Can you help this newbie answer some basic overclocking questions?

Hi, detailed thread here so please stick with it, grab a cuppa morning tea or coffee first and please come back :)

So I recently got into overclocking and I've got my older core i7-920 stable (D0 chip) @ 4.1GHz for about a week now. I used to be anti-overclocking but now that my system is pushing 6 years old now (x58 chipset, asus p6t deluxe v2 mobo), I figured now's the best time to (mostly) overclock the hell out of it and see how much I can get out of it for another couple years or so without killing it. I've got some newer parts as well like an r9-280 gpu and a 128GB Samsung 840 pro (which unfortunately suffers from a sata 2 chipset but I bought it for the 5 year warranty.) So now I have some basic questions to see if my system is fully optimized.

Note: I prefer leaving load-line calibration disabled thus enabling vdroop. From what I've read, this puts significantly less stress on the voltage regulators on the mobo and it really isn't a problem on newer boards anyway as the vdroop isn't as dramatic as it was on pre-x58 boards.

1. Now that my system is stable and everything is appearing to work at minimal voltages, is it worth re-enabling some of the power-saving features like EIST, C1E, and C-States? I don't have dynamic vcore but I *would* like to re-enable some things because summer is coming and dialing back the overclock when it's not on load seems like a good idea. It's (very) hard to glean accurate information from all the forum posts I've read, but it seems that all 3 of these features will dial-back clock speed (and voltage) when not on load (and "load" is a very generous term here since basically doing anything will up-dial it again.) However, drive latency will be increased when these are on which I'm not too happy about. I only have a corsair h60 (single 120mm rad) in push-pull and my hottest core temps (not cpu package temp) will push 82-84c hours into 100% load -- so yeah that will likely only get worse in a hotter ambient room temperature. My new gpu is running like a champ at max OC (120% power limit) at only 63c for hours, no artifacting whatsoever. Curiously, I've noticed it will auto dial-back to stock when browsing some video sites like youtube/vessel but will continue at max OC when the browser tab is closed. Benchmarks are good, here's 1 example:

2. Is it ever worth touching the ICH voltage from auto? DRAM bus over 1.66v (with dV VTT <= 0.5)? The idea being to give my poor ssd a small boost from bottle-necked sata 2 speeds? The benchmarks are terrible. 7zip compression will smother it all the way down to 10MB/s. My other plan is to enable Samsung's "rapidmode" which uses a pseudo-ram drive like environment -- but I only have 6GB of ram. I have a decent UPS so I'm not too worried about sudden power loss to the system as I can get it safely shut down in time. Ram is currently running at 7.2GB/s via memtest86+.

3. I could push it up the bclk a tick higher, but my vcore is for sure going to exceed 1.35 and even at the lower voltages I've tried, my core temps were going over 85c in a short time. Trying out a new bclk would probably take another another day of stress-testing the ram and then even further fiddling the VTT (IMC/uncore voltage) until it's finally stable. Not really something I'm interested in if it's probably going to fail since that jump in tick requires a lot more vcore than what I'm currently running at.

TL;DR: please go back and read it. Being thorough seems to be a rare thing these days.:sarcastic:
-Thank you.
2 answers Last reply Best Answer
More about newbie answer basic overclocking questions
  1. Best answer
    I thought I would give an update. Disabling HW acceleration in Chrome seems to have preserved the gpu overclock. I'm guessing that Google doesn't want to be responsible for crashing your system so it will secretly down-clock the gpu when videos are loaded. Clocks will stay at max even on 60fps videos on 1440p. The video quality does appear to suffer a little as it looks like anti-aliasing is disabled, but the videos are still playable and not otherwise choppy. I didn't test flash though and not really interested in doing so because Adobe is constantly being suffocated in security vulnerabilities. For html5 viewing, I'm assuming it reverts to OpenGL html5 decoding and it does take about 15% cpu utilization (on my quad-core, 8 thread cpu) -- so keep that in mind if you're also running a full cpu load doing other might start getting choppy videos if you're not allowing some slack for OpenGL rendering.

    Also, I've since gotten a couple bsod's during my testing of re-enabling the power features. EIST, C1E, C-States, pci-e link state power management "moderate power savings," and USB selective suspend were re-enabled. I also enabled "rapidmode" which bumped up the ssd benchmarks (255MB/s seq. write -- much better than before but still slow compared to a traditional "ram drive.") I was playing L4D in steam for 35 minutes (video settings max quality on 1200p, gpu still on OC) while listening to a ustream video in the background (HW acceleration on at the time) and then got a code 101 bsod.

    I then disabled C-State which turned off much of the "yo-yo"ing of the vcore (significant fluctuation in the vcore would happen previously even if no apps (other than cpu-z and maybe a couple win7 gadgets) were loaded.) Also turned pci-e link state power management to off. Second bsod (code 9C) happened later when I was typing up a forum post with a bunch of tabs opened in Chrome (hw acceleration still on), gpu was on near-100% load.

    I then ticked up the VTT by 1 notch. In the meantime I'm treating the system as unstable and saving checkpoints every 10 minutes. Testing for stability in this environment is very difficult because you're basically testing the system going from 0% for an extended 100% load intermittently and see how it handles these fluctuations. Someone in another forum suggested leaving the system on overnight with no apps loaded (including gadgets, CCC, Intel storage console, and av scanners), disabling the network adapter, and see if it's still on in the morning lol. :) Theoretically there's going to be some amount of cpu utilization during idle periods due to Window's scheduled tasks and cleanup (aka archive) items it does in the background.

    I'll update if anything changes. You need to have an exceptional level of patience to get stability it seems -- almost "Neo"-like patience. Here's a bonus funny video video if you're read this far, did you know you can override your case power button into a "jet plane" over-clocking button HAHAHAHAHAHA: (Warning: extremely broken "Asian English" spoken dialogue.)
  2. Well I'm pretty dumbfounded after a latest "0xBE" bsod. The only significant app I had open was f@h active and it rebooted in the middle of the night. The driver involved was a dx11 file. The gpu has the latest mantle driver and the dxdiagnostic tool says nothing. Could be the overclock on the gpu but that's the first time it would have crashed despite hours of otherwise gpu benchmarking. I ran 4 passes of memtest86+ this morning and nothing turned up. Last night my usb wifi adapter was acting funny (randomly restarting) so who knows if usb ports had anything to do with anything.

    New items changed: disabled Samsung's rapidmode (if the problem is memory, it could mean too much ram utilization and not enough page file available.) Modified the page file to more consistent 2000-2000MB size. Turned off USB selective suspend and switched around all my usb ports and now have my usb wifi adapter plugged into my keyboard's usb hub. I also set speedfan to enable logging temps and s.m.a.r.t. summary on startup. Time to install OpenOffice again to view/chart .csv log files lel. Making graphs with a butt-load of data is fun!

    My core temps are getting toasty and pushing 85c again, I have a couple Corsair SP120 high performance fans coming in the mail so we'll see if temps go down on my current cpu OC. Considering how little points you get running f@h on the cpu (vs. the gpu), it seems like it's not even worth the heat.

    I hate how bsod codes can mean sooooooooooooooooo many different things. Is it really so hard for Microsoft come up with a scheme that likely points to 1 very specific thing and rules out many different things instead of leaving that to you?

    P.S. Is anyone out there or bothering to read these posts? I have an odd question for anyone out there. Since we're disabling spread-spectrum on the cpu and pci-e bus, is it possible EM radiation from other sources like power-lines, bubble-pack FRS/GMRS radios, smartphones, microwave ovens, wifi, or even the sun/ionosphere can make your OC unstable? I have a fully-aluminum case but I've got a fairly hefty 5w GMRS radio I occasionally use nearby the case...I don't use it at night though. I've noticed my FiiO headphone amp curiously isn't well-shielded and sometimes I can pick up interfernce on the radio...the included usb cable doesn't even have a "Ferrite bead" on its cable -- which is weird considering not to include one on a headphone amp of all things...
Ask a new question

Read More

Overclocking Chipsets Basic