POPPUR爱换

标题: [多图]VIA Nano 火拼 Intel Atom [打印本页]

作者: 平日 时间: 2008-7-30 14:18
标题: [多图]VIA Nano 火拼 Intel Atom
[多图]VIA Nano 火拼 Intel Atom
      ugmbbc发布于 2008-07-30 08:56:39|3433 次阅读字体：大  小  打印预览

在低功耗计算领域,被Intel寄予厚望的Atom遭到了耕耘该市场数年之久的VIA的强力回击,65nm Nano对45nm Atom的战争已经打响.
目前他们已有的产品分别为VIA Nano L2100 (1.8GHz)和Intel Atom 230 (1.6GHz),这两位主攻手将承担第一场性能大战,PC Perspective 与 Hot Hardware已经给出了他们的性能评测数据,目前看来从性能上,VIA更占优势,在能耗上,Atom又更胜一筹一起来看吧!

VIA Nano

Intel Atom

VIA Nano Reference Platform

For testing and evaluating the VIA Nano platform the companyprovided us with a mini-ITX motherboard from a reference design - thisexact product isn't going to be on sale though partners will offerproducts very similar to it.

Themini-ITX platform has officially been adopted by Intel with the Atomproducts meaning that the VIA-created standard should actually become astandard now -- good news for all. For those that haven't seen amini-ITX motherboard yet you'll be scratching your head at seeing sucha compressed design. There are two DDR2 DIMM slots, four SATAchannels, one IDE connection and a single PCI Express x16 slot forgraphics or other add-in cards.

Onthe back of the reference board are a CompactFlash connection (for somesolid-state storage action) and a mini PCI slot as well. These wouldbe GREAT features to have on a retail board as well - I imagine bootingoff of an 8GB high-speed CF drive and using a standard drive forstorage.

Theexternal connections on the VIA reference board include the standardPS/2 connections, USB ports, serial output, dual NICs and VGA output touse in conjunction with the integrated S3 graphics core on the northbridge. There are audio connections as well on the right hand sidethat support up to 6-channels of output.

Hereyou can see the VIA 82375 south bridge under a passive heatsink, fourSATA channels on the motherboard that all support RAID and the singlePCI Express x16 slot that gives the VIA Nano platform such an edge overIntel's Atom for real world user applications. Without a PCIe slot onthe Atom motherboards, OEMs and users are forced to use the inferiorintegrated Intel IGP or a PCI graphics board, really limiting thepossibilities of getting top performance out of the Atom architecture.

VIA's reference design supports up to 4GB of DDR2-667 MHz memory though the controller is just single channel.

Theonly power connection on the VIA Nano motherboard is a 20-pin ATX powerconnector - don't worry your 24-pin power supply connections will workjust fine though.

Underthat single large heatsink with the fan on it lies the VIA CN896 northbridge chip and the VIA Nano processor; in our case the CPU is the Nano L2100 which operates at 1.8 GHz on an 800 MHz front-side bus with a max TDP of 25 watts.

Thenorth bridge handles all the PCI/PCIe and memory controller functionsand also sports the VIA Chrome9 HC IGP that supports DX9 features. Nothing to scream about, but it turns out to be faster than Intel'ssolution.

Intel Atom D945G Platform

Forcomparison testing to our VIA Nano reference platform we got our handson an Intel D945GCLF motherboard with processor package.

You can find this exact retail package for sale on our pricing engine for just about $75- not a bad deal from just the amount of functionality you get insidethe box. Just like the VIA board, the Atom processor is soldered andpermanently affixed.

TheIntel D945GCLF is the same form factor as the VIA Nano referenceplatform: mini-ITX. This board design looks to be much simpler atfirst glance and that is in fact the case as we investigate a bitfurther into the design.

Theonly expansion slot that the Intel Atom motherboard offers is a PCIslot - any graphics expansion you plan to do will have to be relegatedto this older interface. This is really VIA's best selling point -they are allowing their platform to be completely open but Intel isONLY ALLOWING their boards and board partners to ship with PCI slots.

TheIntel board only has a single DIMM slot supporting up to 2GB ofDDR2-667 memory. The ATX power connector is again a 20-pin rather than24-pin but will still fit with 24-pin PSUs. There is a single IDEconnection and two SATA channels for your storage connectivity.

Hereyou can see the processor heatsink is incredibly small and passive onlyand is dwarfed by the chipset heatsink above it. Notice also that theIntel board requires you to use a secondary 4-pin power connection inaddition to the standard ATX power connector.

Theexternal connections on the board are similar to those seen on the VIANano reference board - PS/2, USB 2.0, serial, parallel, a single 10/100NIC, 6-channel audio outputs and a VGA output.

Removingthe heatsink reveals the full array of Intel chips - the Atom processorat the top, 945G north bridge in the middle and the Intel ICH7-M southbridge at the bottom. It is interesting to see how much more coolingthe 945G chipset requires compared to the Atom processor.

[ 本帖最后由平日于 2008-7-30 14:24 编辑 ]

作者: 平日 时间: 2008-7-30 14:19
不是Nano太强，纯属Atom太弱，Intel继续抢钱吧...

[ 本帖最后由平日于 2008-7-30 14:23 编辑 ]

作者: elisha 时间: 2008-7-30 14:34
在atom的芯片组更耗电的情况下，满载比atom高出20w来
而atom1.6的tdp不超过4W

作者: 平日 时间: 2008-7-30 14:50

原帖由 elisha 于 2008-7-30 14:34 发表
在atom的芯片组更耗电的情况下，满载比atom高出20w来
而atom1.6的tdp不超过4W

4w的 atom配945，鲜花插在牛粪上：

Power Consumption Testing

Becauseof the importance of power and efficiency for this particular article,we are going to spend extra time with our power meter. First up, let'ssee our standard idle and load power consumption comparisons:

Atidle the VIA Nano uses just about 3 watts more power but at full loadthat changes dramatically. The VIA CPU spikes the system as high as77.5 watts while the Intel Atom processor JUST crosses the 60 wattbarrier.

I also decided to usethe ability for my power meter to log results to see how powerconsumption looked over time when comparing the VIA Nano and Intel Atomprocessors. What you see below is a graph of power consumption overtime.

Forthe first test I used our benchmark that encodes an MP3 with LAME tocompare power draw.  Notice that for the blue line that represents theVIA Nano processor, you can clearly see where the processing begins: atthe large jump in power consumption from about 60 watts to 77 watts orso.  The Intel Atom processor lives up to its claim of super-lower loadpower consumption and you almost can't tell that the Intel platformincreases power draw during the test - but it does!

Of course as our benchmarks showed you onearlier pages, the VIA Nano CPU is definitely faster though - so howcan we compare the data in this graph?  One way is to see how muchenergy it takes for each CPU to actually complete the work.  You mightremember this theory from Intel's Centrino 2 launch known as the "Hurry up and Get Idle" idea.

Forour MP3 encoding test, the VIA Nano processor used a total of 37,323watt-seconds (Joules) of energy while the Intel Atom processor used38,290 watt-seconds (Joules) of energy.  That is a difference of just2.5% indicating that even though the Atom processor is slower, it's notthat much less efficient than VIA's Nano.

Usingthe same method to gauge the results of our CineBench 10 test, we findthat the VIA Nano used 63,434 watt-seconds (Joules) of energy to renderthe scene while the Intel Atom used 65,893 watt-seconds (Joules) ofenergy - an advantage of 3.8% to the VIA CPU.

These kinds of power comparisons are incrediblyinsightful and you can clearly see how both teams of CPU designers havemade trade offs for the either power consumption or speed. The VIANano L2100 is able to perform these tasks faster (by as much as 30% insome cases) while still using less total energy than Intel's Atom.

作者: Prescott 时间: 2008-7-30 15:02
nano啊，你太弱了吧。{cry:]

作者: zaknafein 时间: 2008-7-30 15:55
这也弱得太没边了...同频相当么...

作者: frankincense 时间: 2008-7-30 16:33
更高的功耗和三倍以上atom的核心面积
与其带来的性能优势不成比例呢……

作者: boris_lee 时间: 2008-7-30 16:56
貌似.........现在市面上的atom都是270把

作者: itany 时间: 2008-7-30 19:06
靠，当初看理论测试还以为Nano能和x2比划比划，现在看，在超线程的掩护下，Aom同频也就是比Nano稍慢一点
测试没找个同频的PM来已经是给Nano很大的面子了……
VIA自己说得好，3发射，乱序执行 vs 双发射顺序执行，最终就是这样一个结果，看来VIA也就是这样一个水平了
Nano估计实际功耗不比TDP低，估计和低电压Penryn功耗应该差不多吧。别忘了Penryn是双核的。VIA的功耗控制也不过尔尔

作者: pjtomtai 时间: 2008-7-30 21:57
NANO的实测性能低的出乎意料.
相反, ATOM在进入32nm制程以及集成南北桥和图形后会很诱人,问题是会在2009年发生吗?

作者: Cameraman 时间: 2008-7-30 22:04
双网卡的主板非常不错

作者: pjtomtai 时间: 2008-7-30 22:17
这些评测还陷入了最基本的误区,用那么大的PSU测低功耗的平台,简直是犯罪.
一个1200W的PSU,就算超级高效率-90%,也有120W给浪费了.

作者: iwannasay 时间: 2008-7-31 00:13
nano上45nm的话，功耗和面积会小一些

作者: elisha 时间: 2008-7-31 00:56
本来不应该是同一个级数的对手的

但是，via，你做到了{sweat:]

作者: boris_lee 时间: 2008-7-31 01:21

原帖由 pjtomtai 于 2008-7-30 22:17 发表
这些评测还陷入了最基本的误区,用那么大的PSU测低功耗的平台,简直是犯罪.
一个1200W的PSU,就算超级高效率-90%,也有120W给浪费了.

效率不是这个算法吧:funk::funk::funk:

作者: 花泥 时间: 2008-7-31 10:19
晕，这年头都流行北桥比U大而且热了。。。。。。

作者: Jason21 时间: 2008-7-31 11:45
Nano也不过如此，面积还那么大

作者: 明镜止水 时间: 2008-7-31 12:25
怎么不拿1.6G 的NANO跟ATOM比较啊。TDP低许多啊。{closedeyes:]

作者: samhrc 时间: 2008-7-31 13:24
实在搞不懂为啥Atom还要搞个南北桥，直接整一个SOC 多好。

作者: Prescott 时间: 2008-7-31 13:32

原帖由 samhrc 于 2008-7-31 13:24 发表
实在搞不懂为啥Atom还要搞个南北桥，直接整一个SOC 多好。

别急，别急，大官僚机构做事情稍微慢一点点，是可以理解的

作者: qq125177530 时间: 2008-7-31 14:55
-_- 我想问一下：VIA的17*17东西不都是很贵的吗

作者: lsir777 时间: 2008-8-6 12:03
这多数关注ATOM和NANO的人,不仅希望CPU的功耗低,更希望整个系统功耗也是低的.
还有就是组成整个系统各个配件的总成本不要太高,要不,不如来个本本更好了

作者: 少年包青天 时间: 2008-8-8 09:24
提示: 作者被禁止或删除内容自动屏蔽

作者: acqwer 时间: 2008-8-8 09:46

原帖由 少年包青天 于 2008-8-8 09:24 发表
via nano的这个成绩差么？把e6600和4600+都屏蔽掉一个核心，再来测试吧。

明年上半年via会推出双核心45nm nano，频率可达3G，amd和intel谁都笑不出来！

很希望nano跟上潮流，废除fsb，倒是当务之急！

PCmark可不是双核成绩就能翻倍的啊。9650的得分才5300，只有E6600的1.4倍。

作者: taylee 时间: 2008-8-8 09:51
intel难免试试水嘛，可以理解

作者: xiaolongzi 时间: 2008-8-8 13:00
ATOM还是很强。毕竟核心面积比NANO小很多。

作者: kimax 时间: 2008-8-28 12:15
感觉ATOM推出并不是针对ITX市场，而应该是手持式微机市场（如手机、PDA之类）。{victory:]

作者: flymop 时间: 2008-8-28 12:23
INTEL没认真，VIA洗洗睡吧

作者: Lynin 时间: 2008-8-29 09:37
提示: 作者被禁止或删除内容自动屏蔽

作者: chfyjn 时间: 2008-8-29 11:47

原帖由 Lynin 于 2008-8-29 09:37 发表
华人的企业，还是要支持的，总比所谓龙芯好吧

龙芯和NANO这两者根本就不具有可比性，架构都不一样的东西，晕死，把东西弄清楚了再来说。

作者: Lynin 时间: 2008-8-29 11:49
提示: 作者被禁止或删除内容自动屏蔽

作者: elisha 时间: 2008-8-29 11:58

原帖由 Lynin 于 2008-8-29 09:37 发表
华人的企业，还是要支持的，总比所谓龙芯好吧

不要热脸贴冷屁股啦，这是纯美国设计的cpu

作者: w56824419 时间: 2008-8-29 11:59
就算VX800难找VX700总不难找吧？这种评测不用看了～～～拿最弱的配置跟另一方最强的配置比～～～低功耗系统本来就是测整套平台的～～～现在到来玩田忌赛马阿～～～

作者: spinup 时间: 2008-8-29 12:10

原帖由 elisha 于 2008-8-29 11:58 发表

不要热脸贴冷屁股啦，这是纯美国设计的cpu

我看谁的牌子远比谁做的重要得多。
在intel,amd,nvidia做的华人再多，都是米国人的东西。做c8的人就算没有一个华人，说起来还更得意。

作者: bill_max 时间: 2008-8-29 12:48
不是Nano太强，纯属Atom太弱，Intel继续抢钱吧...

这句话说得好～

作者: elisha 时间: 2008-8-29 13:43

原帖由 bill_max 于 2008-8-29 12:48 发表
不是Nano太强，纯属Atom太弱，Intel继续抢钱吧...

这句话说得好～

Atom 2W, Nano 20W
这两个凭什么站在一起比{huffy:]

作者: w56824419 时间: 2008-8-29 18:53

原帖由 elisha 于 2008-8-29 13:43 发表

Atom 2W, Nano 20W
这两个凭什么站在一起比{huffy:]

N230是4W～～～

顺序执行功耗本来就底，用了45工艺如果跟乱序的NANO差不多的话Intel就可以回老家种地了～～～这个评测拿65工艺NANO TDP最糟糕的一款型号出来并选择性忽略VX700/800芯片组，来跟45工艺的ATOM比功耗很有意义？！在此提醒下～～ATOM目前唯一的芯片组只有945G，但NANO的选择面很广

VX700的TDP是3.5W，VX800的TDP是3.5W-5W

还是那句话～～～田忌赛马～～～

作者: elisha 时间: 2008-8-29 19:05

原帖由 w56824419 于 2008-8-29 18:53 发表

N230是4W～～～

顺序执行功耗本来就底，用了45工艺如果跟乱序的NANO差不多的话Intel就可以回老家种地了～～～这个评测拿65工艺NANO TDP最糟糕的一款型号出来并选择性忽略VX700/800芯片组，来跟45工艺的ATOM比功 ...

就算是4W，另外一个却要高19W，这两个有可比性？而且945明显功耗更高的
还有，怎么看两者同频性能也差不多，乱序执行3发射的Nano真的好去死了

PS. Atom在移动设备上有专用的芯片组的，2W

[ 本帖最后由 elisha 于 2008-8-29 19:06 编辑 ]

作者: w56824419 时间: 2008-8-29 20:36

原帖由 elisha 于 2008-8-29 19:05 发表

就算是4W，另外一个却要高19W，这两个有可比性？而且945明显功耗更高的
还有，怎么看两者同频性能也差不多，乱序执行3发射的Nano真的好去死了

PS. Atom在移动设备上有专用的芯片组的，2W

你知道NANO常规TDP是多少吗？你以为所有的NANO都是这个TDP规格？不会的话要不要我教你用下google？

你说的Poulsbo几个月前就知道了～～～能出货了再说～～TDP是2.3W～～VX700和VX800实实在在有现货的～～～～

你要说火星产品的话那就没啥好说了～～～

台积电的工艺开始赶上intel了，intel的工艺优势能保持几年难说了，同制造工艺下在功耗还有功能上Intel不会有优势的，从系统集成角度出来说，音频和视频是VIA的优势也是Intel的软肋，Intel的优势在于强大的制造能力和成本压缩能力

BTW：你看PCMARK看不出什么的，这个东西明显存在猫腻，同样是浮点测试，LAME Encoding这样的成绩在PCMARK上根本就没有体现出来，为什么？需要明说吗？有时候实在是最简单的测试手段最能说明问题

还有就是这份测试报告太侧重浮点运算了

数据，可以解读出不同结论的，关键看你怎么读怎么摘了

作者: spinup 时间: 2008-8-29 20:39

原帖由 w56824419 于 2008-8-29 20:36 发表

你知道NANO常规TDP是多少吗？你以为所有的NANO都是这个TDP规格？不会的话要不要我教你用下google？

你说的Poulsbo几个月前就知道了～～～能出货了再说～～TDP是2.3W～～VX700和VX800实实在在有现货的～～～～
...

{lol:] 其他不说，nano其实浮点/sse暴强。

作者: Prescott 时间: 2008-8-29 20:43

原帖由 spinup 于 2008-8-29 20:39 发表

{lol:] 其他不说，nano其实浮点/sse暴强。

改YY VIA了？个人觉得还是YY AMD比较有前途

作者: 66666 时间: 2008-8-29 21:19

原帖由 kimax 于 2008-8-28 12:15 发表
感觉ATOM推出并不是针对ITX市场，而应该是手持式微机市场（如手机、PDA之类）。{victory:]

4W的东西还想做手机？超小型笔记本才是它最合适的地方

作者: itany 时间: 2008-8-29 21:20

原帖由 66666 于 2008-8-29 21:19 发表

4W的东西还想做手机？超小型笔记本才是它最合适的地方

Atom是最高4W的

作者: itany 时间: 2008-8-29 21:23

原帖由 w56824419 于 2008-8-29 20:36 发表

你知道NANO常规TDP是多少吗？你以为所有的NANO都是这个TDP规格？不会的话要不要我教你用下google？

你说的Poulsbo几个月前就知道了～～～能出货了再说～～TDP是2.3W～～VX700和VX800实实在在有现货的～～～～
...

别YY台积电的工艺了
多少nm并不代表性能、功耗，以及成本

现在Nano连AMD的低功耗Sempron都比不过，还有什么YY的
VIA的GPU和声音优势？这个还真有趣。用G45相比，VIA这点GPU还是自己挖个坑别出来了

另，移动版的Core2 Duo现在遍地都是25W的版本，平均下来单核心多少W呢？

[ 本帖最后由 itany 于 2008-8-29 21:25 编辑 ]

作者: itany 时间: 2008-8-29 21:26

原帖由 spinup 于 2008-8-29 20:39 发表

{lol:] 其他不说，nano其实浮点/sse暴强。

暴强是相对于自己的整数而言么？

作者: acqwer 时间: 2008-8-29 22:41
某些人都把AMD的老总当BC了，居然不用领先intel 65nm半代接近intel 45nm的TSMC的55nm工艺做CPU，传说中的CPU代工首选居然还是特许。

欢迎光临 POPPUR爱换 (https://we.poppur.com/)