Phenom 9500 w/ MSI K9A2 Platinum

[KTE]

Here's some more testing.

Well, look what happened today out of the blue. Set 240HT ref. in BIOS, booted up fine.
CPU stock, HT stock and NB +0.1V => Prime95 1hr plus stable+temps @ 240HT x 11 = 4 cores 2640MHz, NB at 3120MHz and HT at 2160MHz:


CPU: 47C
Ambient: ~20C

Could've done hours Prime easy but never had time. Ignore EVEREST frequencies. They are obviously higher than 2420MHz I ran in the tests earlier as you will see further down so EVEREST is very buggy if I change my NB/HT speed in BIOS. All settings were only BIOS changed.

More benchmarks at that setup. Note: no tweaks, I had full services, AV and background processes running as on a usual system you would have. Also RAM is higher latency this time and still single channel.

BENCH SETUP


3DMark06: http://img233.imageshack.us/img233/953/dm06vq7.png
Cinebench 10: http://img233.imageshack.us/img233/704/cbqq1.png
EVEREST read: http://img233.imageshack.us/img233/37/evnf4.png
EVEREST write: http://img231.imageshack.us/img231/4567/evwoz9.png
EVEREST copy: http://img231.imageshack.us/img231/5828/evcce9.png
EVEREST latency: http://img231.imageshack.us/img231/9015/evlon2.png
CPU Queen: http://img231.imageshack.us/img231/3564/evqueenfv1.png
CPU PhotoWorxx: http://img228.imageshack.us/img228/9849/ev1mp0.png ( lower than 2420MHz )
CPU ZLib: http://img228.imageshack.us/img228/5886/ev2bi8.png
CPU AES: http://img228.imageshack.us/img228/5944/ev3fo6.png
FPU Julia: http://img228.imageshack.us/img228/8749/ev4tl7.png
FPU Mandel: http://img230.imageshack.us/img230/1108/ev5qc0.png
FPU SinJulia: http://img230.imageshack.us/img230/3424/ev6ak4.png
Fritz: http://img230.imageshack.us/img230/8113/fritzgn4.png
ID2: http://img215.imageshack.us/img215/5813/idcgt2.png
Lightsmark 2007: http://img215.imageshack.us/img215/5634/lmrt4.png
Nuclearus: http://img215.imageshack.us/img215/6342/numb4.png (look at the frequency it picks up! Soldner you could've at least fixed that )
POV-Ray 3.7-SSE2: http://img215.imageshack.us/img215/6220/ppsta7.png
RealStorm2006: http://img215.imageshack.us/img215/8/rsag1.png
Sandra Arithmetic: http://img215.imageshack.us/img215/4859/safu9.png
Sandra Multimedia: http://img222.imageshack.us/img222/5180/smtm2.png
Sandra Cache/Memory: http://img222.imageshack.us/img222/4320/scfo6.png
SPi 1M: http://img222.imageshack.us/img222/3200/spidd9.png
Vecmat: http://img236.imageshack.us/img236/3912/vecmatut1.png

I'll make a table of the scores later for easier comparison but I won't put scaling down because you can't work out CPU frequency scaling with so many variables. But it does scale very well overall I would say.

Then I did 245 x 11 = 2695MHz and threw ALL voltages to MAX which is around 1.55V each.

Here's the maxmimum power draw of the system during Prime95 (which is more than in anything else):



330W AC = ~273W DC

Yup, it shoots up very high, near my PSU limit.

Also, the maximum power consumption pulled in a ms was 331W AC and no bench got close to it apart from EVEREST FPU Julia. It pulls the highest power.

HERE-> what effect does Core speed/NB speed have on L1/L2/L3 cache latency? Compare:

240HT x 11 = 4 cores 2640MHz, NB=3120MHz, HT=2160MHz, RAM=480 5-5-5-5 2T:


220HT x 11 = 4 cores 2420MHz, NB=1980MHz, HT=1100MHz, RAM=440 4-4-4-4 2T:


Very noticeable difference.