From what I understand, to use faster memory I will need to overclock the uncore clock speed of the processor to two times the memory speed I want. Is this completely separate from the base clock speed (can I still run my 920 at 2.66ghz and just overclock the uncore speed to 3.2ghz to use 1600 ram), or will I want to overclock the base clock as well?
I don't have an i7, so I can't help you. What I can do though is share this how-to-overclock-the-i7-compilation I posted a while back: Intel Core i7 Overclocking How-to
It's your computer, and you're perfectly entitled to use/configure it as you please. But IMHO, you're wasting time and effort in overclocking memory - and as a bonus plan, leaving your system open to instability - for little or no return in real world performance.
In My Humble Opinion, you'd be best off by far leaving your memory speeds at/near stock. If you want to Overclock then feel free to work on the Processor itself, where you'd see noticable gains.
------------------------------Which Chip? Well, it depends on which set of thieving b@stardz you choose to support: The ones who use insider trading to enrich themselves while running their company into the ground, or the ones who illegally pay vendors to not support the first group.
Reply to Scotteq
The memory on your DRAM chips isn't even accessed for the vast majority of memory operations because of the very large caches in the CPU. Because of this, the CPU is pretty insensitive to changes in DRAM speeds.
For example, if 90% of the memory operations are handled by cache, that leaves only 10% which involve DRAM. That means that only 1 in 10 memory accesses is affected by DRAM speeds. If you increase DRAM speed by 10%, you could expect a net performance increase of only around 1%.
There are some workloads which have poorer cache hit rates than average, and they'd benefit more from faster memory. But in general, tweaking memory is not all that productive in terms of improving overall system performance.
The memory on your DRAM chips isn't even accessed for the vast majority of memory operations because of the very large caches in the CPU. Because of this, the CPU is pretty insensitive to changes in DRAM speeds.
For example, if 90% of the memory operations are handled by cache, that leaves only 10% which involve DRAM. That means that only 1 in 10 memory accesses is affected by DRAM speeds. If you increase DRAM speed by 10%, you could expect a net performance increase of only around 1%.
There are some workloads which have poorer cache hit rates than average, and they'd benefit more from faster memory. But in general, tweaking memory is not all that productive in terms of improving overall system performance.
This is for my CAD workstation. Could CAD present those workloads that I would benefit having more ram?
This is for my CAD workstation. Could CAD present those workloads that I would benefit having more ram?
Benefit from 'More' memory? Surely. The larger and more complex the objects being worked, the greater benefit from having more resource available. "Faster" memory? As relayed above - any gains would be minimal at best.
You didn't indicate earlier this was for a workstation. Most of the posters here are enthusiasts/hobbyists, rather than having a professional need, so that's generally my mindset when I answer something. Given that, I withdraw my recommendation for overclocking the Processor. Next, I'd like to direct you to the writer of your CAD software, and strongly urge you speak to one of their technical reps regarding the ideal configuration.
Message edited by Scotteq on 07-24-2009 at 09:57:40 PM
------------------------------Which Chip? Well, it depends on which set of thieving b@stardz you choose to support: The ones who use insider trading to enrich themselves while running their company into the ground, or the ones who illegally pay vendors to not support the first group.
Reply to Scotteq