I'm about to buy the following hardware for heavy graphics and video editing:
ASUS Sabertooth X58 mobo
Core i7 980X
Thermalright Silver Arrow (if I can find one)
nVidia GTX480
WD and Samsung HDDs (7 in total)
CoolerMaster Silent Pro Gold 1200W
My problem is the choice of RAM. I need a minimum of 12GB for the video editing. This could be one of the following options:
1. Two kits of 3x2GB (same spec and ideally close serial numbers)
2. One kit of 6x2GB (not so many around)
3. One kit of 3x4GB (apparently higher CAS latencies)
From what I've read, the advantage of option 2 over option 1 would be in overclocking, since the DIMMs would supposedly be more closely matched. I will probably try some non-extreme overclocking to improve rendering times. So is it likely to be easier to overclock one bank of triple-channel memory (option 3) than two banks (option 1 or 2)?
I've also read that the performance of X58 mobos is best when only one bank of the triple-channel memory is populated. Why is this? One article indicated that the second bank controller was too weak and that voltages may need to be increased for stability. This seems very odd - why offer six slots if they can't be used without tweaking voltages? Or is this performance degradation to do with the increased quantity of RAM, rather than the fact that 3 or 6 slots are being used? I had a look at the Intel datasheets and it seems there's only one integrated memory controller, so from a clock cycle / latency point of view, what's the difference between addressing 12GB on 3 DIMMs and 12GB on 6 DIMMs? Is this to do with extra processor transactions, increased track lengths / impedence, or something else?
Various video forums indicate that lower CAS latencies are more important than high frequencies for video work, saying that there is little point spending money on memory with higher frequency than that recommended for the CPU (Intel state 1066MHz for the 980X/X58 - here). Is this correct?
I'd be grateful if someone could explain what happens to CAS latencies when two kits of 3x2GB DIMMs are used. For example, the Mushkin Ridgeback 998826 3x2GB kit (review by OC3D) is rated at 6-8-6-24, 1600MHz stock. If I used two of these kits for 12GB, would the latency settings still be correct for this amount of memory? If so, shouldn't these 6 sticks be faster than one kit of the Mushkin Ridgeback 998776 3x4GB (OC3D review), which has 9-9-9-24 timings at stock 1600MHz?
Has anyone successfully used the above Mushkin combinations on the ASUS Sabertooth X58 mobo?
Sorry for the length of this post and the number of questions, but I guess that if someone can answer them, there will be many people out there who will be as pleased as me.
Thanks, Charles.
ASUS Sabertooth X58 mobo
Core i7 980X
Thermalright Silver Arrow (if I can find one)
nVidia GTX480
WD and Samsung HDDs (7 in total)
CoolerMaster Silent Pro Gold 1200W
My problem is the choice of RAM. I need a minimum of 12GB for the video editing. This could be one of the following options:
1. Two kits of 3x2GB (same spec and ideally close serial numbers)
2. One kit of 6x2GB (not so many around)
3. One kit of 3x4GB (apparently higher CAS latencies)
From what I've read, the advantage of option 2 over option 1 would be in overclocking, since the DIMMs would supposedly be more closely matched. I will probably try some non-extreme overclocking to improve rendering times. So is it likely to be easier to overclock one bank of triple-channel memory (option 3) than two banks (option 1 or 2)?
I've also read that the performance of X58 mobos is best when only one bank of the triple-channel memory is populated. Why is this? One article indicated that the second bank controller was too weak and that voltages may need to be increased for stability. This seems very odd - why offer six slots if they can't be used without tweaking voltages? Or is this performance degradation to do with the increased quantity of RAM, rather than the fact that 3 or 6 slots are being used? I had a look at the Intel datasheets and it seems there's only one integrated memory controller, so from a clock cycle / latency point of view, what's the difference between addressing 12GB on 3 DIMMs and 12GB on 6 DIMMs? Is this to do with extra processor transactions, increased track lengths / impedence, or something else?
Various video forums indicate that lower CAS latencies are more important than high frequencies for video work, saying that there is little point spending money on memory with higher frequency than that recommended for the CPU (Intel state 1066MHz for the 980X/X58 - here). Is this correct?
I'd be grateful if someone could explain what happens to CAS latencies when two kits of 3x2GB DIMMs are used. For example, the Mushkin Ridgeback 998826 3x2GB kit (review by OC3D) is rated at 6-8-6-24, 1600MHz stock. If I used two of these kits for 12GB, would the latency settings still be correct for this amount of memory? If so, shouldn't these 6 sticks be faster than one kit of the Mushkin Ridgeback 998776 3x4GB (OC3D review), which has 9-9-9-24 timings at stock 1600MHz?
Has anyone successfully used the above Mushkin combinations on the ASUS Sabertooth X58 mobo?
Sorry for the length of this post and the number of questions, but I guess that if someone can answer them, there will be many people out there who will be as pleased as me.
Thanks, Charles.
