5 reasons eGPU setups might be making a comeback

When first introduced to the market, external graphics cards or eGPUs were hailed as the savior of gaming on the go, able to increase the graphical power of the best laptops by offloading the processing to the eGPU. While this was a fantastic idea in theory, the practical applications and slower connectivity standards of the time conspired to make them lackluster at best, and the market went stagnant. Most of the eGPU enclosures available now were released years ago, and the manufacturers haven’t really been in a rush to release updated versions.

But technology goes on, and sometimes it’s a situation of the right idea but at the wrong time. Since the first eGPUs, connectivity standards have improved, both in terms of bandwidth and the protocols that can transmit over them, and USB4 brought Thunderbolt-like capabilities to many more devices. Add in the explosive popularity of PC gaming handhelds, and the conditions are right for a resurgence of eGPUs.

5

The rise of PC gaming handhelds

Handhelds are crying out for more power when in docked mode

While other technologies make eGPUs viable in theory, one section of the gaming market makes them viable commercially. One big reason eGPUs never quite took off is that they needed to be connected to laptops, and the vast majority of laptop users don’t care about gaming. They have laptops for work, they do work with them, and that’s the end of things. But the popularity of gaming handhelds means there is a captive audience for eGPUs. The Nintendo Switch famously gives you more performance when used in docked mode, and PC gaming handhelds have a variety of existing eGPU solutions to give a similar boost to graphical performance.

The ROG Ally used the ROG Xg Mobile interface to connect to the ROG Xg Mobile eGPU dock, which offered several options for GPU performance and added features like Ethernet, more USB ports, and HDMI and DP for external monitor use. The dock might have been overpriced, and Asus switched the port to USB4 instead with the ROG Ally X, but it showed that eGPU docks were viable for mobile gamers.

The Xg Mobile dock was one of the first things many of my friends got when they ordered their ROG Ally, and I can’t remember the last time anyone I know bought an eGPU for their laptop. Then there’s the crop of mini PCs and handhelds that use the newer OCuLinK standard, which promises multiple times the bandwidth of Thunderbolt 4.

The other common thread here is that every PC gaming handheld, from the Steam Deck to the Legion Go, has a desktop mode, so they can be used as computer replacements with the right dock. That dock might as well increase the graphical power while it’s increasing the number of viable ports for peripherals and monitors, giving eGPUs a new audience.

Related

I tried an OCuLink external GPU — is it actually better than USB4?

Could OCuLink rahert USB4 and Thunderbolt obsolete?

4

Thunderbolt 5 is here

Okay, it’s kind of here, but it’s available and brings lots of bandwidth

Early eGPU solutions used Thunderbolt, but the early versions of Thunderbolt were limited in bandwidth, limiting the effective power of the graphics cards installed in the enclosures. In 2017, we tested an Nvidia GTX 1080 inside a Thunderbolt 3 enclosure, and the performance was anywhere between 12% and 20% lower when installed in the enclosure. It’s not terrible, but it’s also not impressive, considering you had to spend $300 for the enclosure and whatever the cost of your graphics card was.

But Thunderbolt 5 is now here, and among the other things it promises, you can get up to 120Gbps of dynamic bandwidth in one direction. Perfect for high-performance storage options, sure. But it’s even more exciting for gaming use because it means more bandwidth for eGPU use. Not as much as plugging that GPU into a PCIe slot in a desktop PC, for sure, but enough for mid-range GPUs to shine. Flagships like the GeForce RTX 5090 and AMD RX 7900 XTX are going to need more bandwidth than even TB5 can supply, but maybe future versions will close that gap.

Related

3 reasons Thunderbolt 5 is making external GPUs more viable than ever

Thunderbolt is back and it’s better than ever, but could we see a resurgence in eGPUs?

3

Smaller and more powerful GPUs just arrived

Nvidia’s RTX 5000 FE models are small form factor ready

Over the last few generations of discrete graphics cards, the average size has trended up, leading to cards that almost hit a four-slot thickness. Most eGPU enclosures support a GPU that’s 2-slots width, at the most. I’m sure you can see what the issue here is, and it’s a big one because it limits the power of the GPU that you can even choose, before you get to the issue of performance drops from using an eGPU enclosure. The RTX 4090 FE is a three-slot solution (barely), but even that’s too much unless you’re making your own eGPU that doesn’t have an enclosure.

However, Nvidia’s GeForce RTX 50 series promises smaller, more powerful GPUs. At least, if you can get one of the Founders Edition models, because the company has said they’re two-slot and small form-factor ready. That means they’ll happily pop into the eGPU enclosures on the market, although you might need to upgrade the PSU that they come with, depending on which RTX 50 series card you’re looking at.

Related

Don’t forget these 5 things before building a small form factor PC

An SFF PC can be an exciting project, but keep these things in mind before you start building one

2

The rise of soldered-on components

Laptops solder on components to save space but then you run out of performance

Whether it’s to make slimline laptops, decrease RAM latency by putting the memory chips on the same PCB (or in some cases, on the CPU), or to save space in PC gaming handhelds or mini PCs, soldered components are everywhere. That means CPUs, memory, storage, and GPUs on the same mainboard, with little or no option to upgrade when new technologies come along or if you run out of storage space.

Which is fine, really, as you’re trading future upgrades for actually being able to carry your laptop around all day. But when you’re back at home, the original promise of eGPUs can come into play, giving you more GPU grunt, optional speedy storage space, and more ports that wouldn’t fit on the thin edge of your laptop. It won’t let you plug in more RAM, but even the base model MacBooks come with 16GB of memory now, so you should be okay for some time on that front.

Related

4 reasons why repairability should be the next big thing for laptops

Easy-to-repair laptops need to become the norm

1

They weren’t fast enough before

Bandwidth issues from earlier connectivity options made them subpar

When the Nvidia GeForce RTX 50 series hits shelves later this month, we’ll have the first PCIe 5.0 graphics cards (finally!). But that means an increased amount of bandwidth required for peak performance, and that’s just not possible with any of the current cable-connected connectivity standards. Thunderbolt 5, OCuLink, and USB4 aren’t up to the task of comparing them to a physical PCIe slot on a motherboard, and neither were any of their predecessors.

The current crop of GPUs use PCIe 4.0 x16, which gives a theoretical maximum of 31.5GB/s of bandwidth, roughly eight times more bandwidth than Thunderbolt 4 can supply. Thunderbolt 5 closes that gap somewhat with three times the bandwidth of Thunderbolt 4 when the dynamic mode is enabled, but it’s still not quite at PCIe 4.0 x16 levels.

OCuLink gets closer, with the first iteration bringing a theoretical 3.9GB/s and then a revision to 16GB/s, and OCuLink 2 gets all the way to 32GB/s of bandwidth in every direction when implemented as a four-lane implementation. That’s just as much as an internal PCIe 4.0 x16 slot, making even flagship GPUs viable. The only question is how much performance is lost with overhead as the conversion to over-cable PCIe happens. That’s a question that doesn’t currently have an answer.

eGPUs are ready for a resurgence

Technological advances in connectivity and improvements to the protocols used in DisplayPort over cables make external GPUs more relevant than they ever have been. Plus, Thunderbolt 5 increases charging capacity to 240W over the same cable, making almost any laptop able to charge while docked. Add the runaway success of PC gaming handhelds and mini PCs, and you’ve got the right conditions for eGPUs to thrive.