was strange.The XG Mobile setup lowered quite a bit at1080p when connecting an external screen.I’d love to retest it and confirm, but unfortunatelyASUS needed it back a few weeks ago so I wasn’table to double check.Due to this suspicious result, the Thunderboltenclosure was ahead at 1080p for one of thefew times.Microsoft Flight Simulator was generally behindon the Thunderbolt setup, unless you’redoing 4K gaming where it was reaching a 24%higher average frame rate, but otherwise at1080p and 1440p, the XG Mobile was doing better.God of War was reaching basically the sameaverage FPS at 1440p, but check out the differencesin 1% lows, the XG Mobile is simply offeringa more stable experience.Like many other games, the Thunderbolt setupwas able to come out on top at 4K.Forza Horizon 5 on the other hand was betteron the XG Mobile at all resolutions, and thedips in the 1% lows from the Thunderbolt setupwere pretty bad compared to what the XG Mobileis offering.On average over all 12 games tested, at thelower 1080p resolution the XG Mobile was reachinga 42% higher average frame rate.Only one game, Metro Exodus Enhanced, wasgiving higher average FPS with the Thunderboltsetup, but as mentioned I wasn’t able toretest that one.The XG Mobile winning at 1080p was my expectationgoing in, because the higher bandwidth availableto the XG Mobile matters more at lower resolutions.That said, some of the crazy results likeRed Dead Redemption 2 and Watch Dogs Legiondefinitely weren’t expected, something iswrong with the Thunderbolt setup there.Stepping up to the higher 1440p resolutionand the XG Mobile was now 24% faster comparedto the Thunderbolt plus desktop graphics cardsetup.We can see this number is primarily beingheld up by the top 2 or 3 games though.Then at the higher 4K resolution the Thunderboltenclosure was finally able to offer an improvement.The XG Mobile was now almost 8% slower interms of average FPS out of the 12 games tested.This doesn’t include 1% lows though becausemany games tested didn’t measure that, soit’s harder to directly compare, but aswe saw those could often be far worse withthe Thunderbolt setup.GPU heavy tasks outside of gaming are lessof an issue though, take 3DMark for example.The more powerful desktop graphics card inthe Thunderbolt enclosure was ahead regardlessof the test being run.In a GPU heavy stress test, the Thunderboltsystem was able to use substantially morepower, and that’s why these GPU heavy tasksare able to perform better with the desktopcard.It’s just that for whatever reason, whetherthat be PCIe bandwidth limits, power deliveryissues to the CPU, or some other problem thatthe XG Mobile often does better in actualgames.This wasn’t always the case though, especiallyin content creator workloads.SPECviewperf was generally faster on the desktopGPU, but as we can see this wasn’t alwaysthe case.Blender was faster on the desktop GPU in theMonster and Junkshop tests, but the Classroomtest was doing better with the XG Mobile.Maybe it can benefit more from the higherVRAM capacity, I’m just speculating, I don’tknow enough about these specific workloadsto say.DaVinci Resolve is generally quite GPU dependent,but the XG Mobile was able to consistentlyscore higher despite the Thunderbolt setuphaving the more powerful desktop graphicscard.The Thunderbolt setup has an even bigger defeatin Adobe Premiere, and then a smaller differencein Adobe Photoshop, but regardless it’sa win for the XG Mobile in these three creatorapplications.So despite the desktop card being more powerfulon paper, I just can’t recommend this setupto anyone looking to purchase a Z13.The only exception would be if you’re superserious about 4K gaming and maybe you’replaying games that don’t have poor 1% lows.But even then you’d still be dealing withall of the other negatives mentioned.I just don’t think it’s worth it for anyone.Are ASUS doing this to sell more XG Mobiles?Probably not.I mean they’re extremely difficult to actuallyget your hands on for one thing.Offering a device like this that can use eithereGPU setup is an extremely niche product.Is it possible that ASUS can fix the remainingissues with the Z13 and a Thunderbolt eGPUsetup?Absolutely, but at this point I’m not reallytoo confident.I’ve spent months of time troubleshootingthis stuff with them and I’m just kind ofover it, and I mean this is hardware thatyou can buy right now.The problems don’t seem to be just for meeither.A member of my Discord server mentioned thatthey bought the Z13 and tried to use it withan RTX 3090 in the Aorus Gaming Box, and theycouldn’t get it to work at all.I believe they ended up returning it, andthat matches my experience too, all sortsof random problems with Thunderbolt.Now that said, the XG Mobile from ASUS workedflawlessly in my own testing.So yeah long story short, I would definitelyrecommend that over a Thunderbolt solution.So now that we know that the XG Mobile isthe way to go, ASUS actually offer it in bothAMD and Nvidia configurations.Find out which is better in this video overhere next.I’ve compared AMD’s best RX 6850M XT againstNvidia’s RTX 3080, but again like this video,there are problems to be aware of, so I’llsee you in that one next!\n"
