Nano crushes Atom

Printable View

Show 100 post(s) from this thread on one page

07-29-2008, 10:47 PM
JumpingJack

Quote:

Originally Posted by Hornet331

its time that reviewers use PSUs with higher efficiency for such review... heck there are even PUS out there that have 92% efficiency @ 20% load

http://www.80plus.org/manu/psu/psu_r...50W_Report.pdf

if im gona rebuild my HTPC this baby is gona power it. :yepp:

Now there's the ticket!
07-29-2008, 10:47 PM
savantu

Quote:

Originally Posted by villa1n

Maybe i don't fully understand, but wouldn't an overpowerful psu, be a detriment to both systems... or are you saying it is unfairly skewing the results in favour of the nano?

The higher the power consumption the lower the variability is.

With a 600w PSU , a 4w CPU like Atom is in the noise region.We're talking a granularity yet to be found in PSUs.

JumpingJack posted a chart comparing the difference between a 600w and a 220w PSUs on an Atom system.The difference is on the order of 20w.
07-29-2008, 10:53 PM
JumpingJack

Quote:

Originally Posted by savantu

The higher the power consumption the lower the variability is.

With a 600w PSU , a 4w CPU like Atom is in the noise region.We're talking a granularity yet to be found in PSUs.

JumpingJack posted a chart comparing the difference between a 600w and a 220w PSUs on an Atom system.The difference is on the order of 20w.

20 watts out of ~55 ... which is ~35% error :)

Anyway -- the reviewers are not comparing the products that are actually competing in this space. They should be looking at Nano vs the Celeron 220, in the D201GLY2A --- it is about the same performance, but the Celeron 220 consumes about 20% less power.

Jack
07-30-2008, 12:34 AM
STaRGaZeR

Quote:

Originally Posted by Donnie27

So if it is getting the job done sooner, it is really using less power. If it ain't buggy, I easily go with the Nano for low power and performance=P PCI-E slot? For CUDA or something?

Much rather use a MATX and an E2180 or something.

If you read it carefully it seems that they entire comparison is a bug. Also in your quote they're comparing the entire platform, not the CPU only. In perfomance/watt the Atom wins, in absolute perfomance the Nano wins.

Either way I still don't get WTF they've paired a 2,5W CPU with a 6W chipset. Shrink it or something :stick:
07-30-2008, 12:55 AM
Hornet331

Well i think intel did silverthorn/menlow more as a proof of concept, i dont think they expected that it is such a hughe demand for it. :p

I think you will see the real performance/watt king with moorestown. :yeep:
07-30-2008, 12:56 AM
Shintai

The 8800GTX aint used for the Nano or Atom. Hothardware have been a huge failure basicly.

Also even tho some dont seem to accept it. The power consumption of the Atom is alot lower than the Nano. Even tho this is a desktop model and they still have the crappy 945 paired. It will be good to see its next incarnation.

I also think Atom is quite abit cheaper than the Nano boards. Here a Nano board cost about the same as a Q9550.

Nano´s segment is also mainly targetted as an alternative to LV/ULV "standard" mobile CPUs.
07-30-2008, 01:20 AM
ownage

I think 55watt is to much for an Atom at full load. VR-zone managed to pull only 39,8 watts.
Anyway I'm going to test Atom power consumption myself.
07-30-2008, 02:41 AM
Hornet331

Quote:

Originally Posted by ownage

I think 55watt is to much for an Atom at full load. VR-zone managed to pull only 39,8 watts.
Anyway I'm going to test Atom power consumption myself.

just look at the results JumpingJack provided, they have just used a far to big psu.
07-30-2008, 03:16 AM
R101

Isn't there a 'Poulsbo' motherboard that works with 'Z' Atoms and uses way less power?
07-30-2008, 03:49 AM
jcool

I'm sorry but in my book, both platforms suck. 0 performance and yet rather high power consumption.
My IBM X60 drains below 15W idle with wlan and Bluetooth enabled, and that's with a 12" screen, rather strong Yonah Dualcore and 2Gigs of ram.
Of course it costs 5x as much, but the old rule still applies I guess.. what you get is what you pay for ^^

The power drain of either Atom or nano platforms are ridiculously high in any case. Of course, hothardware's measures are incorrect as always. You won't get under 30W idle though, which is still 3x higher than it should be.
07-30-2008, 04:46 AM
XS Janus

^ +1
07-30-2008, 05:05 AM
Shintai

Quote:

Originally Posted by R101

Isn't there a 'Poulsbo' motherboard that works with 'Z' Atoms and uses way less power?

Yes, these are "desktop" types. Also the CPU used here cost 29$.
07-30-2008, 05:50 AM
Donnie27

Quote:

Originally Posted by JumpingJack

20 watts out of ~55 ... which is ~35% error :)

Anyway -- the reviewers are not comparing the products that are actually competing in this space. They should be looking at Nano vs the Celeron 220, in the D201GLY2A --- it is about the same performance, but the Celeron 220 consumes about 20% less power.

Jack

Bang up job on all of the info and thank you very much! Yet I'll still say I'd rather have an E2180 and MATX:up:

Quote:

Originally Posted by STaRGaZeR

If you read it carefully it seems that they entire comparison is a bug. Also in your quote they're comparing the entire platform, not the CPU only. In perfomance/watt the Atom wins, in absolute perfomance the Nano wins.

Either way I still don't get WTF they've paired a 2,5W CPU with a 6W chipset. Shrink it or something

Here's something for you to read carefully, no need for the stick since it looks like Jack explained it to both of us;)
07-30-2008, 06:12 AM
STaRGaZeR

Quote:

Originally Posted by Hornet331

Well i think intel did silverthorn/menlow more as a proof of concept, i dont think they expected that it is such a hughe demand for it. :p

I think you will see the real performance/watt king with moorestown. :yeep:

Yes, I think the same. The real thing will come when everything (CPU, NB and SB) is included inside the same chip. Until then... is just a 'temporal' solution if you know what I mean :p:

Quote:

Originally Posted by Donnie27

Here's something for you to read carefully, no need for the stick since it looks like Jack explained it to both of us;)

:yepp:
07-30-2008, 06:12 AM
Donnie27

Quote:

Originally Posted by Shintai

Yes, these are "desktop" types. Also the CPU used here cost 29$.

I think one site said the board and processor sold for $79.
07-30-2008, 06:22 AM
Miwo

All the reviewers should be using a PicoPSU for something like the nano or atom.....
07-30-2008, 06:40 AM
RTB

I'm kind of amazed that in 66 posts, nobody asked how these buggers overclock? Even taking into account that they're supposed to be low power only.
07-30-2008, 06:49 AM
Donnie27

Quote:

Originally Posted by RTB

I'm kind of amazed that in 66 posts, nobody asked how these buggers overclock? Even taking into account that they're supposed to be low power only.

Overclocking low performance processors meant for low voltage, that's a new one:up:
07-30-2008, 07:05 AM
Quest_7F

Ok, that feature comparison chart, can someone please explain to me what to be concerned with there.
07-30-2008, 07:38 AM
Gogeta

Like other posters have stated, competition is exactly what makes this industry/overall market a more enjoyable place. Go VIA, go.

Quote:

Originally Posted by NapalmV5

even the 700/900mhz @ eeepc are snappy critters

what all these low-watt cpus need is fast/er hdd/ssd

The lightweight OS doesn't hurt either. :D
07-30-2008, 09:28 AM
villa1n

Quote:

Originally Posted by JumpingJack

It is unfairly skewed toward Nano on the power measurements. You could power both these boards with a 1200 W power supply if you wanted... why you would do this is beyond me... in fact, why you would use even a 200W power supply is beyond me. But that is what these numnuts did... they used a high rated power supply to power a low wattage board. Just because a PSU is rated at say 500 Watts does not mean that it outputs to that wattage to the board. The PSU will only output what the load is requesting.

Just a quick lesson in power supplies. Power supplies come rated in the total peak power they are able to deliver, sustained power is a bit lower....

Power supply makers are also interested in quoting you power efficiency, which is essentially the % of the power input that makes it to power output at a given load.

For example, say I measure at the wall socket 100 Watts, but the efficiency is only 80%, the actual power delivered is only 80 Watts.

Ok... now, here is the kicker. The efficiency quoted for a power supply is between 20% and 80% of it's rated power. Above or below that, the efficiency is drops of rapidly.

Here is an example Anandtech did of a recent 900 W PSU:
http://www.anandtech.com/casecooling...spx?i=3364&p=7

Notice at 200 W the efficiency starts to drop off rapidly, going to 73% at 90 W load... now look at how steep that curve is.... at 20 W (which is about what these boards are pulling) efficiency is down near the 40 or 50% range.

Ok... now think about it. If 30 or so W of the raw numbers in these reviews is wasted by the PSU, how much penalty does that put on a 2-3 W CPU vs a 17 W CPU?

Example ... 17 out 70 is much less than 17 out 30 (percent wise), but 2-3 out of 50 is nothing, you are essentially vs 2-3 out 30 is quite a bit different.

If you are thinking of building a very low wattage nettop like computer --- the last thing you want to buy is a 650 W thermaltake toughpower to power you Nano or Atom board. This is a much better power supply to build the system: http://www.mini-box.com/picoPSU-90-power-kit

And after all is said and done, the peak atom draw will be in the 30-35 range.

Not 60 like they review sites are telling you. In fact, their data is so far off... you should completely disregard any conclusions they make about power efficiency or performance per Watt. I am writing it off as either laziness or groteseque incompetency.

Jack

Ahh oh, i wasnt aware of the lower limits affecting efficiency of PSU's since i m usually only ever pushing the upper limits :P So basically the margin of error is ridiculous at that point, and it is not really a measurement of what the board is doing, but the powersupply. Gotcha.
07-30-2008, 09:54 AM
ToTTenTranz

Quote:

Originally Posted by Shintai

Yes, these are "desktop" types. Also the CPU used here cost 29$.

Actually, the Centrino Atom (Atom Z + Poulsbo) is announced only to be integrated in MIDs, at the moment. The netops are more cost-oriented than power-oriented, hence the different packaging between Atom Z (lower power) and Atom N (cheaper).

We can only assume it's going to replace the current 945G+ICH7 because Intel can surely make a more profitable single bridge @ 65/45nm than sticking two 130nm bridges. At least in the long run.

And the Poulsbo may actually never come to nettops, since it has no SATA controller (it could use one PCI-e 1x lane for an external controller though).
07-30-2008, 10:20 AM
PetNorth

http://arstechnica.com/reviews/hardw...o-review.ars/8

250W PSU

http://media.arstechnica.com/reviews...edia/Power.jpg
07-30-2008, 10:32 AM
perkam

Quote:

Originally Posted by jcool

I'm sorry but in my book, both platforms suck. 0 performance and yet rather high power consumption.
My IBM X60 drains below 15W idle with wlan and Bluetooth enabled, and that's with a 12" screen, rather strong Yonah Dualcore and 2Gigs of ram.
Of course it costs 5x as much, but the old rule still applies I guess.. what you get is what you pay for ^^

The power drain of either Atom or nano platforms are ridiculously high in any case. Of course, hothardware's measures are incorrect as always. You won't get under 30W idle though, which is still 3x higher than it should be.

There is some truth to that. Although, X61 > X60 anyday :p:

Perkam
07-30-2008, 12:41 PM
Kingcarcas

It looks like Jumpingjack has won teh Internets :up: Someone get that man a cookie STAT! Anywho, Atom is better but the platform is f-ed, Intel needs to get a better chipset in there.

Show 100 post(s) from this thread on one page