In-Depth Review: Gigabyte's J1800N-D2H Mini-ITX Motherboard
Intel's new Bay Trail-D family of processors have been released, and motherboard makers are putting out products with the chips soldered on. Join me as I take Gigabyte's entrant, which boasts a dual-core Celeron J1800, for a not-so-brief spin.
This review is broken down into 9 parts:
Part 1: The CPU
Part 2: The Motherboard
Part 3: The BIOS
Part 4: Benchmarking Methodology
Part 5: Benchmarking Performance
Part 6: Power Consumption
Part 7: Temperatures
Part 8: Fan Control
Part 9: Conclusion
(Part 1: The CPU)
The low-budget, low-power, low-performance Intel Atom family, or at least the subset of it targeted at desktops and
laptops netbooks, first came out in 2008. Originally a three-chip design with the original 45nm Diamondville CPU, an update in early 2010, dubbed Pine Trail, reduced that to two chips and lowered power consumption. Later, 2011's Cedar Trail saw a die shrink to 32nm.
Despite the platform updates, all of these systems have stuck with the original Bonnell CPU core (renamed to Saltwell with the 32nm shrink) with few changes – mainly just clock frequency increases. As a simple in-order architecture designed to sip power, the performance was merely passable in 2008, and very sluggish here in 2014.
So, at long last, Intel released the successor to the Bonnell core late last year. Promising somewhere around double the performance, Silvermont is an out-of-order architecture built on Intel's 22nm manufacturing process. The core is being rolled out into various families of chips designed for different markets, including the Merrifield platform for smartphones, and the desktop Bay Trail-D platform, a member of which we'll be looking at today.
The first thing you might notice about the Celeron J1800 is its brand name. Indeed, the Atom name is being attached to all of the Silvermont-based parts except for the ones meant for desktop PCs and laptops. Why? We're not sure. Perhaps Intel is trying to avoid a stigma instigated by the slow performance of the older parts.
Anyhow, let's take a look at the Celeron J1800's specifications.
||Intel Celeron J1800
|Cores / Threads
||2 / 2
|Clock Frequency / Burst
||2.41 Ghz / 2.58 Ghz
|L1 Cache (Data / Instruction)
||2 x 24 KB / 2 x 32 KB
||1 MB (shared between cores)
|Advanced Instructions||MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, VT-x|
|Max. Memory Capacity||8 GB|
|Memory Type / Channels||DDR3L-1333 non-ECC / 2|
|Integrated Graphics||Intel HD Graphics (Gen7, 4 EU) w/ Intel Quick
|Graphics Frequency / Burst||688 Mhz / 792 Mhz (660 Mhz / 768 Mhz reported
by HWiNFO32 4.36)
|PCI Express Revision / Lanes||2.0 / 4|
||USB 3.0, USB 2.0, 2x SATA 3Gbps
|Thermal Design Power||10 watts|
The J1800 contains two Silvermont cores, each normally running at 2.41 Ghz. These cores can boost to 2.58 Ghz when temperatures allow, and drop down to 1.33 Ghz during idle periods. Unlike Bonnell and Saltwell, HyperThreading is unsupported: each core executes one thread per tick.
Meanwhile, the J1800 packs a downsized version of Intel's Gen7 graphics core, also found in the firm's “normal” Ivy Bridge processors. The DirectX 11, OpenGL 4.0, and OpenCL 1.1 graphics APIs are supported. QuickSync hardware video encoding/decoding is supported on newer revisions of the chip. Sadly, the chip I have on hand is an older revision.
As with most integrated solutions, the HD Graphics shares system memory. On that note, the Celeron J1800 contains a dual-channel memory controller, supporting 1.35v DDR3L modules at speeds of up to 1333 Mhz. I intend to see how big an effect the dual-channel memory capability has on the chip's performance; stay tuned for the benchmarks.
So far, all of the products based on Silvermont cores are single-chip “system-on-chip” (SoC) designs. Bay Trail-D is the first time this approach has been used in the desktop “Atom” platform. Accordingly, the chips contain I/O blocks for several common interfaces, including PCI Express 2.0, SATA 3Gbps and USB 3.0.
Finally, the collective thermal design power (TDP) of the Celeron J1800 is just 10 watts. “Normal” desktop processors tend to be rated closer to 65w. This is as expected from Intel's ultra-low-power platform.
In the interest of keeping this review relatively short, there is a lot more to the Silvermont architecture that I'm leaving out. Those who are interested in reading more can check out The Tech Report's excellent writeup on the subject.
(Part 2: The Motherboard)
Now we get to the Gigabyte motherboard. Historically, embedded CPUs, including the Intel Atom family, have been non-socketed setups, soldered permanently onto their boards in order to further reduce costs. The Celeron J1800 is no exception, so I can't really review the chip without also reviewing the board it shipped on.
A variety of manufacturers are beginning to put out motherboards with Bay Trail-D chips, and the one we've got on the bench today is Gigabyte's GA-J1800N-D2H model. We'll start with the box.
The box is positively tiny for a motherboard box, and the online store I ordered this from shipped it with padding on only one side. D'oh! Thankfully the board survived the trip in perfect working condition.
Front of the box. We see that the J1800N-D2H is covered under Gigabyte's Ultra Durable moniker, which certifies quality onboard parts (polymer capacitors and the like) and good quality board construction.
Other motherboard makers make similar claims with their boards, though. Gigabyte is hardly the only one out there doing this.
Here's the identification sticker on the box, which gives us a serial number (blocked to protect the innocent) and a brief set of specifications for the board within.
Back of the box. There's some extra Ultra Durable-related marketing here, along with more specifications.
Time to unpack. First thing we see after opening the box is the J1800N-D2H itself, packaged in an anti-static bag.
Box contents include the motherboard, a user manual, a driver DVD, an I/O shield for the case, and two latching SATA cables.
Now let's get a look at the board itself. See that big, black aluminum heatsink? Our Intel Celeron J1800 SoC of the day sits underneath that. Its 10w TDP is low enough that Gigabyte didn't see the need for a fan. I elected to not remove this heatsink in the interest of benchmarking the setup in its pristine, out-of-box state.
Like many embedded CPU boards, the J1800N-D2H fits into the mini ITX form factor. This is close to the smallest motherboard standard that is still compatible with ATX, which came out in 1998 and is still in use with today's PCs. Mini ITX is tiny enough to allow only one expansion slot, but large enough to fit the I/O panel and our board's other essentials.
Speaking of the I/O panel, here it is. We have two PS/2 ports for mouse and keyboard, a VGA d-sub video output, an HDMI output, a USB 3.0 port, four USB 2.0 ports, a gigabit wired LAN jack, and audio jacks for microphones, speakers, and line-in input.
Now we'll zoom in to the board surface and see what we can see. At the bottom left is a PCI Express 2.0 x1 slot, good for 500 MB/s of bandwidth for supported cards. Opposite is a mini PCIe slot, which looks good for a wireless card. Or, perhaps, one of these, to augment the Celeron J1800's support for only two SATA 3Gbps ports.
We also see one of the two fan headers, this one supporting 4-pin PWM fans.
Here we see the CMOS battery and a pair of laptop-style RAM slots. These can handle up to 4GB each for 8GB of maximum system memory. Only 1.35v modules are supported, though the Corsair sticks I used, despite being 1.5v, worked just fine. Either way, double-check the supported memory list before purchasing to avoid heartache.
The HD Graphics reserve a varying amount of memory for its own use. However, with Intel's DVMT (Dynamic Video Memory Technology), it won't take more than 64MB (changeable in the BIOS) unless it needs to.
As I mentioned earlier, I'm going to benchmark this board with both Corsair modules inserted, and then do it all over again with just one. We'll see how much performance the boards like Biostar's J1800NH leave on the cutting room floor for only implementing one of the two memory channels.
At the top of the Gigabyte we find the 24-pin ATX and 4-pin ATX12V power connectors. That 4-pin connector seems a little redundant given the low-power CPU, but the board won't power up without it – I tried. Sandwiched between these connectors is the front panel LED and switch header. On the far left is a legacy RS232 serial port header (labelled COM) for those who want it – avid InventorSpot followers, perhaps.
Below these, we see part of the two-phase power circuitry for the Celeron SoC. The PWM controller is an Intersil 95836. True to Gigabyte's Ultra Durable branding, the entire board surface is studded with polymer capacitors. (The black-print ones are from APAQ; I can't figure out who made the blue ones.)
Behind the I/O panel and next to a legacy LPT header, there's an ITE IT8620E Super I/O chip (handling the PS/2 ports, legacy headers, and hardware monitoring), an ASMedia ASM1442 TMDS chip for the HDMI, the last fan header, a mysterious DEBUG connector, the sole front panel USB 2.0 header (for two ports), and... a Genesys GL850S USB hub.
How interesting. This chip connects to one USB 2.0 port and turns it into four. While that might be well and good, all four will share the bandwidth of the first port. The question is, which of the J1800N-D2H's USB ports are attached to this hub? Gigabyte's specifications table doesn't tell us, but luckily I have a truckload of USB devices and a special program. The answer:
Indeed, all four of the USB 2.0 ports on the J1800N-D2H's back I/O panel are attached to the hub. Running an HD Tune read test on one of the external hard drives I connected yielded roughly 36 MB/s, normal for USB 2.0. However, running the test on both drives at once had the rates for both hovering at 19 MB/s.
Is this a problem? Maybe, if you're regularly going to use the back panel USB 2.0 ports to access multiple drives at once. The front panel header's ports don't seem to be affected, and there's also that lone USB 3.0 port to connect storage devices to.
Bandwidth issues aside, I didn't ever have any problems with peripherals (including a keyboard) attached to the hub'd ports, even while navigating the UEFI. Your mileage may vary.
We find ourselves back where we started, having made a full circle of the board surface. The gigabit Ethernet controller is a Realtek RTL8111F, and the 7.1 channel audio codec, a Realtek ALC887.
A front-panel audio header sits next to a two-pin S/PDIF output header, the latter presumably for a bracket that's sold separately. Meanwhile, the BIOS chip is soldered on and unfortunately not user-replaceable.
Flipping the board over, we discover that the mammoth heatsink is bolted on using a pair of spring-loaded screws. For your information, these extend about 4mm off the board surface. The screws themselves are positioned about 62mm away from each other, for those who feel like getting an aftermarket heatsink.
While I'm at it, the heatsink itself measures 21mm high off the board surface. Compare this with the I/O panel's highest point (the audio jack block), which is 33mm.
Based on that, I don't think you'll have much trouble fitting the Gigabyte GA-J1800N-D2H into even the smallest of ITX cases, including the Mini-Box M350. These have almost no breathing room past the top of the I/O panel. The heatsink here is a good deal shorter than the I/O panel.
Now, let's boot this up and get a look at the BIOS setup interface. You can find the Gigabyte GA-J1800N-D2H on Amazon.
(Part 3: The BIOS)
Pressing the <esc> key – rather than the usual <del> key, for whatever reason – on this here boot screen does the trick.
One small note before I start, though. My Gigabyte GA-J1800N-D2H shipped with the F2 BIOS, but a newer F3 version was available on their site, so I flashed it with a bootable USB key. (An even newer F4 version has since appeared.) These screenshots come from the F3 revision after the “Restore Defaults” setting was invoked.
Here's the Main page. This board uses one of those newfangled UEFI interfaces, but you wouldn't guess it just by looking at it. While some of Gigabyte's other boards have fancy graphics and mouse support, the J1800N-D2H's firmware looks just like the one on that old Dell Dimension in your closet.
As you can see, I have 4GB of RAM installed (two modules), and I've set the system date and time properly. Honest.
This is the Advanced tab. The J1800N-D2H's UEFI has a fair amount of options in general, but those who enjoy tweaking their systems might be astonished by the complete lack of overclocking and fan control options.
Indeed, here's the hardware monitoring screen, where the fan controls usually live. Nothing here. Connected fans are temperature controlled though; more on that later.
CPU Configuration screen under the Advanced tab. “Power Technology” is usually set to “Energy Efficient,” but I've set it to “Custom” here to display all of the possible options.
Though it won't matter to most people, the presence of Virtualization Technology, even on as cheap a chip as the Celeron J1800, is nice to see.
PPM Configuration screen under the Advanced tab. The “Max CPU C-State” option seemed interesting (C-states are processor power states), but setting it to either C6 or C7 seemed only to lock the Celeron to 2.41Ghz and prevented it from idling at 1.33Ghz.
The Compatibility Support Module Configuration screen under the Advanced tab. By default, you can boot the J1800N-D2H with legacy devices as well as with UEFI boot devices (such as the Windows 8 setup disc).
This screen is noteworthy because “CSM Support” was, by default in the F2 BIOS, set to disabled (which removed all of the options underneath, and set the board to only use UEFI to boot). This might be the reason behind all the bad reviews this motherboard was getting from early adopters, who were having trouble installing any OS other than Windows 8.
You need to enable CSM and set the options to “Legacy first” in order to use a bootable USB key to flash the new BIOS. This can also help with installing things like Linux and, ultimately, Windows 7. It sure did for me.
IGD Config screen, under the North Bridge Config screen under the Chipset tab. I know I'm flying through these pages, but a lot of the options I'm skipping will be very familiar to anyone who has installed a motherboard in the past few years. It's the usual “enable/disable onboard audio” and “enable/disable AHCI on the SATA ports” kinds of stuff.
Anyway, these here are options that have to do with the Celeron J1800's HD Graphics. You can turn off the Turbo Boost option, as well as change settings relating to the DVMT memory allocation.
Some of these options are real question marks though – “GFX Boost,” for example, sounds promising, but enabling it didn't affect any of my graphics benchmarks, and it's neither explained in the UEFI nor in the manual. Again, I used the default settings for my testing, so the option was disabled for the entire run.
The Security tab. The usual setup passwords can be configured here, along with Secure Boot for those operating systems that support it (including, once again, Windows 8).
The Boot tab. You can set which connected drives (including USB-connected drives) to boot off of first, as well as enable or disable the boot logo.
We end at the all-important Save & Exit tab. You can also boot directly off of an attached drive through this page.
(Part 4: Benchmarking Methodology)
So, at long last, we're preparing for benchmarking. Here is the setup.
Indeed, I'm testing the GA-J1800N-D2H without a case, mounted on four nylon pegs from the Baby AT days. I made sure that the heatsink wouldn't see any actual airflow, though, to emulate the conditions it might face in a cramped, fanless case.
That's what that business card next to the power supply is for – it's blocking one of the vents. Don't worry, it'll be fine. To provide a point of comparison, I will also be running my set of benchmarks on a system based on a single-core, 2009-era desktop AMD Sempron processor.
Both setups will use the same power supply, storage drive, monitor, keyboard, mouse, and operating system. I'll be comparing general performance and power consumption between the two setups. After that, I'll be checking how well Gigabyte's giant heatsink cools the Celeron J1800, as well as how the board's invisible fan controls work.
In addition, while the Sempron setup will run its integrated graphics throughout its session, I will, at a few points, include a few benchmarks with an ATI Radeon HD 4350 discrete graphics card instead. This will be to see how well the Celeron J1800's integrated graphics compare to a budget graphics card from a few years ago. I'm anticipating some interesting results.
All of that said, here are the detailed specifications (with driver versions where applicable) for the four total setups we'll be testing today. I'll make a note right here that the Celeron J1800 sustained its turbo clockspeeds for both CPU and GPU throughout the entire test session.
||Intel Celeron J1800
||AMD Sempron 140
|Motherboard||Gigabyte GA-J1800N-D2H Rev 1.0
||Gigabyte GA-MA78LMT-S2 Rev 3.4
||N/A (Intel INF v18.104.22.1686; Intel
||AMD 760G+SB710 (SB Driver 13.12)
|Memory||2x Corsair CMSO4GX3M2A1333C9 2GB DDR3||1x Corsair CMSO4GX3M2A1333C9 2GB DDR3||2x Kingston KVR1333D3N9/2G 2GB
|Memory Timings||1333 Mhz, 9-9-9-24-1T
||1066 Mhz, 7-7-7-20-1T|
|Storage||SanDisk SDSSDP-064G-G25 64GB
|Graphics||Intel HD Graphics (Driver
||ATI Radeon 3000 (Driver 13.9)
||ATI Radeon HD 4350; 512MB DDR2 [Sapphire 11142-07-20G] (Driver 13.9)|
||SeaSonic SS-300ES 300w
||Samsung SyncMaster 170N:
1280x1024, VGA d-sub input
||Microsoft Windows 7 Home Premium
Here are the benchmarking apps and their versions that I will be running. Each will be run three times with default settings and reboots between each attempt. The online tests will additionally have the involved browser clean-reinstalled between attempts.
- 7-Zip 9.20 x64
- Cinebench R11.5 11.529 x64
- Cinebench R15 Build RC83328DEMO x64
- PCMark 8 2.0.228
- Internet Explorer 11.0.9600.16428
- 3DMark 1.2.362
- TrueCrypt 7.1a
- SiSoftware Sandra 2014 SP1c
- Google Chrome Portable 34.0.1847.116
- Audacity 2.05
- LAME for Audacity v3.99.3
- x264 r2409
- Prime95 27.9
- FurMark 1.13.0
The copy of Windows 7 I'm using has Service Pack 1 already baked in. The only software installed are the benchmarking apps above that use installers, the hardware drivers (downloaded from Gigabyte's site in the J1800's case, as the driver DVD doesn't include Windows 7 drivers), and whatever updates are downloaded and installed with that version of Internet Explorer 11.
In regards to Windows 7 settings, I switched off the timers for “Turn off the display” and sleep mode, disabled Windows Update, and hid all of the Action Centre notifications. That's all. Only the browser tests and Futuremark benchmark suites were done with a TP-LINK WN727N USB wireless module attached.
Right, let's begin.
(Part 5: Benchmarking Performance)
Maxon's Cinebench uses the CPU to render a photorealistic 3D scene. It can use one thread or all available threads, and we're starting here with the former. We see that the Sempron 140, based on an architecture not designed for low power consumption above all else, performs better than the Celeron J1800 here.
Start using its second Silvermont core, though, and the Celeron outpaces the solo Sempron. Note that dual-channel memory isn't really having much of an effect on performance as of yet.
Cinebench also has a GPU test, where it renders a 3D car chase scene using the OpenGL API. Impressively, we see that the little Celeron's integrated graphics core, especially with two memory modules, just about touches the Radeon HD 4350 discrete card.
A newer version of Cinebench provides us generally the same CPU results as before.
Interestingly, the Celeron's HD Graphics don't do so well with the updated car chase scene. The ageing discrete card has a much more sizable lead here, and the Celeron ends up trading punches with the Sempron's integrated Radeon 3000.
TrueCrypt is an open-source disk encryption app with a built-in benchmark. TrueCrypt supports AES-NI hardware acceleration built into some new CPUs, but neither of our two subjects today support it.
Again, by virtue of its two cores, the Celeron can boast being roughly 31% faster at handling the AES algorithm than the Sempron 140. The difference widens to about 45% with the Twofish algorithm.
7-Zip, a popular multi-threaded file compression program, also has a built-in benchmark. Predictably, the Celeron J1800 is in the lead for compressing and decompressing files using the .7z format.
SiSoftware's Sandra is an extensive benchmarking program with tons of test patterns. Here we're looking at the results given by its CPU Arithmetic tests.
Dual-channel memory subtly increases the Celeron J1800's scores throughout. The single-cored Sempron finds it difficult to compete overall, especially in the Dhrystone Integer Native test.
I also ran Sandra's memory bandwidth tests.
Predictably, the Celeron J1800's bandwidth drops like a stone with one memory channel. When both are enabled, the Celeron speeds up dramatically. Partly by virtue of its clocking the memory 267 Mhz faster, it knocks the Sempron's own dual-channel configuration out for first place.
The Sempron 140 gains a moderate amount of memory bandwidth with a discrete card installed, but, again, it's not enough to topple the Intel Bay Trail-D family member.
Time for a slightly more real-world test. Here I'm using Audacity 2.05, equipped with the LAME MP3 library, to export a sizable WAV file as a 320kbps (constant) MP3. The WAV file used is a rip of The Doors' When The Music's Over from disc one of their 1985 anthology, The Best Of The Doors.
The LAME library is not multi-threaded, so we find the Sempron finally eking out a win, churning out that eleven minute MP3 a solid 30 seconds or so faster than the Celeron ever could.
Using the x264 h.264/AVC encoder, I've tasked the setups to encode the one-minute long, 1080p, M2TS format Samsung Earth From Above clip from this page using the following command line switches:
--output NUL --profile high --preset medium --crf 18 --video-filter resize:1280,720 –force-cfr
The encoder is multi-threaded and supports a variety of extra CPU instructions. Unsurprisingly we find the pipsqueak-that-could coming out on top. At a glacial seven frames per second, though, it's clearly not suited for this kind of job.
Moving on to Futuremark's Peacekeeper online benchmark, which, with Google Chrome, examines how well systems can cope with new web standards and techniques, including HTML5 video playback.
Interestingly, the single-core AMD Sempron 140 did very well here, at up to 82% faster. I will say, though, that some of the graphical tests, including the videos, had some very slight hitching to them, both with the chipset graphics and the discrete card. The Celeron J1800, on the other hand, rendered the same tests smoothly and without anomaly.
One thing is for sure, though. When opening lots of tabs, the dual-core Celeron copes better with the extra work. The Sempron's one core quickly loses steam and slows down when faced with the same workload.
Futuremark has traditionally offered a wide-ranging benchmark that tests many aspects of a system's performance in PCMark. It includes everything from video chat to casual DirectX 9 games to web browsing.
The Celeron SoC scores just under the AMD IGP setup with one memory channel active. It speeds up with both memory slots populated, probably because of the boost this provides to its graphics core.
PCMark 8 also has an “OpenCL Accelerated” version of the Home 3.0 benchmark, but the resulting scores were essentially unchanged from the non-accelerated ones above, so they aren't included.
The latest version of 3DMark is apparently cross-platform, with versions available for both PCs and smartphones and tablets with different operating systems. I'm a little sceptical of just how comparable the scores are across platforms, but anyway.
The Celeron J1800's HD Graphics try hard here and manage to outpace the discrete card, both with and without dual-channel memory. Nice. The leaderboard's the same for both the Ice Storm and Cloud Gate tests, which use DirectX 9 and DirectX 10 features, respectively.
The intensive Fire Strike test wouldn't run on the AMD setup, what with neither of its graphics options supporting DirectX 11. The Intel HD Graphics do, but it predictably scores very low.
I know, I know... never trust what the Windows built-in benchmark tells you; it's not a replacement for real empirical testing, and it often gets things wrong. However, for curiosity, here's what the setups score with Windows 7's incarnation of the Experience Index.
Backing up the hate for WEI, we find that disabling the ATI Radeon 3000 integrated stuff in favour of a Radeon HD 4350 actually lowers the Sempron 140 setup's base score. Don't say I didn't warn you.
So, with that, we wrap up our performance testing. In summary: the single-thread performance of Intel's Silvermont cores within its Celeron J1800 is a little pokey. It's a good thing there are two cores, then, and the extra thread helps with multitasking. This is why the Sempron 140 system feels, at least to me, more sluggish in day-to-day use than the Celeron, despite a higher clockspeed and a faster CPU architecture.
Meanwhile, dual-channel memory appears to pay more dividends to the Celeron J1800's integrated graphics than for much else, though admittedly I haven't run many benchmarks. Memory bandwidth drops steeply with one module, but most of the CPU-hungry tests weren't phased much by that. Hmm. Overall, the HD Graphics competed well with the circa-2008 ATI Radeon HD 4350.
Despite up-to-date API support, the HD Graphics won't handle newer games very well, but at least it works much better than the Radeon 3000 IGP, which can still be found today on many contemporary AMD motherboards.
(Part 6: Power Consumption)
Throughout this section, I'll be using a Kill-a-Watt metre attached directly to the power supply. The only peripherals attached to the setups will be the storage drive, USB keyboard and mouse; and monitor. There will be four “work states” that I'll be measuring: idle, CPU load, GPU load, and CPU+GPU load. Each state will be left alone to stabilize for 30 minutes before recording figures.
The CPU load is the Prime95 stress test with the “in-place large FFTs” option, and the GPU load is FurMark and its default settings. Both run at the same time for the combined load.
I'll also have HWMonitor 1.24, HWiNFO32 4.36, and the Windows Task Manager opened to keep an eye on vital signs. I'm saving some time, here, measuring both power consumption and temperatures in one fell swoop. You'll see the temperature results in the next section.
When doing almost absolutely nothing, the Gigabyte GA-J1800N-D2H motherboard and the Celeron J1800 SoC collectively pull just 14w at maximum. That's pretty low. The Sempron 140 takes substantially more, and more still with the discrete graphics card installed.
Both processors lowered their frequencies and voltages through the stabilization period. The Celeron J1800 was down at 1.33 Ghz with the core voltage at 0.984v. The Sempron's clock dropped lower, all the way down to 800 Mhz, but clearly that wasn't enough to save it. Meantime, the Celeron's integrated HD Graphics clock dropped to 660 Mhz.
When actively doing work, the Celeron J1800's Silvermont CPU cores, in true Intel Atom fashion, only drew 3.9w more than they did when idle. They ran at their burst frequency of 2.58 Ghz with a core voltage of 1.056v.
With those cores idling at 1.33 Ghz with 1% utilization through the GPU load state, it seems the HD Graphics are the more power-hungry component of this SoC, pushing a 6.5w rise above idle at 768 Mhz.
At most, with both components loaded and holding their burst clocks, I only got 23.8w (or 9.8w above idle) out of the Gigabyte GA-J1800N-D2H, and 1-2 watts less throughout with only one memory module installed.
Those figures make the rating on the puny 300w power supply look... substantially un-puny.
Indeed, keep in mind that traditional power supplies are much less efficient at low loads like these, even despite the 80Plus Bronze certification on this particular unit. A PicoPSU or equivalent with a power brick would be much more appropriate for this setup. You might shave a couple extra watts with the improved efficiency, and it'll take up less space with fewer unused connectors to hide.
Anyway, the Sempron 140 setup comes off as a watt guzzler here, hitting up to 65.9w with integrated graphics and 77.5w with the video card. Requiring two fans (one each for processor and graphics card) to stay sane, the setup in either configuration also made extra noise. These are moderately typical figures for normal desktop PCs, though.
(Part 7: Temperatures)
Now for temperature testing. Same story as yesterpage in regards to setup and 30-minute stabilization periods, though it's time to say goodbye to the AMD Sempron 140. From here on, we'll be focusing on the Gigabyte motherboard and Intel Celeron J1800 chip only. The ambient room temperature throughout this session was around 24 degrees Celcius.
Temperatures were measured using the sensors displayed by the HWMonitor utility. These include the twin CPU core and graphics core temperatures reported by the Celeron SoC itself, plus a mysterious “TMPIN0” display. I assume that's from a thermistor positioned near the chip on the board surface.
I additionally used my IR thermometer on the giant heatsink for each of the four load states.
When just sitting there, temperatures were fairly cool, around forty degrees with and without both memory banks populated. This trend of slightly higher temperatures with dual-channel mode enabled will persist through the rest of this page.
With both processor cores in use by Prime95, the temperatures rose and stabilized in the upper fifties. Not too much else to say here, other than watch your fingers.
Temps rose even higher with FurMark loading the GPU, adding more strength to my theory that the Celeron SoC's HD Graphics take more power than both the Silvermont CPU cores combined.
With the full-on combined load going, we see chip temperatures stabilizing right at eighty degrees. Yikes. Amazingly, as mentioned on the last page, both the CPU and integrated graphics held their burst speeds of 2.58 Ghz and 768 Mhz, respectively, with no signs of throttling.
Also, while that is a scary looking temperature figure, Intel specs the Celeron J1800 as being able to hold onto its lunch until five degrees above boiling point, so I wouldn't worry too much. As well, both Prime95 and FurMark are artificial loads – you probably won't be seeing temperatures that high during day-to-day use.
In short, the Celeron J1800 is definitely capable of running entirely fanless with Gigabyte's giant heatsink.
(Part 8: Fan Control)
Now, some words about the fan speed control. I mentioned earlier that there's nothing in the BIOS for configuring fan speed, but the board does temperature-control fans to a degree. Pun not intended.
So, I grabbed my Adda AD1212UB-A7BGL 120mm, 4-pin PWM fan and left the board off for awhile to cool it down. After, I plugged the fan into one of the two fan connectors per turn and ran my Prime95 and FurMark CPU+GPU combined load. I kept an eye on temperatures and the fan's rotational speed using HWMonitor, looking for a pattern.
And a pattern there was. The results are in this here table:
|TMPIN0 (Degrees Celcius)
||CPU_FAN 3-pin (RPM)
||SYS_FAN 3-pin (RPM)||SYS_FAN 4-pin (RPM)|
It seems like that TMPIN0 variable calls the shots here, and both headers appear to control almost exactly the same, regardless of whether my fan is connected with three pins (voltage control) or four pins (PWM control).
And, yes, I yanked the fourth wire from the Adda fan's connector to see if that four-pin SYS_FAN header could voltage-control fans. It can, and it all but mirrors the 3-pin CPU_FAN header's behaviour in this setup.
Anyway, the fan, in all three fan header configurations, remained at a low speed of just 288 RPM until about the 32 degree mark. From there, the speed increased with each subsequent TMPIN0 degree up to just under 1000 RPM at 45 degrees.
Why did we stop at 45 degrees? Well, as we saw on the last page, the Celeron J1800 simmers at nearly eighty degrees when TMPIN0 hits 45, and neither got any hotter than that. Interestingly, this particular Adda fan model has a maximum speed of 2500 RPM. It was running at less than half that when we reached the temperature limits here.
So, while your mileage may vary, any fans you plug into the Gigabyte GA-J1800N-D2H should always stay at a relatively low speed, unless your setup is configured in a way that somehow makes the TMPIN0 value rise higher than 45 degrees.
(Part 9: Conclusion)
So, we're just about full circle with this review of Gigabyte's GA-J1800N-D2H and the CPU that calls it home. We've tested performance, power consumption, temperatures, and the fan speed control. We've also doted on some strange quirks that are hopefully unique to this board. With all that in mind, one can get a reasonable answer to the question of “what are these parts good for?”
Well, if you're looking to build an inexpensive, silent, low-power PC, Intel's Celeron J1800 is generally up to the task. It'll handle light web browsing with your browser of choice just fine. YouTube videos up to 1080p quality are doable, along with most Facebook games. Productivity apps should give this platform no trouble, along with Skype calls and mass-Torrent downloads of... uh... public domain music.
However, the relatively limited CPU performance can hinder multitasking. Having tons of Chrome tabs open (complete with Flash ads), plus most of the apps noted above, alongside a couple of notoriously bloated apps (such as iTunes) can feel sluggish at times. As well, while 1080p YouTube videos play relatively smoothly, you won't want too much other stuff open.
Stay under the performance limits and the Celeron will work well, and with negligible impact to the electricity bill. In other words, this chip is living up to its Intel Atom heritage: a focus on providing “just enough” performance for most daily computing tasks while using as little power as possible.
Meanwhile, hobbyists can appreciate the platform as a whole. The GA-J1800N-D2H's fanless heatsink is up to the task. There isn't much connectivity over what the Celeron SoC supports, but what's there should be enough for most users. The board itself looks and feels well-made and durable. Nothing crashed or malfunctioned at any point of the testing.
Readers of this website might see the GA-J1800N-D2H as a beefier Raspberry Pi alternative, being able to run full-fat Windows with x86 capability, unlike the Pi, while still remaining relatively power-efficient and small. The board also has more of things like USB ports, and even has legacy I/O headers. Perhaps this board will form the basis of your next big project, or – dare I say it – invention.
The board could also form the backbone of an HTPC, what with the HDMI 1.4 port, mini ITX form factor for tiny cases, and silent operation, though you won't be able to attach many storage drives. I also didn't get to testing media playback. However, the integrated graphics support hardware acceleration for several common video codecs at up to 4K resolution, which is promising.
I have some gripes, though. Mainly, even the cheapest PicoPSU model, at 80w, provides far more power than most GA-J1800N-D2H setups may ever take. Yet the 4-pin ATX12V connector, which that PicoPSU model lacks, needs to be plugged in to even boot up. It's inexpensive to add the 4-pin connector to that PicoPSU or go up to the next size, but it still feels like a waste.
Less serious are my other gripes. I'm not sure how many USB 2.0 ports the SoC provides in total, so I don't know if I can ding Gigabyte for the back-panel hub. On the OS front, only Windows 7 64-bit and Windows 8/8.1 in 32 and 64-bit trim are supported with drivers, and the former requires a BIOS update to even install. However, newer batches should ship with the new firmware.
And anyway, you might not even need a Windows license after all. I briefly tried a handful of Linux distributions, including Lubuntu 13.04 32-bit, Xubuntu 13.04 64-bit, Fedora 19 64-bit, and Linux Mint 13 64-bit. All booted into their live CD sessions properly from an external drive configured with the YUMI USB Multiboot Creator. I couldn't boot off the drive without setting the BIOS's CSM “Storage” setting to “Legacy first,” though, just so you know.
All of that said, we get to the value proposition. The GA-J1800N-D2H is priced at around $70 as of this writing. For not much more you can get Intel's Celeron G1610 processor and a motherboard for it. That's a pairing with more expansion options (including upgradeable processors), less annoying operating system support, and vastly better performance overall.
Those parts will suck down more power, though, and will have at least one fan running by default. Tiny ITX cases that can hang off the back of a monitor are out of the question, and you won't get ports like USB 3.0 or HDMI without spending extra. Those who want such features and don't mind a somewhat underpowered CPU will find a decent compromise with the Gigabyte GA-J1800N-D2H.