Installing Windows, Testing and Results
We’ll start with stock clock speeds and voltages for the CPU, memory and GPU:
The CPU clocks and voltages shuffle themselves around to match the load, generally idling at 825MHz and 0.66v. In the shot above I caught it halfway between the two states. It also has quite a few mid range positions, in case 825MHz isn’t enough but 1650MHz would be excessive.
I ran ATTO on the SSD as well, both at the stock queue setting (first screenshot) and at queue 10:
This is a real SSD! It isn’t exactly a high IOPS monster, but it lays down some solid numbers. You can tell when you’re using the ZBox too, very little time is spent waiting for storage.
Now for installing Windows. Without an optical drive or an additional computer you would be up a creek. I have an addtional computer, so I used WinToFlash’s free version to peel Windows 7 off its DVD and plop it on a USB flash drive. Once that has been done the USB drive can be treated just like the Windows install DVD, and things installed happily.
It is worth noting that the blue USB3.0 ports require a driver that is not built into windows, so they cannot be used for mouse/keyboard/USB Drive during windows installation without some more advanced fiddling with the windows installer to fold the driver in.
With that out of the way you can use the other computer to copy the driver DVD onto a USB drive as well and install the drivers off it. I had to go to Zotac’s driver download page to get a driver for the USB WiFi doodad, the mini-DVD did not please my DVD drive in the slightest.
Once installed, windows boots surprisingly quickly. From hitting the power button to a functional desktop takes about 24 seconds! Impressive.
With that out of the way we can start looking at performance, we’ll start with the CPU, as measured by CineBench 11.5:
It won’t win any awards for absolute speed, but for a CPU with a TDP of less than 12 W it’s not bad.
Just to be abusive I ran HWBot’s version of Unigine Heaven DX11, a very demanding DX11 benchmark. The ZBox Nano XS AD11 Plus ran it without complaint, though the frame rate was best described as masochistic at about one frame per second. On the plus side the ZBox recorded the all time fastest run for a 6320M GPU in that benchmark, so while the 6320M may not be up to gaming at extreme settings, the ZBox’s setup is faster than the average e-450.
My go-to low end computer game is Minecraft, it’s horribly addictive and therefor quite dangerous to test with, especially when the computer being tested can run it. Unfortunately FRAPS and Minecraft don’t get along so I don’t have a nice graph. However with all the detail settings turned on and Minecraft running full screen at 1680×1050 with the view distence on Far the frame rate varied from around 20 to around 38, generally in the high 20s to low 30s. Quite playable, if not butter smooth.
This being a HTPC built for streaming I did a fair amount of that too, attached to a 1080P LCD TV. I found that the USB WiFi dongle is very, very picky about antenna angle and what else is in the area. The thing drove me nuts trying to find a setup that worked well. Once I did however it streamed Youtube 1080P videos without complaint, as well as ESPN’s high definition live video and everything else I threw at it.
Similarly it was quite happy to decode and display 1080P video from a USB key.
I did all of this with the ZBox plugged into a Kill-A-Watt for power consumption monitoring as well as with a pair of thermocouple temp probes hovering just over the intake and exhaust vents. Here is what I found:
CPU Temp |
GPU Temp |
Intake Air Temp |
Exhaust Air Temp |
Wall Draw |
|
Idle |
55cc | 54c | 22c | 40c | 13.5W |
Prime95 |
74cc | 74c | 23c | 49c | 22.1W |
Heaven DX11 |
70c | 70c | 22c | 47c | 23W |
Furmark |
76c | 76c | 22c | 53c | 27.3W |
Heaven DX11+Furmark |
84c | 84c | 23c | 57c | 32.3W |
Youtube 720P |
74c | 73c | 23c | 45c | 18-24W |
Youtube 1080P |
74c | 73c | 23c | 46c | 18-24W |
Pandora Music Streaming |
62c | 62c | 23c | 45c | 16.1W |
I decided to go fairly light on the gaming tests as this unit is most definitely not made for such things. My favorite low end game is Minecraft and that is what I used for testing. I set the details (for what they’re worth) on high, with snazzy lighting and a “far” view distence. This is actually a surprisingly heavy load. FPS varied between 15FPS and 45FPS. On average it was able to keep things rolling at 27FPS up on the surface, if you’re running quickly into totally new territory that slows things down. Underground the frame rate stays much higher, as there isn’t nearly as much to render.
During all testing the fan was audible, at idle it is very quiet, while at heavier loads it speeds up a fair bit. At full tilt it can be heard in a silent room from 12-14 feet away from the unit. With any ambient noise at all you have to listen hard to hear it if you’re 4-6 feet away and further than that it disappears. If you’re right next to it for some reason you’re stuck hearing the fan noise unless there is a fair amount of ambient noise. The fan noise itself is far from offensive to me, it is a simply whoosh type noise without any rattling or clicking or whining.
General use found the ZBox Nano XS AD11 Plus to be fairly impressive for such a small package. It won’t beat my 24/7 system in snapping windows around, but it does a good job. It’s fast enough that I did not find it frustrating, despite being used to near-instantaneous action from my 2600k+6970+240GB MaxIOPS main system. I’m impressed!
I found the ZBox quite easy to use, but the included USB WiFi N dongle was balky, sometimes it worked great for quite a while, other times the driver became convinced that there were no networks in the area. I found the range to be surprisingly short as well. That said it’s still the best USB WiFi dongle I’ve run into. It also may be an issue with my specific part too of course.
The USB IR reciever works perfectly and has good range when used with the included remote, that is a very cool system. I was unable to convince the IR reciever built into the power button to function or accept a driver.
After testing I took the unit apart and replaced the factory thermal interface material with Arctic Alumina Cermaique 2, I use AA-C because it’s cheap and effective, there are other compounds (Arctic Cooling MX2/3/4 particularely) that give better temps. The result was a 5-6 degree drop in CPU core temps across the board! Whatever the factory stuff is, it is not very good.
Being an overclocker and a bencher I took a crack at that on this system. Unfortunately Zotac and the ZBox resolutely refuse to give even simple OCing options, the only setting in that part of the BIOS is ram voltage. I can understand this descision as the last thing a tiny unit needs is more heat, but it would have been fun to play with. The PLL module is inside the FCH as well and inaccessable. With that project failed I turned to undervolting instead. I found that I could drop the vcore substantially from the 1.3v it ran at stock. It turned out to be capable of running Prime95 at a mere 1.1vcore. The temperature diference was significant. Where at 1.3v the core temps were running in the mid 70c range, at 1.1v they dropped to 68-70c! 4-6 degrees cooler is a nice plus, especially because that means the fan doesn’t spin as fast either so it runs quieter.
There are two programs that can be used for this, one is K10Stat, the other is BrazosTuner. Both work fine. Despite both claiming to allow you to overclock neither actually can, you can use them to lower the clock speeds (more power/temp savings), but while CPUz will report the overclocking it is not actually happening.
Underclocking and undervolting are not nearly as hazardous as overclocking and overvolting, the worst case scenario when underclocking and/or undervolting is that the operating system becomes corrupted and you have to reinstall.