Crucial Ballistix DDR3-1600 + Asus P5K3 Deluxe (anandtech style)

[eva2000]

So far, with the short amount of time I have had to play with DDR3, I have found that it is hard to find a set of comparable cpu multiplier/fsb/memory speed configurations to match and compare with DDR2 benchmarks. It's even harder to find comparable DDR3 results at similar cpu/multiplier/fsb/memory speeds and timings as everyone is running at different DDR3 platform settings.

So decided to take a look at Anandtech.com's recent DDR3 testing and do some DDR3 benchmarks of my own but at similar settings as Anandtech.com's DDR3 review so as to have something to compare against.

Anandtech.com's Super Pi 2M, Sandra Buffered/Unbuffered benchmarks is what I will be concentrating on. But I'll throw in 32M Super Pi and Everest Ultimate bandwidth tests as well.

http://www.anandtech.com/memory/showdoc.aspx?i=3121&p=4
http://www.anandtech.com/memory/showdoc.aspx?i=3121&p=5

• 9x333FSB 5:6 400Mhz 5-3-3-5 2T at 1.7v
• 9x333FSB 5:8 533Mhz 5-4-3-9 2T at 1.7v
• 9x333FSB 1:2 666Mhz 6-5-5-12 2T at 1.8v
• 6x500FSB 5:8 800Mhz 7-6-5-15 2T at 1.8v

Unfortunately, I can't hit DDR3-2000Mhz with either Crucial Ballistix PC3-12800 or OCZ PC3-14400 Platinum memory, it could be a memory limit or more likely a Asus P5K3 Deluxe P35 chipset limit for this particular motherboard ? On Asus Blitz Extreme has gone as high as DDR3-1900 8-7-6-20 1T at 2.08v with 1.73v vMCH. But can't hit those clocks on Asus P5K3 Deluxe.

Tabulated results

[eva2000]

Screenshots

9x333FSB 5:6 400Mhz 5-3-3-5 2T at 1.7v

Super Pi 2M & 32M & Everest


Sandra Buffered


Sandra Unbuffered

[eva2000]

9x333FSB 5:8 533Mhz 5-4-3-9 2T at 1.7v

Super Pi 2M & 32M & Everest


Sandra Buffered


Sandra Unbuffered

[eva2000]

9x333FSB 1:2 666Mhz 6-5-5-12 2T at 1.8v

Super Pi 2M & 32M & Everest


Sandra Buffered


Sandra Unbuffered

[eva2000]

6x500FSB 5:8 800Mhz 7-6-5-15 2T at 1.8v

Super Pi 2M & 32M & Everest


Sandra Buffered


Sandra Unbuffered

[jugeen]

nice job! congrats!

[eva2000]

thanks... seems this Asus P5K3 Deluxe isn't clocking my DDR3 memory as high as on Asus Blitz Extreme - ~20mhz less max memory clock on P5K3 Deluxe compared to Blitz Extreme. But P5K3 Deluxe has higher FSB max of 525-530FSB versus Blitz Extreme 485-495FSB max.

[JNav89GT]

Hi Eva
I've noticed my P5K3's are very finicky! I have a P5E3 coming tomorrow
Write to Precharge in bios set to 10-11 and
Write to read set to 9 seems best for me with some "fallout" dimms I've been testing. By fallout I mean they didn't meet bin spec for proposed line this maker is putting out. I've been able to get dual channel stable at 1840mhz orthos for over 1hr, and dual super pi etc.... This is at 7-7-6-20 1T 2.15v, and 1.7MCH.
I did note that I was stuck at about 1700mhz until I set MCH to 1.7 and loosened up the Write timings as stated above. I do believe if I had the balls to volt mod the MCH and put in about 1.75v I'd be able to go a bit higher.

Nice timing at 1600mhz btw

Also, I did cool the NB and dimms at this level as they were getting a bit warm. Don't think it made a diff in max clocks, but made me feel better .

[eva2000]

Thanks

Yeah these DDR3 boards are a bit finicky heh. I know what you mean about lack of balls (and lack of soldering skills) to do volt mods hehe.

Hopefully Asus sends me some X38 boards soon

[3oh6]

nice, a discussion on the P5K3 and memory clocking related to MCH voltage, this thread is going to get run over by this...i can just tell .

i can weigh in from the side of the fence with the 'balls' AKA soldering skills to volt mod the MCH. i haven't gone over 2.01v with my Ballistix yet so my higher clocks are at 8-7-6 but i basically hit a wall at 940MHz with the MCH maxed in the BIOS for dual 32M SPi and 934MHz prime/HCI memtest stable before the vMCH mod.

since the MCH mod, aside from being able to bench at 510+FSB PL6 @ 1.95v vMCH, i have also raised that 8-7-6 dual 32M up to 950MHz vMCH at 1.95v but i probably doesn't need to be that high...i really need to turn that down when i start clocking this memory again.

as well, for benching, there was no way i could get near 1000MHz 32M stable before the vMCH mod and as you can see from the 32M low clock challenge thread (http://www.xtremesystems.org/forums/...postcount=1214) 1010MHz 8-7-6 PL6 was a piece of cake for single 32M. i basically ran that setup for 3-4 hours straight doing 32M runs messing with a couple different OS's. the highest i ran was 1014MHz 8-7-6 PL6 but i am sure i could go higher.

the NB cooling was just a Noctua NC-U6 with a 120mm fan blowing on it and the memory. so hopefully that adds some insight to added vMCH on the P5K3 but i think the P5K3 is still luck of the draw. just look at BAs results...http://www.xtremesystems.org/forums/...=155626&page=2 952MHz dual32M PL6 at a decent FSB an no mention of any kind of super cooling or voltage on the NB, assuming stock cooling and max volts in BIOS.