Is an Atom better than an Athlon XP 2400+?
in [UK Linux]
|
Prev: Ogg in Firefox
Next: ACPI says temp is 256C
From: alexd on 11 Sep 2009 16:26 Daniel James wrote: > I'd say that ion is fairly frugal, and while it does draw more power > than the intel chipset under most condition it doesn't draw much more, > and does perform better. OTOH the 945 is so bad that it shouldn't be > too hard to design a chipset that outperforms it AND draws less power > -- the ion isn't that chipset. Is it possible to underclock a chipset + CPU to achieve the desired power level? -- <http://ale.cx/> (AIM:troffasky) (UnSoEsNpEaTm(a)ale.cx) 21:25:28 up 13 days, 23:21, 2 users, load average: 0.14, 0.20, 0.18 Qua illic est accuso, illic est a vindicatum
From: Daniel James on 12 Sep 2009 18:47 In article <h8d456$ovl$1(a)news.eternal-september.org>, Chris wrote: > That's right. The cpu draws about 8w (Atom 330) whereas the chipset > draws about 20w. That's what you get for coupling a 45nm CPU with a 130nm chipset ... > [I wrote] > > The ION, on the other hand, is a fairly high performance chipset > > and -- however efficient it may be -- it does end up drawing more > > power than the 945 under most conditions. > > That's contrary to what I've seen and the reason why I'm keen on it: > http://www.tomshardware.com/reviews/nvidia-ion-atom,2153-10.html > http://www.anandtech.com/video/showdoc.aspx?i=3562&p=51 Thanks for those ... That Anandtech link doesn't work for me (just goes to a "Search" page at Anandtech.com). The Tom's article seems confused to the point of misinformation ... they harp on about having had to use different power supplies to test the different boards AND about working from published efficiency ratings for the PSUs rather than measurements of their own units; they produce some results that seem to me to show the Ion performing surprising well compared with the 945; and then they express surprise that it didn't do even better based on some comparison of published TDPs (which have nothing to do with actual power consumption of any given component). I find their figures -- and their apparent lack of understanding of what they're talking about -- strangely unconvincing. > I wonder where the disparity comes in, then. They do mention > different power supplies in your second link which could easily > be where the extra draw comes from. I have seen benchmarks of poor > quality PSUs being highly inefficient. The xbitlabs review -- like the Tom's one you cited -- is based on the nVidia reference/prototype box for the Ion, which takes a DC power supply from an external brick rather than using a standard PC PSU. Any review comparing that kit with a standard PC board is going to have some uncertainties the PSU efficiency. The PSU used for the 945 system in the xbitlabs test was a "Power Man IP-AD120-2" (whatever that is) I don't know how efficient that might be? Given that the Ion box was an nVidia reference platform they may have cherry-picked components to give good performance in the review, I imagine (but don't know) that the 945 system involved was off-the-shelf. > If you mean the xbitlabs benchmark, I doubt the 1w differences are > significant. No, indeed ... but taking the figures at face value the (which I agree may not be meaningful) the best we can say is that the Ion uses about the same or a little more power than the 945 under load, and perhaps a little less at idle. I wish I could find the comparative review (of one of the Zotac Ion boards) that I was reading a few weeks ago -- IIRC it found a higher power consumption than intel under all conditions, and by a greater margin. Of course, a commercial board may perform less well than a manufacturer's reference sample ... or it may be that the Zotac board was tested with an extra RAM DIMM, or something, which would add a few Watts across the board. Note, also, the point made in the xbitlabs review that their comparison was between the Ion platform and a desktop board using the 945GC chipset, and that the 945GSE chipset used in netbooks is much more efficient (by 16W or so at idle). A desktop board using 945GSE would be interesting ... > > The ion is also expensive ... �100+ for every Atom 330 board I've > > seen, while the D945GCLF2 is around �65 (plus VAT, in both cases). > > You'd have to run the ion at idle for a long time to make up the > > difference in saved power ... > > That is true. Although the power difference isn't the only reason > for choosing the Ion, that is a big price difference. hmmm... This > isn't as obvious as I first thought. In fairness, though, a board like the Zotac IONITX-D-E has some definite advantages (apart from the chipset) over a D945GCLF2. For one thing it has two DIMM slots and can take twice the RAM (4GB vs 2GB) and presumably interleave accesses to the two DIMMs for greater RAM throughput. It also has twice as many SATA ports. It DOES cost twice as much, though. > I want a useable desktop system (with X, Gimp, OOo and iPlayer > functionality); a Geode won't give me that, from what I've seen. Probably not, no ... as I said: it's a question of which compromises you're prepared to make. I have a system running a 1GHz VIA C3 (in a VIA M10000 board with 512MB RAM) which certainly runs X and OOo acceptably ... I haven't tried Gimp or iPlayer (though my box does have a DVB card and makes a passable TV) ... I suspect Gimp would quickly run out of steam without more RAM and iPlayer would be OK. Cheers, Daniel.
From: tinnews on 15 Sep 2009 13:22 Gordon Henderson <gordon+usenet(a)drogon.net> wrote: > In article <h82ufn$10q$1(a)news.eternal-september.org>, > chris <ithinkiam(a)gmail.com> wrote: > >I'm wanting to replace my ageing home desktop with a new small machine > >that I can leave on most of the time. It's for very basic email, web, > >openoffice use with the occasional light photo editing. Eventually would > >like to add audio streaming. > > > >Something like this could be ideal: > >http://www.asrock.com/nettop/spec/ION%20330.asp > > > >My question is how much difference (better or worse) would I notice when > >compared to my current athlon XP 2400+? In particular, is the iPlayer > >cpu or gpu dependent. > > I replaced my XP2400+ with an Intel Atom dual-core mobo some 6 months > back. Not really noticed much difference - mplayer still works, audio > is OK, X is fine too, but I don't do anything that demanding. > > >Ideally, I want something that uses a minimum of power when idling, but > >with enough grunt to do the above day-to-day stuff. > > My box idles at about 42 watts including a single SATA drive. > You can do better than that with an Intel Core 2 Duo processor, Tom's hardware has two machines with one down at around 30 watts by choosing an efficient motherboard and power supply. The Atom comes with very inefficient chipsets in the main. -- Chris Green
From: Gordon Henderson on 15 Sep 2009 15:59 In article <h8oihc$9nq$1(a)news.eternal-september.org>, <tinnews(a)isbd.co.uk> wrote: >Gordon Henderson <gordon+usenet(a)drogon.net> wrote: >> In article <h82ufn$10q$1(a)news.eternal-september.org>, >> chris <ithinkiam(a)gmail.com> wrote: >> >I'm wanting to replace my ageing home desktop with a new small machine >> >that I can leave on most of the time. It's for very basic email, web, >> >openoffice use with the occasional light photo editing. Eventually would >> >like to add audio streaming. >> > >> >Something like this could be ideal: >> >http://www.asrock.com/nettop/spec/ION%20330.asp >> > >> >My question is how much difference (better or worse) would I notice when >> >compared to my current athlon XP 2400+? In particular, is the iPlayer >> >cpu or gpu dependent. >> >> I replaced my XP2400+ with an Intel Atom dual-core mobo some 6 months >> back. Not really noticed much difference - mplayer still works, audio >> is OK, X is fine too, but I don't do anything that demanding. >> >> >Ideally, I want something that uses a minimum of power when idling, but >> >with enough grunt to do the above day-to-day stuff. >> >> My box idles at about 42 watts including a single SATA drive. >> >You can do better than that with an Intel Core 2 Duo processor, Tom's >hardware has two machines with one down at around 30 watts by choosing >an efficient motherboard and power supply. > >The Atom comes with very inefficient chipsets in the main. And oddly enough I know this... Now. I did recently build a server for a client - core 2 duo, 4GB of RAM and a pair of the fast WDC drives (640GB?) It's a busy little LAMP type server, but I thought I'd enable all the speedstep/low power stuff anyway - it idles at about 55 watts. (And the drives are about 10 each) I'm not that impressed with the atom motherboard in terms of power, but I've got it now, so hanging on to it! My main aim was to see how the Atom was going to fare as a cheapish server type of board (LAMP and VoIP) and I'm impressed enough to use it for just that... One thing though - on 4 Atom boards I've bought (3 dual core, 1 single), then fans have siezed up. One is running in an air conitioned data centre without the fan quite hapilly (and the cpu does get thrashed!) the others I've replaced the fans for decent ones. These are all Intel boards... I'm just about to put together 2 single core fanless Atom mobo's in one 1U rack case for a little project too.. Gordon
From: alexd on 16 Sep 2009 16:30
Gordon Henderson wrote: > I'm just about to put together 2 single core fanless Atom mobo's in one > 1U rack case for a little project too.. There was a story on /. today about multi-core, 2GHz ARM chip design. It'll be interesting to see if that shows up on any desktop- or server-flavoured motherboards. -- <http://ale.cx/> (AIM:troffasky) (UnSoEsNpEaTm(a)ale.cx) 21:23:07 up 18 days, 23:18, 2 users, load average: 0.06, 0.14, 0.19 Qua illic est accuso, illic est a vindicatum |