Introduction and Setup Impressions

Over the last couple of years, mini-PCs in the ultra-compact form factor (UCFF) have emerged as one of the bright spots in the troubled PC market. Zotac is no stranger to this segment. Even though the Intel NUC is credited with kicstarting the UCFF trend, the Zotac nano xs units actually came to the market before them. With the nano xs, Zotac redefined the small form-factor (SFF) PC. The ZBOX C Series nano units were launched to cater to the demand for fanless mini-PCs. The nano series units use slightly bigger motherboards, but they are still small enough to mount discreetly behind monitors with the supplied VESA mounts.

Most of the mini-PC / UCFF PCs in the market are based on Intel CPUs. Interestingly, Zotac launched an AMD-based SKU with the C series - the Zotac ZBOX CA320 nano. Based on a Temash APU (originally meant for the tablet market), the product does stand out compared to the other alternatives at a similar price point (based on the Intel Bay Trail SoCs).

Traditionally, Zotac samples PLUS units for review, and the CA320 nano was no different. The ZBOX C Series is one of the few lineups for which a PLUS model makes sense - the bundled 2.5" drive is a SSD, and there is only one SO-DIMM slot. The fact that PLUS models come with only one memory stick is not an issue here. Other than being ready to go out of the box in terms of hardware, the unit does come barebones (no OS installed). The specifications of our Zotac ZBOX CA320 nano review configuration are summarized in the table below.

Zotac ZBOX CA320 nano Specifications
Processor	AMD A6-1450 (4C/4T x 1.0 GHz, 28nm, 2MB L2, 8W TDP)
Memory	1x 4GB DDR3L-1600
Graphics	AMD Radeon HD 8250
Disk Drive(s)	FORESEE 64 GB 2.5" SSD
Networking	1x Gigabit Ethernet, 1x1 802.11ac/Bluetooth mPCIe
Audio	Capable of 5.1/7.1 digital output with HD audio bitstreaming (HDMI)
Operating System	Retail unit is barebones, but we installed Windows 8.1 Pro x64
Pricing (As configured)	$270
Full Specifications	Zotac ZBOX CA320 nano PLUS

The Zotac ZBOX CA320 nano kit doesn't come with any pre-installed OS, but does come with a CD and a read-only USB key containing the drivers. In any case, we ended up installing the latest drivers downloaded off Zotac's product support page. In addition to the main unit, the other components of the package include a 40 W (19V @ 2.1A) adapter, a US power cord, a VESA mount (along with the necessary screws), a single 2.4 GHz / 5 GHz antenna for the Wi-Fi feature, a driver CD / read-only USB key, user's manual and a quick-start guide. The unit is quite easy to open up (even without a screwdriver), but it is not necessary to do it unless one wants to replace either the memory or the disk drive.

The gallery below takes us around the hardware in the unit.

In the table below, we have an overview of the various systems that we are comparing the Zotac ZBOX CA320 nano against. Note that they may not belong to the same market segment. The relevant configuration details of the machines are provided so that readers have an understanding of why some benchmark numbers are skewed for or against the Zotac ZBOX CA320 nano when we come to those sections.

Comparative PC Configurations
Aspect	Zotac ZBOX CA320 nano	Zotac ZBOX CA320 nanoIntel D54250WYKHGIGABYTE GB-BXBT-1900GIGABYTE GB-BXi7-4500Zotac ZBOX CI540 nanoZotac ZBOX OI520 PlusECS LIVA
CPU	AMD A6-1450	AMD A6-1450
GPU	AMD Radeon HD 8250	AMD Radeon HD 8250
RAM	Crucial CT51264BF160B (Micron 8KTF51264HZ-1G6J1) 11-11-11-28 @ 1600 MHz 1x4 GB	Crucial CT51264BF160B (Micron 8KTF51264HZ-1G6J1) 11-11-11-28 @ 1600 MHz 1x4 GB
Storage	FORESEE S600S064G (64 GB; 2.5in SATA 6Gb/s; MLC)	FORESEE S600S064G (64 GB; 2.5in SATA 6Gb/s; MLC)
Wi-Fi	Intel Dual Band Wireless-AC 3160 (1x1 802.11ac - 433 Mbps)	Intel Dual Band Wireless-AC 3160 (1x1 802.11ac - 433 Mbps)
Price (in USD, when built)	$270	$270

Performance Metrics - I

The Zotac ZBOX CA320 nano was evaluated using our standard test suite for low power desktops / industrial PCs. We revamped our benchmark suite earlier this year after the publication of the Intel D54250WYK NUC review. We reran some of the new benchmarks on the older PCs also, but some of them couldn't be run on loaner samples. Therefore, the list of PCs in each graph might not be the same.

Futuremark PCMark 8

PCMark 8 provides various usage scenarios (home, creative and work) and offers ways to benchmark both baseline (CPU-only) as well as OpenCL accelerated (CPU + GPU) performance. We benchmarked select PCs for the OpenCL accelerated performance in all three usage scenarios. These scores are heavily influenced by the CPU in the system. The main system for comparison is the ECS LIVA, a fanless Bay Trail PC (that is alot cheaper, but comes with soldered DRAM and eMMC intead of a SO-DIMM slot and ability to take a 2.5" drive). Interesting, the AMD A6-1450 compares quite favourably with the Bay Trail Celeron J1900-based GIGABYTE BXBT-1900, even though the latter is actively cooled. Other than that, there is no surprise in these graphs - particularly, the performance lag when compared to a Haswell-Y CPU with similar TDP levels in the ZBOX CI540 nano.

Miscellaneous Futuremark Benchmarks

One of the interesting aspects of the 3DMark 11 benchmarks was that we were unable to get Bay Trail-based units to complete any run other than the Entry-level workload. The Temash-based ZBOX CA320 nano had no trouble processing both the Entry and Extreme workloads. Both graphs are presented here. Temash's HD 8250 seems to be no match for the IGP in the Haswell Core i5-4210Y, but it handily surpasses the IGP in the Bay Trail SoCs. The latter aspect is the important point - Temash-based units are competing at the same price points as the Bay Trail-based ones.

3D Rendering - CINEBENCH R15

We have moved on from R11.5 to R15 for 3D rendering evaluation. CINEBENCH R15 provides three benchmark modes - OpenGL, single threaded and multi-threaded. Evaluation of select PCs in all three modes provided us the following results.

Single-threaded performance for Temash seems to be quite a bit worse compared to even the Bay Trail-based Celeron N2807 in the ECS LIVA. However, multi-threaded performance is better for the Cinebench workload, thanks mainly to the presence of double the number of cores / threads (4 vs. 2) in the A6-1450. The GPU capabilities as exposed by the OpenGL run (and also seen in the 3DMark benches) is also better for Temash compared to Bay Trail.

Performance Metrics - II

In this section, we mainly look at benchmark modes in programs used on a day-to-day basis, i.e, application performance and not synthetic workloads.

x264 Benchmark

First off, we have some video encoding benchmarks courtesy of x264 HD Benchmark v5.0. This is simply a test of CPU performance. As expected, the quad-core A6-1450 can give a good fight to the dual-core Bay Trail-based Celeron N2807 in the ECS LIVA. However, it is no match for the quad-core Celeron J1900 at much higher clocks (but, one which also has a much higher TDP and is actively cooled).

7-Zip

7-Zip is a very effective and efficient compression program, often beating out OpenCL accelerated commercial programs in benchmarks even while using just the CPU power. 7-Zip has a benchmarking program that provides tons of details regarding the underlying CPU's efficiency. In this subsection, we are interested in the compression and decompression MIPS ratings when utilizing all the available threads.

TrueCrypt

As businesses (and even home consumers) become more security conscious, the importance of encryption can't be overstated. The AMD A6-1450 APU does have accelerated AES capabilities, but the lower clock speeds tend to pull down the performance a bit. TrueCrypt, a popular open-source disk encryption program can take advantage of the AES capabilities. The TrueCrypt internal benchmark provides some interesting cryptography-related numbers to ponder. In the graph below, we can get an idea of how fast a TrueCrypt volume would behave in the Zotac ZBOX CA320 nano and how it would compare with other select PCs. This is a purely CPU feature / clock speed based test.

Agisoft Photoscan

Agisoft PhotoScan is a commercial program that converts 2D images into 3D point maps, meshes and textures. The program designers sent us a command line version in order to evaluate the efficiency of various systems that go under our review scanner. The command line version has two benchmark modes, one using the CPU and the other using both the CPU and GPU (via OpenCL). The benchmark takes around 50 photographs and does four stages of computation:

• Stage 1: Align Photographs
• Stage 2: Build Point Cloud (capable of OpenCL acceleration)
• Stage 3: Build Mesh
• Stage 4: Build Textures

We record the time taken for each stage. Since various elements of the software are single threaded, others multithreaded, and some use GPUs, it is interesting to record the effects of CPU generations, speeds, number of cores, DRAM parameters and the GPU using this software.

Dolphin Emulator

Wrapping up our application benchmark numbers is the Dolphin Emulator benchmark mode results. This is again a test of the CPU capabilities. This benchmark appears to be heavily influenced by the clock speeds - the performance difference between the Celeron N2807 in the ECS LIVA is more pronounced compared to the other benchmarks.

Networking and Storage Performance

We have recently started devoting a separate section to analyze the storage and networking credentials of the units under review. On the storage side, one option would be repetition of our strenuous SSD review tests on the drive(s) in the PC. Fortunately, to avoid that overkill, PCMark 8 has a storage bench where certain common workloads such as loading games and document processing are replayed on the target drive. Results are presented in two forms, one being a benchmark number and the other, a bandwidth figure. We ran the PCMark 8 storage bench on selected PCs and the results are presented below.

The FORESEE 64 GB SSD is a OEM-only part and hasn't been subject to extensive benchmarking. Exact specifications are unknown, though we do have a PDF of the specifications of the 128 GB model from the same company. The 128 GB model indicates sustained read/write speeds of 120/90 MBps. The controller appears to be JMicron's JM612. The 64 GB should have similar (or, slightly worse) specifications. Our PCMark8 storage bench run (storage bandwidth, in particular) seems to track the specifications. In terms of the storage subsystem score, the FORESEE SSD is not a top performer, but it definitely is a better choice compared to a traditional HDD (in the ZBOX OI520 PLUS) or eMMC (in the ECS LIVA).

On the networking side, we restricted ourselves to the evaluation of the WLAN component. Our standard test router is the Netgear R7000 Nighthawk configured with both 2.4 GHz and 5 GHz networks. The router is placed approximately 20 ft. away, separated by a drywall (as in a typical US building). A wired client (Zotac ID89-Plus) is connected to the R7000 and serves as one endpoint for iPerf evaluation. The PC under test is made to connect to either the 5 GHz (preferred) or 2.4 GHz SSID and iPerf tests are conducted for both TCP and UDP transfers. It is ensured that the PC under test is the only wireless client for the Netgear R7000. We evaluate total throughput for up to 32 simultaneous TCP connections using iPerf and present the highest number in the graph below.

In the UDP case, we try to transfer data at the highest rate possible for which we get less than 1% packet loss.

The WLAN component in the CI540 nano, Intel's D54250WYKH Haswell NUC, the ZBOX OI520 PLUS and the CA320 nano are the same - Intel's Dual-Band Wireless-AC 3160. There is no surprise that the TCP throughputs are similar (differences can be attributed to antenna placement / chassis design). However, we are yet to determine the reason behind the lower UDP throughput on the CA320. Despite repeating the test multiple times, we didn't see any change in the observed numbers.

HTPC Credentials

The absence of any moving parts inside the ZBOX CA320 nano enables a completely silent PC irrespective of the workload. This makes it an ideal HTPC. While acoustics form one part of the HTPC story, there are a few other aspects that we will cover in this section.

Refresh Rate Accurancy

AMD and NVIDIA have historically been able to provide fine-grained control over display refresh rates. The default rates are also quite accurate. Intel used to have an issue with 23 Hz (23.976 Hz, to be more accurate) support, but that was resolved with the introduction of Haswell. As expected, the Zotac ZBOX CA320 nano has no trouble with refreshing the display appropriately in the 23 Hz setting. In fact, the accuracy is much more compared to what we have been used to seeing in AMD-based PCs and AMD GPUs over the last few years.

The gallery below presents some of the other refresh rates that we tested out. The first statistic in madVR's OSD indicates the display refresh rate.

Network Streaming Efficiency

Evaluation of OTT playback efficiency was done by playing back our standard YouTube test stream and five minutes from our standard Netflix test title. Using HTML5, the YouTube stream plays back a 720p encoding, while Adobe Flash delivers a 1080p stream. Note that only NVIDIA exposes GPU and VPU loads separately. Both Intel and AMD bundle the decoder load along with the GPU load. The following two graphs show the power consumption at the wall for playback of the HTML5 stream and the Adobe Flash stream in Mozilla Firefox v33.1.1 using Adobe Flash 15.0.0.223 and AMD Catalyst 14.9 drivers. OTT streaming doesn't seem to be as efficient as what we had in the ECS LIVA and the GIGABYTE BXBT-1900. While the HTML5 stream had an average GPU load of 37.68%, the Flash stream loaded it to only 28.44%. 

Netflix streaming evaluation was done using the Windows 8.1 Netflix app. Manual stream selection is available (Ctrl-Alt-Shift-S) and debug information / statistics can also be viewed (Ctrl-Alt-Shift-D). Statistics collected for the YouTube streaming experiment were also collected here. The average GPU load was around 17%.

Decoding and Rendering Benchmarks

The specifications of the A6-1450 APU make is very clear that the HTPC is not meant for usage with madVR. In order to evaluate local file playback, we concentrated on EVR-CP (using MPC-HC v1.7.7) and Kodi (v 14.0-b5). We already know that EVR works quite well even with the Intel IGP for our test streams. The decoder used was LAV Filters v0.63.0 (default configuration) bundled with MPC-HC v1.7.7. We left the video playback settings at default (using DXVA2) in Kodi also.

Zotac ZBOX CA320 nano - Decoding & Rendering Performance
Stream	EVR-CP		Kodi
	GPU Load (%)	Power (W)	GPU Load (%)	Power (W)
480i60 MPEG2	53.82	10.41	29.28	8.84
576i50 H264	47.36	10.63	32.13	10.27
720p60 H264	72.72	11.84	57.51	11.26
1080i60 MPEG2	98.04	12.50	57.86	10.86
1080i60 H264	97.12	12.56	77.64	11.78
1080i60 VC1	96.91	12.45	71.67	11.62
1080p60 H264	57.13	11.23	45.43	10.33
1080p24 H264	23.41	9.71	32.32	9.43
4Kp30 H264	31.65	11.63	22.92	9.26

One of the unfortunate aspects was that the interlaced streams caused too much loading on the GPU and the frame drops during playback was quite obvious. On the other hand, the 1080p60 stream, despite being light on the GPU had stuttered playback, enough to be unwatchable. 4K decoding is not supported by the Temash GPU and the CPU is not strong enough to support software decoding. The 4Kp30 playback was a slideshow. On the other hand, things were a little bit better with Kodi (though deinterlacing wasn't perfect for some streams). Frame drops during playback for the interlaced streams, if any, were not discernible. However, the issues seen with EVR-CP / DXVA2 decoding for the 1080p60 and 4Kp30 streams were also seen in Kodi.

Power Consumption and Thermal Performance

The power consumption at the wall was measured with a 1080p display being driven through the HDMI port. In the graphs below, we compare the idle and load power of the Zotac ZBOX CA320 nano with other low power PCs evaluated before. For load power consumption, we ran Furmark 1.12.0 and Prime95 v27.9 together. The numbers are not beyond the realm of reason for the combination of hardware components in the machine.

We do find that the idle power consumption is lower than the ZBOX CI540 nano, but it doesn't match the numbers set by the ECS LIVA. As expected, the load power consumption tracks the inherent capabilities of the CPU in the system.

We evaluated the thermal performance from a system perspective.  We start with the system at idle, followed by 30 minutes of pure CPU loading. This is followed by another 30 minutes of both CPU and GPU being loaded simultaneously. After this, the CPU load was removed, allowing the GPU to be loaded alone for another 30 minutes. The various clocks in the system as well as the temperatures within the unit are presented below.

At the outset, it must be said that we didn't observe any throttling in action. The CPU cores are advertised for 1 GHz and they maintain that frequency throughout the pure CPU loading stage. The GPU cores idle around 200 MHz. After the GPU load kicks in, the Radeon HD 8250 cranks up to 400 MHz, while the CPU cores move down to the 800 - 900 MHz (and occasionally up to 1 GHz) range to obey the TDP limits. After removal of the CPU load, the cores dial down to 200 MHz (just cranking up once in a while). On the temperature side of things, the core and GPU track each other closely. The junction temperature is around 90 C, and the maximum temperature we encountered in our stress test was slightly above 80 C. Idling temperature was around 50 C.

Another important aspect to keep note of while evaluating fanless PCs is the chassis temperature. Using Seek Thermal's thermal imager, we observed the chassis temperature after the CPU package temperature reached the steady state value in the above graph.

Surprisingly, the chassis temperature reached only 56 C after full loading. Compared to the 75 C+ that we saw with the ZBOX CI540 nano, this is really cool. Some of the other thermal pictures we took of the unit after the full loading process are reproduced in the gallery below.

Final Words

The Zotac ZBOX CA320 nano provided us with the first opportunity to evaluate a passively cooled mini-PC based on an AMD APU. Passively cooled systems are either very costly (particularly if they integrate powerful CPUs) or downright abysmal in performance (when they integrate the low-end / low-power CPUs such as the older Atoms). Zotac's offering with the ZBOX CA320 nano aims to strike a balance. $175 for a barebones configuration is quite reasonable for this type of system. With the bundled SSD and RAM, it is still less than $300.

One of the aspects we were worried about was thermal throttling, but the ZBOX CA320 nano surpassed our expectations. The chassis never got extremely hot (reaching only around 55 C, even after extended thermal stress with a couple of power viruses).

Pretty much the only downside of the unit is the relatively bad single-threaded performance of the AMD A6-1450's CPU cores and the HTPC aspects of the AMD GPU drivers. The clock rates are a bit low. Given the thermal headroom that seems to be available, Zotac could have been a tad more adventurous in overclocking. While the BIOS managed to pull up the DRAM frequency, the APU itself was clocked as per specifications. However, we shouldn't be really complaining since the system seems to operate quite nicely for day-to-day use. The SSD could be a bit better, but that is not an issue if the end-user buys a barebones configuration.

All in all, Zotac manages to deliver a very price-effective passive mini-PC in the ZBOX CA320 nano. Along with the ECS LIVA in the market, the days of users having to spend an arm and leg for passively cooled systems with decent performance are history. Zotac also has a Bay Trail-based ZBOX CI320 nano in the fanless C series, and we will be looking at that system next month.