AMD Phenom II 920 & 940 Full Review

[justapost]

@OP: Thanks for the results, nothing unexpected here.

Plenty of reviews have shown the 45nm C2Ds and C2Qs to have exceptional power usage. Like here where the QX9650 uses less power than a E6750:

http://www.techreport.com/articles.x/14606/12

Or Anandtech's examination of the QX9650 which suggests it probably doesn't hit its TDP till around 3.8GHz

http://www.anandtech.com/cpuchipsets...spx?i=3184&p=2

It's pretty evident that the Intel's mainstream dual-cores use 20-30W under full load while even the QX9650 only consumes about 65W.

What did anandtech use to generate load? techreport used cb10. Use CoreDamage or Core2MaxPerf to get close to the worst case, should be at least 20W AC more.

Still doesn't change the fact the chip sucks 1.35v at default clocks, performs on par with a Q6600.

Can you please run one of the above mentioned apps and post consumption in idel and load, so we can compare once PII is available for all of us?.

Please also mention at what speed PhenomII NB you will be running at in that review 3.1Ghz clock-to-clock review.

Also 3.1Ghz seems to be too low of a overclock , probably 3.6 ~ 3.8 Ghz with PhenomII NB@2.7Ghz would be a nice comparison to do.Also mention the idle temps/load temps and Power consumption at those clock speeds.

I'd vote for 3.0/3.6GHz and 3.8-4GHz. 3GHz at stock all others at lowest possible voltages. I assume no chip will have problems with 3.6GHz at reasonable voltages.
For power consumption it would be nice to have two alternatives generation load. One can be CB10 or povray to simulate an average load and the other should be an worst case tool like coredamage or core2maxperf.

you are right.

Lets say that its not AMD optimized(which i dont know), how many "real" x64 progs do you use on your every day use?

the phenom that i got in my hands, tops out @ 3.8 with ~1.5v.(stable)

@4g im not able even to boot windows with any voltage.

I wonder if the scaling on PII's diminished because the single core runsnow makes better use of the bigger L3 cache but it's not enough to improve the multi core test as well?