Original URL: http://www.theregister.co.uk/2008/08/14/review_via_nano/
VIA Nano ultra-low power processor
Preview The mobo in our pics may look like a VIA SN18000G, which sells for about £150 with a 1.8GHz C7 processor, but it’s more exciting: it's a reference board for VIA's would be Atom-smasher: Nano.
It's designed to demonstrate the new 1.8GHz VIA Nano L2100, to be precise. The CPU's fabbed at 65nm and is set to consume no more than 25W of power. It supports the Intel's SSE 3 instruction set. It runs on an 800MHz frontside bus, and in many respects it's the least interesting model in the Nano range.
VIA's Nano reference board
The 1.6GHz L2200, for instance, has a 17W TDP and yes, we have noted that the higher model number has a lower clock speed. If you think that’s strange, the U2300 runs at 1.0GHz with a 5W TDP; the U2500 has a clock speed of 1.2GHz and a surprisingly precise TDP of 6.8W; the U2350 runs at 1.3GHz and 8W; and finally we have the U2400 with a clock speed of at least 1.3GHz - it's not finalised yet - and the same 8W TDP.
If there is any rhyme or reason to those model codes, it has completely passed us by.
VIA supplied a reviewer’s guide with the Nano kit and this compares the chip with Intel’s Atom, and as we have reviewed the 1.6GHz desktop Atom 230 , we’re all set to put the Nano through its paces.
The Atom left us unmoved despite its low price because the performance is none too great. We reckon the dual-core Atom could have some potential, but in the meantime we'd like Intel to combine its low-power CPU with a chipset that was younger and better than the i945G it's currently saddled with. And while Intel is giving Atom a workover, could it please sort out the silly heatsink on the chipset.
Atom may well be the natural competition for Nano, but heading into this review it seems to us that VIA has set the bar for comparison rather low.
VIA makes a number of claims about Nano - codenamed 'Isiah' - to whet our appetite. For starters; the chip's pin-to-pin compatible with the C7 series of CPUs, which is the sort of thing that will please motherboard manufacturers. Atom is a new design.
Nano has 64-bit, superscalar architecture and can handle out-of-order instructions. According to VIA, Atom can’t do any of those things so it’s worth pointing out that Intel lists drivers for the D945GCLF Atom motherboard for Windows XP x64 along with 32-bit XP and 32-bit Vista.
Other nuggets of information about Nano obscure and inform in equal measure. For instance, there is a dual-core Isaiah on the roadmap, but we don’t know when it's due to be released. We do know that a shift to 45nm is planned for H2 2009, and you might well expect that the dual-core changeover will happen at the same time, but that’s pure speculation on our part. Either way, we’re looking at a full year and that’s a long, long time to wait.
The layout of the Nano reference board is more sophisticated than the Intel G945GCLF, with no sign of an ATX 12V connector that Intel decided to include. VIA has added a PCI Express expansion slot and supplies two DDR 2 memory slots where Intel felt that one would suffice.
When the Nano system starts the CPU fan spins at 4300rpm by default which seems awfully fast and noisy. Happily, you can change the fan setting from Enabled to Auto or - God help you - Disabled. By default, the Auto setting sets a lower temperature limit of 65°C and an upper limit of 75°C, and once we had the fan on Auto, the Nano system was transformed. Initially, the fan doesn’t turn until the system is under load at which point it picks up when the heatsink hits 50°C or so. It works perfectly.
VIA's Nano reference board: top down
We installed the same 32-bit Windows Vista Ultimate Edition SP1 that we had used on the Atom and ran 3DMark06, PCMark05 and POV-Ray. In its reviewer’s guide, VIA picks and chooses benchmarks with a great deal of care. It highlights the CPU element of PCMark05 and claims a score of 1418 for the Atom and 2089 for the Nano. Those figures are very similar to our own results of 1379 and 2052, but that’s only part of the story.
After we published our Atom review, a number of readers pointed out that it’s all well and good for the 1.6GHz Atom 230 to have a 2.5W TDP but this was dwarfed by the power draw of the chipset - which is the reason why Intel stuck an active cooler on the chipset and a passive heatsink on the CPU.
We completely agree, and have to consider the system as a whole both for Atom and Nano. The Overall score in PCMark05 for Atom is 1683, compared to the Nano's 1538. That’s a small win for Intel because the graphics and memory scores on the Nano platform are truly dire. We have to hope that VIA will update its CN896 North Bridge and Chrome9 HC IGP graphics as they are letting Nano down.
Longer bars are better
Note, no ShaderModel 3.0 scores
Longer bars are better
POV-Ray is a pure test of CPU rendering power and it shows that the 1.8GHz Nano has significantly better performance than Atom. The VIA chip completed the test in 19 minutes; the Atom took 27 minutes.
Time in Seconds
Shorter bars are better
The problem is that the Nano system draws more power than Atom – 80W against 60W – which rather blows a hole in VIA’s argument that it has a 47 per cent greater fuel efficiency than the Intel offering. That figure is achieved by averaging work loads and power draws, but we very much doubt that mobile systems with the same battery would yield a significantly longer runtime with the Nano.
Power Draw Results
Power Draw in Watts (W)
The first reason for this is that both the Atom and the Nano desperately need better chipset support than the two systems that we've seen to date.
The second is that neither the Atom nor the Nano has enough performance to convince us of their appeal and we really, really want a dual-core version ASAP.
Intel vs VIA: D945GCLF (left) and Nano reference
Third – and this is the killer – we have no idea how much Nano will cost. VIA address the point thus in its FAQs:
What is the anticipated price point for systems using the VIA Nano processor?
VIA does not set the price for systems or boards… pricing will be decided by our system and board partners…
As VIA SN hardware sells for £150 we see no reason why the same won’t be true for Nano, and that suggests Nano might be three times the price of Atom. Ouch.
If so, it'll be roughly three times the price of a comparable Atom. Yes, the Nano has twice the performance of the Atom - according to PCMark05 - but the premium you pay makes it a much less attractive choice.
Let's hope VIA mobo partners will offer Nano packages for a lot less than £150. If they do, we'd raise the 50 per cent score we'd give the Nano at that price to well above the 55 per cent score we gave to the desktop Atom.
From our initial survey, it's clear VIA's Nano is a promising processor. But we were unimpressed by the performance of the CN896 chipset and Chrome9 graphics, and the prospect of a £150 price fills us with horror. VIA, say it ain’t so!