Netbook platforms: VIA Nano v. Intel Atom
Summary: Today's netbooks are mostly Intel Atom-powered, but that could change with the advent of VIA's Nano processor, which outperforms Intel's CPU in several areas.
Introduction
Despite the rapidly cooling economic climate, netbooks sold like hot cakes last year. The boom in small, affordable, notebooks looks likely to continue this year. Most netbook manufacturers use Intel's Atom processor along with the somewhat dated G945 graphics chipset.
The main criticisms levelled at netbooks centre on poor performance. AMD chief Dirk Meyer predicts the netbook's demise, while Nvidia president Jen Hsun Huang has criticised the weak performance of the Atom platform and announced a graphics chipset that's designed to address the Atom's deficiencies. This Nvidia chipset, called Ion, could give a significant boost to Atom-based netbooks. Ion has a clear edge over any graphics option than Intel can currently muster. On top of that, Ion can use Nvidia's CUDA parallel computing interface to perform calculations for certain non-graphics applications. The arithmetic performance of the graphics chip far exceeds that of the main processor.
Some manufacturers view the VIA Nano processor as a viable alternative for netbooks. Samsung has just released the NC20, which is equipped with a particularly energy-efficient variant of the Nano chip. The benchmark tests presented here examine the VIA and Intel chips. Nvidia's Ion is not yet available, but is due to appear in April.
Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.
Talkback
This review is incompetent
You should be testing a 1.3 GHz Nano against a 1.6 GHz Atom using netbooks with batteries removed to test the true performance and power consumption. It would also be fair to test Atom clocked to 1.9 GHz on many Netbooks.
You should also do more multi-threaded testing because the Atom does so well with Hyperthreading.
For reference...
Where's the 'gross incompetence' in having a reference point for the netbook benchmarks? Surely this is useful additional information that helps to put the Nano and Atom results into context?
Certainly Not Incompetent
As for testing the Atom over-clocked, why on earth would they want to do that? First, how many netbook buyers are going to know how to over-clock a CPU, or even know that it can be done? Second, what is the "expected use" of a netbook - is it likely to be used for something that is so CPU intensive that over-clocking will be worthwhile? Third, and perhaps most importantly, in my experience one of the largest problems with netbooks is the amount of heat that they generate, so why on earth would one want to over-clock the CPU and thus produce even more heat?
jw 11/2/2009
The incompetence is that you used 1.8 Nano results to represent a Netbook
Furthermore, you didn't even run an adequate suite of multi-threaded benchmarks nor did you test Atom at 1.9 GHz which is a valid clock speed for some atom based Netbooks.
The realworld performance is 1.3 GHz for the Nano Netbooks
So yes, the article is misleading and incompetent.
Look up the MSI Wind, it officially operates at 1.9 GHz
More details needed
The power consumption figures are surprisingly high 48W Atom 68 Nano. I am writing this from dual core 1.9GHz Athlon notebook which maximum consumption is 66W (from AC power adapter). This is Dell Latitude D531.
Everest 4.60:
FPU Julia 1545
SinJulia 819
Mandel 879
CPU AES 4739
Zlib 19946
PhotoWorxx 4761
Queen 6409
Memory
copy 4022
read 4481
write 4393
Since I get better performance and lower power consumption What the netbook platforms that you tested have to offer?
The article is clear about what it's testing, processor-wise
If we could, I'd like to run benchmarks over a much wider range of operating conditions - virtualised systems, ones biassed towards graphics, ones biassed towards video editing, ones at extremes of power, memory, processor or networking configurations. The really interesting things come out when you push the limits. But there's so much else to do aside from benchmarking, that it's just not possible.
We're lucky enough as it is to have Kai and ZDNet Germany with their emphasis on real-world testing, together with Charles and Toby in the UK who really understand the nuts and bolts.
Stuff like this is both straightfoward and useful. It's not comprehensive, nor can it be.
The article CLAIMS it is testing Netbooks in the title
Wrong comparison
Same goes for the Atom 230. It's not used in any netbook (never has and never will), because it's Intels nettop-aimed processor. The Atoms for netbooks are either the Nxxx (most popular is the N270) or the Zxxx (most popular is the Z520 & Z530), of which the latter are a whole 0.5W "greener" (2.0W compared to 2.5W).
The results of this review should be ignored if one is searching how these CPU's would compare in actual netbook products.
Bananas, Apples and Oranges
It would be more appropriate to look at battery life, AC power consumption, heat production etc. Time how fast a test webpage gets downloaded through the Wifi (or whatever) connection and then rendered. Those are the specifications that are more meaningful then mega-flops per second. You're not testing a "gamer" system guys. Now if its fast AND low-powered that might be something to talk about.
The biggest specification that has the most impact on run-time on batteries is how fast the CPU can shutdown sub-sections of itself to prevent power consumption. Savings there directly impact usability of the system.
One other missing specification is the power consumption of the RAM modules used. Nobody seems to talk about that at all. RAM is possibly the third ranking power hog in the computer. First and second are the screen and backlight and then the CPU and IO subsystem.
On a netbook a more important specification is the user interface and nobody tests that quantitatively. How well does the keyboard, touchscreen (if it has one) mouse or mousepad if not and the screen perform? If the box is fast but a b-----h to use it doesn't matter what the specs say if the user isn't going to be able to type on it effectively.
We're talking about devices that are going to be used and carried around all day and plugged back into a charger at night. None of the specifications being bandied about here directly relate to how the user is going to USE the netbook.
MSI Wind? Uses 1,9 GHz processor?
Official Capable
It's a turbo button
Biased or inept, you be the judge
The real problem for Nano is that, aside from power consumption owing to the core logic and devices, it has roughly THREE TIMES the power consumption of Atom but offers only 20% ~ 30% better performance, at the same clocks.
Read it again: THREE TIMES the power consumption, 20% ~ 30% better performance. That is not an acceptable trade-off by any measure and shows just how inferior Nano truly is. Your tested configurations have Nano consuming nearly as much more over Atom as E5200 consumes over Nano; 20W v. 23W (keeping in mind that 945GSE adds 10W to this disparity in favor of Atom). And yet Nano fails to deliver anything near the performance increase that E5200 delivers with its 23 watts. Epic fail.
The move from 65nm to 45nm only netted 10% ~ 12% decrease in power consumption for Intel Core 2 Duo with the same clocks and core configuration. Nano must overcome a HUGE disparity here and 45nm won't remotely be enough.
Nano cannot compete with Atom WATT FOR WATT, and power consumption is all important in this class. The only exception to this is the AES benchmark, which of course Nano would run-away with because it is the only one of the three sporting a hardware encryption engine, a feature that is almost completely worthless to the netbook class.
As soon as Intel finally ships an appropriate purpose-built chipset for Atom (which is on the way as we speak), Nano will be dead to OEMs unless VIA can slash the price enough to keep it attractive purely from a BOM standpoint. Sure, you might get Nano in some 800MHz variant that will be competitive with the best-performing Atom in power consumption, but won't be able to run Pong with tolerable performance.