Atom 330 is benchmarked, fares slightly worse than expected
By Laura June 
posted Nov 13th 2008 5:31AM
posted Nov 13th 2008 5:31AM
PC Pro's given Intel's dual-core, 1.6GHz Atom 330 (coupled with a 7200 RPM SATA hard drive and 1GB of DDR2 RAM) the benchmark run-through, and they've got some conflicting details to pass on to you. Overall, the testers found the Atom to be, as expected, faster than the N270, but only by 16 percent. In specific tests, the 330 ran Office 2003 slower than both a 2GHz VIA C7-D and the single-core Atom; PC Pro actually performed the test several times just to be sure it wasn't a glitch... and it wasn't. The 330 performed better running 2D graphics, outpacing the N270 by 41 percent, and it also outperformed its competitors in encoding and multitasking. Not enough details for you? Hit the read link for the full-on benchmarking experience.
Before anyone wonders, Dual-core Atoms also have Hyperthreading, making it show up as 4 CPU cores to the OS.
Here's AMD's chance to get back in the game!
Hopefully AMD's netbook cpu's are at least 10% faster than the Atom 330, cost about the same, and are widely available before Christmas.
AMD will have the advantage no matter how you see it. The integrated memory controller will help them a lot, which the Atom will only have in its next generation.
Regarding the benchmark procedure:
I'm assuming that the motherboard on which this thing is soldered upon is not the best one to compliment it.
Also most of the tests are not what Benchmark enthusiasts usually look for!
Performs Office 2003 better?
Better 2D?
I mean who gives a ****ing *****!?
The common Office (2003?) user doesn't want faster CPUs as long as the software works!
Faster encoding and multitasking are more meaningful to benchmark enthusiasts but something tells me that those who need FAST CPUs in these areas have already invested in Quad Cores and unless the Atom can compete with Quads they don't give a damn!
How exactly should a lowvoltage CPU be comparable to a Quadcore?
Uhm I'd like a faster processor so my netbook runs faster (office included)... It's not really that complicated.
I dont really care about Office, I would most likely be using google docs anyway, but the multitasking bit is what it's all about!
Engadget use png for screenshots!!
Yeah, because engadget took that screenshot
Engadget editors don`t seem to know much about computers.. it`s enGadget, not enPC after all ;)
You can`t compare a dualcore CPU with a single core in performance in a single-thread application...
Well, actually you can.
The only thing is that they shouldn't do, is be surprised when it does perform the same/slightly worse.
Schwing! Good one substance, although lately you'd think they're more like enMac than enGadget or enPC. And from a guy who has much Apple love with no fanboyism, that should say enough ;)
This comes as no surprise.
Doubling the number of cores on a chip while keeping the same Front Side Bus speed reduces the bandwidth available to each core, which in turn can hurt single threaded performance.
Any test that is multi threaded will still see a performance boost, and multi tasking will also be faster so the computer will "feel" faster.
Until Intel transition the Atom over to QPI (probably a long long time from now) its going to suffer the same scalability problems as the Pentium 4. They only got around it in the Core 2 Duos by raising the FSB to insane levels and that's not feasible in a low power environment.
Well observed sir.
You are somewhat correct. However, the FSB isn't 'halved' to each core. If both cores are MAXED, then they will see only half being used. If one core is completely idle and the other is maxed, the one should see near-full utilization of the FSB. Now, the Atom 330s aren't 'true' dual cores to begin with. They're multiple dies on the same silicon, that haven't been split apart. That's why the Thermal Footprint is exactly 2x that of a single core (same with the Core2Quad vs Core2Duos). They don't have a L3 cache to speak to one another, therefore must speak to each others' L2 cache, they must utilize the FSB, decreasing overall bandwidth..
Um, is anyone else thinking they wish they could vote @George and @Precurse to be editors on Engadget? Haha.
That said, I really want to thank you guys for taking the time to explain these things. I built one of these for a friend who needed a low-power, very cool-running little internet box for complex browsing but not power-use. To be honest it's quick but I was disappointed. There were few instances when it felt more than just slightly faster than my single Atom netbook. BUT the times when it was significantly snappier was running like say a dozen programs in OSX. On the single core you'll notice some sluggishness begin in Firefox3 and VMware whereas on the 330 it's still chugging along like it's on caffeine. I do realize that having an extra core is SIGNIFICANT when running virtual environments but I still was surprised at the difference between the two in that category.
Anyway, again, thank you for explaining how these work. I didn't really have a good understanding of how these tiny dual atom mobos work before now.
On a separate note, for anyone doing a benchmark on a processor as unique as the Atom, I think they should also look in to what the processors will likely be used for. That said, the Dual-Core Atom is likely too underpowered to run Vista comfortably on a netbook, while also being too over-powered to be sold with an XP license. This is why if you look on any of the websites selling Atom330 mobos or if you look on the forums you will find that most of the people running these little cheetahs are using them for personal home servers. This is ideal because of their ultra low power draw and that they run extremely cool. And how likely do you think home server people are going to be using virtual environments? Isn't that almost what that technology was designed for? Anyway, I'm just thinking out loud here but it makes a lot more sense to me now that these dual-core super low wattage rigs would be ideal for a home server, and I think the chipset handles that pretty well. Let's hope we can see some benchmarks verifying that online here very soon.
To some up what george said.... Atom processors suck.
There were too many variables in this comparative benchmark. Different machines (Shuttle PC chassis vs Samsung netbook), different hard disks, different XP set-up, different CPUs, etc. If the Samsung ran Office faster than the Shuttle, then the conclusion is that the Samsung *system* is faster. It could be due to the chassis, hard disk, CPU, XP set-up, anything. The headline could equally say "7200 RPM disk is benchmarked, fares slightly worse than 5400 RPM disk"...
Well played.
How does a chassis make a computer faster? Flame decals?
Yeah my newegg sticker give me an extra 3fps in crysis.
LOL about the go-faster stripes! I meant, of course, that the different chasses (mini desktop vs netbook) contained different motherboards...
Well, I bought one of these ready soldered into an Intel D945GCFL2 motherboard last week. Mini ITX, Intel chipset, GMA950 video, HD audio, 2x SATA, USB, gigE for about 60UKP new.
I've just built a system around this - added 2GB RAM, a boot drive and 2x500GB SATA drives in RAID1, installed OS X 10.5.5 on it and created a Hackintosh.
It's bloody great - home NAS, web/ftp/webDAV/file server, Time Machine backup, redundant storage - and the whole thing draws about 45W of power.
Beat that, Time Capsule.
And the moral of the story is . . . . don't use Office. - or Windows for that matter.
The VIA Nano has never been beaten by the Intel Atom N270 at any performance benchmark except for at power consumption, where the Atom uses slightly less power (but only when running at full speed). I know this is a comparison of the VIA C7-D, but why compare the Atom N270, which is used for laptops, against VIA's previous generation tech? No shit the Atom beat the old tech. The VIA Nano has been out for over half a year.
They compared it to a VIA cpu that you can actually purchase, which is a pretty fair criteria. Exactly where have you seen a Via Nano CPU for sale (embedded or otherwise)?
For some benchmarks with real figures you can relate to:
http://www.mini-itx.com/reviews/atoms/default.asp?page=7
Wouldn't it have been kinda funny/cool if the dual-core Atom chips were called Molecules?
What I'd like to know is if the extra core is sufficient that CoreAVC combined with Haali's Renderer are now sufficient for 720p playback. CoreAVC is pretty good at taking advantage of dual-core processors, so it should see a pretty healthy boost in performance on the 330, possibly enough to get 720p content working. And Haali's Renderer (part of Haali's Splitter, which is maintained by an employee of the guys making CoreAVC) supports raw-frame (and audio) buffering, which goes a LONG way towards smoothing out CPU spikes when your computer is borderline in decoding/playing back content.
For a point of comparison, my Core 2 Duo laptop with ffdshow has serious trouble (we're talking 5FPS) with some high-motion scenes in 1080p content. CoreAVC has no problem playing back the scene at full speed, and Haali's keeps things ticking when another app grabs a burst of CPU time every so often.
A single core Atom is already sufficient to decode 720P material with CoreAVC if you tweak it right. Instead of trying to use Haali to give it a big buffer you should try setting the buffer in MPC up to 100 or 200MB. I usually overclock my Atom up to 1.9Ghz anyways but I've tested it at 1.7Ghz and even tough to decode scenes come out smooth. You can even look at dropped frames stats to check whether it's enough. Most of the time 1.6Ghz will be fine anyways with enough of a buffer since most scenes aren't really that demanding.
I actually think the Haali renderer might be using more CPU power on your system than the regular overlay renderer would and making it harder for you to decode 720P.
Is this faster than a 2.66GHz Pentium 4?
And how does it compare to a 1.2 Ghz 45nm Core 2 duo ULV ??