VIA EPIA N10000 Nano-ITX Review

Posted on Feb 10, 2006

Go to: Introduction Luke CPU and South Bridge Full Specifications Box Contents and Heatsink Back Panel Connectors Connectors and Jumpers Temperature, Noise, Power Sandra Benchmarks AES Benchmarks Video Playback Tests HDTV and 6 Channel Audio Summary and Conclusion

Benchmarks

The EPIA N10000 isn't targeted as a performance machine, but we do expect a certain level of performance from it. Certainly a 1GHz machine should be able to run an operating system with a GUI smoothly and perform most tasks thrown at it. We were also particularly interested in the media playback abilities of the board, as many potential buyers will be considering it for this task.

Mini-ITX owners will understand when we describe the the power of the N10000 as very similar to an M10000, but with the advantages of the newer CN400 chipset. No, those aren't typos - VIA chose 'N' for the Nano-ITX range, whilst 'M' is the moniker for their popular range of Multimedia Mini-ITX boards, released in 2003.

We tested in comparison with the M10000 Mini-ITX board, a stalwart of many a Mini-ITX project, and still popular today. This was the nearest thing we could use a yardstick - there being currently no other Nano-ITX boards in existence to benchmark against.

Sisoft Sandra

Sisoft's Sandra is a comprehensive diagnostic, analysis and testing package. We tested the most relevant benchmark modules using the Standard 2005 SR2 version. Benchmarks do not always represent a true real-life performance, but they are useful to compare the speed of various CPUs, and elements of the system. We've used this program in previous Mini-ITX reviews, so had good figures to compare against.

Estimated Performance Rating

Sandra can quickly estimate a performance rating for a system, based on various factors such as CPU speed, RAM size etc. The rating of 900 puts the board at 75% of the speed of the M10000, but as we discovered later - this was not entirely indicative of real world use. However it's certainly not going to win any awards for folding - today's most powerful CPUs can achieve 10 times this figure.

CPU Benchmark

The Dhrystone benchmark is a long standing industry benchmark used to measure CPU performance using a standardised sample of mainly numerical operations. The result is given in MIPS (Million Instructions Per Second). The N10000 produced approximately 1.5 MIPS per 1 MHz of CPU speed - about 93% the speed of our M10000 yardstick.

The Whetstone benchmark measures FPU (Floating Point) performance, although many modern processors have a number of newer features such as out-of-order execution, pipelining and SSE2 which are not tested using this benchmark. We're traveling at 99% of our M10000 with this one.

CPU Multi-Media Benchmark

The CPU Multi-Media Benchmark uses all the Multimedia Extensions available to the CPU to draw a Mandelbrot fractal. Multimedia Extensions are additions to the x86 instruction set designed to make repeated or parallel operations run faster. Digital imaging or streaming video applications can make good use of these extensions, which use Single Instruction Multiple Data (SIMD) techniques. Changing the contrast of an image or MPEG decoding and encoding all require a large amount of data to be manipulated by the same instructions. This benchmark is therefore a reasonably good test of raw and MPEG2 recording and playback functionality. The N10000 achieved 3623 here, approximately 93% of the M10000.

The Floating-Point results were identical to the M10000 at 3653, unsurprising given the earlier Whetstone performance. Real world results will depend on the particular SIMD instructions used by a particular task.

Memory Benchmark

Sandra's Memory Benchmark creates several large arrays in memory and performs simple memory-bound arithmetic computations on them - thus reading and writing memory broadly independent of the CPU. It is slightly more objective than simply reading and writing to a large block of memory. The results of 756 Integer and 420 FPU were both significantly faster than the M10000, as you would expect when testing a 400MHz SODIMM against a 266Mhz DIMM. The SODIMM used in the N10000 runs faster than the DDR DIMMs used in the M board.

Rip WAV to MP3

Next we used dBpower AMP Music Converter to convert a 70MB WAV file to an MP3, using the default settings. We've still got the same file from all our previous tests, so have a direct comparison. By ripping from the hard disc and not a CD, we removed the CD from the test and concentrated more on the raw power of the CPU. A similar task would be compiling a Linux kernel. Our file was converted in 3 minutes, exactly the same as the M10000 achieved when it was on the bench. A modern Intel or AMD CPU will of course do the same test in under a minute.

Benchmarks (2) - AES Encryption -->

*Advert* World's Smallest 12V DC-DC ATX Power Supply now at the Mini-ITX store! *Advert* The picoPSU is now available at the Mini-ITX.com Online Store. We serve the UK, Europe, USA and beyond. Order in-stock components before 7.00PM GMT and we'll ship same day!