Crucial Ballistix PC3-12800 2x1GB @ O2

[3oh6]

very hot off the presses, as in, posted not more than a few minutes ago; is my official review of a very nice Crucial Ballistix DDR3-1600 2x1GB kit of memory. a handful of benchmarks and plenty of 24/7 stable overclocking results can be found in the review here:

Crucial Ballistix PC3-12800 2x1GB @ O2

but a few of the 'not-so-official', AKA fun, overclocks and benchmarks will go up here in this thread. feel free to ask any questions here as well, i won't mind talking about this kit

to get things started, here is a teaser of the 'fun' benches i was just referring to:

click for full size including Memset...

[sswilson]

Holy Crap!!!!!!!!!

DDR3 2000!?!?!?!?

Thats what folks are waiting for before they jump into a new memory format!!!! :)

Double holy Crap!!!!!!!

(1T)?????

I haven't had a chance to run through the original review (pretty busy working on a review of my own... :) ), but did you get a chance to compare bandwidth between (1T), and (2T)?

It used to make a huge difference on the S939 platform, but doesn't seem to be supported on any Intel chipset motherboards so I'm not sure how big of a difference it makes for C2D systems.

[enaberif]

holy crap 3.5ghz on a e6600 stable? I can only achieve 3.4 :(

[SKYMTL]

Not bad at all. Comparable to what I am achieving with my Dominator DDR3 modules but the 1T pretty much threw me for a loop.

On the other hand...maybe someone can explain to me what the point is of measuring ANY voltage on a motherboard with a DMM. I'd love to hear an answer because I can think of at least 3 reasons why it would give a very inaccurate reading.

[3oh6]

Originally Posted by sswilson

(1T)?????

I haven't had a chance to run through the original review (pretty busy working on a review of my own... :) ), but did you get a chance to compare bandwidth between (1T), and (2T)?

no comparison in the review but ask and ye shall receive:

rightmark 8-7-6-20 1T DDR3-2000 / min=6555.95 max=18093.51 avg=8936.81

rightmark 8-7-6-20 2T DDR3-2000 / min=6459.37 max=18052.63 avg=8818.52

Everest 4.10 - 3 consecutive runs each

Timing   Read        Write       Copy        Latency
1T       11639       9123        9877        59.2
1T       11605       9124        9858        59.2
1T       11611       9120        9843        59.3
====================================================
2T       11577       9125        9837        59.8
2T       11555       9094        9790        59.8
2T       11541       9118        9791        59.8

ScienceMark 2.0 - Memory Bench

Timing    Bandwidth
1T        9301.62
1T        9324.22
1T        9300.87
===================
2T        9187.33
2T        9135.21
2T        9139.58

SuperPi 32M - 1T= 13m 29.844 / 2T= 13m 33.781 (all secondaries exactly the same and method for running was identical from a clean reboot)

click for full size including memset...


As you can see, there isn't a significant difference between the two but all the bandwidth measurements and SPi indicate an advantage to 1T. so there is definitely a change between 1T and 2T but is itsn't as large as say on AMD with the integrated memory controller.

Originally Posted by enaberif

holy crap 3.5ghz on a e6600 stable? I can only achieve 3.4 :(

sorry for your misfortune...i'm 3.7GHz Orthos Blend priority 9 stable for hours on this badboy with a Thermalright Ultra-120, but more importantly, there is no wall until 560FSB on air.

Originally Posted by SKYMTL

On the other hand...maybe someone can explain to me what the point is of measuring ANY voltage on a motherboard with a DMM. I'd love to hear an answer because I can think of at least 3 reasons why it would give a very inaccurate reading.

please do explain, measuring voltages directly from the motherboard is more than an accepted practice in overclocking circles and has been for more years than i know by electrical engineers in the OC scene.

how else are you suppose to measure the voltage going to the component you are adjusting with the variable resistor? and why is it so in-accurate?

i admit, your knowledge from an electrical point of view is greater than mine but my knowledge of hardware and how it relates to overclocking is much better and i have been reading from motherboards since the day i started with remarkable consistency.

interested to hear your thoughts on this subject Michael.

[SKYMTL]

please do explain, measuring voltages directly from the motherboard is more than an accepted practice in overclocking circles and has been for more years than i know by electrical engineers in the OC scene.

how else are you suppose to measure the voltage going to the component you are adjusting with the variable resistor? and why is it so in-accurate?

Unfortunately, whenther we like it or not, not all multimeters are made alike. The higher end ones can be calibrated to give spot-on accurate readings but that is not what most overclockers use. I'll give you a quick example:

The Fluke 187 I have is calibrated every 2 months considering its relatively light usage.

For the heck of it the other day I borrowed an uncalibrated Fluke 187 and hooked it up to the same PSU connector as my 187. Lo and behold, my Fluke (freshly calibrated) showed 12.15V on the +12V rail while my buddy's showed 12.23V. While it may not seem like much, a difference of 0.08V is ALOT. Remember, these are $500CAD multimeters, not the $40 Canadian Tire-like stuff many OCers use and even the uncalibrated one was wrong.

While I am not saying that your readings are WRONG, I am saying that they could be off by quite a bit depending on the state of your multimeter. Knowing you, I would say that your readings are pretty accurate because I know the care you put into your testing and equipment. On the other hand, I have always had an issue with overclockers saying "measured by DMM" like it was God's Own truth when they have no clue how to properly maintain a good multimeter.

So, there is really no way of knowing with even 50% certainty that the readings are correct unless the overclocker has calibrated their multimeter on a regular basis.

Hope that all makes sense...sometimes I tend to rant....

[3oh6]

okay, i thought you might have been referring to something else in how taking a reading from a motherboard was going to be influenced, where readings were taken from, or how the voltage is actually outputted; not just the multi-meter itself.

it goes without saying that multi-meters can be off but i would much rather hear someone indicate that they measured with a DMM (calibrated or not) than someone say it was X voltage because that is what is selected in the BIOS. in the end, both of our opinions have fallacies in them but that is why it is just an opinion.

***edit***
oh yeah, i simply note that the voltage i used as reference is measured by DMM because the AI Suite only reports what is selected in the BIOS which had been accepted as incorrect. with a calibrated Fluke i 'borrowed' from work, this board tested .05v~.06v higher than selected through the entire voltage range. if i went by the BIOS selected voltage, then these numbers are that much better :)

[SKYMTL]

Yeah, this had nothing to do with your measurement but rather people holding DMM measurements to a higher standard than they are warranted. If the measurement is done with a properly calibrated DMM then the numbers are infallable but if they are taken with an el-cheapo DMM, they may as well be taken from the BIOS...

It is good that you had a reference point of a calibrated DMM so you can base the difference between it and your own DMM and compensate accordingly. Many overclockers do not have this luxury.

[3oh6]

yup, and without having these things calibrated daily laying around at work, i wouldn't have a reference for my crappy DMM. the important thing in any kind of numbers whether it be for CPU or memory is how the memory scales with the voltage increments...take the actual voltages with a grain of salt, like CPU temperatures.

and thanks to SS, i have updated the graph in the review to reflect that the overclocked results are all at 1T command rate. it completely slipped my mind that i stuck with 1T. 2T does provide a slightly higher clock at a given voltage but it also produced some wierd stability issues with certain combinations of memory ratios/strap/FSB. leaving the command rate to 1T provided a more consistent platform to work with. perhaps i will spend some more time clocking at 2T to see if i can narrow down the discrepencies and why it happened. either way, the differences are very very small between 1T and 2T.

[3oh6]

alright, i had a goal and a few hours late last night to work on my best sub 13M time. this is stemmed from the XS 32M Low Clock Challenge. if your into 32M SPi benching...this is THE thread for you to start hanging out in. the setup is the same as in the review with the addition of avMCH volt mod, here is what i came up with...DDR3 power at its finest.

3535.1MHz SPi 32M DDR3-2020 8-7-6 = 12m 59.953s
click for full size including memset...