i7 makes me insane
Ok, anyways. I am making an effort to overclock my i7 rig (Specs in sig, disregard speeds because they're being changed around). Right now I'm working on getting the memory to run at it's correct timings. It boots fine at 9-9-9-24 (2000MHz), voltage at 1.66 (no 1.65 option), and QPI/VTT to auto. I personally can't get a setting for the QPI/VTT that will let me boot, but the board on auto is auto setting it to a whopping 1.56V for me to boot. Changing it to 1.50V in Windows with EasyTune6 seems to still be stable, and a couple of hours ago that's what the board was auto setting it to until I restarted my system.
Another observation is that my board's Loadline Calibration seems to do jack shit. With it disabled and CPU at 1.375V, the system reads 1.34V. With Loadline Calibration enabled, it reads 1.34V :blarg: As well, with the RAM set to 1.66 in BIOS, the system reads 1.65V (which is actually good I guess). This all leads me to believe I'm probably getting some pretty bad vdroop for the QPI/VTT as well, which is causing the board to set such high voltages.
Checking all these threads on overclocking, mostly everyone is using 1600MHz RAM and setting their QPI/VTT to around 3.4V. I don't know what it *should* be around for 2000MHz RAM, but I still think 1.5V is high. My memory multiplier is 10x (200BCLK) and QPI is 20X (4000MHz), so that I know I have done right.
Hopefully someone can shine some light on my situation before I go completely insane.:help:
(I miss the ease of FSB:sad:)
If you're overclocking, you should drop your memory divider down to 2:6 or 2:8 toeliminate that as a source of instability and just focus on the CPU first.
My board, an eVGA 758 A1 does not like overclocking with vdroop/LLC on.
C0/C1 or D0?
Your chip might just not be able to run that high a memory frequency.
2000 MHz is generally a really high frequency for RAM. Those RAM sticks are very risky to buy because they may or may not be stable. There comes a point where you just don't need insanely fast speed out of RAM. DDR3-1600 would do perfectly fine for maybe 0.5 less fps?
Exactly, running such high frequencies on memory would net extremely insignificant gains for most usages. Furthermore, it may require higher NB and VTT volts to get it stable. Most D0's at least reach the max around 4000 uncore and usually requires more VTT which isn't worth the amount. Why not just run lower frequencies and try to get the best timings for reasonable voltages? Those 2000mhz sticks should do 1600mhz 7-7-7 easily or maybe a little better if you tweak it here and there.
I know they're high speed sticks, I got them because even though they're 2000MHz, they're just a tiny bit more expensive than most 1600MHz kits. Now, I have been thinking of running at 1600MHz because that's the next lowest multiplier I can go at 200BCLK. (8x 200 = 1600MHz, 10x 200 = 2000MHz) The BCLK may have to change depending on what my CPU clocks end up being (hoping for at least 3.8GHz, but a Google search shows me that most chips from my batch were crappy overclockers. It's a C0/C1 as well, gotta check which exact one though)
I'll work on overclocking my CPU next, I was just hoping I could get the RAM running how it should.
QPI is at 4GHz? Bet that's your problem right there. Many people running high BLCK are using the x18 multi, specifically because the QPI tends to choke at the 4GHz mark. I'm not at all familiar with your board and it's options, but I'd look closer at people who are running high BLCK and managed to push their QPI past the 4GHz mark.
Alright, with the QPI at 3200MHz and RAM at 1600MHz with horrible timings for the sake of CPU overclocking, I seem to be able to get to 3.8GHz. It ran Linpack for over an hour without crashing, so I'm going to do a longer stress test later tonight. The system refuses to go fully into Windows at anything over 200BCLK, giving me a 124 stop error, which to my knowledge usually means more QPI/VTT voltage. I don't feel like putting it higher (1.4V for the time being, going to work on getting it lower, based on what I get for RAM timings when I get to that.)
And one thing I have learned which is really pissing me off, running the GPU OCCT throws me a full system shutdown. Not overclock or anything like that related, it seems me going i7, adding 6 more fans(up to 19 total:shok:) and a pump has pushed my amp draw over what my power supply can give out. No combination of cables, modular or not, have been able to give me enough amps dedicated to my video card to not shut off immediately after starting the GPU OCCT test. Just a glaring reminder to me to get a single rail next time. :blarg:
Anybody have a HX750 for sale?:censored:
You could try more vcore.
I have your gigabyte mobo, ver 1.0 which is good for ddr3 2000+ memory. If you have the ver1.6 board, you are good for 2100 memory. So, looks like you are at the margin. I would just overclock the cpu first to make sure it is stable for the speed that you want. I use P95 and I run it overnight for 9-10 hours. I had the loadline calibration disabled. You can see my detailed settings, as well as a few others, in this thread:
After you are happy with your cpu speed, cpu temp, fan speed and noise, etc. You can then slowly increase your memory speed to see how high you can go. If you have not done so, you can run memtest86+ to stress test your memory as its own to see if it can do 2000 at rated spec. BTW, I guess you already know the limit for vtt is 1.65v.
My past overclock experience is AMD64 4000+ on a DFI board. It was a lot of work compared to overclocking this i7 and gigabyte mobo. I find this really easy.
|All times are GMT -7. The time now is 10:18 PM.|