Remember LAN parties? In some parts of rural America, we still don’t have fancy DSL or cable internet connections, and we certainly didn’t back in the mid-90s. Everyone was happy to lug their full-tower ATX PCs and 50-pound CRT monitors across the county just to play a few hours of Quake II or Starcraft. That’s all ancient history now for most, but we’re still having LAN parties, and so are folks elsewhere.
The thing is, the machines you’re likely to find at a LAN party these days are less Chieftec and more like that Gigabyte P57X gaming laptop we just reviewed. Gaming laptops have come a long way in the last few years, and even more so with the ridiculously-efficient mobile graphics chips from Nvidia that have recently become available. While the complete package of a mobile gaming laptop might not perform as well as a full-blown desktop, it comes close. You can get laptops with adaptive-sync displays, SLI, and even external liquid-cooling these days, too. Life is good if you need to take a powerful machine with you.
Even so, laptops still have a lot of limitations. You have to worry about saving space for the battery, and you have to make sure it doesn’t get so hot it will scorch the user. You also need to account for the display’s power usage, and the whole thing has to fit within a certain thickness and weight envelope. What if you wanted a PC that was small enough to carry around anywhere, yet had some really serious gaming chops? Like ultra-settings-at-four-megapixels gaming chops? Say hello to Zotac’s Magnus EN1070, the successor to the Magnus EN970 we reviewed a while back.
From the pictures, I really thought this tiny terror would be larger. I have some experience with Zotac’s mini-PCs, and the machines I’ve had using AMD low-power APUs and ULV Intel chips have been pretty dang small. Given the hardware on offer and without much visual context, I was expecting this machine to be more the size of a Playstation 4 or Xbox One.
But it isn’t! It’s flippin’ tiny. It has around the same footprint as a mini-ITX motherboard, but it’s barely thicker than a Wii U. It’s about as thick as four stacked DVD cases. This thing is so small my best lady-friend can stick it in her handbag. The solid copper cooling hardware inside makes it a bit heavy for that to be comfortable, though. My analog postal scale puts it at right around four pounds. The actual dimensions of the machine are 8.3″ x 8″ x 2.4″ (21 x 20.3 x 6.2 cm.)
Despite the tiny size of this box, Zotac packs a desktop Skylake quad-core CPU and a GeForce GTX 1070 inside. The GTX 1070 is the same mobile graphics card in the Gigabyte P57X we tested recently, with 2048 Pascal shaders running at a base clock speed of 1442 MHz and a boost range of 1645 MHz. Meanwhile, Intel’s Core i5-6400T handles more general-purpose math. Those in the know will already be aware that this is a low-power version of the base-spec, sixth-generation Core i5 series. It starts out at 2.2 GHz, and it can Turbo up to 2.8 GHz if thermals permit.
The CPU gets a pair of DDR4 SO-DIMM slots for RAM, and storage can slip into an M.2 socket and a 2.5″ SATA drive bay. The M.2 socket supports SATA and PCIe drives, although it didn’t like the OCZ RD400 we stuck in it when we first put together the EN1070.
After some correspondance, Zotac pinned the EN1070’s reluctance to boot the RD400 to a hardware or firmware problem with our pre-production EN1070 that shouldn’t be present in production units. In the RD400’s place, the company lent us a HyperX Predator PCIe AHCI SSD, which went nicely with the pair of 8GB DDR4-2133 CL13 SO-DIMMs HyperX provided for use with the machine.
I want to take a moment and say thanks to Toshiba for the RD400 SSD, as well as to HyperX for the memory we used in the machine. The RD400 is one of the fastest SSDs we’ve ever tested, and I sorely missed its high capacity during testing. Meanwhile, the low-latency HyperX Impact memory is on sale at Newegg right now. Without these contributions from OCZ and HyperX, this review wouldn’t have happened.
Let’s get a closer look at this mighty mite.
Like a record, baby
The front of the EN1070 hosts the usual headphone and microphone jacks, flanked by a USB 3.1 port on either side. The left-side port is a standard USB Type-A, but the right-side port is a Type-C connection. Unfortunately your reviewer doesn’t own any USB 3.1 devices, but I can at least confirm that both ports work correctly in 3.0 mode. An SD card slot serves as sort of a modern-day floppy drive, and I used it to install Windows 10 Pro on our test machine purely for the novelty of doing so.
One thing that isn’t completely clear from the pictures is that all four sides of the EN1070 are aluminum, save for the plastic insert in the front. The top and bottom portions of the casing are plastic, but the rest of the external casing comes apart in four aluminum pieces. The first time I took it apart, I accidentally disassembled the entire case. Getting it back together more or less requires a clamp (or an extra pair of hands) to hold the parts together while you screw them down. Let that serve as a warning to anyone who takes one of these apart.
Around the backside, we have a whole bunch of paired ports. USB 2.0, USB 3.0, HDMI 2.0, DisplayPort 1.3, and gigabit Ethernet all get two jacks each. Both of the Ethernet connections are powered by Realtek chips, and all four display connections are hooked up to the GTX 1070. You can use all four simultaneously, although I did have a brief issue getting the fourth display connected to the second HDMI 2.0 port. After a reboot all was well, so it was probably a janky connection or similar transient issue.
Also on the back of the EN1070 are the antenna connection for the Wi-Fi connection and the barrel plug for the power adapter. The 180W power adapter that Zotac includes is a wide-and-flat laptop-style brick with a normal-sized AC plug. I don’t normally use Wi-Fi aside from my smartphone, but for completeness’ sake I did test it briefly. Pulling files from my buddy’s high-performance NAS I was able to manage upwards of 25 MB/sec. Considering I was connected to an access point at the far end of his home in a home filled with electronics (during a LAN party), I felt like this was a pretty decent showing for the little five-decibel antenna on the Zotac.
Digging in
Zotac sells the machines pre-configured, or as a bare-bones system without RAM or storage (as we received.) Folks who purchase the barebones machine will have to open it up to add their RAM and storage devices. Fortunately, like most other Zboxes I’ve seen, all that’s required to open up the Magnus EN1070 is the removal of two thumbscrews and a tiny bit of lateral force. The bottom cover slides free, and you’re greeted with the bottom of the motherboard.
In the picture above you can see the paired SO-DIMM slots on the left side, the 2.5″ SATA bay on the right side, and the M.2 socket right in the middle. Off to the right behind the SATA bay is the M.2-2230 Wi-Fi and Bluetooth adapter. The EN1070 impressed me with its intuitive internal layout. This is as far as any user should have to go to get their unit up and running.
This clean design even extends to the top of the board. Simply remove five screws, disconnect the front panel cable, and then the motherboard lifts right out of the chassis. You do have to be careful not to damage the Wi-Fi antenna that’s attached to the outer casing, though. On the top side of the motherboard we can see a small daughtercard for the ASMedia USB 3.1 controller and two huge centrifugal fans that cool solid copper heatsinks. It’s no secret why this machine is so heavy.
Looking at the top of the motherboard also confirmed one of my suspicions about this machine: it uses a socketed CPU. Checking Intel ARK, the Core i5-6400T is a socketed CPU, but it seemed strange to me that Zotac would utilize some of the precious space inside the EN1070 for a CPU socket and the supporting hardware, such as the retention bracket. I suppose this choice could have a number of advantages for Zotac, the foremost being that the company can use the same motherboard for various system configurations.
Now that we’ve seen the EN1070’s guts, let’s see how it performs.
The moment of truth
The Magnus EN1070’s box says “Mini PC or gaming powerhouse? Why not both?” I wanted to put that statement to the test. To tease out any weaknesses in the machine, I chose to benchmark with one title that hits the CPU hard and then one that would load up the graphics card. To that end, I picked out Grand Theft Auto V for CPU-intensive testing and the 2016 version of Doom for a primarily graphics-card-bound game. I tested both games with the fans at stock settings and with the “full speed” option from the system’s firmware to see whether this box is thermally limited. I also gathered some subjective impressions from Overwatch, Dark Souls III, and Warframe.
Doom
When I think of gorgeous games right now, id Software’s latest Doom title is the very first thing that comes to mind. While legendary graphics programmer John Carmack is no longer at the helm there, his shoes have been filled by the extremely capable Tiago Sousa, previously the graphics R&D lead at Crytek. Tiago and team’s stellar work has resulted in a game that looks incredible when it’s cranked, but remains playable even on meager hardware. Fortunately for us, the hardware on the bench today isn’t exactly “meager.”
Doom is absolutely stunning with the settings turned up, and that’s how I tested it: with every single setting slammed to the ceiling at 2560×1440. I stuck with OpenGL API for the EN1070, as our previous testing has shown that Vulkan doesn’t help Nvidia graphics cards much. For my testing, I played through an easily-repeatable 60-second portion at the beginning of the third level (“Foundry”) while recording frametimes with PresentMon.
Not a bad showing for the little box that could. Even with Doom cranked up all the way, the EN1070 pumps out over 80 FPS on average. Moreover, the frame-time plot and 99th-percentile frame times above show that frame delivery is quite smooth. Playing Doom at these settings on the Magnus EN1070 is nothing less than a joy. Notably, while cranking up the fans all the way didn’t do much for the average FPS, it seems to have improved the worst-case performance (as measured by the 99th-percentile frame time) slightly. It also increased the amount of noise the machine made under load, but I’ll get to that later.
I carefully monitored CPU and GPU usage while informally benchmarking Doom, as well. The game spread its CPU load evenly across the Core i5’s four cores, and none of them were consistently maxed out. That’s because the GPU was constantly power-limited, according to GPU-Z’s monitoring window. The mobile GeForce’s clock rate varied a lot—as GPUs do, these days—but it mostly stayed around 1350 MHz. That’s below the GPU’s base clock and well below the specified boost clock.
If you’re just joining us here at The Tech Report, you might be confused about this business immediately above. The “time spent beyond X” graphs tell you how much real time our test system spent rendering frames that took longer than 16.7ms, 8.3ms, and so on. 8.3ms corresponds to a framerate of 120 FPS, 16.7ms is roughly 60 FPS, 33.3ms is roughly 30 FPS, and 50ms is 20 FPS. This gives us a good way to pick out stutters and hitches during gameplay. If you drop below 20 FPS during gameplay, even for a moment, it’s going to cause a noticeable stutter, and our “time spent beyond” graphs aggregate that roughness.
We won’t be using those 50-ms and 33.3-ms graphs here, though, because the EN1070’s performance is quite solid in Doom. This system spent barely more than three-hundredths of a second under 60 FPS during our run, regardless of fan configuration. By my reckoning, that should satisfy all but the most demanding gamers. The game ran so well, in fact, that I tried to step the resolution up to 3840×2160. That was too much for the mobile GTX 1070, though. Frametimes lurched into the mid-40-ms range, with spikes over 50ms. That’s sub-20-FPS territory on average, and that’s not a good time in Doom.
Grand Theft Auto V
Grand Theft Auto V is over three years old now, but the PC release is just over a year and a half past its debut. Rockstar’s magnum-opus-to-date is unmatched in its combination of breadth and depth; there are still very few games that even approach the type of broad-ranging and comprehensive simulation seen in GTA V. Tracking physical simulation and artificial intelligence for dozens of entities simultaneously across a massive, fully-open game world takes a lot of compute horsepower. That made it a natural choice for my CPU-heavy gaming test.
To benchmark the game I ran Franklin through the standard TR test in offline mode. This test won’t maximize the CPU load compared to the game’s online mode, but it does provide standard, reproducible results.
Perhaps thanks to GTA V‘s heavier CPU load, its 99th-percentile frame times don’t quite mesh with its roughly 60-FPS average. Even so, the game rarely drops below 50 FPS on average during our test run, and that’s still a commendably smooth gaming experience.
Like with Doom, I monitored the machine’s resource usage during some informal benchmarking. I correctly surmised that GTA V would load up the Core i5-6400T’s four cores harder than Doom would. Most of the time I was playing GTA V, the CPU was over 90% loaded across all four cores. The graphics chips was working hard too, and it was still hitting its power limits according to GPU-Z. The GTX 1070 did hit higher clock speeds in GTA V, at around 1400 MHz, but it still frequently dipped below the Nvidia-specified base clock of 1442 MHz. Dips down into the roughly 1300 MHz range were more common than the rare moments it would spike over 1500 MHz.
As you can see from the enormous settings composite above, I was testing the game on very nearly the highest possible settings. It’s possible I could have smoothed things out by reducing some of the settings, particularly with regards to distance scaling. Alternatively, enabling MSAA or raising the resolution would tip the load balance over onto the GPU, although they would also reduce the average framerate.
The run with the fans maxed-out arguably turned in a worse performance than with the fans at stock speeds, for whatever reason. I wouldn’t put too much stock in the difference that ramping up the fans makes, because the temperature on both CPU and GPU was within a few degrees C either way. If GPU-Z is correct, the cooling capacity of the EN1070 isn’t the issue; it’s more that the system is being hamstrung by its 180W power supply.
For fun, I kicked the resolution up to 3840×2160 on GTA V, too. Performance worsened significantly, as you might expect, but GTA V surprisingly ran better at 4K than Doom did. I didn’t benchmark it extensively, but Fraps’ overlay was showing around 35 FPS on average. That’s impressive for a little machine like this. On the other hand, considering that our sample of the desktop GeForce GTX 1070 managed a 20.1ms 99th-percentile time in 4K, it looks pretty lame. That’s about the same performance that the EN1070 turned in at 2560×1440. Admittedly, our graphics test rig has much more forgiving thermal and power envelopes, but it does show the lines that Zotac had to color in to make the EN1070 a reality.
Power, limited
Our gaming results suggested that the EN1070 seems to be pushing the envelope of what’s possible within the power and thermal budget of a tiny air-cooled gaming system, so I dived in with some testing and monitoring tools to figure out just where its limits were.
When I loaded up the EN1070’s Core i5-6400T CPU, its core clock sat around 2.5 GHz, which is the expected all-core Turbo speed for this model. Skylake CPUs are some of the quickest around, to be sure, but 2.5 GHz is relatively slow compared to even the 51W Core i3-6100 and its 3.7 GHz clock. In fact, GTA V warned me that the EN1070’s CPU didn’t meet the game’s minimum requirements when I first launched it. Even then, we didn’t see performance issues that would have suggested thermal throttling on the CPU side, so we dug in some more.
Even though the CPU of the EN1070 isn’t thermally limited, the box does appears to be toying with both thermal and power limits for the GTX 1070 inside. The hottest graphics-card temperature I recorded was 82° C—not coincidentally the target maximum temperature that Nvidia bakes into its Founders Edition cards with GP104 GPUs on board. Given the sub-base-clock results we sometimes saw around that temperature, it would seem that the EN1070’s graphics-cooling system is merely adequate. Thermal limits weren’t the only things hamstringing the EN1070’s graphics chip, though. For what it’s worth, the GPU-Z utility indicated that the GPU was power-limited during most of our testing, and especially in Doom.
Thermal design power, or TDP, isn’t a standardized or even generally cross-comparable measurement, but you’ll forgive us some ballpark math to see why the EN1070 was hitting its power limits. Nvidia specifies a 150W TDP for its mobile GTX 1070, and with the 35W rating of the CPU, we’re already brushing up against the 180W the power adapter can supply. We haven’t even started talking about power for the memory, storage, dual Ethernet adapters, Wi-Fi, USB devices, and so on. The laptop GTX 1070 almost assuredly draws less power than the desktop version, but the chips aren’t that different on paper. In any case, there’s just not a lot of extra juice to go around.
Despite the EN1070’s apparent envelope-pushing, the box at least remained quiet while doing so. As I stated earlier, I have some past experience with Zotac’s mini-PCs. The other Zboxes I’ve used have noise profiles that could charitably be described as annoying. In contrast, to actually hear the fans on the Magnus EN1070 with the box on my desk, I had to turn down the game volume, turn off my personal PC and the air conditioner, close all the windows, and wait for my refrigerator’s compressor to kick off. Impressive stuff, and a far cry from other premium mini-PCs we’ve tested. Manually cranking the fan to a 100% duty cycle did produce a considerable amount of noise, but even then the sound is a distinctly “whoosh”-y kind of noise. I didn’t find it grating or even particularly bothersome. If you put this system in a living room or shared space, it should be a pleasant companion.
The power of Hertz
To figure out where the Core i5-6400T fits into the pantheon of lower-power chips, we ran some directed CPU tests. First up is AIDA64’s benchmark suite. The system information and testing tool has been a staple of the TR toolbox for ages, and it includes among its many functions an extensive suite of CPU and memory benchmarks.
That Photoworxx benchmark makes heavy use of the latest AVX instructions, so it strongly favors newer chips. It’s a good argument for grabbing the latest chip you can when buying new hardware. The i5-6400T bests the evergreen Core i5-2500K despite a gigahertz of clock-speed deficit, but it trails the 84W Core i5-4590 by a wide margin.
7-Zip’s built-in benchmark is only slightly less synthetic than the results from AIDA64, but it’s easy to run and so we have a lot of data points to compare with. Notice how, in this benchmark, the 6400T falls behind not only the 2500K, but also AMD A10 APUs and a Haswell Core i3. 7-Zip as a benchmark is well-suited to lots of threads, but watching the desktop Core i5 battle it out with the 28W Core i7-5557U is a little disheartening.
Finally, we have some browser benchmarks. SunSpider is an older Javascript benchmark that is no longer maintained. As a result, this is probably the last gasp for it in our testing. The single-threaded nature of SunSpider means that the 6400T performs toward the back of the pack, though we could also chalk that up to the age of the benchmark. We’d not put too much stock in this test.
Mozilla’s Kraken is based on SunSpider, but adds several more intensive tests based on common tasks in searching, audio processing, image filtering, and other things we expect web browsers to do these days. The Core i5-6400T fares much better in this test.
All told, the Core i5-6400T delivers a somewhat mixed performance picture, but given how hobbled the GeForce GTX 1070 is in this system, that CPU seems like an OK match for it. In a game that stresses both the CPU and the graphics card, however, like GTA V‘s online mode, it could pose a challenge.
I noticed much rougher performance from the EN1070 playing GTA Online than I did in the game’s single-player mode. To try and put some numbers to that experience, I ran three supercar races on a private server with a group of friends. While we’d be extremely cautious about reading too much into these numbers due to the intensely variable nature of online gameplay, loading up the EN1070 this way produced the rather fuzzy frame-time plot you see above.
GTA Online still ran at about 60 FPS this way, but its 99th-percentile frame time dropped to about 30 ms. That was a noticeable decrease in smoothness, and my much beefier desktop didn’t exhibit similar roughness to my eye. Given this result, we’d be curious to see what the EN1070 could do with a beefier CPU, a higher-wattage PSU, and perhaps a more aggressive cooling setup. For now, though, we’d classify the EN1070’s CPU choice as merely “adequate.”
A quick VR test
While I don’t have access to TR’s VR headsets, it wasn’t an issue to fire up the SteamVR Performance Test to see whether the Magnus EN1070 lives up to Zotac’s VR-ready sticker on the box. This simple test runs on the Source 2 engine and mostly measures graphics-card readiness, but it’s still an OK rule of thumb for determining whether a PC is VR-capable.
Given the GTX 1070’s desktop-VR prowess, it’s not surprising to see that the EN1070 is VR-ready. Given its size, this Zbox could be a good candidate for a portable VR demo rig. While the Core i5-6400T might fall a bit short of Oculus’ and HTC’s recommended CPU specs, it’s not that far off. Just be ready for warning messages in Oculus’ and HTC’s dashboards regarding non-compliant system specs.
Out and about in the world
So we know the CPU is a little slow and the GPU is power-limited, and we know what that means for core gamers looking for serious gaming. Frankly speaking, it’s a little hard to imagine anyone buying the Magnus EN1070 with other intentions. More and more folks are picking up gaming as a casual hobby, though. Setting a machine like the EN1070 in the living-room and using it mainly for Netflix and Facebook could seem like a gross waste of processing power, but when you want to play a game and you want it to look good, you need to have that performance on tap.
In my time with the Magnus EN1070, it came along with me quite a bit. I set it up outside on my woodworking bench and kicked back with a Sunny D while playing Dark Souls III in the midday sun. I brought it to my parents’ house and hooked it up to their big 60″ UHD TV to grind some Warframe. I also took it along to a LAN party and hooked it up to a 17″ CRT (through an HDMI-to-VGA convertor box) to fail hilariously at Overwatch. Throughout all of this, it performed admirably, even though I was cranking the settings on these less-demanding games to the roof.
In a real-world situation, for someone who isn’t connecting a 144Hz monitor or gaming at a high-level of competitive play, the Magnus EN1070’s performance is just fine. Its relative silence and unobtrusive noise profile makes it a pretty reasonable candidate for HTPC duty, too, and it’ll kick the pants off any current console for sheer gaming enjoyment. The ability to carry the Magnus EN1070 around and set it anywhere is incredibly convenient. Of course, the same can be said of a fancy gaming laptop.
Conclusions
When hardware nerds get together to talk shop, the refrain surrounding small-form-factor systems has always been “Pick two: fast, cheap, small.” Everyone appreciates a little computer because it’s easy to transport and because it simply doesn’t take up much space. You can stick it anywhere, and unlike a laptop, you don’t pay extra for a display or battery you aren’t likely to need. That’s Zotac’s Magnus EN1070 in a nutshell.
As we saw, the mobile GeForce GTX 1070 in the EN1070 can provide some impressively solid gaming performance at 2560×1440 as long as it isn’t being CPU-limited. Even when it is, the fast graphics chip offers the opportunity to crank up visual settings without impacting framerates much. The EN1070 can deliver smooth gameplay in some of the most demanding games on the market now, and you can be sure it will stomp all over less-taxing titles. If you can swallow its price tag, this machine offers true console-killing power in a console-size footprint.
That said, the GTX 1070 in this box doesn’t seem like it’s being used to its full potential. We didn’t observe more than the most vanishing of boost-clock headroom for the graphics chip in our tests, and in some cases, the GTX 1070 can’t even maintain its specified base clock due to the strict power limits of the EN1070. More aggressive cooling and a larger brick capable of 200W or so may have at least kept the GeForce at its full base clock. Gigabyte supplies a 200W adapter with its P57X laptop, and its GTX 1070 has no problem boosting well past even Nvidia’s specified boost range.
That old refrain about mini-PCs applies to Zotac’s micro-monster, too. It’s small, and it’s pretty fast, so it’s gotta be pretty expensive. The Magnus EN1070 costs $1200 as a barebones. Our machine as configured comes out to a little over $1750 at retail after you throw in the memory, storage, and Windows 10 Pro license, and that’s before a monitor or input devices go in the shopping cart.
While it’s true one could forego the fancy NVMe SSD in our build, $1750 can build a frighteningly quick and portable Mini-ITX gaming machine these days. A Core i7-6700K and a screaming-hot GTX 1080 could easily fit into that budget. If true mobility is essential, you can get a much more powerful CPU and a more boost-happy GeForce GTX 1070 in Gigabyte’s P57X v6 gaming notebook for not much more money, as well. The laptop comes with Windows, a keyboard, and a monitor “for free,” too. Still, none of those systems will be quite as compact as the EN1070, and if you have a monitor and keyboard at your destination, the Zbox and its power adapter can go in a small bag without issue. You can’t say that of a Mini-ITX desktop or a 17.3″ notebook.
Even with its limitations, I really like this little PC. It makes all but the smallest small-form-factor PCs look like gargantuan superstructures, and yet it has the stones for VR and 4K gaming in Warframe. My buddies at the LAN party all liked it, too, since it let me leave my extremely noisy tower at home. In fact, the Magnus EN1070 is the densest size-to-performance package I’ve ever seen. The fact that this tiny box—barely larger than a carton of cigarettes—could match and sometimes even outpace my more traditional (and much larger) gaming rig is pretty neat. The fact that it stays cool and quiet during gaming sessions is a real testament to Zotac’s engineering prowess.
At the end of the day, whether the Magnus EN1070 is worth it depends on how many cubic centimeters you need a PC to occupy. If you’re looking for a relatively beastly gaming PC in a system whose footprint isn’t much larger than a Mini-ITX motherboard alone, the Magnus EN1070 offers more-or-less unparalleled gaming power in a perfectly totable form factor. As with every other mini-PC we’ve tested, however, every cubic centimeter you can spare will increase the amount of performance you can get per dollar, and that’s a calculation that aspiring small-form-factor PC owners will need to weigh for themselves.
Share
