Gigabyte’s Fusion-infused GA-E350N-USB3 motherboard

Manufacturer Gigabyte
Model GA-E350N-USB3
Price (street) $150
Availability Now

Four and a half years have passed since AMD acquired ATI and announced its Fusion plans for world domination. That’s how long it’s taken for the first accelerated processing unit (APU) to emerge. Make that two APUs: Ontario and Zacate. The former is a 9W part best suited to inexpensive netbooks like Acer’s Aspire One 522. Zacate uses the same silicon but offers slightly more power within a 18W thermal envelope that’s perfect for budget ultraportables, lightweight desktops, and home-theater PCs.

These first Fusion APUs are more Atom killers than true world beaters. They don’t have nearly enough horsepower to keep up with desktop CPUs, including budget models. However, they offer snappy performance for basic tasks and an integrated graphics processor capable of fueling smooth HD video playback. The on-die Radeon can handle light gaming provided you stick to smaller indie titles. In the more powerful Zacate-based E-350, the Radeon HD 6310 is will even produce playable frame rates in older big-name games.

The E-350 is teamed with AMD’s Hudson M1 platform hub, a sort of miniature but very modern core-logic chipset. In addition to four second-generation PCI Express lanes (complementing four lanes in the APU), Hudson M1 serves up a quartet of 6Gbps Serial ATA ports and a boatload of USB 2.0 ports. That foundation is a great basis for Mini-ITX motherboards like Gigabyte’s GA-E350N-USB3, which adds USB 3.0, Gigabit Ethernet, and a BIOS loaded with overclocking options. This marriage of a low-power core with high-performance peripherals is sort of a second degree fusion—like a pulled-pork sushi roll that’s been popped into the deep fryer.

As a sucker for all things deep fried, and for Mini-ITX, I had to check out the E350N for myself. Does its APU have much overclocking headroom? Can the board’s high-performance peripherals keep up with those on full-sized desktop mobos? Is this the perfect starting point for a mini HTPC? Let’s find out.

A tour of the board

Here’s a quick look at the E350N’s rap sheet:

APU AMD E-350 w/Radeon HD 6310 IGP
Platform hub AMD Hudson M1
DIMM slots 2 DDR3-1333
Expansion slots 1 PCIe x16 (x4 bandwidth)
Storage I/O 4 6Gbps SATA RAID via Hudson M1
Audio 8-channel HD via Realtek ALC892
Ports 1 PS/2 keyboard/keyboard

1 DVI

1 HDMI

1 VGA

2 USB 3.0 via NEC D720200

4 USB 2.0 w/ 4 headers
1 RJ45 via Realtek RTL8111E

1 analog front out
1 analog bass/center out
1 analog rear out
1 analog surround out

1 analog line in

1 analog mic in

1 optical S/PDIF out

That’s quite a lot of hardware for something that costs $150. All you need is an enclosure, hard drive, and memory to form a complete system.

Lest you think Gigabyte has adopted a generic black-and-blue color scheme across its entire lineup, there are plenty of models sticking with the old-school turquoisey blue. Amen. I especially dig how the heatsink’s dark-grey metal contrasts with the board’s light-blue accents.

Beneath the hunk of finned metal sits a power-efficient Brazos tandem that runs quietly in ultraportable notebooks like HP’s Pavilion dm1z. Despite the fact that the E350N is surely headed for much airier confines, Gigabyte has equipped the heatsink with a 40-mm cooling fan. The fan is all but inaudible from a few feet away—today. However, as we’ve often noted, tiny fans like this one tend to develop an annoying whine over time. That’s the last thing you want coming from a home-theater PC in the living room or a lightweight desktop tucked into the corner of a dorm or bedroom. I’d be more forgiving if the board were more cramped, but there’s loads of room for a larger passive cooler.

Gigabyte hasn’t let the E350N escape the clutches of its Ultra Durable 3 branding. That’s just a fancy way of saying that the board uses higher-quality electrical components and thicker copper layers. Those little touches are common on enthusiast-oriented desktop boards but less often found in the Mini-ITX realm.

Space constraints also prevent some Mini-ITX boards from offering full-sized DIMM slots. Not so with the E350N, which can host a pair of DDR3 modules at speeds up to 1333MHz. Users can dedicate as much as a gig of system memory to the APU’s integrated Radeon.

The presence of a PCI Express x16 slot would usually get me excited about the prospect of building a pint-sized gaming rig. However, even with a powerful graphics card riding shotgun, the E-350’s CPU cores are simply too slow to keep up with modern games. At least the PCIe slot gives the board some flexibility on the expansion front. Look closely, and you can see that the slot has only four lanes of electrical connectivity. All of those lanes stem from the E-350 APU.

Just beyond the PCIe slot in the picture above sits a cluster of SATA ports linked to the Hudson M1 chipset. 6Gbps connectivity is probably overkill for a system of this caliber, but it’s nothing to complain about. The lack of RAID support might upset folks considering the E350N for a home storage server, though.

Of course, burying this board inside a closet with a bunch of hard drives would be a waste. The DVI and HDMI outputs are there for a reason, and multi-channel digital audio can be passed over HDMI to a compatible television or receiver in fancy formats like TrueHD and DTS-HD. You also get a separate S/PDIF audio output, although the board’s audio codec doesn’t provide real-time multichannel encoding or headphone virtualization.

Gigabyte kicks in a couple of USB 3.0 ports, and the addition of Gigabit Ethernet is a nice touch. However, I’m a little bummed that there’s no Wi-Fi onboard. Wireless networking can always be added via a USB dongle, but I’d rather see it integrated—especially on a product with living-room aspirations.

The E350N’s BIOS has a lot more overclocking options than one might expect from a Mini-ITX board. I have to question Gigabyte’s priorities, though. Here we have a BIOS that lets users increase the base clock speed by up to 20% and take the GPU clock from a default of 500MHz all the way up to 2GHz. There are voltage controls for no fewer than five system variables, including the USB 3.0 chip. All the important memory timings can be tweaked, and you can choose between a couple of memory bus speeds.

For some reason, however, fan speed controls are nowhere to be found. That makes about as much sense as putting an adjustable spoiler and suspension on your mom’s grocery getter… and then ripping out the climate control. I guess Gigabyte thinks more people are interested in overclocking a Brazos board than putting one inside a quiet desktop or media box. Odd.

Our testing methodsWhile it may be entirely inappropriate to compare the GA-E350N-USB3 to full-sized desktop motherboards, we’re gonna do it anyway. You see, we already know that the E-350’s Bobcat cores are no match for modern desktop CPUs. We don’t, however, know how its Hudson platform hub stacks up. There’s no better way to find out than throwing Hudson into the mix against cutting-edge desktop chipsets riding the latest enthusiast motherboards.

To ensure that the E350N at least has someone its own size to pick on, we’ve also included some memory and application performance results from the MSI E350IA-E45 motherboard in Scott’s E-350 review. The MSI board has the same 1.6GHz APU and should offer equivalent performance. However, while the MSI board was run with 4GB of 1066MHz memory at 7-7-7-20-2T timings, the Gigabyte had 8GB of RAM cranked to 1333MHz with 9-9-9-24-1T timings. Scott also used a different hard drive and PSU, so I haven’t included his results in our power consumption tests.

To focus our attention, I’ve greyed out the desktop motherboards in some of our performance graphs. The E350N, its MSI twin, and our chipset results have all been color-coded to make the graphs easier to read.

With few exceptions, all tests were run at least three times, and we reported the median of the scores produced.

Processor AMD E-350 1.6GHz AMD Phenom II X6 1090T 3.2GHz Intel Core i7-870 2.93GHz Intel Core i7-2600K 3.4GHz
Motherboard Gigabyte GA-E350N-USB3 Asus M4A89GTD PRO/USB3 Gigabyte GA-P55A-UD4P Asus P8P67 PRO Intel DP67BG Gigabyte GA-P67A-UD4P MSI P67A-GD65 Zotac H67-ITX
Bios revision F1 1606 F14 0907 10J.86A.1780 F5R E7681IMS.152 A166
Platform hub AMD Hudson M1 AMD 890FX Intel P55 Express Intel P67 Express Intel P67 Express Intel P67 Express Intel P67 Express Intel H67 Express
South bridge AMD SB850
Chipset drivers AHCI: 1.2.1.269 Catalyst 10.12 Chipset: 9.2.0.1019

RST: 10.1

Chipset: 9.2.0.1019

RST: 10.1

Chipset: 9.2.0.1019

RST: 10.1

Chipset: 9.2.0.1019

RST: 10.1

Chipset: 9.2.0.1019

RST: 10.1

Chipset: 9.2.0.1019

RST: 10.1

IGP: 15.21.5.64.2266

Memory size 8GB (2 DIMMs) 8GB (2 DIMMs) 8GB (2 DIMMs) 8GB (2 DIMMs) 8GB (2 DIMMs) 8GB (2 DIMMs) 8GB (2 DIMMs) 8GB (2 DIMMs)
Memory type Corsair Vengeance DDR3 SDRAM at 1333MHz Corsair Vengeance DDR3 SDRAM at 1333MHz Corsair Vengeance DDR3 SDRAM at 1333MHz Corsair Vengeance DDR3 SDRAM at 1333MHz Corsair Vengeance DDR3 SDRAM at 1333MHz Corsair Vengeance DDR3 SDRAM at 1333MHz Corsair Vengeance DDR3 SDRAM at 1333MHz Corsair Vengeance DDR3 SDRAM at 1333MHz
Memory timings 9-9-9-24-1T 9-9-9-24-1T 9-9-9-24-1T 9-9-9-24-1T 9-9-9-24-1T 9-9-9-24-1T 9-9-9-24-1T 9-9-9-24-1T
Audio Realtek ALC892 with 2.55 drivers Realtek ALC892 with 2.55 drivers Realtek ALC889 with 2.55 drivers Realtek ALC892 with 2.55 drivers Realtek ALC892 with 2.55 drivers Realtek ALC892 with 2.55 drivers Realtek ALC892 with 2.55 drivers Realtek ALC892 with 2.55 drivers
Graphics Radeon HD 6310 with 8.792.0.0 drivers Asus EAH5870 1GB with Catalyst 10.12 drivers
Hard drive Raptor WD1500ADFD 150GB
Power Supply PC Power & Cooling Silencer 750W
OS Microsoft Windows 7 Ultimate x64

We’d like to thank Asus, Corsair, and Western Digital for helping to outfit our test rigs with some of the finest hardware available. Thanks to each of the motherboard makers for supplying their boards, too, and to AMD and Intel for providing the CPUs.

We used the following versions of our test applications:

The test systems’ Windows desktop was set at 1280×1024 in 32-bit color at a 60Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.

All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.

Memory performance

Don’t even bother comparing the memory performance of the desktop boards to our Fusion offerings. It’s not going to be pretty.

Even the Zotac H67 board running with its integrated graphics enabled offers gobs more bandwidth and much lower latency than the E-350 boards.

The Gigabyte board scores much higher than the MSI in the bandwidth test thanks to its faster memory clock.  MSI does support a 1333MHz memory clock, too, but we didn’t use it. Our disparate test configs have given Gigabyte a bit of an unfair advantage. Technically, we’re overclocking the memory controller on the Gigabyte board, since AMD doesn’t officially sanction 1333MHz memory speeds on the Brazos platform.

The gap between the two motherboards is only a couple of nanoseconds quicker when we look at memory access latencies, though.

Application performance

Think the E-350 can hang with Sandy Bridge, Lynnfield, and a six-core Phenom in a handful of application benchmarks?

Think again. As expected, there’s no difference in performance between the MSI and Gigabyte E-350 boards. The Gigabyte config’s faster memory doesn’t deliver higher performance in these tests.

Serial ATA performance — HD Tune

Even without RAID support, Hudson M1’s 6Gbps Serial ATA controller looks pretty impressive on paper. Let’s see how it handles a couple of 6Gbps drives starting with the fastest mechanical model of them all: Western Digital’s latest VelociRaptor.

Hudson’s burst speeds aren’t quite fast enough to catch the P67 Express. That said, the Brazos platform’s SATA controller is quicker to the draw than the P55 and 890GX.

Things continue to look promising for Hudson in HD Tune’s sustained transfer rate tests.

Wouldn’t you know, the first Fusion Controller Hub has impressive SATA access times, too. So, what happens when we plug a 6Gbps Crucial RealSSD C300 into the board?

The P67 Express gets even faster, while Hudson maintains a healthy lead over the P55 and 890GX.

We see wider gaps in HD Tune’s sustained read and write speed tests when they’re running on the SSD. Again, Hudson is second only to Sandy Bridge’s core-logic sidekick—you know, the one with broken SATA ports.

SSD access times are very quick indeed. Hudson has a nice lead with writes but pulls up a little slower than the Intel chipsets with reads.

Serial ATA performance — IOMeter

Next, we tackle IOMeter, which hammers drives with increasing I/O loads. We’ve restricted our testing to IOMeter’s workstation and database access patterns because they’re slightly less irrelevant to the sorts of workloads likely to face a Zacate-based system. IOMeter makes good use of the Native Command Queuing capability built into the AHCI specification, and as one might expect, it loves the quick access times of solid-state drives. First, though, let’s see how things shake out with the ‘raptor.

Things are looking good for Hudson in IOMeter.

However, the board does burn through more CPU cycles than the other platforms. Considering Zacate is going up against full-fat desktop CPUs, I’m surprised the gaps aren’t larger. Perhaps the picture will change when the C300 is crunching our IOMeter workloads.

Not really—the performance of Hudson’s Serial ATA controller once again shadows that of the desktop chipsets. The P55 falls a little short because its SATA interface is limited to 3Gbps.

Overall CPU utilization is much higher this time around, especially for the Fusion platform. Yikes. Chewing through nearly 20,000 IOps requires a fair amount of CPU power, and there’s only so much on tap in the E-350.

Serial ATA performance — TR DriveBench

TR DriveBench simulates disk-intensive multitasking sessions by playing back pre-recorded traces of all disk activity associated with different workloads. You can read more about DriveBench on this page of our latest SSD round-up. We’re only using the file copy and virus scanning traces today, since they’re the most demanding of the workloads we’ve concocted. In fact, both are likely workloads for a Brazos-based system.

Interesting. The Fusion chipset performs well with our mechanical hard drive, but it falls a little behind the leaders with the SSD. Even in its poorest showing, Hudson is only about 10% slower than the P67 Express.

USB 2.0 performance

AMD’s Hudson M1 platform hub may not include SuperSpeed USB connectivity, but it does have a handful of old-school USB 2.0 ports. How fast are they?

Pretty quick, all things considered. Hudson’s read speeds are a little slower than its desktop competition. However, the AMD chipsets have quicker writes. 13% CPU utilization isn’t even that bad considering that the six-core Phenom is at 7%.

PCI Express performance

We now test PCI Express performance with an auxiliary storage controller that has a PCIe 2.0 x1 interface plus two 6Gbps Serial ATA ports. To those ports, we attach a pair of SandForce-powered SSDs configured in a RAID 0 array. This RAID card is usually plugged into a motherboard’s PCIe x1 slot, but we’ve used the x16 slot on the E350N because that’s the only expansion option available. Since the slot’s PCI Express lanes are coming directly from the APU rather than from the Hudson platform hub, we’ve labeled the results “E-350” in the graphs below. Our 890GX board has PCIe slots connected to PCI Express lanes stemming from its 890GX north-bridge and SB750 south-bridge components, and we’ve tested both.

Excellent. The E-350’s PCI Express lanes have no trouble keeping up with those offered by the P67 Express.

Power consumption

We measured system power consumption, sans monitor and speakers, at the wall outlet using a Watts Up Pro power meter. Readings were taken at idle and under a load consisting of a Cinebench 11.5 render alongside the rthdribl HDR lighting demo. We tested with Windows 7’s High Performance and Balanced power plans.

Motherboard makers usually ship their boards with energy-saving features that promise to lower power consumption without resorting to CPU throttling that might hinder performance. We’ve tested the Asus, MSI, and Gigabyte desktop boards with their power-saving features enabled and disabled. The Intel and Zotac boards aren’t equipped with such features, and neither is the E350N. Those boards were tested with only Windows handling power management.

You’ll notice two E350N configurations in the graphs below. One uses the same PC Power & Cooling Silencer 750W PSU as the other systems. 750W is definitely overkill for a Zacate-based system, so we’ve also tested with the 200W PSU from Thermaltake’s budget Element Q Mini-ITX enclosure (look for a full review soon). Those results are labeled “200W” in the graphs. For our purposes today, you’ll want to look at how the E350N compares with the Zotac H67 IGP config, which is the only other one not equipped with a discrete graphics card.

Surprisingly, the 750W Silencer is more efficient than the low-wattage Thermaltake PSU. The Thermaltake case-and-PSU combo costs quite a bit less than the standalone Silencer power supply, though. For what it’s worth, Scott observed lower idle power consumption from the MSI E-350 board using a different system configuration and power meter. He also used a laptop-style 85W PSU.

The key takeaway here is that Zacate-based systems have much lower power draw than enthusiast desktops. While the Zotac Mini-ITX board draws to within striking distance at idle, the very same system pulls nearly 2.5 times more power than the E350 under load.

Overclocking

Many have been curious about how AMD’s first Fusion APU would respond to overclocking. With that in mind and a dashboard full of clock-speed and voltage controls at my fingertips, I couldn’t resist turning the screws on the E350N. To keep our memory modules out of the equation, the DRAM multiplier was lowered to yield a 1066MHz memory speed with the default 100MHz base clock. With that done, I turned my attention to increasing the base clock speed.

A 110MHz base clock yielded a CPU speed of 1.76GHz according to CPU-Z. The system was perfectly stable at stock voltages with a combined workload of Prime95 and the rthdribl HDR lighting demo, so I pushed a little harder. Windows managed to boot with a 115MHz base clock speed, and our stress test ran without a hitch. However, the system blue-screened when I tried to reboot and was never perfectly stable with a 115MHz base clock again. Fiddling with the numerous voltage options didn’t help, either.

After probing the limits of the E-350’s CPU cores, I started on the integrated Radeon. Everything went smoothly up to an 800MHz GPU clock. At that speed, the system crashed during our stress test and failed to recover gracefully. Stability remained elusive even after backing off the GPU clock to 750 and 700MHz—speeds that had been stable before our initial crash. Discouraged and frustrated, I gave up after confirming that the board still worked properly with its default 500MHz GPU clock.

As is always the case with overclocking, your mileage may vary. I didn’t bother with our usual virgin sacrifice this time around, and that might’ve upset the gods with influence over such matters. Based on the fact that Zacate and Ontario were designed to be low-power parts with relatively modest clock speeds, I wouldn’t expect either to have substantial overclocking headroom.

Motherboard peripheral performance

Our last stop on the testing front is the wonderful world of onboard peripherals. Can the E350N’s onboard goodies keep up with what’s offered by the big boys?

  HD Tach USB 3.0 performance
  Read burst

speed (MB/s)

Average read

speed (MB/s)

Average write

speed (MB/s)

CPU utilization

(%)

Asus 890GX 153.2 125.8 48.3 12
Gigabyte P55 163.4 153.7 55.2 1
Asus P67 182.2 169.5 55.6 2
Gigabyte P67 224.5 179.3 61.2 2
Intel P67 195.8 170.1 53.0 2
MSI P67 215.8 175.8 60.0 2
Zotac H67 164.7 163.9 55.5 3
Zotac H67 (IGP) 160.5 162.7 55.9 3
Gigabyte E-350 154.0 143.5 49.0 29

So far, so good. The E350N’s USB 3.0 transfer rates aren’t quite as speedy as those of our full-sized desktop mobos, but they’re in the same ballpark. The one obvious exception is CPU utilization, which is understandably higher on the E350N.

  HD Tach USB 2.0 performance
  Read burst

speed (MB/s)

Average read

speed (MB/s)

Average write

speed (MB/s)

CPU utilization

(%)

Asus 890GX 33.2 30.7 25.2 7
Gigabyte P55 34.3 32.5 22.5 3
Asus P67 36.6 34.6 24.6 4
Gigabyte P67 37.6 36.8 26.2 2
Intel P67 34.9 32.4 22.9 2
MSI P67 36.4 34.3 23.9 2
Zotac H67 37.5 34.4 24.2 2
Zotac H67 (IGP) 37.5 33.7 24.1 1
Gigabyte E-350 31.2 29.4 24.2 13

We’ve graphed most of these USB 2.0 results already, so I’ll be brief. The E350N’s USB 2.0 performance is solid as long as you can live with slightly higher CPU utilization.

  HD Tune Serial ATA performance – VelociRaptor
  Read Write
  Burst (MB/s) Average (MB/s) Random 4KB (ms) Burst (MB/s) Average (MB/s) Random 4KB (ms)
Asus 890GX 199.6 125.5 7.3 199.8 124.0 2.4
Gigabyte P55 213.9 129.9 7.0 213.8 128.1 2.7
Gigabyte P55 (Marvell) 240.9 129.9 7.2 240.8 127.7 2.7
Asus P67 283.4 129.7 7.3 252.4 123.4 2.7
Asus P67 (Marvell) 201.8 129.7 7.2 203.0 92.3 2.6
Gigabyte P67 292.6 129.9 7.0 295.1 126.1 2.7
Intel P67 260.3 129.8 7.3 256.2 123.0 2.6
MSI P67 235.2 129.4 7.2 232.8 124.8 2.6
MSI P67 (Marvell) 189.8 128.5 7.3 187.0 87.0 2.6
Zotac H67 281.8 129.7 7.3 241.6 123.6 2.6
Zotac H67 (IGP) 286.1 129.6 7.2 263.8 124.0 2.7
Gigabyte E-350 239.3 129.8 7.2 232.9 124.4 2.3

Hudson’s 6Gbps Serial ATA controller offers respectable performance whether you’re using a mechanical VelociRaptor…

  HD Tune Serial ATA performance – RealSSD
  Read Write
  Burst (MB/s) Average (MB/s) Random 4KB (ms) Burst (MB/s) Average (MB/s) Random 4KB (ms)
Asus 890GX 165.2 225.9 0.22 165.0 201.6 0.34
Gigabyte P55 181.4 236.8 0.13 180.7 200.2 0.43
Gigabyte P55 (Marvell) 196.2 258.6 0.17 190.8 177.7 0.38
Asus P67 198.0 292.8 0.15 194.5 210.6 0.36
Asus P67 (Marvell) 172.2 258.5 0.18 172.6 116.0 0.39
Gigabyte P67 242.2 317.2 0.13 245.8 215.8 0.40
Intel P67 243.2 307.3 0.14 222.4 215.2 0.40
MSI P67 183.3 287.0 0.17 186.5 206.6 0.38
MSI P67 (Marvell) 165.9 215.4 0.20 165.6 109.7 0.36
Zotac H67 196.0 287.6 0.16 196.0 208.7 0.36
Zotac H67 (IGP) 193.4 288.7 0.16 195.3 209.4 0.37
Gigabyte E-350 199.0 246.5 0.18 207.1 209.5 0.23

Or a fancy solid-state drive.

  NTttcp Ethernet performance
  Throughput (Mbps) CPU utilization (%)
Asus 890GX 940.3 11
Gigabyte P55 (1) 936.8 2.7
Gigabyte P55 (2) 945.4 2.1
Asus P67 924.9 1.9
Gigabyte P67 935.0 3.2
Intel P67 941.4 2.0
MSI P67 935.9 3.8
Zotac H67 938.8 3.8
Zotac H67 (IGP) 935.8 3.7
Gigabyte E-350 848.6 31.1

The E350N’s Gigabit Ethernet throughput is lower than one might expect given the performance of the other boards. I’m not sure whether Hudson’s PCIe interface is to blame for impeding throughput or if the E-350’s CPU cores are responsible for slowing performance in this test.

  RightMark Audio Analyzer audio quality
  Frequency response Noise level Dynamic range THD THD + Noise IMD + Noise Stereo Crosstalk IMD at 10kHz Overall score
Asus 890GX 5 4 4 5 3 5 5 5 5
Gigabyte P55 5 5 5 5 4 5 4 5 5
Asus P67 5 4 4 5 3 5 5 5 5
Gigabyte P67 5 4 4 5 3 5 5 5 5
Intel P67 5 4 4 5 3 5 5 5 5
MSI P67 5 4 4 5 3 5 5 5 5
Zotac H67 5 4 4 4 3 4 5 4 4
Zotac H67 (IGP) 5 4 4 4 3 4 5 4 4
Gigabyte E-350 5 4 4 5 3 5 5 5 4

According to RightMark Audio Analyzer, the E350N’s analog audio signal quality is no worse than Zotac’s Mini-ITX H67 board. Of course, I suspect most folks will want to use the board’s digital audio outs, whether it’s directly via S/PDIF or bundled with a video stream over HDMI.

Conclusions

If you’ve been following along since page one, you’ll already know the answers to the questions we posed at the start of this review. For everyone who skipped over my eloquent prose, carefully constructed charts, and pretty graphs, I’ll recap.

The Hudson M1 platform hub isn’t nearly as flashy as the APU that makes up the heart of AMD’s first Fusion offerings. However, this tiny core-logic component is stacked with 6Gbps Serial ATA ports and second-gen PCI Express lanes. The performance of Hudson’s SATA controller is especially impressive, although the lack of RAID support does hinder the platform’s suitability for closet file servers and network-attached storage. Folks who want redundant storage will have to resort to purely software- or OS-based RAID implementations.

Gigabyte was smart to hang USB 3.0 and Gigabit Ethernet controllers off of Hudson’s PCIe lanes. These integrated peripherals are a little slower on the E350N than they are on full-sized desktop boards, but that’s just it—they’re only a little bit slower. Considering the vast difference in CPU performance between the E-350 and modern desktop CPUs, I can live with peripherals that only lag a step behind.

I’m not so keen on the Gigabyte board’s overclocking potential, though. The BIOS might serve up loads of memory timing, clock, and voltage options, but our sample’s E-350 wasn’t particularly happy when pushed beyond stock speeds. Zacate wasn’t expected to be a potent overclocker, so its reluctance to run at higher speeds isn’t a great disappointment. Unfortunately, Gigabyte appears to have focused its BIOS efforts on exploiting Zacate’s seemingly minimal overclocking headroom while ignoring its obvious suitability for unobtrusive desktops and home-theater PCs. The complete lack of BIOS-level fan controls on this class of motherboard is even more mind-boggling than the use of a tiny little fan on the APU heatsink.

In addition to being poorly tailored for Zacate’s ideal applications, the E350N is also a little pricey at $150 online. MSI’s E350IA-E45 sells for $10 less and lets users set a temperature threshold for its APU fan. If you can live without USB 3.0, ASRock has an E-350 board available for just $110. Asus puts the very same APU on a $175 E35M1-I Deluxe, so the Gigabyte board isn’t the most expensive of the lot.

Given that competitive landscape, I’m a little hesitant to recommend the E350N at present. The addition of decent fan speed controls to the board’s BIOS would make this a much better platform for quiet desktops and home-theater PCs. Zacate is an excellent choice for those kinds of systems. Adding a stack of tweaking options to the BIOS isn’t going to turn it into a compelling overclocking platform, though.

Comments closed
    • burntham77
    • 9 years ago

    There were some snippets of relevant data in this review, but focusing on performance charts seems like a poor choice for reviewing this board. I would have loved to see this thing built into an HTPC system, complete with a nice variety of media to see how it handles the files. OTA HD recorded via WMC would be a good place to start. Maybe toss in some codec performance, let us know if this thing can handle 1080 streams from Netflix? Something so I can decide if i want to replace the 45 watt Athlon X2 in my HTPC with this efficient little guy. As it is, I am still undecided.

    • zorglub
    • 9 years ago

    Do you think I would be better with that little baby rather than my old Pentium D 3.2 Ghz ? In term of processing power, since for the power efficiency there no doubt !

    • flip-mode
    • 9 years ago

    Someone tell me I’m mistaken, but does this Asus come with both integrated wireless AND bluetooth (look at the little dongle plugged into one of the ports in the pictures)?

    [url<]http://www.newegg.com/Product/Product.aspx?Item=N82E16813131698&cm_re=zacate-_-13-131-698-_-Product[/url<]

      • Palek
      • 9 years ago

      That’s a yes. Look here:

      [url<]http://www.asus.com/product.aspx?P_ID=9BmKhMwWCwqyl1lz[/url<] $174.99 is a bit steep for what is essentially a budget platform, but the fanless design and the wireless connectivity are worth the extra cash. [EDIT]BTW, good find! There appear to be two fan headers on the board, albeit of the 3-pin variety. That's more than what the Gigabyte board offers, though. The Asus website also mentions "Q-Fan" fan speed control.

        • Bauxite
        • 9 years ago

        I’m strongly considering building a dead silent streamer out of this board with exactly 0 fans (picopsu brick) and 0 spinning parts (small boot SSD) relying on ethernet for all the actual data in another room.

        I could just get the zotac barebone but the possibility of a tuner card and DIY case is tempting.

        The asus board is pricey for what it is, but a proper passive heatsink and a minicard slot are big bonus points for a lot of itx ideas.

          • MadManOriginal
          • 9 years ago

          Id be very wary of running a truly passive PC regardless of the large heatsink or other components. Something about zero airflow, or only convection, just doesn’t sit well with me, and it’s not just the CPU and I/O hub that need cooling…some airflow is good for the other components as well. I think the Asus board is awesome though (if quite overpriced, I’d pay about $100 for a standard Zacate board w/ USB 2.0 and maybe $130-140 for the Asus) but I’d want at least some kind of very low speed intake or exhaust fan on it.

            • NeelyCam
            • 9 years ago

            It all depends on the components you have. With integrated graphics, SSDs and very low power HDs, it’s fine – there aren’t any components that really heat up that much except the CPU, and those are designed to throttle if they heat up too much.. they’ll last a long while.

            Best part: if the CPU dies, I have a perfect reason to upgrade

          • NeelyCam
          • 9 years ago

          Sounds great.

          I did something similar with i5-670 (idles at 22W), and will repeat it again with i7-2600K in a couple of months. Mine wasn’t 100% silent; I got a 1.5TB WD Green in it.. But it spins off quickly, so most of the time that’s dead silent too.

          Fans are overrated

    • obarthelemy
    • 9 years ago

    Thanks and congrats for a very informative review. The sound quality and I/O perf tests, including I/O CPU usage, are especially relevant, and rarely done. Peaking at 20% CPU usage for I/O is a bit frightening ^^

    The price for those things seems high.
    – For a NAS box, Atom boards are much cheaper, and at least as well suited. Probably better supported by FreeNAS and linuxes.
    – For a desktop box, power usage is less of an issue, processing power more of one. I can get a mini-itx MB *PLUS* a CPU that’s a lot more powerful for about the same price
    – that leaves the HTPC or mixed use case. OK.

      • Palek
      • 9 years ago

      Yeah, I personally would not build an HTPC around a Zacate system. I want my HTPC to handle some light gaming (nothing cutting-edge, just racing games and FPSes from a few years back) at the native resolution of the TV, and Zacate definitely would fail this requirement.

      If the only requirement for a NAS box is file serving, the LinkStation and TeraStation boxes from Buffalo (Mitsubishi Electric subsidiary) do the job. I’m using one of these at home:

      [url<]http://www.buffalotech.com/products/network-storage/home-and-small-office/linkstation-duo-ls-wxlr1/[/url<] They won't beat a file server PC in performance, but they're cheap, compact and very simple to use. The model I use can be configured as RAID0 or RAID1. I get around 20MB/s write speeds over a Gigabit LAN connection. According to PR material from Buffalo these NAS boxes can produce 40MB/s read speeds but I haven't measured read performance yet (I use the box for backups, mostly). This particular model also has Active Directory support, DLNA server functionality, iTunes support, web access, bittorrent support, etc. If you are the adventurous type there's also a pretty dedicated hacking community for the LinkStations offering all kinds of custom firmware builds with extended features.

    • swaaye
    • 9 years ago

    What kind of power usage would this thing have on a pico PSU? 40-50W is still considerably more than one of those networkable external hard disks or a router with a USB port.

    I have a Buffalo WZR-HP-G300NH router with USB and have a 1TB drive attached. I loved the thought of using that 5W router as a NAS too. You know, it’s already always-on. I’ve been using it like this for almost a year now. The problem is that it can only move data at about 11MB/s because the CPU maxes out. It has a 400 MHz MIPS 24k inside.

    Is there such a thing as a NAS box built out of this low power ARM/MIPS hardware that can move data at gigabit speed?

    • flip-mode
    • 9 years ago

    Nice review, Geoff. I am really enjoying the coverage of the Zacate motherboards. The Tech Report is way, way out in front on this effort.

    Based on everything I’ve seen I would only consider the Asus or the Asrock, but the Asus would be out of the running if it absolutely had to be miniITX.

    That fact that most of these are not passive is kind of a bummer.

    I wish mobo makers would take advantage of ever bit of space available in the port cluster. Add another gig-e or add some more USB over top of the USB 3 ports on this Gigabyte board, for instance.

    Between the Gigabyte board and the ASrock board, I’m not so sure I think USB 3 is worth the $40 additional. Dunno, hard call. Again, I’d buy the Asus board rather than the Gigabyte board.

    • Xylker
    • 9 years ago

    I was really excited by this platform, so I bought one as soon as I realized it was available.

    I agree with the conclusions here, but I am not so critical of the fan control and fan size flaws. I figure if I can skip Starbucks for 10 days I can save the extra cost of the Gigabyte board. And, since I have had really good experiences with Gigabyte motherboards going back to the GA7-DXR, I figured I would buy my Zacate from Gigabyte.

    I have been using this as my main PC for the last week+ and I will say it is very good as a basic machine, but I have not pushed it much. My main complaints have to do with the transition issues that are really my fault. My old main system had just under 4 TB of local storage split across 4 drives. With less than 100 GB in the new build I have to keep going back to the other box… Eventually I will migrate all that data to online shares, but I am not there yet.

    Fry’s didn’t have the Antec ISK 300-65 so I settled for the 300-150. The exhaust fan for the PSU seems silent, but the variable speed fan on the side is not. It is not too annoying, but it is not silent, either… (sits on top of desk +/- 70cm from me)

    The bottom line for me is that it works just fine for its intended purpose.

    • mczak
    • 9 years ago

    Hmm still no ddr3-1333 igp benchmarks?
    So this time you even run the board with a 1333Mhz memory clock but skip all tests where it would actually make a difference???

      • FuturePastNow
      • 9 years ago

      I’d like to see some game testing with the faster memory, in particular.

    • vikramsbox
    • 9 years ago

    Would have loved to see the Pentium G9650 and Athlon II 250 in the battle instead of the high end quad and 6 cores. The gap in CPU applications is so large that its not worth analyzing. So the ‘beautiful’ charts are of not much use.
    One other thing that could have been done was to underclock the Athlon II X2 and pentium G9650 down to the E350’s clock and see the relative performance of the architectures. This sort of testing used to be done before, and seems to have been left out now.
    And guys, please let us have an option to see the entire article in one page with a print option.

    • UberGerbil
    • 9 years ago

    [quote<]Folks who want redundant storage will have to resort to purely software- or OS-based RAID implementations[/quote<]Or they could just stick a proper hardware RAID card in that PCIe slot -- 4 lanes of PCIe 2.0 is plenty, and it's not like that slot should be in use for anything else in a storage-oriented appliance.

    • esterhasz
    • 9 years ago

    Just for info, the BIOS level fan control on the Asrock E350M1 is really good (target temperature and preferred rotation speed)…

    • ub3r
    • 9 years ago

    Hmm… this system would make a perfect porn box..

    • NeelyCam
    • 9 years ago

    I’d like to see you test these with PicoPSU, to really see what the idle power is. If your 200W PSU with these loads ends up wasting more power than your 750W PSU, you need better PSUs…

      • NeelyCam
      • 9 years ago

      Also, what’s the deal with the CPU utilization comments in the USB2.0/PCIe section? Graphs show Hudson to have a much higher utilization than others, but comments imply it doesn’t. Also, comments say 13% utilization with PCIe, but the graph says 20%.

      Are the graphs wrong…?

        • Edgar_Wibeau
        • 9 years ago

        CPU load is relative to absolute CPU performance. E-350 has two cores at 1.3GHz K10 speed (with no L3 cache: AMD Neo II K325), so Thuban 1090T has about 7.5 times the raw CPU power of E-350. Thus, 13% Zacate load on a Thuban would result in about 1,7% load.

      • Dissonance
      • 9 years ago

      That 13% line was supposed to be up with the USB stuff. It’s fixed now.

      As for our PSUs, the Silencer is quite good. The 200W unit’s power draw is more an indication of what you get with a budget Mini-ITX case-and-PSU bundle.

    • Palek
    • 9 years ago

    Thanks for the review, Geoff.

    What I would really love to see is a match-up of the E350 against machines a couple generations old. I’m sure many of us have parents and other relatives who are still using P4 or Athlon XP-based machines so it would be nice to know if the E350 can give them a comparable computing experience with much-reduced power consumption and improved video playback. Pretty please? 🙂

    Also, what about popping that fan off the heat-sink to see if its absence would cause a thermal shut-down (or worse)?

      • stupido
      • 9 years ago

      Pretty please? from me too…

      • UberGerbil
      • 9 years ago

      It’s going to be more than a “comparable” computing experience vs a P4 or Athlon XP — you’ve got two cores (which you probably don’t have with the P4 and definitely don’t have with the A-XP) for starters, and those older systems will have an AGP-era GPU that won’t be helping much. Whether those “grandma” users actually notice the difference or not, it most definitely will be there as soon as you push it beyond Freecell.

        • Palek
        • 9 years ago

        I’ll take your word for it (what with you being Ubergerbil and all), but it would be reassuring to see a productivity and browsing benchmark comparison.

        To be more specific about machine specs and why I wonder if upgrading to E350 would make sense:

        I have a 7-year-old Shuttle PC in the family with an Athlon XP 2400+ @2GHz with 512MB RAM and a Radeon 9600 (AGP). This machine outruns my 3-year-old work laptop that has a Core 2 Duo @ 1.6GHz with 1GB of RAM and some Intel integrated rubbish. I know it’s an apples and oranges comparison – the biggest differences being the HDD (7200rpm 3.5″ vs 5400rpm 2.5″) and video (dedicated vs integrated); but I expected this laptop to at least provide acceptable performance in typical office usage. I was wrong. This thing is slower than that ancient Shuttle, even when not using disk-heavy applications.

          • accord1999
          • 9 years ago

          If the Core 2 Duo laptop is so slow, then it’s either you don’t have enough RAM for the OS its running or there something horribly wrong with the HD. CPU wise the Core 2 Duo is considerably faster than an Athlon XP or an E350.

            • Palek
            • 9 years ago

            I know it should be faster, so something is probably not quite right with the machine. The HDD is probably a slower model and I suspect that Symantec Corporate Antivirus is dragging it down as well. If it weren’t a work machine I would do some troubleshooting, but since it’s company property and I need it to do my job I have no desire to mess with it. I’d love swap out the HDD for an SDD, but when I make a pitch for the benefits such as increased productivity I get the “in this economy” speech.

          • Skrying
          • 9 years ago

          There’s something wrong with your laptop, or you’re trying to use either Windows 7 or Vista on it with only 1GB of RAM, which is going to result in a horrible experience. In addition, unless you’re playing games there’s no way the Intel graphics are going to be holding you back. Neither of your system’s have a GPU capable of accelerating video but the Core 2 Duo can handle that on its own, your A-XP can’t.

            • Palek
            • 9 years ago

            Thanks for the feedback. Both systems run Windows XP, so the OS is not much of an issue.

            I’ve been thinking about other factors that are dragging down performance on my work laptop. The number one candidate is Symantec Corporate Antivirus, I think. I run ESET software on our home systems; their products have a solid reputation for being fairly light on resources. Throw in the slower laptop HDD and Symantec could be seriously crippling performance…

          • flip-mode
          • 9 years ago

          Didn’t they already compare it to a Pentium EE 840?

          [url<]https://techreport.com/articles.x/20401/7[/url<] Is it possible to figure out from there?

            • Palek
            • 9 years ago

            Thanks, that review had slipped my mind!

            The E350 actually manages to stay close to the EE 840 in performance, so it’s fairly safe to say it will beat an Athlon XP 2400+. All good, then!

      • Dissonance
      • 9 years ago

      Unplugged the fan and the board has been running Prime95 and the rthdribl HDR lighting demo for a couple of hours now with no hiccups. IR thermometer tells me the heatsink is at 75C, and it’s quite hot to the touch. This is on an open test bench, so it’s not necessarily a reflection of how the board will perform without its fan inside a cramped Mini-ITX enclosure.

        • flip-mode
        • 9 years ago

        Thanks for responding to that request Geoff. I was quite curious to know if that was possible.

        • Palek
        • 9 years ago

        Thanks Geoff, I did not expect you to respond quite so quickly!

        A teensy Mini-ITX enclosure may be too cramped to run this board without a fan, like you say, but a somewhat larger case, such as a Shuttle Cube, would provide more room and better ventilation via even a single rear mounted case fan that pulls air over the heat sink. I’m seriously thinking of gutting my old Shuttle case and putting a board like this inside.

    • JustAnEngineer
    • 9 years ago

    Good review.

    I’m looking forward to Llano for a more powerful HTPC system.

Pin It on Pinterest

Share This