I found this from AMD and they have it broken down like this.
ACP TDP(max power)
105W = 137W
75W = 115W
55W = 79W
http://www.amd.com/us-en/assets/cont...61C_ACP_WP.pdf
Printable View
I found this from AMD and they have it broken down like this.
ACP TDP(max power)
105W = 137W
75W = 115W
55W = 79W
http://www.amd.com/us-en/assets/cont...61C_ACP_WP.pdf
It isn't. The Cinema 4D renderer is CRAP.
Cinebench is only popular for those people who DON'T use Cinema 4D. The former is obsessed with statistics and e-peen, the latter is obsessed with creating something in 3D space.
And now that Autodesk has 3DS Max, Maya and XSI all under one roof, I think C4D's relevance is gonna fall even more.
Thanks for explaining that macadamia. What happened to all the maya benchmarks? Im sure I used to see that used alot more often, even at Anandtech and now all I see is this CineBench.
I think Cray is already using these in some boxes and maybe some of the BIG machines. I can't find the article, atm...
The Tech Report has some 3DS Max benches here.
http://techreport.com/articles.x/15818/11
First tests
http://it.anandtech.com/IT/showdoc.aspx?i=3456
http://www.tecchannel.de/server/proz...ance_shanghai/
http://images.tecchannel.de/images/t...3_1000x700.jpg
atleast the power consumption is pretty good.
Unless something has changed recently AMD always rate their processors by TDP. ACP is just there to give a comparison with intel as that's how intel rate their processors.
I am almost certain that you have this backwards.
Barcelona are certainly rated by TDP so if Shanghai are rated with ACP then that makes them worse than Barcelona and that definitely isn't the case.
EDIT:
Actually the more I think about it the more certain I am that you have this wrong. My 9950 is a 140w TDP model yet it is around 100w in reality, so if that 140w was ACP then the TDP would be what, around 200w. Sorry but there isn't a motherboard PWM that can handle that.
KTE looked right into this not long ago and came out with some astonishing findings.
EDIT I'm now confused myself as to what is going on with TDP vs ACP so I have asked AMD to clear this up for me, so I will update as soon as I have my answer..(see post #62)
As I understand it they are using average figures not absolute worse case scenario.
However I wouldn't bet my life on it and I'm happy to be proven wrong. :)
Also don't forget that intel rate their processors in "families" so it's now getting very confusing...On both sides.
Intel is kind of funny that way. My 45nm dual core is rated for 65W while in reality it pulls like 20-30W under load. :) Weirdness.
But on to the big news: Today is launch day! This is going to be exciting.
OK this is confusing, I have been told that Opterons are using ACP but Phenoms are still using TDP.
So a 75w Shanghai is actually 105w rated by TDP although the actual power draw is ~75w.
What is interesting is how this relates to intel as it's quite obvious that Shanghai are more efficient than Harpertown.
This is becoming a mess and probably needs some kind of standard applied because as it stands you can't compare processors between companies going by the numbers quoted on the specification sheet. Both sets just don't correlate.
Erm.. Cinema 4D and POV-Ray, respectively, followed by BodyPaint 3D, and I suppose you could argue, all the modellers designed for the POV-Ray renderer.
I jest, but only lightly - this does illustrate a point... I am sick and tired of people thinking that <benchmark name>: <high number> is the be-all and end-all of performance - personally, I don't give a flying :banana: about (generic) benchmarks - use whichever chips suit your particular application better...
What is the stepping of these retail Shanghai core Opterons? C0, C1 or C2? :rolleyes:
There's no reason why you can't just set the timings yourself. Even if the BIOS doesn't support it it's no problem with use a memory timings setter since the registers to do this are just plain visible on the PCI bus. sourceforge/icspll has "a64timings" which uses pcitweak and will be easy to adapt to newer AMD CPUs.
The error checking itself has negligible slowdown these days (see my benchmarks in my signature).
Of course clocking a given registered ECC module is more limited than an unreg non-ECC. You have 9 instead of 8 chips that all have to play along and the register chip can be overdriven, too. But in general, it's the memory chips that count. You won't be running much slower putting the same chips into a reg ECC module than running them plain.
The ACP vs TDP thing makes sense. They do ACP in the server arena and so does Intel. I'd guess that the reason is somewhat reactive (do it because Intel does it) and somewhat marketing (yay power efficiency). Desktop chips have always been TDP and I doubt that'll change.
Has anyone spotted the new Opterons available at any decent online stores in the US?
Thought you all mite want to see this http://www.fudzilla.com/index.php?op...10482&Itemid=1
I must say, I am impressed. Our HP rep for our company had a lot of good things to say. We might actually sell some multi socket AMD based servers in 2009. That is a shocker, since our company has refused to use AMD for 8 years now. We looked at the specs and were impressed. We have a "green" innitiative and these show promise.
:D :banana::banana::banana::banana: :yepp:
this is awesome news, now Im just hoping some of this translates over into deneb results =) I want it to be a competitive chip also!
Hopin the best for AMD here...I love AMD =) Still holding onto my opty 165 waitin to upgrade to AM3
Five days after the launch, Newegg, ZZF, Directron and ChiefValue still don't have 45nm Opterons. The only places that do are stores I've never heard of, and they cost about $400. I guess I'll go 65nm this time around.
And it is consistently cheaper than the Barcelona parts :up:
http://www.newegg.com/Product/Produc...nd&Order=PRICE
http://www.newegg.com/Product/Produc...e=&srchInDesc=
No 8 series Shanhais yet but they are probably not far behind.