Please Help With Overclocking CPU (i5 2550k) and DDR3 1600 RAM in Intel Visual Bios

[LetoAtreidesII]

I recently "upgraded" to a new motherboard, CPU, CPU Cooler, and PSU after I damaged my old CPU and motherboard. It wasn't much of an upgrade because I wasn't sure what components were damaged and I wanted to swap as few parts as possible so I tried to stick to my old LGA socket. Unfortunately, it turned out that I had to replace both the CPUand motherboard (nearly everything). Still, I did net an upgrade from an I5 2400 to an i5 2550K CPU.

I've never overclocked before, but I'd like to now to get the most out of this CPU. I also learned that the RAM which I had was underclocked by default and I'd like to "overclock" it to run it at its stock rates. My new motherboard is an Intel DZ77BH-55K whose BIOS I updated to the latest version which is Intel Visual Bios. Intel Visual Bios seems to be designed to make overclocking easy. But as I've never overclocked before and I'd like to be as careful and informed as possible, I'd like as much advice and information as possible first.

For CPU overclocking the Intel Visual Bios overclock assistant provides a simple slider which ranges from 3.8 GHz (default) to 4.5 GHz. As near as I can tell, however, overclocking with this slider ONLY affects the Turboboost speed. Is this a good way to overclock or the best way to overclock - to only change the Turboboost speed? If not, how would I go about it? I plan to run Prime95 for 30 minutes afterwards to test stability and use Speccy to monitor temps. Is there anything else I should do as part of the overclock process (I've never overclocked before)? Also, should I increase fan speed in Bios (and how do I do this - there is a cool simple setting in visual bios I'm considering)? At stock speeds it's already running in the 70's under load.

This is my RAM: 4x4GB https://www.newegg.com/Product/Product.aspx?Item=N82E16820233180
There is a slider in BIOS for RAM which I could slide to 1600. But there's also a Performance Memory Profile drop-down box which has an option for an XMP-1600 profile. That profile and the slider at 1600 have very different values. Which should I use?

Finally, how much will these overclocks increase my power usage? I only care because I actually downgraded my PSU from 750W to 650W (b/c a 650 was the spare I had lying around). Fortunately, I think 750 was way more than I needed before and 650 might be more than enough now. Thanks all for any advice you can give.

 

[LetoAtreidesII]

These are my BIOS settings for my CPU (all default except for VR Droop Control):

Processor VR Droop Control - High, Medium, or Low - defaults to High, I set to Low
Voltage Offset (mV) 0
Turbo Ratio 0
Maximum Non-Turbo Ratio 34
Host Clock Frequency 100

Runtime Turbo Ratio - Checked
Intel Turbo Boost Technology - Checked
Burst Mode Power Limit - 118 Watts
Sustained Mode Power Limit - 95 Watts
Sustained Mode Time - 1.5 seconds
TDC Current Limit Override - 97 Amps
Active Processor Cores - ALL
Processor PLL - 1.85V
Internal PLL Voltage Override - Unchecked
Processor Idle State - Low Power
Processor VR Droop Control Low V-droop (performance)
Processor I/O (V) 1.050
PCH Core (V) 1.050

You can see all the settings I have here:
https://sites.google.com/site/visual...ance/processor

One modification I made was to change VR Droop from High to Low since saving power isn't a priority for me. Was this a good idea?
I take back what I said earlier about the OC assistant slider only affecting Turbo Boost. It doesn't. It actually changes the Voltage Offset, Burst Mode Power Limit, Sustained Mode Power Limit, and TDC Current Limit Override. Someone recommended that I overclock by modifying primarily the core ratios and voltage offset only a little bit (+.1V). If I use the slider, the turbo core ratio stays at 0 no matter where I move the slider. So, should I overclock with the slider or by manually adjusting the ratio/voltages? And what voltages? As you can see above there are lots of different voltages and power limits. Overclock guides reference stuff like Vcore (lots of people talk about this but I don't see it in my BIOS), QPI/VTT(VCCIO), and System Agent Voltage (VCCSA) that I don't see in my BIOS.

 

[BonzaiDuck]

So you're new to OC'ing or just unfamiliar.

If the Intel board has a Z77 chipset -- looks like it does and I'm not going to do a web-search -- you should be able to wring some extra clocks out of it.

I may advise against tweaking within a Windows application, but now I see the BIOS screens for your board and I understand what you mean.

First, make sure your memory is set to spec speed, timings and voltage. You can leave all the other settings on auto initially, and I suggest you set the multiplier (ratio) from its stock turbo value to about 4.2 Ghz. This will tell you how far you can go with VCORE on auto settings.

After that, I'm totally unfamiliar with your motherboard, so I would only say that once you find a point of instability with auto-VCORE, you'll need to adjust it manually.

On many of the Z68 and Z77 boards, you could either OC with a fixed VCORE setting, or use the Offset feature. Some of those boards, like my Z68, added a third alternative which was placed in the wrong sub-menu on my ASUS boards. It was labeled something like "Extra voltage applied to Turbo" and essentially provided an "Adaptive" mode for OC-ing. In my opinion, "fixed-VCORE" is only a preliminary tool, and the worst choice among the three. Playing with the Offset was better, but could lead to "idle instability:" your system would be stable in a stress test, but leaving it idle for some longer period of time might result in a BSOD or unannounced reset. So it is best if you have the adaptive option under whatever name they give it, and try to keep the Offset as close to zero as possible or at its default.

You should also control the amount of vDroop your system allows for the CPU voltage under load. This is the built-in protective feature that results in a lower voltage at turbo under stress than what appears as turbo voltage without stress or at idle. You would control vDroop by changing what most boards label as "LLC" or "Load Line Calibration." Your board describes it more directly. Between "none" and the highest "Extreme," you want to probably pick the median or middle notch in the range offered. I can only say that what you want to end up with adjusting LLC is an approximate 2 to 4 millivolts of remaining droop. You would recognize this with your stress testing by noting the highest TURBO voltage applied without load/stress, and comparing to the monitored VCORE under the sort of stress you'd get with IntelBurnTest on it's "Maximum" setting.

You say you have an i5-2550K processor. Its behavior should just fall short of my 2600K in terms of expectations of voltage and speed. For my 2600K, I could get to a solid and stable 4.6 Ghz for a VCORE traversing 1.36 to 1.38V under turbo conditions between load and idle.

It should not complicate your efforts to leave EIST enabled during all this, but at non-turbo idle, you'd see the processor speed drop to about 1.6Ghz with a VCORE between 0.90V and 1.0V.

Do some more reading. Proceed without hurry. You should be able to wring some more clocks out of that puppy.

As for RAM. This can be a lot more tedious. You could TRY to raise your DDR3-1600's to 1866 Mhz by loosening the timings. But my best advice would be simply to buy yourself a set of DDR3-1866 or 2133 RAMs compatible with the board.

 

[LetoAtreidesII]

I'm completely new to OC - never overclocked before at all.

So, I haven't overclocked my CPU at all yet, waiting to install a new cooler as someone recommended and still researching/gathering info.

There is one alarming thing I discovered. As I said, I still haven't overclocked my CPU at all. Running at BIOS defaults for my CPU, CPU-Z reports my Core Voltage at 1.32 V under load (while running CPU-Z's own Stress CPU test from its Bench tab). It fluctuates between 1.312 V and 1.32 V but is at 1.32 V most of the time. Temps gradually rise to about 75 C. Like I said, this is at stock settings - 3.8 GHz Turbo @0 voltage offset. I didn't change anything in BIOS related to CPU except Vdroop. So I thought maybe my Vdroop change was to blame, so I changed Vdroop back to the default value of High. There was no difference - it was still at 1.32 V while running the CPU-Z stress test.

The reason this bothers me is because these Vcore values are basically overclock Vcore values for other people with the same processor. Also, since it's pretty close to the max safe Vcore, it doesn't leave me with much room to increase voltages for when I do overclock.

So, my questions are, is this normal? Is this good or bad? What Vcore values should I be getting at stock (non-overclocked) settings for my i5-2550k? Since Vcore is already so high, does this mean that I can try increasing my core multipliers without increasing voltage?

 

[Kenmitch]

Many moons ago I had a 2550k.

Try setting vcore to normal and offset to 0.000v and see how vcore responds. Id imagine you'd be able to just up the multiplier with 1.32v vcore. Trying to remember if normal and offset of 0.000v locked vcore when upping the multiplier or not. Kind of thinking vcore ups with multiplier setting....Was long time ago. Fixed vcore with zero offset would work if so.

2550k brings back memories from the best OC'ing adventure I've ever had....So far.

Was during my crazy OC days. Chip was beastly!

 

[LetoAtreidesII]

Many moons ago I had a 2550k.

Try setting vcore to normal and offset to 0.000v and see how vcore responds. Id imagine you'd be able to just up the multiplier with 1.32v vcore. Trying to remember if normal and offset of 0.000v locked vcore when upping the multiplier or not. Kind of thinking vcore ups with multiplier setting....Was long time ago. Fixed vcore with zero offset would work if so.

2550k brings back memories from the best OC'ing adventure I've ever had....So far.

Was during my crazy OC days. Chip was beastly!

Kenmitch, 1.32 Vcore IS at 0.000v voltage offset. That would be the only setting in my BIOS related to Vcore as far as I can tell (though maybe Burst Mode Power Limit and Sustained Mode Power Limit have something to do with it as well). That's what has me a little concerned. I literally cannot lower it below 1.32 Vcore. This is the baseline default. Of course, this is with testing with a CPU-Z, but that seems to be what many people use to measure Vcore.

 

[BonzaiDuck]

Kenmitch, 1.32 Vcore IS at 0.000v voltage offset. That would be the only setting in my BIOS related to Vcore as far as I can tell (though maybe Burst Mode Power Limit and Sustained Mode Power Limit have something to do with it as well). That's what has me a little concerned. I literally cannot lower it below 1.32 Vcore. This is the baseline default. Of course, this is with testing with a CPU-Z, but that seems to be what many people use to measure Vcore.

Unless I misunderstood from your description, even with the offset as you have put it, the adaptive feature would certainly increase VCORE on its own unless you also set that feature as well.

These boards -- even the Z68 and Z77's back when -- tend to overvolt the processor with "auto" settings. the board "partners" want to assure their boards will run at stock settings with any processor off the Intel assembly-line. And even though the processors may not vary that much, the board-makers' desire to avoid RMA for such a reason overrides that possibility or likelihood.

If you haven't done so, I'd try upping the multiplier by integer 1 until the system crashes in a stress test -- and without changing any voltage settings. Those processors would allow for a multi up to 4.2 Ghz with the Auto settings -- some more -- some less. At that point, you should begin adjusting voltage together with each step up in the multiplier. But I would guess the i5 processor is probably good to go at 4.5.

Then, there's the fear of the unpublished "maximum safe" threshold. The processors had 32nm lithography -- same as previous-gen Nehalem -- or the later version of Nehalem. And they had published that spec. It was approximately 1.375V as I remember.

I've had two Sandy Bridger i7-K processors volted so that (extreme LinX) Turbo load takes them to about 1.38 to 1.39. One of them had been running (literally) 24/7/365 for almost exactly six years. We recently began to see critical error Event ID 41's in the logs, and I still suspect it may be the PSU. It seems to occur occasionally when the system is coming out of hibernate, and I'm testing it at this moment. The problem machine has been OC'd to 4.6 Ghz -- about par for a 2600K. [And I probably shouldn't have turned it loose on my elderly Moms. She can't remember that returning from hibernate and cold-booting may be two different things. But she denies using the system at 8:34AM this morning, so I will fix this . . . Anyway -- she doesn't NEED 4.6 Ghz; she'll be more than happy with 3.8.]

There are two schools of OC'ing since around Sandy times.

Some folks will get their systems stable with a fixed voltage in the BIOS. And some of those people will then enable EIST, other power-saving features afterward. Some others -- holdouts from a bygone era -- leave the BIOS set for a fixed voltage.

Then there's the other group. As long as you take notes on monitored voltages at stock-optimum settings so you have a baseline, it is no more difficult to leave EIST enabled, set the offset (+ recommended with minimal or 0 mV), and then adjust the Adaptive voltage or "Extra Voltage for Turbo" if you have such a feature in BIOS. Others would simply raise the Offset voltage if that feature wasn't available, and some boards had it, others didn't.

To give an idea, for my 2600K system, I had Offset defined as "+" and maybe 5mV, and my "Extra Turbo" at 20mV. YMMV, YMMV, and my processor is not only likely different, it's a different model with hyperthreading. But -- it's a Sandy Bridger K.

Another thing I've noticed -- may not be true for your CPU -- is that the SB-K's were just a little on the toasty side in idle. So I'd witness temperatures between 35 and 42C once the clocks were set. Of course, my fans are thermally controlled, and I can fix that somewhat, but the temperatures at idle never fall much below 35C.

I could give a dissertation on the use of Stress-testing programs. But I'll leave off here. You apparently embrace your homework.

 

[LetoAtreidesII]

Unless I misunderstood from your description, even with the offset as you have put it, the adaptive feature would certainly increase VCORE on its own unless you also set that feature as well.

These boards -- even the Z68 and Z77's back when -- tend to overvolt the processor with "auto" settings. the board "partners" want to assure their boards will run at stock settings with any processor off the Intel assembly-line. And even though the processors may not vary that much, the board-makers' desire to avoid RMA for such a reason overrides that possibility or likelihood.

If you haven't done so, I'd try upping the multiplier by integer 1 until the system crashes in a stress test -- and without changing any voltage settings. Those processors would allow for a multi up to 4.2 Ghz with the Auto settings -- some more -- some less. At that point, you should begin adjusting voltage together with each step up in the multiplier. But I would guess the i5 processor is probably good to go at 4.5.

Then, there's the fear of the unpublished "maximum safe" threshold. The processors had 32nm lithography -- same as previous-gen Nehalem -- or the later version of Nehalem. And they had published that spec. It was approximately 1.375V as I remember.

I've had two Sandy Bridger i7-K processors volted so that (extreme LinX) Turbo load takes them to about 1.38 to 1.39. One of them had been running (literally) 24/7/365 for almost exactly six years. We recently began to see critical error Event ID 41's in the logs, and I still suspect it may be the PSU. It seems to occur occasionally when the system is coming out of hibernate, and I'm testing it at this moment. The problem machine has been OC'd to 4.6 Ghz -- about par for a 2600K. [And I probably shouldn't have turned it loose on my elderly Moms. She can't remember that returning from hibernate and cold-booting may be two different things. But she denies using the system at 8:34AM this morning, so I will fix this . . . Anyway -- she doesn't NEED 4.6 Ghz; she'll be more than happy with 3.8.]

There are two schools of OC'ing since around Sandy times.

Some folks will get their systems stable with a fixed voltage in the BIOS. And some of those people will then enable EIST, other power-saving features afterward. Some others -- holdouts from a bygone era -- leave the BIOS set for a fixed voltage.

Then there's the other group. As long as you take notes on monitored voltages at stock-optimum settings so you have a baseline, it is no more difficult to leave EIST enabled, set the offset (+ recommended with minimal or 0 mV), and then adjust the Adaptive voltage or "Extra Voltage for Turbo" if you have such a feature in BIOS. Others would simply raise the Offset voltage if that feature wasn't available, and some boards had it, others didn't.

To give an idea, for my 2600K system, I had Offset defined as "+" and maybe 5mV, and my "Extra Turbo" at 20mV. YMMV, YMMV, and my processor is not only likely different, it's a different model with hyperthreading. But -- it's a Sandy Bridger K.

Another thing I've noticed -- may not be true for your CPU -- is that the SB-K's were just a little on the toasty side in idle. So I'd witness temperatures between 35 and 42C once the clocks were set. Of course, my fans are thermally controlled, and I can fix that somewhat, but the temperatures at idle never fall much below 35C.

I could give a dissertation on the use of Stress-testing programs. But I'll leave off here. You apparently embrace your homework.

Thanks for all the great info! Regarding adaptive features and extra voltage for turbo, I don't see anything like that in my BIOS except maybe Burst Mode Power Limit, Sustained Mode Power Limit, and TDC Current Limit Override. These settings would suggest that there is such a feature, but it only sets the limits rather (I think) than having any direct control over feature itself. When I use the OC assistant slider to increase my processor speed, it increases all three of those in addition to turbo ratio and voltage offset.

So anyway, what I plan on doing once I install my new cooler is to increase my turbo ratios manually without increasing anything else - probably to 4.4 Ghz or 4.5 Ghz to start, and then testing in IBT and/or P95. Any advice on whether I should use IBT or P95 or both?