Comparing OC-rated memory for Mobo versus purchasing faster RAM.
Comparing OC-rated memory for Mobo versus purchasing faster RAM.
I have a GD45Plus motherboard with 32GB of 1600Mhz RAM at its standard configuration, paired with an i7 4820K processor. I'm considering two options: either tweak the existing RAM and attempt to overclock it... or install new 2400Mhz RAM that's ready to use. Alternatively, does the overclocking requirement mean I need to adjust the motherboard settings and boost the CPU as well? I've read Tom's guide on RAM overclocking, but honestly, I prefer saving time—so I'd go for swapping in new RAM if it's feasible. Right now I only need about 16GB, and I'm not doing heavy database work on the system. (Although since it's a quad channel setup and I don't see many 4GB SIMMs, I might end up with more than 32GB again).
The GD45 Plus supports DDR3 speeds up to 2133Mhz and 2400Mhz (with overclocking). My main question is whether it's possible to simply install 2400Mhz RAM with minimal effort. I think the answer is likely no!
Background:
I have a build designed for general database tasks—Twin 250GB SSDs for Windows and system applications, plus 2x3TB 7200RPM drives for user databases and media. I also had some gaming setups originally with an R9 280X, which has now been upgraded to a VR rig featuring an MSI GTX1080 (currently overclocked to 1800MHz) and an HTC Vive. I upgraded from socket 2011 for more flexibility as my needs evolved—if the database performance was solid, I'd swap to an Xeon; otherwise, it would be a gaming rig. I'm now launching a Mobile VR business... databases are just background work ;-) So, I'm leaning toward gaming.
From forums, it seems my memory might be limiting performance when playing Project Cars. Other VR games run fine, but since PCars is a full experience rather than a small demo, I can only maintain around 45fps with frame reprojection instead of the full 90fps the HTC Vive requires (which causes double images if you look out the side and jitter when moving). I've observed the GPU running below 50% in MSI after burn-in, though it's hovering around 75% max. CPU was at 25-30%, but I disabled hyperthreading and saw a significant boost to 60%. Previously, I couldn't use two GPUs and had to keep all settings low. Now I can use two GPUs with a 1.3x Supersample setting in SteamVR, achieving medium shadow detail and car detail. It's functional but not ideal for smooth 90fps or high resolution. Forum discussions on hyperthreading are mostly about disabling it—trying it out might give better results.
Interestingly, the GTX1080 can push Pcars to 80fps with ultra settings on a HD monitor, adding significant overhead for stereo images and a 640x480 spectator display.
"(*OC)" Translates to "Put the RAM in and enable Intel X.M.P.". At high speeds like 2400MHz (with DDR3 at least) they can't guarantee 100% stability at stock settings however. It's basically their way of saying that roughly 99% of the time you can just slot it in and run fine, but there's that 1% where you may have to do stuff like bump up the voltage to the CPU's memory controller or boost the DRAM voltage (or adjust timings if you want to avoid touching voltages).
"(*OC)" Translates to "Put the RAM in and enable Intel X.M.P.". At high speeds like 2400MHz (with DDR3 at least) they can't guarantee 100% stability at stock settings however. It's basically their way of saying that roughly 99% of the time you can just slot it in and run fine, but there's that 1% where you may have to do stuff like bump up the voltage to the CPU's memory controller or boost the DRAM voltage (or adjust timings if you want to avoid touching voltages).