P45 Memory Bandwidth & Speed testing

[Tekxpert]

By CryptiK

Yes, it's 1.93v for 556 MHz CL5 stable with DDR2. It's not bad but I hardly think it's a record of any kind, I've yet to achieve anything even remotely noteworthy. The ram is nice, it needs 1.98v for 600 MHz CL5 totally stable, and is SPI32M stable at 640 MHz CL5 at 2.24v on my M2F and 650 MHz CL5 at 2.24v on my P5Q-Deluxe. It appears to be about as good as some of eva's Ballistix kits, but as I don't have the same boards as him, and am not even in the same ballpark skill wise, I can't directly compare them.

Wow, those sticks crap on my Patriot PC-9600 DDR2. They need 2.30v @ 600Mhz 5-5-5-13 to be totally stable.

[GenAn]

The Cellshock 8000 C4 are amazing, it's too bad MSC is ceasing the Cellshock memory brand.

[eva2000]

By CryptiK

Yes, it's 1.93v for 556 MHz CL5 stable with DDR2. It's not bad but I hardly think it's a record of any kind, I've yet to achieve anything even remotely noteworthy. The ram is nice, it needs 1.98v for 600 MHz CL5 totally stable, and is SPI32M stable at 640 MHz CL5 at 2.24v on my M2F and 650 MHz CL5 at 2.24v on my P5Q-Deluxe. It appears to be about as good as some of eva's Ballistix kits, but as I don't have the same boards as him, and am not even in the same ballpark skill wise, I can't directly compare them.

could be the motherboard holding you back - Asus is known for tightening latencies to get better clock for clock performance at the expense of higher mem clocks

By Tekxpert

Wow, those sticks crap on my Patriot PC-9600 DDR2. They need 2.30v @ 600Mhz 5-5-5-13 to be totally stable.

sounds like you got the older D9s which needed more vdim. Then or you're using Gigabyte board and using a divider which needs more vdimm - EP45-Extreme needed more vdimm (~0.1-0.15+) clock for clock than other boards I have to reach similar clocks/performance.

By GenAn

The Cellshock 8000 C4 are amazing, it's too bad MSC is ceasing the Cellshock memory brand.

woah that's news - such a shame

[CryptiK]

I'm pretty sure it's the boards holding me back, I'd be very surprised if my ram could do 650 MHz @ 2.24v SPI32M stable and not be able to go any higher remaining stable regardless of the vdimm. When going above 650 MHz things get unstable fast as I increase the speed, and It's like I just run into a wall at 666 MHz.

[radaja]

about the cellshocks pc-8000cl4's it is a shame,im just grateful that cryptik recommended these to me before memoryc sold out,cause they are an awesome kit. and i was lucky enough to get a second set used that were only used for 3 weeks.same manufacturing date as the first set.

[GenAn]

I wish I hadn't bought the CAS5 kit initially, as it can't really compare to the CAS4 kit I got later. Actully I take that back, one of the C5 sticks is in the same league, but the other one in the pair is crap. What I really wanted was the DDR2-1150 C5 sticks

[CryptiK]

The reasoning behind my choice of the 8000 CL4 sticks was that I had always heard GKX was higher speed binned than GMH (667 vs 800 stock from micron) and GKX usually did better at low latency (ie CL3/CL4) clocking than GMH. I already had some GMH kits, so I wanted GKX that was binned for low latency, rather than GKX that was binned for CL5 performance like GMH seems to excel at.

I know there can be very little difference between GKX and GMH but after you play with both you can definitely tell that they have different characteristics and behave differently. Some also behave similarly just depends on the kits you have.

I essentially wanted some classic GKX to complement my GMH. Glad I picked them up when I did too.

[CryptiK]

I finally got around to some SPI32M runs with large system cache enabled, and it made a significant difference to the times - thanks for the tip eva However it seemed to stress the ram and NB a little harder than with it disabled, I had to slightly increase vdimm, but the vNB was sufficient as is, due to it being set higher than necessary anyway.

SPI32M @ 500 FSB, 600 MHz, PL8



Everest @ 500 FSB, 600 MHz, PL8 (same settings as SPI32M run)



SPI32M @ 534 FSB, 640 MHz, PL8



Everest @ 534 FSB, 640 MHz, PL8 (same settings as SPI32M run)