Phew! After eight months of hosting my sister-in-law’s family, it’s hard to believe they are moving out of our basement and into their new home this week. When dealing with a home invasion of such magnitude, I think most people would find a coping mechanism of some kind. It turned out that mine was chronicling the adventure in the form of a three-way router review. In all seriousness, it’s been great to have so much family so close by, but having that many people under one roof did put a lot of pressure on our already-a-bit-more-taxed-than-average home network.
After the first few weeks, it became exceedingly apparent that my trusty Asus RT-AC66U, purchased in late 2013, wasn’t cut out for the new demands we’d placed upon it. I did the best I could to keep it in the game. It was already running the Merlin firmware to resolve a DDNS-related WAN disconnect problem with the stock firmware. I also disabled the RT-AC66U’s QoS so that NAT acceleration could do its thing. Otherwise, one person downloading from Steam would kill everyone else’s connection.
Despite those measures, the RT-AC66U was swamped. I started looking into the many new wireless options on the market, trying to decide what would be the best fit for a household packed with 12 bandwidth hogs, surprisingly linearly distributed from ages 1 to 36.
Did I simply need the modern equivalent to my RT-AC66U? Or did that many people in one house mean that something a bit more souped-up was called for? Maybe a mesh Wi-Fi solution would spread the load and yield the best result. Choosing the way to go wasn’t an easy decision, and the longer I waited to get something new, the grumpier everyone was getting.
Thankfully, just like it did with the TR BBQ a couple years ago, Asus was willing to help out. The company sent me a router to match up with each question above, so I could figure out the answers myself and share them with you. It wasn’t long before I took delivery of the Asus Blue Cave, Lyra Trio, and ROG Rapture GT-AC5300 routers.
Everybody’s going through the Rapture
Let’s kick things off with the eight-legged elephant in the room: the ROG Rapture GT-AC5300. I could say a lot about its aggressive styling, but you’d be better served by heading over to Amazon and reading the hilarious Q & A banter about it over there. Honestly, once the Rapture was perched in place up on the shelf by my TV, it didn’t look nearly as strange as the top-down isometric press photos. In fact, the tiny white indicator LEDs on the front are more subdued than any of the other routers’ lighting packages.
This router really ties the room together.
Under the hood, the word subdued does not apply. The router sports a Broadcom BCM4908 quad-core processor running at 1.8 GHz. Obvious architectural differences aside, that’s the same number of cores and 75% of the clock speed of a Core2 Quad Q6600 from days gone by. Let that sink in. Coupled with 1 GB of RAM and a Broadcom BCM4366E CPU dedicated to each of the Rapture’s three wireless bands (one 2.4 GHz and two 5 GHz), we’re talking about a serious effort to shut up anyone that blames net lag for being terrible at online gaming.
Making room for all that hardware and mounts for the Rapture’s eight antennas means the router itself is quite large, bigger even than some mini-ITX PCs I’ve built. Asus takes advantage of the space by filling out one side of the unit with eight Gigabit Ethernet ports instead of the usual four, as well as a pair of USB 3.0 ports for storage and printer sharing instead of just one. Two of the Ethernet ports are described as “gaming ports” and traffic running through them gets extra-special QoS treatment.
On the software side… wow. The Rapture throws in everything except the virtual kitchen sink (and, of course, the interface is official gamer black and red). We’ll go over the highlights. The built-in Trend Micro security suite can block all sorts of internet badness as an up-front layer of protection to all the devices on your network. You might think you’ve transcended such frivolities, but we all know how mothers everywhere feel about layers, so don’t write it off just yet. In a similar vein, parental controls and filters—not to mention per-device internet access scheduling—are also available.
Why not take care of these upfront if you can?
The Game Boost menu is where the QoS details can be configured. There’s a wizard for setting overall priorities for gaming, streaming, and surfing, or you can rank activities in your own order. If you enable Game Boost, you can even drag and drop all the devices connected to your network in exactly the order you want their traffic priority to be. From this menu you can also see the real-time upload and download bandwidth for each device and what protocols it’s using.
We’ll talk about the test that created this load shortly.
There’s a bit more gaming-centric functionality to cover. The Rapture includes support and a lifetime subscription for the WTFast Gamers Private Network, a VPN service that attempts to route traffic through optimal servers for your location. I’ll admit upfront that I’m just not sensitive enough to lag to make this a compelling feature to me. With so much else to test, WTFast didn’t make the cut, but it might excite the aspiring professional gamer in your household. The Game Profile menu contains a list of pre-baked profiles for setting up port forwarding (powered by portforwarding.org, no less) but I didn’t use it since I manually configured port forwarding in a different menu. Lastly, Games Radar lets you select from a short list of games and see what the ping time is to their various servers around the globe.
Moving on to the more general-purpose menus and settings, there’s a sophisticated traffic analyzer that can be interesting to dig into after you’ve had it running for a couple weeks. I didn’t turn on any history for web-tracking, but the analyzer was still able to give me a general idea of what devices in the house were using the most bandwidth and what services they were using to do it. Holy moly, do people ever love YouTube. Who knew?
Move-out week is a slow week, but you get the idea.
From here we get into the utilitarian stuff like basic LAN and WAN configuration, DHCP settings, wireless configuration, the firewall, guest networks, port forwarding, and options for the use of the USB ports. You know the drill. The only exceptions from the mundane are the recently added Asus AiMesh support and the Amazon Alexa and IFTTT support. We’ll talk about the Alexa support a bit later, but AiMesh will have to wait for another article.
A trip into the Blue Cave
Asus’ Blue Cave looks like something you might get if the Portal gun could dispense miniature front-loading washing machines. I really dig the aesthetic. I’m surprised by how much I appreciate the complete lack of active indicator lights in favor of just a glowing blue ring (for which you can adjust the brightness or turn off entirely). The Blue Cave just sits stoically on the shelf and does its job.
I’m a sucker for how this thing looks.
To do that job, the Blue Cave employs an Intel AnyWAN GRX350 SoC. In the web interface, the SoC appears as a three-core CPU paired with 512MB of RAM, but the chip that powers it is advertised as just a dual-core affair. Regardless, it’s clear that the brains of the Blue Cave are not just a variant from one of the usual suspects in the market. You’ll find four Gigabit Ethernet ports and one USB 3.0 port hooked up to the semi-mysterious guts of the router.
The Blue Cave has a similar compliment of software features compared to the ROG Rapture. However, as clearly indicated by the change of color scheme, it drops some of the gaming-specific ones, or at least renames them. The Trend Micro security system, the basic QoS implementation, and the nifty traffic analyzer all stick around. There’s also Amazon Alexa support and AiMesh support was added in a recent firmware update. Of course, all the fundamentals are in place too.
Load testing hits the Blue Cave hard too.
Meshing up the Lyra Trio
Asus may have broken the mold with the Blue Cave, but they re-used an old one for the Lyra Trio. The design of the Lyra Trio nodes was clearly lifted from the Asus EA-N66 (just like how I remember some of my 80s Dino-Riders toys magically becoming Jurassic Park toys in the 90s). The diminutive stature of Trio nodes means that there’s only room for two Gigabit Ethernet ports on each one, and no room for USB at all. Any Trio node can be used as the primary node that connects to the WAN interface. The other nodes can use their ports to connect other devices to the network or for wired backhaul.
Crammed inside each of the tiny pyramids is a Qualcomm QCA9563 SoC. The QCA9563 is a single-core affair that runs at 750-775 MHz and has access to 128 MB of RAM. Clearly, these little guys are working with different constraints than the Rapture and the Blue Cave. The features available in the web interface are similarly constrained. The traffic analyzer I’ve became a fan of on the other routers is gone. With no USB ports on the nodes, there’s obviously no mention of attached or cloud storage. However, the Lyra Trio still supports Alexa integration and the lifetime Trend Micro security offering.
Ignorance is bliss?
The Lyra Trio advertises PC-free setup, thanks to the Lyra app for iOS and Android. I stuck with the traditional through-the-web-interface setup process, though. I didn’t feel like I was missing out on anything critical, but if you like apps, it’s an option. The web interface doesn’t provide information about CPU usage or allow you to have separate SSIDs for your 2.4-GHz and 5-GHz networks, although that may be an artifact of band-steering being on by default in the Lyra’s interface. It could also be because one of the 5 GHz channels is already dedicated to communicating between the nodes.
Merlin brought this router a lot of time.
For reference, here’s some specs from my old router. It’s running a pokey 600 MHz Broadcom BCM4706 CPU and has 256 MB of RAM. The USB port on the back is just the 2.0 variety. As I mentioned previously, it’s running the Merlin firmware, version 380.70, the latest version this model will ever see. Thanks to the Merlin firmware, this old router had many of the features of the newer models and, despite its age, was in the best shape of its life during testing.
All of the latest Asus routers boast support for Amazon Alexa integration. I wasn’t particularly enamored with the options afforded by it. Call me a Luddite, but I’m just not ready for a world where voice commands can kick off a firmware update or pause all internet traffic. Now, if I could group someone’s devices under a friendly name and say, “Alexa, shut so and so off from the internet” then maybe I’d be more impressed. The Blue Cave and ROG Rapture do have IFTTT support, though, which seems a lot more practical even if I didn’t have reason to make use of it.
The supporting cast
Since router testing isn’t our bread and butter at TR, I don’t have a ready-made table to drop down full of specs to remind you what our standard test system is. Without further ado, the most important introduction is to my faithful laptop, a Clevo W110ER that has proven itself a great platform over the years. The only replaceable parts in the system that are the same as when I first put the barebone kit together back in early 2012 are the i7-3610QM CPU and the memory. That said, the Intel Wireless-AC 7260 it’s rocking dates to when I got the router that these routers are taking the place of. So, it’ll be interesting to see just how much, if at all, the RT-AC66U was holding back its contemporary Wi-Fi adapter.
Not wanting to only test a Wi-Fi solution that was five-years old, I picked up an Edimax AC1200-Nano to run duplicate testing on the routers. The Edimax’s combination of popularity, user reviews, price and features (including MU-MIMO) made it a logical choice. I wanted to see if it was any good too. So, this router review ends up throwing in a review of the tiny adapter as a freebie. I used the AC1200-Nano in my W110ER for testing, and I disabled whichever adapter wasn’t being used in Device Manager during tests.
Behold my mighty testing rig!
For my wired testing, I used my primary desktop (which you can read about here if you’d like). Suffice it to say that it’s a modern system powered by an i7-7700K with an Intel I219-V driven NIC. The other desktop that was involved in testing was the i7-2600K system that my current system replaced. The Sandy Bridge desktop and its Realtek 8111E based NIC performed the role of iperf3 server for all the tests. I paused Windows Update for the duration of testing, so all the PCs involved were running the Fall Creators Update with all the weekly updates that came out before the April Update during testing.
Finally, the last piece of the puzzle is what I used to test the USB attached storage afforded by most of the routers. I plugged a spare Samsung 840 Pro 120-GB SSD into a StarTech USB3S2SAT3CB adapter and mapped the resulting share as a network drive on my test systems.
StarTech/840 Pro results direct from a USB 3.0 connection.
The lay of the land
When I did my previous review of Asus networking gear, I made a few notes about what to do differently next time. Leading up to this review, I solicited advice from our readers about extra things to test. The new testing environment and multiplicity of routers affords new opportunities and imposes new demands on testing as well. It also restricts some tests. Finally, since I’m operating out of my home base instead of a remote location this time, I wanted to establish a test “suite” I could potentially reuse in the future with different hardware. All that said, here’s a summary of my testing process.
- Internet speed testing from seven locations
- iperf3 speed tests to a wired PC with SSD storage from seven locations
- CrystalDiskMark testing to USB-attached SSD storage from seven locations
- Steam In-Home Streaming and PS4 Pro-to-Vita Remote Play from two locations
- Extensive note-taking of subjective impressions with QoS and other features enabled
The locations that tests were performed in were selected because of their specific combination of obstacles and distance between them and the router. The list of friendly names below is your legend for interpreting the test results. I’ve ordered them roughly according to their difficulty to get a signal to. Distances are approximate.
- Comfy Chair – router is in the same room with direct line of sight to the chair about ten feet away
- Stacy’s Desk – router is in the same general space, but separated by various obstacles and about 15 feet away
- Wren’s Room – router is on the same floor, one wall and about 25 feet away
- Ellie’s Room – router is on the same floor, two walls and about 40 feet away
- Laura’s Studio – router is on the floor above, diagonally separated through the floor and a wall by about 25 feet
- Cat Room – router is a floor above, diagonally separated by two walls, concrete, and stairs by about 35 feet
- Movie Shed – an outbuilding diagonally separated from the router by three walls and a couple hundred feet
Hopefully, the logic behind those descriptions speaks for itself. Some of the locations should be a cakewalk, while others were known to be challenging for my existing router. In fact, the Movie Shed location started out testing as just “Shed” but—spoilers!—some of the new routers worked so much better at range than my old one, I was able to find a new use for an otherwise-neglected space. Oh, and in case you’re wondering why I’d care about the quality of the Wi-Fi signal where the cats’ litter is, it’s also my basement workshop space.
I would be remiss not to mention that the primary Lyra Trio node was located exactly where the other routers were placed during testing: centrally located in the upper floor of the house. The extra nodes were placed as far to the north and south of the primary node as I could put them while still staying indoors. That meant one of the nodes was actually in Ellie’s room, against an exterior wall. The other was in the laundry room, directly above the cat room location and against the exterior wall closest to the movie shed. Keep an eye on the results for those locations from the Trio.
Internet speed testing
Before we dive into my test results, I want to note that wireless network testing in the real world is, by definition, unscientific. I’m the first to admit that my testing environment is a far cry from properly controlled lab conditions. Everybody’s surroundings and RF-interference conditions will be different, and my house is no exception. For each test I’m reporting what I’ll call the “average-mode”—which is to say that I’m looking for similar numbers that repeat frequently, and then averaging multiple instances. This should help clean up the variance in my messy testing environment while still gauging things fairly consistency.
Additionally, to account for the time that passed between running a series of tests, I ran the prior test sequence again to check if the results were similar before continuing. Finally, I ran testing again any time I noticed an outlier, because there are a couple unexpected results that needed to be triple-checked. Did I mention this review has been six months in the making?
I’m also not chasing peak theoretical numbers or results you can precisely duplicate. I’m throwing each router into the deep end to see if it will sink or swim in a tortuous household with about 40 devices connected to the network. I avoided obvious pitfalls, like not running my internet speed test while someone was downloading a game from Steam, but there’s an element of uncontrolledness at work nonetheless.
I can say with confidence that my various tests were the primary source of traffic on the network while they were running, but that’s not an assurance that they were the only source. As you’ll read, simultaneous benchmarking and real-world usage did bring to light weakness I wouldn’t have known about otherwise. All tests were performed with the Trend Micro network protection features turned on and QoS set to prioritize media streaming.
Now, let’s look at the first set of numbers. Data was obtained from the Speakeasy speed test. I always pointed this test to Chicago servers.
There’re a few trends that start right away and will persist through the other tests, so we’ll address them first. The older Intel Wi-Fi card in my laptop typically outperforms the Edimax USB adapter. This is especially true for 2.4 GHz performance and as we progress to the more challenging locations. By the time we get to the cat room and movie shed, the Edimax can’t connect to my old router at all in some cases. In others, it’s too slow to be usable.
Something else to note is that even in the best-case scenarios the 2.4 GHz network can’t deliver my internet connection’s full 100 Mbps speed. That’s with the adapter reporting a 144 Mbps connection. Would a better adapter get us closer? Maybe. The moral of the story is that, except for fringe cases, you want to make sure you’re on 5 GHz. Not exactly a huge surprise, but there it is. (Editor-in-Chief Jeff Kampman, who is also a stickler for solid wireless connections, totally agrees on this point.)
Another thing you may have noticed is that the ROG Rapture’s “#2” 5 GHz connection regularly beats the other 5 GHz channel, at least on the Intel side. The SSID for that extra channel has a password I didn’t share with anyone during testing, so no other devices were connected to it except for my test laptop. Even without significant active usage, it seems like a less burdened channel can perform measurably better.
Looking at the performance overall, it’s clear that my old router is truly outclassed at some of the more extreme distances. The Lyra Trio is showing signs of weakness at range, as well—the opposite of what you’d expect. It clearly has some disagreements with the Edimax adapter compared to the Intel one, too. I also found it odd that the Lyra Trio couldn’t deliver the full speed of my internet connection even to my wired desktop.
Ultimately, the clear winner of our hardest test is the Intel card connected to the second 5 GHz channel of the ROG Rapture. Interestingly, the Lyra Trio, Blue Cave, and the first 5 GHz channel on the Rapture clock in at very similar speeds at the same range.
iperf3 LAN speed testing
Let’s see how iperf3 testing lets the routers further distinguish themselves. I ran my tests with the parameters below.
iperf3.exe -c [server IP address] -f M -i 2 -t 10 –logfile [test name].rtf
Not capped by my internet connection, the routers get a chance to stretch their legs a bit. However, at best, they are only able to attain wireless speeds of about one third that of the wired testing. So much for the 867 Mbps connection that Windows claims, but this is Wi-Fi. I don’t think anyone expected anything different.
As you flip to the various testing locations, we can see a more significant impact on throughput than we were able to with the internet speed tests. All the routers start to lose speed as soon as they lose line of sight. The Lyra Trio still has curious results. It seems to fare worse than its single-device counterparts, even when the test is performed within just a few feet of one of the auxiliary nodes. The Lyra Trio and the Edimax continue to have a love-hate relationship, as well.
Looking all the way out to the cat room and movie shed, you can see just how far the numbers fall from their peaks. The Rapture and Blue Cave are the only ones that remain tolerable if you think of these as file copy times, which is important if you want to play a .mkv from your local network instead of stream a movie from Netflix or Amazon.
Attached storage performance with CrystalDiskMark
CrystalDiskMark is primarily part of my testing because I was curious about how much the USB ports on each router would bottleneck the storage device you plugged into them. In my experience, it can be dramatic. It’s also interesting to run on network drives because the multiple queue depths involved, 4KiB madness, various controllers, the drive itself, and our wireless signal quality pretty much make it a torture test. This is going to be fun.
Whoa! Score one for the Lyra Trio, right? Not exactly. The Lyra Trio is cheating here. It doesn’t have a USB port, so those test results are from the Startech SATA adapter and Samsung 840 Pro SSD plugged into the same PC that served as the iperf3 server (an i7-2600K system, in case you forgot). Throughout these results, think of the Lyra Trio’s numbers as an example of why you may want to consider forgoing the USB ports on your router and mapping a network drive to an always-on PC in your home instead, if you have one. It’s also worth noting these results are at QD32, an uncommon condition in client storage use. QD32 will let NAND flash devices show off their very best performance.
Even so, these sequential transfer speeds are probably my favorite numbers to come out of my testing. Not only do you get to see the impact of signal quality on transfer rates, but it becomes very clear which routers have the muscle to keep up with processing lots of data. The wired read results for the ROG Rapture destroy everything save the cheating Lyra Trio. The Rapture’s write speed is half as quick as its read speed, though.
The wired results and the comfy chair results make it clear the that the Rapture is in a league of its own when it comes to quasi-NAS duty. Meanwhile, the Lyra Trio and many of the Edimax tests continue to be wonky and get downright ugly in the more distant locations.
Let’s kick things up a notch with some random performance tests.
The first thing you’re likely to notice about these numbers is that my old router and the Blue Cave are remarkably consistent until you flip all the way out to the cat room and movie shed. That suggests we’re hitting a processing bottleneck, not a bandwidth one. On the other hand, the Rapture’s CPU has time to kill while it waits for more packets to devour.
Other than that, we’re seeing patterns we’ve already seen above: the inexplicable fondness followed by fallout between the Lyra Trio and the Edimax adapter, and the general dominance of the Intel adapter over the Edimax despite the latter’s MU-MIMO chops.
Not much to say here. The results of the random Q8T8 results are quite like the Q32T1 results.
QD1 storage testing is the most typical use case for any storage device, and QD1 random performance really separates the wheat from the chaff. Not only are we looking at the performance characteristics of the underlying storage device, we’re also likely to be showing off the per-core performance of each router CPU in our wired tests. If you’ve ever run this test on a vintage SSD or even a hard drive (you monster), you’re probably impressed by how well the wired ROG Rapture holds up. Even the cheating Lyra Trio backed by its PC-attached SSD is crushed by the relentless load. Silly as it may be, I’ve got a feeling this test could be a good indicator of which routers can hold up under the most duress, in general.
Next up, some real-world benchmarks. I tested two forms of game streaming with our routers: Steam in-home streaming and Sony’s Remote Play streaming. For the Remote Play streaming, I used my dusty PS4 Pro and a used PS Vita 2000 that I picked up specifically to perform this testing. On the Steam side, I streamed my perennial favorite, DayZ. The crazy amount of leaves and grass in the game are a compression artifact’s best friend, making it a theoretical challenge to stream well. Since it’s a game I have a lot of time invested in, I knew I would notice a problem if one showed up. For the Remote Play side of things, I picked up Rocket League on the PS4. It’s another game I’m well acquainted with, and I was stoked by the prospect of being able to play on the Vita.
The host and client settings I used for Steam streaming are shown below. I’m not sure when the “prioritize network traffic” box got added, but of course I had to check it. I left the bandwidth limit to automatic to see if any of the routers would allow the client to auto-negotiate different rates with the host. The W110ER that I used as my client only has a 1366×768 screen, but I verified using a 2015 model Asus Zenbook UX305 that the same 1920×1080 stream’s worth of data was streaming to the smaller screen before it was downscaled on that client. The host PC was my primary i7-7700K desktop with its GTX 980 Ti.
Now, I’ve played around with Steam streaming since it first came out, but I’ve never really used it for more than just fooling around. I have to say, though, that it’s come a long way since the last time I checked it out. Overall, it works splendidly, with none of the glitches or gotchas that I remember from a few years back. Specifically, the “display performance information” option is greatly appreciated, especially within the context of this review. You can see below that all kinds of gerbil-approved numbers are on-screen. Input latency, display latency, ping time, incoming and outgoing bitrates… and there’s even a real-time frame-time graph. It’s beautiful.
When it comes to the performance of each router, though, the numbers are not that interesting. I learned very quickly that trying to stream on the 2.4 GHz band is ill-advised (seriously, just don’t). On a 5-GHz channel, however, all the routers including my old RT-AC66U handled Steam streaming essentially flawlessly. I was quite surprised. The approximately 14-Mbps stream that Steam defaulted to on all the routers is more than double the bandwidth of a conscientious Twitch streamer. Image-quality-wise, that’s firmly in good enough territory for me, and the input lag was imperceptible. In the last image above, you can see a frame time spike that slowly tapers off. That’s from my old router and I was lucky to get that screen shot because that minor problem was rare even on my dated hardware.
Based on my other performance numbers, there’s clearly room to try and extract more image quality by raising the bandwidth limit on the client side, especially if you’ve selfishly kept the ROG Rapture’s second 5-GHz channel all to yourself like I did. Everything I just said holds true from my comfy chair and Ellie’s room, and for both the Intel and Edimax Wi-Fi adapters. It’s all very impressive.
Unfortunately, things aren’t so rosy for Sony’s Remote Play offering. Before I start complaining, though, let’s acknowledge upfront that real-time video game streaming is probably one of, if not the most, difficult tasks you can ask a home router to perform. Unlike normal streaming, there’s no buffer and packets are traveling both directions. One could argue that it’s a great place for superior routers to distinguish themselves. However, we just saw how Steam’s near-perfect take on streaming was essentially router-agnostic. Sony’s appears to be router-agnostic, as well, but in a bad way.
Streaming to the PS Vita was my focus, and its 2.4 GHz-only Wi-Fi connection is immediately suspect. However, there’s also a Remote Play client for Windows, and, in my experience, the issues I’m about to describe apply whether we’re talking about a gigabit Ethernet connection to a PC or the lowly Wi-Fi of the Vita. The problems came in different forms. There were full-on pauses to the stream that lasted for about a second and happened a couple times a minute. I saw inexplicable and progressively worse image quality over time, and good-old-fashioned artifacting like you might see from poor signal quality on satellite or over-the-air digital TV. I witnessed all those problems on all the routers to varying degrees when streaming from my PS4 Pro.
The PS4 Pro sending data to the Blue Cave for consumption by the PS Vita.
Given those issues, I got to troubleshooting. Both my PS4 Pro and Vita were updated to the latest firmware available. The PS4 Pro enjoyed a wired connection to the routers. While troubleshooting, I read that some folks got better Remote Play performance by disabling Wi-Fi power saving on their Vita, so I did that as well. Changing the image quality from 360p with low framerate and 540p with high framerate didn’t make any difference, either. The high-end option was only using about 5 Mbps, well within the bandwidth capabilities of the 2.4 GHz testing we’ve already done for the locations in question. I even fiddled with and turned off all the QoS and security features that I had on for my other tests just in case they were interfering somehow, but it didn’t help. I came to the conclusion that Remote Play just wasn’t as robust as Steam in-home streaming. Maybe Sony’s poor performance has something to do with the power of hardware that’s doing the heavy lifting, or maybe it just needs another software update or two to work out the kinks.
If I had to rank the routers based on Remote Play performance, the rankings wouldn’t be separated by much, but my old RT-AC66AU would be in last place, just beat out by the Blue Cave and Lyra Trio. I’d have to call a tie between Asus’ non-ROG routers without having more hard data to go by. I should note that the Blue Cave is noteworthy in this lineup for being the only router that occasionally caused my Vita to tell me the connection was too slow to use when first initiating Remote Play. However, a couple retries is all it would take to eventually get the Vita to connect. I had high hopes for the Lyra Trio in this test, especially in Ellie’s room where it had the advantage of offering up a 2.4 GHz signal to the Vita from a node just a couple feet away. Sadly, it didn’t seem to make a significant difference.
I felt that the Remote Play performance of the ROG Rapture was the best of the bunch, even if it wasn’t perfect. Remote Play displayed all the problems with the ROG Rapture that it did on the other routers, but they did seem to happen less frequently on the more powerful device. It was as if anything else the routers had to do was more important than serving up the Remote Play stream. Maybe Sony could use a prioritization checkbox like the one Steam offers. The video above was made while streaming from the ROG Rapture. Whatever you perceive as wrong with this stream, imagine it being about 25% worse on the other routers. Sorry, but that’s just the best I can quantify it.
Subjective impressions, post-testing notes, and conclusions
In addition to the more formalized benchmarking and game-streaming tests, each router was also subjected to use as our home’s exclusive path to the internet for at least a month. As I’ve alluded to, our household of late has been pretty demanding of networking devices. 12 internet users, each with their own collection of devices, can really bring the pain. We’re talking about a few TB of data each month, easily.
We’ve got kids’ tablets left lying around streaming Netflix, YouTube, or Amazon video even while no one is watching. Fortnite, PUBG, and Diablo III are frequently all being played at the same time across Origin, Steam, and Battle.net. Someone’s practically always in Discord, and they’re probably sharing their screen with their group of friends. Even all the OTA TV we watch hits the network (courtesy of a HDHomeRun tuner). That’s before you consider automatic things like game patches, Windows updates, and 24/7 RTSP streams from baby cams. You get the idea: it’s no walk in the park even when I’m not benchmarking. Naturally, I learned some things as we went along.
The only day-to-day problem that I encountered across all the routers had to do with viewing Twitch streams. Like the PS4 Remote Play issues, I have a hard time blaming the routers for it, but it happened consistently, and I think it’s worth mentioning. Basically, Steam downloads wrecked Twitch streams. Regardless of my QoS settings, the browser I used, the combination of PCs, wireless or wired connections—whenever Steam was allowed to download un-throttled, Twitch would not play smoothly. It always buffered up to about 20 seconds and then played that buffer down to nothing before starting over again. The buffer recovered quickly after a couple seconds, but it was like the Twitch player just didn’t ask for more data while the stream was playing.
Other downloads or traffic didn’t cause this problem—only Steam. The only solution was to throttle Steam downloads within the Steam client. Asus couldn’t duplicate this problem, but it persisted even after I switched modems. No other streaming video services that I tested had this problem. For what it’s worth, Jeff notes that his significant other is a big Twitch watcher and immediately notices when Steam starts downloading anything in the background on other networked PCs in his house, too. Weird.
Something else I wanted to mention is that all the routers outdid my RT-AC66U in the YouTube upload department. YouTube uploads would kill my downstream traffic on the old router, but thankfully none of the new routers had that problem when someone decided to send a file to the video giant’s servers.
Circling back to the USB attached storage performance and how demanding it was, I can’t help but feel like it’s a much better idea to share a drive from an always on PC or buy a proper NAS for your network. At least those solutions will save you from the wrath of teenagers obsessed with their server pings if you have the audacity to start slinging files around from your router. I don’t know about you, but I’m the only nerd in my house that even uses local network storage in the first place (a 4TB RAID 5 array in an aging Ivy-Bridge home server that you’ll be reading more about in a future story). Even then, I use OneDrive more often than not. I’m curious to know, just how many people out there plug a drive into their router in the first place?
Usage from left to right, un-throttled Steam with Twitch, throttled Steam with Twitch, and just Twitch.
Verdict: ROG Rapture
I’ve been waiting this entire review to write the next sentence. I freaking love this thing! Finally, a router where a Steam download doesn’t peg a CPU core. Sure, I’ve mentioned a couple small annoyances the Rapture has, but those are shared with all the other routers, as well. I don’t think they’re the router’s fault. I have complaints about the other access points I tested, but not with the ROG Rapture. In short, nothing I threw at it made it stumble. No one complained about net-lag or lost connections while the ROG Rapture was calling the shots, not even when I was running my worst torture tests. It never crashed or needed to be rebooted. Simply put, it’s a beast.
Remember this when you look at the next image.
Look at the image above to drive the point home. it’s one of my wired test-runs to USB-attached storage on the Rapture. Those results are with a Steam download happening in the background. The results are nearly identical to what I recorded in the official benchmark results with nearly idle network traffic. The ROG Rapture didn’t even break a sweat.
Overall, the ROG Rapture is an easy Editor’s Choice for me. It was the only router that could take a licking and keep on ticking in our demanding testing environment. At $380 retail, it’s not cheap at all, but nothing else had what it took to do the job. It’s unapologetically extreme and all the better for it.
Verdict: Blue Cave
While it’s still fresh in your mind, let’s look at the same Steam-download-plus-CrystalDiskMark test result from the Blue Cave. It’s a good illustration of some of the other problems I found with Asus’ square donut.
Yeah, those numbers don’t line up quite so well with the nearly idle network traffic numbers in the earlier graphs. The good news is that my Steam download proceeded at full speed, but at the expense of USB attached storage performance. However, people in the house complained about lag whenever I ran these tests, so while the raw download speed was unhindered, there were other latency-related penalties caused by this workload on the Blue Cave. How much this weakness should worry you depends on how highly you value quasi-NAS functionality from your router. In the image below, you can see that it’s more than just a synthetic problem, though.
Ignore the Twitch stats, this screenshot was taken before I realized the Twitch buffering glitch was universal.
In that image, I’m downloading a game from Steam, watching a Twitch stream, and copying a file to the USB-attached storage. Both the Steam download and file copy function alone were observed to peg one of the CPU cores at 100%. With both running at the same time, the file copy speed crashes. For some reason, core three on the Blue Cave always took the workload while cores one and two figuratively took the day off. One other bit of weirdness when it comes to the USB attached storage that’s specific to the Blue Cave: I could only see the drive from my PC when I enabled SMB 1.0. That’s not exactly a recommend practice these days.
Unfortunately, I have one more problem with the Blue Cave to share. Every couple of days, it would completely drop the WAN connection and require a reboot to bring it back. LAN traffic was unaffected. Asus was very responsive troubleshooting the issue, but nothing I tried was able to resolve it. Asus reviewed my log files but could not replicate the problem. The best guess the company could offer was an incompatibility with my modem, so I replaced it with a model Asus recommended: the Arris Surfboard SB6190. I tried a second Blue Cave and updated firmware after I got the new modem, but the problem persisted.
Seriously, the Blue Cave looks dope.
Combined, those two issues are a real bummer for me. The transfer speed issue may not matter to most users and the WAN drop issue may not happen everywhere, either, so rendering a verdict on this router is a bit tricky. For day-to-day usage the Blue Cave performed admirably in our strenuous environment, all for half the price of the ROG Rapture. I’ve actually recommended it to a couple people that asked me for router advice while this review was ongoing because I felt it unlikely they would run into the same problems I did. Still, I can’t endorse the Blue Cave unequivocally, even though it may deserve it.
The Blue Cave would probably earn a TR recommended award if it wasn’t for the repeated WAN dropping at my home. I also don’t like its attached-storage performance, although I feel like the answer to my question about how many people use router-attached storage is “not very many.” Since the WAN dropping is quite possibly specific to something in my setup, most folks would have very little to complain about. For just $150 at the time of writing, the Blue Cave is much cheaper than the ROG Rapture without costing you much in the way of performance. If you’re willing to chance some annoyances, the Blue Cave could be a solid and stylish addition to your home network.
Verdict: Lyra Trio
You’re not going to read about high CPU usage or problems with USB-attached storage when it comes to the Lyra Trio. That’s because the Lyra system is so simple it doesn’t show its CPU usage or include a USB port on any of its nodes. That’s probably for the best, going by what we’ve seen from the more powerful Blue Cave. The Lyra Trio needs to save its strength for just routing.
The Lyra Trio I got was not set to English out of the box. The setup process was so simple, though, that I was able to complete its steps and eventually get to a screen where I could pick English as my preferred language without any problems. Pairing the nodes took me a bit longer, but that was on me for not taking the time to consult the quick-start guide and get the sequence of button presses and timing just right. You can tell the nodes are properly paired when their lights switch to a solid cyan color. You can also see the nodes in the web interface when they’re connected, although they show up as a wired connection despite the 5-GHz wireless backhaul.
The Edimax adapter I used for testing had a strange relationship with the Lyra Trio. It was frequently able to best the Intel adapter built into my notebook in some tests and locations, a feat that practically never happened with the other routers. When testing the Lyra Trio, I gave the adapters a couple minutes of idle time when I turned them on at a new location, just to make sure they had a chance to connect to the best node. The Intel adapter would always and instantly connect at 5 GHz to whatever node was automatically selected. However, the Edimax always connected at 2.4 GHz initially and then took anywhere from a few minutes to over 30 minutes to eventually flip over to 5 GHz. I don’t blame the Trio for this, but I did wait to perform my tests until the Edimax was on properly settled.
Since the Trio wouldn’t let me split the 2.4 Ghz and 5 GHz SSIDs, I believe anyone using the Edimax would see this behavior unless they changed the settings of the adapter to only connect at 5 GHz on the client-side. It left me wondering how smart the overall system was at making sure the fastest nodes were being used and got me thinking about the poorer than expected results of the Lyra Trio at the longer ranges like the Movie Shed where it should have had a significant advantage. I can only report on what I found though, because there weren’t any configuring options available to tweak. Maybe the system is a bit too simple.
I’m not going to harp on the next point of discussion, because by the time the Trio got its time in the hotseat I’d already already started thinking it wouldn’t be able to take the heat. Long story short, the Lyra Trio locked up just about daily. It wasn’t just a WAN drop like the Blue Cave, but a complete freeze of the primary node. The latest firmware (which has to be updated on each node individually, by the way) didn’t help. In between resets, multiple members of the household grumbled about connection problems just as much as they did when my RT-AC66U was in place. Maybe things would be just peachy under different circumstances, but I think a single-core router with only 128 MB of RAM just wasn’t cut out for what it was being asked to do, even if it had friends.
Overall, the $260 Lyra Trio kit is a tough pill to swallow. It seems to me that its internals are better suited to extender-only duty while something a little beefier should be tapped to do the heavy lifting. Ultimately, I didn’t come away any better off for having a three-node network than just one. Stay tuned though, because with my small army of routers to play with, including a first-generation Lyra that I haven’t cracked open yet, there could be more meshing to come.