What's more stable 2000 mhz or 1600 mhz ram? I'm building a 3d render workstation for a friend. It's a water-cooled i7 930 poot with a Quadro card. Now I know proper workstations use mobos with two xeon cpus and registered ram. But this is what he wanted. I plan to OC the bugger a bit but don't intend to sacrifice stability for speed. So naturally I'll divide the ram to get it as close as poss to a 1:1 ratio with the fsb and lower the cas latency (I won't push that too hard as to risk too much instability) before raising the fsb (again, not going overboard). The mobo is a Asus P6T. However, I am unsure as to which ram would be best for the job though I suspect 2000 mhz is the one to go for. Any thoughts and expertise would be most welcome. Thanks.
You shouldn't need 2000mhz RAM to overclock. If I remember rightly you can still do 200mhz base clock with 1600mhz memory, which for an i7 930 is, I think, 4.2Ghz.
Hey Sam, how are you? Well, how does that go for sustained 3d rendering? And, if I was to OC a little, wouldn't it be more stable to use 2000mhz ram? Or would the lower speed 1600 have better cas outcomes with lower OCing. Perhaps I should not OC at all if sustained stability is negated by it. I would have thought some was possible. II admit to being unfamiliar with OCing an i7, I know there's more to it than OCing a Quad core and I have some info on it I have to read up. I get conflicting remarks from other sites.
No, why do you think faster RAM makes overclocking more stable? There's no scientific basis for that at all. The only reason you'd ever need faster RAM is to improve memory performance (and beyond 1600mhz the gains are so small you'd never notice the difference) and in case you're overclocking so far you run out of memory multipliers, which as I said before, gives you at least 4.2Ghz on an i7, and I think it's actually higher than that. Sustained stability can always be ensured when you overclock a CPU (within reason), the only thing you need to do is make sure the memory runs within spec, and that the voltages to the relevant components are high enough.
Thanks mate. Chatting with my nephew today ~ he did electronic design for chips ~ he states all these ram chips are made under the same process then individually tested and rated for their quality/tolerance/stability/etc. This is what determines whether they will be placed in a 2000, 1866, 1600 or 1333 MHz timed module. This sort of thing he does usually know about. For the record, he pretty much agrees with your last sentence above and concerning unbuffered (non-registered) 1600mhz ram. This is a link I found concerning OCing a i7 930, what do you think?: http://lifehacker.com/5580998/a-beginners-guide-to-overclocking-your-intel-processor
I believe that's correct for memory, it's usually the same stuff with higher/lower tolerances, which of course means it's quite likely that low-spec memory will run fine at higher speeds. I've not overclocked an i7 but I have overclocked an i5. Literally all I did was check the memory multiplier, raise the CPU voltage and then up the baseclock (which is how you actually overclock). That's all I needed to get to 4.12Ghz. Your mileage may vary, but I'd start there before tinkering with anything else, as it's likely you'll get all the speed you need without needing touch the more complex stuff.
Around 4 to 4.2 Ghz sounds fine for him I reckon. I worked out a strategy. We'll try running an over-night render. If it buggers up, then the next day I could try upping the ram voltage a 0.1v ~ rinse and repeat. The whole side of the case is a 32cm fan btw. ----------------------------------------------------
