Sign in with
Sign up | Sign in
Your question

6 gigs but not 12 gigs ram i7

Last response: in Motherboards
Share
February 12, 2009 4:24:53 PM

cant get my i7 940 Gigabyte EX58-UD4P to run at an overlock with 6x2 gigs ocz 12800 ram. Running 3x2gigs 12800 it will do 4.1ghz plus on air and ram will do 1400-1600 depending on timing. These are with stock voltages (ram 6.4-1.66). Although the board does up them slightly.


yet stick all sticks of ram in and it will only run at around stock values and that means for the ram jdec values of 1066. wont boot with any relevant clock over this on cpu and/or ram. So its a major performance loss to go with 12 gigs. Any ideas here? BTW with all sticks of rma i did manually up all voltages to vcore 1.35..qlick 1.3 v and ram 1.66....

More about : gigs gigs ram

February 13, 2009 12:11:49 AM

I don't think you can run more than one channel (3x2) @ 1,600. Try all 12 at a lower speed. Check the website for memory compatibility, that's where you should find the one channel rule.
February 13, 2009 7:29:22 AM

the point here being you cant run with anything overcloked. The cpu will do 4.2 GHz. But i cant even run 12 gigs with the cpu at over 3 hgz and the memory down at 800 mhz!!!!
Related resources
February 13, 2009 3:10:33 PM

I would recommend hitting up OCZ Technology Forums, they have a wealth of good Core i7 overclocking info. That being said, have you upped the QPI/VTT Voltage? What multiplier are you running? If it's 20, try 21 or 19 instead, 20 can be a little finicky for some reason when overclocking. Also be sure the Uncore multiplier is at least 2x the memory multiplier or it won't boot. Disabling HyperThreading may make higher overclocks easier to achieve as well. Let us know if any of that works, good luck.
February 13, 2009 7:25:54 PM

i have it on turbo, im not actually even sure what the settings do. How do extreme and turbo differ?
February 13, 2009 9:42:47 PM

smellthefear said:
i have it on turbo, im not actually even sure what the settings do. How do extreme and turbo differ?


It should set your memory clock and memory voltage higher. On my UD5 it sets it to 1.66volts.
February 14, 2009 8:35:03 AM

ok well i set that manually anyway. Given up trying to run awith 12 gigs as this board and memory wont run with the cpu overclocked even if i bump up qpi to 1.35 and dram to 1.66. Sweet points seem to be 170x23 (using the turbo multiplier)
with ram either 8x170 7-6-6-14 1t or 6x170 6-5-6-10 1t
but given that this ram is rated to run 8-8-8 at 1600 and my 3x1 gigs of the same stuff does 1800 cl9 then im surprised 2x3 gigs wont perform better
a c 177 V Motherboard
February 14, 2009 5:20:07 PM

This likely will work, it will just take a lot of patient experimentation and tweaking; you have, essentially, a beta MOBO running a beta CPU, and you want to push it to the physical limits of capacity; if you didn't expect this, you bought the wrong stuff! As I've pointed out before, it's kinda like deliberately shooting yourself in the foot, and then complaining because you have to change the bandage twice a week...

Even on 775 MOBOs, putting in the whole possible 8 or 16 G takes fiddling, and, for some ungodly reason, often getting 4x1 to work is much more difficult that 4x2; you've got a brand new memory technology, a brand new memory controller technology, and a brand new CPU mask - the only reason to subject yourself to this is to really need to do development specifically for this new technology. Now, with 64 bit addressing, the address pin limitations on RAM are, pretty much, gone; now, the physical limits to current, clock speed, segmenting, timing, and the like, are the stumbling block. Certainly, these things will fade away in response to Moore's law, but the early experiences are always bloody! Everyone else - wait a while: one memory manufacturer has announced an x3 (three bits per cell) technology that is pretty close to accessible, as it's based on existing litho masking abilities; a second has announced a very low power x4 based on entirely new masking techniques; give 'em a year or two, and we'll be straining at terabytes!
a c 177 V Motherboard
February 14, 2009 5:36:00 PM

BTW, though I've, to the largest extent possible, sworn off attempting to help with any i7/x58 combinations, I will start the search to try to help with this - it's, well, technically challenging, and my whole purpose in helping here is to expand my own knowledge. The place I will comb (and you might benefit by doing some searches there too) is TweakTown - the MOST astute and robust GB forum in existence...
http://forums.tweaktown.com/f69
February 14, 2009 6:43:17 PM

thanks. BTW the board claimed to support 12 gigs at 1333 so i figured with ram rated at 1600 this wasn't asking too much :) 
a c 177 V Motherboard
February 14, 2009 7:50:39 PM

Well (as you've certainly learned the hard way), claimed support, and getting the little bugger to actually do it, are two very different animals! One problem that I've seen the result of over and over again is that the memory support list appears to be compiled once, at board release, has lots of stuff we've never even heard of here (though I'm sure you can pick it up at any gas station in Taiwan), and is never revisited during the life of the MOBO; leaves us to poke around the forums for bits and pieces of what's working for whom, and who's consistently b#*&^ing about what isn't working... My policy regarding builds is: let the gamers and the 'bleeding-edgers' pay the big bucks, and do the legwork debugging; wait for the market to sort out the good MOBO implementations from the junk, and the price to drop in half; and then strike! I originally settled on the 9550 as an excellent solution (well, actually the 9450, but the 95 dropped past it in price for a higher clock): 4 cores, good OC ability, and lots of cache, back when it was well over $600; said "I'll buy when it drops through $300"; did, and by that time, had enough background info to 'plop it in', and 'dial-in' the OC, knowing what memory would work well at speed... Now, they're down under $280, and falling monthly. I won't even think about i7s until: the hex cores are released; they're on a 'D' or an 'E' stepping, and both the RAM and MOBO manufacturers have gotten the kinks out of properly supporting the new three phase memory technology... Obviously, there are great gains to be had by finally having an on-chip memory controller, with better management and better core connects; but, it will take a while to reap those benefits - and, in the meantime, RAM/MOBO makers are charging you a premium to help with their DDR3 development work!
February 15, 2009 12:49:08 AM

bilbat said:
... Obviously, there are great gains to be had by finally having an on-chip memory controller, with better management and better core connects; but, it will take a while to reap those benefits - and, in the meantime, RAM/MOBO makers are charging you a premium to help with their DDR3 development work!


I was going to do exactly what you described, and build a high end Core2Quad system, since everything was "so much cheaper than the Core i7 gear." After looking carefully though, any decent motherboard with DDR3/SLI support for the C2Q was ~$300 (Besides gaming, I do quite a bit of HD video editing and content creation, so memory bandwidth was a big priority for me). This pegs the RAM as the only savings over the Core i7 platform, but if you break down what you're paying for 2 channel DDR3 kits per stick and compare to what you're paying vs 3 channel DDR3, the savings aren't that enticing. I agree that the platform isn't cheap, but if you're looking to build a high end C2Q system, a Core i7 920 based system is not going to cost much more.
a c 177 V Motherboard
February 15, 2009 5:06:20 PM

See - there's where we differ; I still won't pay for DDR3 - once you calculate the actual nanosecond 'RAMBUS cycle time', times the ungodly higher latencies, you wind up paying really a lot for really marginal throughput gains (same as going to DDR2-1200)... I admit, my MOBO won't allow me to take advantage of the vast improvements that will eventually materialize along the DDR3 avenue, but, neither was I required to pay more for a 'full board' of RAM than I shelled out for the CPU; nor, among all the problems inherent in a serious development system, am I trying to sort out unknown BIOS, RAM, and CPU problems/interactions. By the time the 'possible' advances in memory technology are realized, and become affordable, Intel will have my D0 step six core i7, the MOBOs will be $225, and I'll be willing to gut my system for a rebuild...
February 15, 2009 5:10:54 PM

and i will want to know why i cant run 1024 gigs of ram......
a c 177 V Motherboard
February 15, 2009 5:53:51 PM

Yeah LOL - you and me both - it will come though, and probably quickly... I'm an old fart: my first was a Commodore 'PET' with a 'chicklet' keyboard and a casette recorder for storage; followed by a KayPro that people oohed and ahhed about because it had (mostly hand soldered) 64K of working RAM, a 20Meg HD (they only shipped with 10s, I found a source for Seagate 225's - it had to be partitioned into three partitions, as the limit then was 8M!), and a whole meg of RAMdisk (which I fortunately can't remember what I paid for)!

I think we won't start making real progress at what passes for 'machine intelligence' until we get to these levels - I think I read that the human nervous system (I'd have to dig into my research to be exact) has either 30 or 300 trillion synapses (not neurons, but inter-neuronal connections), and until we get to within, at least, a few orders of magnitude of that, all our attempts will remain at the heuristic response level...

Ahh, reminds me of the only science fiction joke I know:

In the far future, when humanity has colonized the galaxy, every planet around every inhabitable star system is run by a master computer, intelligent and aware, that runs and optimizes everything... Finally, they discover a form of quantum communication, that is instantaneous, removing the light-speed lag. They decide to link the computers together into one massive intelligence. The head of the galactic science federation is to be both honored, and challenge with the first question to this vast intelligent network. He steps up to the control interface, throws the 'on' switch, and, after a dramatic pause, asks "Is there a God?" The machine thinks for a millisecond, and responds "There is now..."
February 15, 2009 6:02:51 PM

what if it you hit the reset button? :-)
to be honest i dont think theres a connection between the memory capacity of the brain (100 terabytes i believe) or the number of instructions it can carry out per nanosecond since computers are already able to achieve tasks at a huge magnitude greater loevel than the brain..self aware and what passes for intelligence i believe will simply be a matter of clever programming ..yes it may require hug amounts more computing power than is currently achievable but i dont think theres any comparison to the human brains capacity for an indicator of required hardware,,...
a c 177 V Motherboard
February 15, 2009 7:12:08 PM

Quote:
self aware and what passes for intelligence i believe will simply be a matter of clever programming

There is an important (and often neglected, by the AI community) difference here; 'self aware' and 'passing for intelligence' are two totally different animals; this is what I meant by "our attempts will remain at the heuristic response level". I do some of that 'clever programming'; I do machine systems - I have done machines that make several pieces per second, so when they go down, they are basically hemorrhaging money onto the ground. The machine I have in mind, particularly, is controlled from a touch-screen interface that is a bird's-eye view from above, and has eight or ten operating stations around a central, rotating part carrier. When a problem crops up, the offending station flashes in red, and pops up a window describing the faulty step; touch the red area, and you are taken to an AutoCAD drawing of that station with the possible offending sensors, again, highlighted in red; touch one of those, and you are taken to a well-illustrated Word document that tells you how to physically manipulate the machine to test the sensor, and gives you acces to a diagnostic control and analysis screen; touch the button that says you've located the problem, and you go to a screen, again illustrated from AutoCAD, that tells you how to remove, replace, and re-align the sensor - with the necessary controls to manipulate the assembly; touch 'complete', and it takes you to a control screen that tells you how to re-initialize, and put the thing back into service... It 'seems' awfully intelligent, but it's actually just heuristics and 'clever programming'!

We will make machines (probably fairly soon, say, ten or twenty years) that can adequately pass the Turing test - but they won't be aware...

Dennet, in 'Darwin's Dangerous Idea', I believe, quotes an Italian philosopher (whose name, currently, escapes me): "Yes, we have a soul. But it is made up of lots of tiny robots..." We, and our much-vaunted self-awareness, are made of millions of bundles of: pattern recognition filters, recognized and stored patterns; pattern 'super-recognizers/connectors', pattern connections, and, mainly, feedback loops, that, mostly, operate 'out-of-the-limelight', beneath what passes for our conciousness... If we were actually sitting somewhere in our forebrain, running things, so to speak, why in god's name would there be Hummers, or 'Smiling Bob', and his 'Enzyte'???

Doug Hofstadter (in a book I'm sure you'd really enjoy (searly eventy-something Pulitzer, but still infinitely relevant), called "Escher, Bach, and Göedel - the Eternal Golden Braid" - he talks about everything and nothing, from Göedel's incompleteness theorem, and it's influence on mathematical philosophy, to recursion, and its' relationship to conciousness, to pattern recognition and its' 'disassembly', to musical counterpoint and its' mathematical expression, to [on, and on and on], and illustrates, in deference to Aristotle and Lewis Carrol, with little vignette conversations between the tortoise, the crab, and Achilles) talks here about the difference between the operations of conciousness as looked at from the 'symbol' level versus the 'signal' level:
http://rapidshare.com/files/198523271/HofstadterAntFugu...

As you say - we are slow processors, but we're awsomely 'wide'; I still think it'll be a while... ;=)

Bill



February 15, 2009 8:23:47 PM

very interesting read. let me ask you this, are you agreeing that human are truly self aware because i couldnt quite grasp where you stood between quotes. of course the next step would be to ask self ware of what!
a c 177 V Motherboard
February 15, 2009 9:14:17 PM

Oh sure, we're self-aware, but what 'exactly' is self-awareness? Is a dog self-aware? A cat? Almost certainly! Is a really good chess playing machine? Certainly not - although it might kick both our asses at a game that is built around the 'intentional stance' (I intend to get his queen by sacrificing my knight...). Two stories - one from a neurophysicist, and one from my own experience:

Dawkins, I think, quotes a story related by a cognitive researcher, who has a dog that is only allowed to lie in one 'human' chair in the house - a particurlarly comfortable one in the front room. He's sitting in the chair, and the dog is whining around because he wants it... Finally, the dog 'changes tacks', and goes to the door to bark to be let out; the guy gets up to let out the dog, and the dog turns around and makes a dash for the chair! Pretty much certain that the dog told an intentional lie, which, of course, entails the intentional stance, that is basic to the concept of self-awareness...

Second is simply amazing cognitive ability; my Mom used to have a dog, that, for some unknown reason, was upset by English accents - any time someone would come on the TV sounding British, the dog would bark its' fool head off (which was an ongoing problem, as she liked to watch 'Vets in Practice', recorded in England)! Now - here's the amazing part: if she'd hear Hugh Laurie on TV as 'House' - she was fine with it; but, if she'd hear him interviewed on, say, Jay Leno (where he'd speak in his native English accent), she'd raise a ruckus! How did this work - what was going on in her head???

Take a look at this:



Before reading on, try it. You can probably figure it out even if you don't really know the old German Fraktur typeface—and even if you don't know German! Look again, closely. Did you get it? Impressive stunt! You quickly determined, didn't you, that this fragment must be part of a German translation of some lines from an Elizabethan tragedy (Julius Caesar, act III, scene ii, lines 79-80, to be exact). Once you think about it, you realize that it could hardly be anything else! The odds against this particular sequence of German letters getting strung together under any other circumstances are Vast. Why? What is the particularity that marks such a string of symbols? And how does our pattern recognition mechanism have a wide enough grasp to see it; when we have a machine intelligence that can perform this little feat by scanning a page image, then we'll be close to machine intelligence...

If you'd like to further amaze yourself with you brain's ability to 'pattern filter', first take a look here:
http://www.foundalis.com/res/diss_research.html for a general description of Bongard problems, and then go here:
http://www.foundalis.com/res/bps/bpidx.htm
to enjoy a bunch...

!