VIA EPIA CN13000G review

EPIA CN overview

Mini-ITX motherboards in general form an excellent basis for multimedia devices and other computing appliances because of their low power consumption, specialized cryptography and video decoding abilities, and small size. They're unique enough that VIA has created its own unofficial small formfactor standard -- Mini-ITX -- that is interoperable with most ordinary PC peripherals and accessories, and fits into any MicroATX, FlexATX, or Mini-ITX chassis.

There have been many EPIA models over the years, each with its own collection of features that aim for specific computing appliance markets. Specifically, the EPIA CN series was designed for maximum multimedia performance in a Mini-ITX formfactor. Toward that goal, EPIA CN motherboards have the following features:

• VIA C7 CPU (either 1.0Ghz or 1.3Ghz as of this writing).
• VIA CN700 northbridge chipset, which supports DDR2-400 and DDR2-533 memory.
• VIA VT8237R-Series southbridge chipset, which supports both Serial ATA and parallel ATA hard drives.
• VIA/S3 UniChrome Pro AGP with Chromation. This is a competent integrated graphics chip that includes hardware acceleration for Windows Media Video and MPEG-2 (DVD) decoding as well as full HDTV support.
• VIA PadLock Security Engine, which provides hardware acceleration for the AES, SHA-1, SHA-256, and RSA cryptographic algorithms.

Electrical design

The EPIA CN is a Mini-ITX formfactor motherboard with an integrated VIA C7 CPU. It has a single DDR2 slot, which will accept a single 256MB, 512MB, or 1GB stick of DDR2-400 or DDR2-533 memory. While the CN700 chipset can technically handle 2GB sticks, the EPIA CN is not configured to accept anything above 1GB. Realistically this should be enough RAM to do anything that the board was designed to.

Rear I/O panel connectors consist of PS/2 keyboard and mouse, four USB 2.0, one 9-pin serial, one 15-pin SVGA, one S-Video, one RCA, one RJ45 10/100 network, and three audio jacks. There are two disappointments here: there is no parallel port, and the audio connectors are too few to properly do 5.1 surround sound. It's not that you can't do surround sound -- indeed you can -- but in order to hook up 5.1 speakers, you lose the ability to use the microphone jack. I don't think the lack of a parallel port and the "either-or" status of the sound output will affect many people, but you should be aware nonetheless.

Internally there are wiring blocks for USB 2.0, a 3-pin system fan, a parallel port (though an actual port is not included with the motherboard -- you have to buy one separately if you need a parallel port), and frontpanel audio. Disappointingly, the act of connecting the frontpanel audio jacks disables the rear ones. That's one negative point of the EPIA CN -- multimedia connectivity is something that this board should do well above all other things.

There are two SATA and two IDE connectors, but no floppy drive. Who needs one these days anyway? The power connector is a standard 20-pin ATX, and does not require or support additional power to the voltage regulator like ATX motherboards do.

The entire motherboard has only one fan, and it's on the northbridge chipset, not the CPU, though both share the same elongated aluminum heatsink. It's a standard cheap Sunon chipset fan, and doesn't make much noise. Still, previous experiences with EPIA boards in enclosed spaces has taught me that a good 80mm or 120mm case fan is necessary despite the lack of a CPU fan.

The C7 processor is available in either a 1.0Ghz or 1.3Ghz configuration at the time of this writing. You shouldn't expect the level of performance from this chip that you would get from a more modern, high-powered design like the Pentium D or Athlon 64 X2. If your goal is to use the EPIA CN in a desktop computer, you should calibrate your expectations accordingly, and select an operating system and desktop software that do not put a heavy strain on the CPU or video device. The most logical choice for that scenario would be a GNU/Linux distribution that is optimized for single core i686- or Pentium4-class CPUs.

A frequently overlooked feature of the C7 is its PadLock Security Engine. As mentioned above, this is the name for VIA's integrated encryption capabilities. When performing AES, RSA, SHA-1, or SHA-256 encryption or decryption, the C7 outperforms any other x86-based processor in most instances (details and test results are below).

Lastly, the integrated UniChrome graphics chip may not be much good for high-end 3D gaming, but it is specifically equipped to provide better video playback, especially on high definition televisions (HDTV). The graphics chip takes much of the load off of the CPU for video decoding, though it does not possess DVD decryption (DeCSS) capabilities, so you'll still need a software DVD decoder if you want to view commercially produced DVD movies.

Putting it to the test

The EPIA CN13000G, while among the most speedy and capable of the EPIA series, is still not well-suited to anything above everyday desktop computing. 3D games and other CPU-, RAM-, and graphics-intensive tasks are better left for ATX or MicroATX machines.

With that in mind, I put the CN13000G into an Antec Fusion chassis with a DVD writer and a Western Digital Raptor SATA 36GB hard drive with the intention of building a digital video recording device. The build went very well; VIA hardware is always easy to install and configure.

At first I spent a few days installing and configuring Gentoo Linux on the EPIA machine, but couldn't get the integrated video chip to work properly with the latest Gentoo kernel. I read through all of the available documentation and a number of forum topics and outdated wiki entries, tried alternative drivers, and experimented with different kernel options and X.org configuration settings, but never made any progress.

Next I tried Freespire 1.0.13, which has excellent proprietary video driver support for Intel, ATI, and Nvidia graphics chips. I'm happy to say that Freespire proved its worth once again, providing direct rendering and GLX through the OpenChrome video driver. The downside to using Freespire is the relatively high amount of system overhead that it requires; I'd wanted to use Gentoo because it generally has outstanding hardware support, the most modern software, and can be easily configured for a lightweight GUI. Freespire wasn't unusable with the CN13000G, but it was a little sluggish.

Previous EPIA models included a decent paper manual that had diagrams of the onboard connectors and a rundown of all of the BIOS options. That's more or less what I would expect to find included with any new motherboard these days, but the EPIA CN13000G had nothing of the sort. In order to properly connect the chassis frontpanel options to the CN13000G, I had to get online and download the real manual in PDF format from the VIA Web site.

To measure the basic 2D graphics rendering performance in X.org, I used x11perf. Since there are so many tests in the x11perf program, most of them being small and insignificant, I created a script to use only the tests with the largest graphical tests. The table below shows the VIA EPIA CN13000G in Freespire against some other hardware configurations. Higher numbers mean more responsive 2D graphics rendering:

VIA EPIA CN13000G, 515MB DDR2-533, integrated graphics with 64MB shared video RAM, OpenChrome driver	MSI K8T Neo2-FIR motherboard, Athlon 64 4000+, 1GB DDR400 RAM, XGI Volari V5 128MB AGP8X, VESA driver	MSI K8T Neo2-FIR motherboard, Athlon 64 4000+, 1GB DDR400 RAM, Albatron GeForce FX5700U3 256MB AGP8X, Nvidia proprietary driver
1000 reps @ 0.9625 msec ( 1040.0/sec): 500x500 stippled rectangle (161x145 stipple)	300 reps @ 4.9622 msec ( 202.0/sec): 500x500 stippled rectangle (161x145 stipple)	3000 reps @ 0.4698 msec ( 2130.0/sec): 500x500 stippled rectangle (161x145 stipple)
1000 reps @ 1.0766 msec ( 929.0/sec): 500x500 tiled rectangle (161x145 tile)	2000 reps @ 0.9062 msec ( 1100.0/sec): 500x500 tiled rectangle (161x145 tile)	3000 reps @ 0.4307 msec ( 2320.0/sec): 500x500 tiled rectangle (161x145 tile)
1000 reps @ 0.9560 msec ( 1050.0/sec): 500x500 tiled rectangle (216x208 tile)	2000 reps @ 0.8324 msec ( 1200.0/sec): 500x500 tiled rectangle (216x208 tile)	3000 reps @ 0.4103 msec ( 2440.0/sec): 500x500 tiled rectangle (216x208 tile)
90000 reps @ 0.0115 msec ( 86700.0/sec): 500-pixel line segment	30000 reps @ 0.0475 msec ( 21000.0/sec): 500-pixel line segment	300000 reps @ 0.0041 msec (245000.0/sec): 500-pixel line segment
90000 reps @ 0.0112 msec ( 89300.0/sec): 500-pixel line	30000 reps @ 0.0476 msec ( 21000.0/sec): 500-pixel line	300000 reps @ 0.0041 msec (245000.0/sec): 500-pixel line
10000 reps @ 0.1063 msec ( 9410.0/sec): 500x500 rectangle outline	40000 reps @ 0.0273 msec ( 36700.0/sec): 500x500 rectangle outline	70000 reps @ 0.0154 msec ( 65000.0/sec): 500x500 rectangle outline
10000 reps @ 0.1311 msec ( 7630.0/sec): 500-pixel circle	10000 reps @ 0.1509 msec ( 6630.0/sec): 500-pixel circle	10000 reps @ 0.1294 msec ( 7730.0/sec): 500-pixel circle
2000 reps @ 0.7890 msec ( 1270.0/sec): 500-pixel filled ellipse	3000 reps @ 0.3903 msec ( 2560.0/sec): 500-pixel filled ellipse	8000 reps @ 0.1289 msec ( 7760.0/sec): 500-pixel filled ellipse
600 reps @ 1.7623 msec ( 567.0/sec): Fill 300x300 stippled trapezoid (161x145 stipple)	600 reps @ 1.8014 msec ( 555.0/sec): Fill 300x300 stippled trapezoid (161x145 stipple)	1600 reps @ 0.6714 msec ( 1490.0/sec): Fill 300x300 stippled trapezoid (161x145 stipple)
240000 reps @ 0.0044 msec (227000.0/sec): Char in 60-char line (9x15)	480000 reps @ 0.0021 msec (476000.0/sec): Char in 60-char line (9x15)	2400000 reps @ 0.0005 msec (1930000.0/sec): Char in 60-char line (9x15)
12800 reps @ 0.0818 msec ( 12200.0/sec): Char in 30-char rgb line (Charter 24)	160000 reps @ 0.0064 msec (156000.0/sec): Char in 30-char rgb line (Charter 24)	960000 reps @ 0.0013 msec (753000.0/sec): Char in 30-char rgb line (Charter 24)
2000 reps @ 0.7882 msec ( 1270.0/sec): Scroll 500x500 pixel	2000 reps @ 0.7883 msec ( 1270.0/sec): Scroll 500x500 pixels	3000 reps @ 0.4023 msec ( 2490.0/sec): Scroll 500x500 pixels
1600 reps @ 0.7700 msec ( 1300.0/sec): Copy 500x500 from window to window	1600 reps @ 0.8103 msec ( 1230.0/sec): Copy 500x500 from window to window	2800 reps @ 0.3978 msec ( 2510.0/sec): Copy 500x500 from window to window
200 reps @ 5.8221 msec ( 172.0/sec): Copy 500x500 from pixmap to window	1600 reps @ 0.8277 msec ( 1210.0/sec): Copy 500x500 from pixmap to window	2800 reps @ 0.3984 msec ( 2510.0/sec): Copy 500x500 from pixmap to window
80 reps @ 12.2607 msec ( 81.6/sec): Copy 500x500 1-bit deep plane	80 reps @ 15.1707 msec ( 65.9/sec): Copy 500x500 1-bit deep plane	2400 reps @ 0.4921 msec ( 2030.0/sec): Copy 500x500 1-bit deep plane
4 reps @ 2569.9386 msec ( 0.4/sec): PutImage XY 500x500 square	4 reps @ 2752.8096 msec ( 0.4/sec): PutImage XY 500x500 square	80 reps @ 12.9997 msec ( 76.9/sec): PutImage XY 500x500 square
4 reps @ 1539.5263 msec ( 0.7/sec): GetImage XY 500x500 square	4 reps @ 1933.8253 msec ( 0.5/sec): GetImage XY 500x500 square	4 reps @ 3730.2001 msec ( 0.3/sec): GetImage XY 500x500 square

The results show that the UniChrome Pro AGP graphics chip actually does quite well for an integrated solution on a low-power small formfactor motherboard. In fact it consistently outperformed the XGI card, which had twice the RAM, but was using a reduced-performance X.org VESA driver. It's hard to beat an Nvidia card of that caliber, though, even at 2D rendering.

I attempted to test the MPEG-2 accelerator, but for days of trying, I did not get it to work in a way that I could appropriately measure.

Since the EPIA CN13000G includes PadLock, I figured I'd use the openssl speed command to show just what a difference it makes in OpenSSL cryptographic speed. Unfortunately, after performing the test on several different operating systems, I couldn't seem to harness the PadLock encryption engine through OpenSSL. I even tried OpenBSD, which supports more software and hardware encryption options than any other operating system, and fully supported the previous version of PadLock.

Below I have listed the speed test numbers as they happened. Obviously the PadLock engine was not in use, and I could not figure out a way to make it work within the week's time I had to do performance testing. As I understand it, I need a Linux kernel newer than 2.6.16, and I have to add a special patch to it. I tried that, but it still didn't work, so I think it's safe to say that at very least the majority of people who buy the EPIA CN13000G will not be able to take advantage of its integrated encryption engine in GNU/Linux or *BSD in the near future.

I have included speed test numbers from the same Athlon 64 4000+ used in the X.org tests above. You can also easily run the OpenSSL speed test on your own machine at home if you wish to see how the CN13000G compares to your computer. First, let's take a look at how the EPIA CN13000G did:

OpenSSL 0.9.8a 11 Oct 2005 built on: Mon Jan 23 20:11:02 PST 2006 options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial) blowfish(idx) compiler: gcc -fPIC -DOPENSSL_PIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DL_ENDIAN -DTERMIO -O3 -march=i686 -mcpu=i686 -Wa,--noexecstack -g -Wall -DOPENSSL_BN_ASM_PART_WORDS -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM available timing options: TIMES TIMEB HZ=100 [sysconf value] timing function used: times The 'numbers' are in 1000s of bytes per second processed.

type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes md2 292.01k 629.96k 838.56k 904.30k 965.49k mdc2 0.00 0.00 0.00 0.00 0.00 md4 3297.68k 10450.38k 26855.72k 47547.98k 62574.04k md5 2857.41k 9564.47k 27727.83k 52736.00k 65591.19k hmac(md5) 3485.93k 12086.92k 32932.52k 52849.74k 64671.13k sha1 2567.17k 7891.25k 16761.44k 23684.98k 28535.47k rmd160 2096.07k 6188.94k 13056.77k 16831.23k 18560.90k rc4 25196.98k 28844.01k 27766.35k 27658.67k 29400.18k des cbc 7768.34k 7559.21k 7510.76k 7959.72k 8157.87k des ede3 2541.80k 2548.89k 2718.13k 2794.28k 2568.92k idea cbc 0.00 0.00 0.00 0.00 0.00 rc2 cbc 3066.83k 3171.74k 3436.73k 3457.35k 3202.59k rc5-32/12 cbc 0.00 0.00 0.00 0.00 0.00 blowfish cbc 12994.73k 14993.56k 15299.18k 14307.00k 14161.67k cast cbc 6034.88k 6450.33k 6101.99k 6127.29k 6614.96k aes-128 cbc 7286.31k 10136.48k 11054.58k 10465.14k 10437.00k aes-192 cbc 6933.39k 8756.04k 8700.70k 8847.93k 9527.79k aes-256 cbc 6317.12k 7115.81k 7519.26k 8169.07k 8225.75k sha256 1546.72k 3361.47k 5843.38k 7603.86k 8448.58k sha512 1004.44k 3946.53k 7300.41k 11166.24k 12276.38k sign verify sign/s verify/s rsa 512 bits 0.004229s 0.000285s 236.5 3506.4 rsa 1024 bits 0.017188s 0.000724s 58.2 1381.8 rsa 2048 bits 0.088019s 0.002192s 11.4 456.3 rsa 4096 bits 0.522222s 0.007173s 1.9 139.4 sign verify sign/s verify/s dsa 512 bits 0.002787s 0.003295s 358.8 303.5 dsa 1024 bits 0.006894s 0.008736s 145.0 114.5 dsa 2048 bits 0.021276s 0.026351s 47.0 37.9

And for the Athlon 64 4000+ computer:

OpenSSL 0.9.7e 25 Oct 2004 built on: Thu May 26 12:20:08 EST 2005 options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial) idea(int) blowfish(idx) compiler: i686-pc-linux-gnu-gcc -fPIC -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DOPENSSL_NO_KRB5 -DL_ENDIAN -DTERMIO -Wall -O2 -march=pentium4 -pipe -fomit-frame-pointer -Wa,--noexecstack -DSHA1_ASM -DMD5_ASM -DRMD160_ASM available timing options: TIMES TIMEB HZ=100 [sysconf value] timing function used: times The 'numbers' are in 1000s of bytes per second processed.

type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes md2 1972.43k 4168.21k 5798.06k 6426.28k 6635.52k mdc2 5760.49k 6564.10k 6793.22k 6858.07k 6850.25k md4 19628.92k 68563.37k 196675.16k 368591.19k 495938.22k md5 17130.05k 59096.75k 169058.56k 314997.76k 425473.37k hmac(md5) 23604.52k 74008.53k 197261.27k 339661.82k 430830.93k sha1 17479.47k 56888.85k 145963.09k 239197.18k 294557.01k rmd160 14445.16k 42099.41k 93096.19k 133608.11k 153138.52k rc4 191472.19k 213686.31k 218618.20k 221005.82k 221640.02k des cbc 56039.15k 61279.66k 62657.19k 63102.29k 63255.89k des ede3 20636.86k 21323.07k 21541.21k 21596.84k 21613.23k idea cbc 42682.09k 45844.16k 46854.31k 47114.92k 47188.65k rc2 cbc 21029.18k 21966.81k 22171.82k 22229.67k 22249.47k rc5-32/12 cbc 177728.25k 199504.81k 204492.71k 206903.30k 206955.50k blowfish cbc 95399.50k 101648.68k 103123.63k 103719.25k 103890.94k cast cbc 46513.56k 49863.98k 50752.00k 51018.41k 51079.85k aes-128 cbc 52871.24k 53593.86k 54562.05k 54757.72k 54790.83k aes-192 cbc 46108.03k 46761.24k 47343.02k 47491.07k 47529.98k aes-256 cbc 40442.60k 41259.07k 41572.19k 41811.63k 41842.01k sign verify sign/s verify/s rsa 512 bits 0.0005s 0.0000s 2161.0 24897.0 rsa 1024 bits 0.0020s 0.0001s 489.5 9602.4 rsa 2048 bits 0.0114s 0.0003s 87.8 3073.1 rsa 4096 bits 0.0729s 0.0011s 13.7 892.3 sign verify sign/s verify/s dsa 512 bits 0.0003s 0.0004s 2886.5 2368.8 dsa 1024 bits 0.0010s 0.0012s 1052.0 858.8 dsa 2048 bits 0.0030s 0.0037s 330.6 268.2

Power consumption

Obviously a small formfactor board with a low-power CPU and few peripherals is going to draw significantly less power than traditional desktop computers. But how much less? Below is a comparison with two relatively powerful, modern desktop computers.

Each system used an Antec TrueBlue 480 power supply; 1GB of either DDR2-533 or DDR400 RAM in two modules; one Seagate SATA-V 160GB hard drive; a Matrox G550 1X PCIe video card; and a Lite-On 52X CDRW/DVD-ROM drive. The Intel machine was based on an Asus P5WD2 motherboard, and the AMD machine used an Asus A8N-E. The EPIA had one stick of DDR2-533 memory, an Antec TruePower 430 power supply, the aforementioned Lite-On optical drive, and a Western Digital Raptor 36GB hard drive. Power readings were measured with a Watts Up Pro meter. The numbers indicate how many Watt hours were consumed during the 15 minute test period, what the average monthly Kilowatt hour usage would be based on 15 minutes worth of data, and how much it would cost each month to run it constantly if the cost per KWh were U.S. $0.0731.

Conclusions and manufacturer recommendations

The VIA EPIA CN13000G is a great integrated small formfactor motherboard, and a credit to the EPIA product line. VIA has taken its intelligent approach to expanding small formfactor performance without making the CPU's power consumption unmanageable to the next level by adding support for DDR2 memory and SATA hard drives. With Windows XP Home, Professional, or Media Center Edition, you can expect much more from the EPIA CN13000G than you can with other operating systems at this time. Hopefully PadLock will be fully supported in the Linux and *BSD kernels soon, but as of this writing, no shipping operating system has full support for the new PadLock engine.

Low power consumption is still the EPIA's main selling point. You could run four EPIAs for the cost of running one Intel Pentium D desktop computer. For the average non-3D-game-playing user, this could mean significant energy savings with virtually no noticeable loss in performance.

Here's what I'd like to see in future EPIA CN models:

• Open drivers. OpenChrome is a wonderful first step, but for the EPIA to really have value to DVR builders, the MPEG-2 accelerator needs to have flawless and invisible support in the Linux kernel and userland programs like MPlayer and Xine. If you could harness the power of this accelerator chip for encoding and decoding video files from the command line (ala mpeg2enc and dvdrip), that would also be a huge benefit.
• Better audio. VIA has found some success with isolating processes to their own processors. Why not have a fully-fledged audio chip with true 5.1 surround sound outputs, too? The current integrated sound solution could stand to be improved.

Device	Mini-ITX motherboard/CPU
Manufacturer	VIA Technologies
Device support	PS/2 keyboard and mouse, four USB 2.0, one 9-pin serial, one 15-pin SVGA, one S-Video, one RCA, one RJ45 10/100 network, and three audio jacks for either stereo or surround sound. Internally there is support for one DDR2 RAM module, two SATA drives, and two IDE connectors that can support up to two drives each.
Market	DVR system builders, small formfactor enthusiasts, electricity-conscious consumers
Price (retail)	US ~$185
Previous version	N/A
Product Web site	Click here