Anymore headroom to overclock?

[mrbunny]

EDIT I wouldn't recommend using my settings I post in this thread. A lot were experimental, but not too aggressive Read the stickies! Know your own bios, be careful following stray settings
http://images.uesp.net/c/c4/Fishystick.jpg

Better Guide-http://www.tomshardware.com/forum/265056-29-2600k-2500k-overclocking-guide


Read the rest of this thread to see my mistakes, and helpful links. Thanks for the help!



Hello!
This is my first time overclocking without using built in settings. I would like to know if anything can be improved or changed. I may have to much voltage for some settings?

Memory: 1866mhz 10-10-10-27 1.65v Gskill Ripjaw ddr3 pc12800 4x4gb

Cpu: 4.5ghz 45x100base 1.3v 1.25v i/o Intel i5 2500k

Gpu:
950/1900mhz 2150mhz @ 1.15v EvGA 2x560ti 448 UltraClassified
core/shader/memory with Msi Afterburner

Mobo: MSI z77 MPower
PSU: OCZ 750w
Cooler: CM 212+

Temperatures stay below 55 c under normal usage, cards can get up to 70 c under stress
The rest of the voltages and settings were left to auto in bios.
Any advice would be helpful. Thanks!

 

[bebop460]

It can be time consuming to go past ~4.5ghz, with only a small performance increase.

Personally, I would keep the CPU under the Intel spec of 72C max, definitely no more than 80C when stress testing.

I would mostly consider your temps as the limiting factor, but for voltage, on the conservative side, many people say 1.3-1.4 vcore max. On the flip side, some people run 1.5v or more, but I'd definitely keep it under 1.47v.

 

[mrbunny]

I had more time to test the overclock, and backed off a few settings. Here it is for anyone with a similar setup.

Not Efficient

Memory: 1866mhz 10-10-10-30 1.65v Gskill Ripjaw ddr3 pc12800 4x4gb

Cpu: 4.4ghz 44x100base 1.35v 1.25v i/o Intel i5 2500k

Gpu:
950/1900mhz 2150mhz @ 1.15v EvGA 2x560ti 448 UltraClassified
core/shader/memory with Msi Afterburner

Temperatures are not bad, but I have way too many 120mm's

You were right about 4.5 being a harder to achieve. Just backing off 100mhz made it completely stable. Changing the latency increased stability as well.

 

[bebop460]

Nice overclock.

You may be able to get 4.4ghz+ stable with less vcore if you try offset voltage.

Also, supposing you overclocked your RAM, you may want to have a look at JackNaylorPE's post here http://www.tomshardware.com/answers/id-1795015/system-memory-overclocking-cpus-die-memory-controllers.html about calculating actual RAM NS latency. This post illustrates that, when you consider latency, its possible to overclock your memory and actually make it run slower.

For example, 1866mhz C10 is slower than 1600mhz C8. I've tried everything from 1600mhz C8/C9, 1866 C9/C10, and 2133 C11, and found no difference except in synthetic benchmarks. For that reason, IMO it's best to leave RAM at XMP and save the stress on your hardware.

I don't know which kit you have, but If you still want to overclock the RAM, you may want to try 9-9-9-28 1866mhz 1.57v. That was the sweet spot for my kit (http://www.gskill.com/en/product/f3-12800cl8d-8gbxm ....F3-12800CL8D-8GBXM; XMP is rated at 1600mhz 8-8-8-24 1.5V), though I dont run it overclocked anymore.

EDIT: Oh, and theres nothing wrong with 120mm fans really. You can get some nice high CFM/static pressure fans in 120mm if you want to drop temps a bit more. Silverstone and Koolance in particular make some very nice 120s.

 

[mrbunny]

That is a lot of great info! Thank you I have a hard time finding what i'm looking for.
Memory model number is f3-12800cl9-4gbxl
If It would help I can provide more info.

It looks like I'll revise it a little more, those are interesting articles. I may fiddle with the timings at 1600 MHz, if the 9-28 doesn't work.
I'll have a look at offset voltage and see if it can be changed, and if vcore could drop.
We'll see what happens, it sounds good.
Cheers

 

[bebop460]

Being that your XMP speed seems to be C9, you may not be able to bump it to 1866 C9 and be stable. If you want to overclock it, Its worth a shot though. You may have better luck with 1600mhz.

I'd definitely back everything up before tinkering with the RAM. I corrupted my OS twice overclocking it.

But again, in my experience the difference is only seen in synthetic benchmarks. I ran a comparison with 1600 C8 (5ns delay) and 1866 C9 (4.8ns delay) with 3DMark, Heaven, Aida disk bench, OCCT, and MetroLL bench. The aida bench was better and +~2 minimum fps in the metro bench and a few more in OCCT (Heaven and 3D were identical). I concluded that it was not at all worth it for me, with negligible performance gains, but perhaps others may achieve greater benefit.

 

[mrbunny]

So it appears I should stick to the XMP profile. I am not really satisfied with the memory's performance. At 1600mhz, 8-8-8-28 or 8-9-9-28, 1.65v, 1.60v 1.57v it still crashed, perhaps less volt would work I don't know. I'm not going to increase the voltage more, for such little gains. It was a low priced kit at the time though.

Could you explain offset voltage to me? I found information online to be confusing. I believe it is I/O voltage in my bios? I set that to 1.15v based on a guide I found, but changed no other Offset related settings. V core I left at 1.35v

I was able to achieve a higher cpu clock,4.6ghz, although it isn't stable enough. 4.5ghz works fine now. Heat isn't a problem yet, under normal load. I am still messing with the settings,trying to get it more stable at 4.6ghz.

Worked, but high voltages
After messing around and reading guides, I came up with these changes
4.6ghz@1.35vcore, ?
VDROOP 100%
I/O 1.15v
Turbo, Enhanced Turbo Off

It Hasn't crashed yet.

 

[bebop460]

Yeah, sticking to XMP is usually best, but if your not satisfied with the RAM performance, I would bench/try both 1866 C10 and 1600 C9 for awhile and see if you note a difference. Is your CR (command rate) set to 1? That (theoretically) will give it a bit of a boost too. You could try 8-8-8-24 1600mhz, but 1866 C10 is probably as good as it gets, and the difference between the two is minimal.

As far as offset goes, I preferred to play with it, as the different methods were ineffective and tedious for me. I'm not certain, but I believe it is a +/- in relation to auto voltage. So if auto is, say, 1.25V (at 4.5ghz), a -0.05V offset should provide 1.20V under load. Yet, it didn't work that way for me, as 0.05V increment configs wound up being anywhere from ~0.03V-0.08V.

However, some have luck with the calculation listed here: http://www.overclock.net/t/968053/official-the-sandy-stable-club-guides-voltages-temps-bios-templates-inc-spreadsheet/2240#post_14466483

Voltage can be confusing. Lots of jargon and different settings there. You may have tinkered with these before; they can help achieve stability, especially past 4.5ghz: VCCSA (0.925-0.971v. not too helpful. auto or the stock setting of 0.925 seems to be best usually), VCCIO (also known as QPI or VTT, 1.08-1.12v, 1.15 max), and PLL voltage (~1.6-1.89v) seem to have a sweet spot thats unique for each motherboard and CPU and can help in stability. PLL voltage in particular can have a big impact, with less voltage often helping with stability. LLC has a big impact as well. I'd suggest running medium LLC for the best voltage curve.

Additionally, when overclocking with offset voltage it is usually necessary to disable C3, C6, and package C sleep states. C1 can be left enabled.

 

[mrbunny]

Wow, my actual voltages in use may have been a little higher than I thought. I didn't completely understand what those settings did.

Now I'm running 1.352v max with hwmonitor in Vcore, but 1.381v in VID? It's a good thing I didn't manually push the voltage higher than 1.35v in bios. I backed off to 4.5 GHz until I understand how the voltage settings work for my bios. It was doing prime 95 at 4.6 GHz, but was getting a little warm. I'll have to do some more reading.
VDROOP to 75%
Changed power limit to 250/250
Should speed stepping be disabled if for this kind of overclock?
I left C-State and C1E enabled as well.

I left my memory set to the XMP settings; 1600mhz 9-9-9-24 1.5v except for CR, which is 1T
This was at 4.5 GHz http://i.imgur.com/nLDgdhU.png?1 The room got too warm, I'll see if its actually a 24/7 o.c.
Thanks, I never thought I could get the 2500k to 4.6ghz

 

[bebop460]

The VID only represents how much vcore the CPU is requesting, therefore it is rather useless outside of using it for calculations. Vcore is the actual amount of voltage being delivered.

Your temps (according to the screenshot) look fine. I wouldn't worry unless you exceed 72C/162F (the Intel stated maximum) or more when stress testing. Personally, 80C is as hot as I'll allow mine to get. It will likely never get that hot during normal use. I'd watch the VRM temps too.

4.5ghz at 1.35v is quite a bit of voltage for an i5. I'd try it with less vcore. Your vdroop and power settings should be good.

Speed step allows the processor speed to go faster or slower depending on its workload. I'd definitely leave it enabled.

If you are using an offset voltage and are unstable, I'd definitely try disabling all C states except C1. This can also increase HDD/SSD performance slightly, in addition to helping stabilize offset overclocks.

 

[schmuckley]

There should be some more headroom there.

http://hardforum.com/showthread.php?t=1578110

I have no experience with MSI though..sorry.

 

[bebop460]

I agree. 4.5ghz is a good place to be at, and a nice overclock, but personally, I'd keep bumping the voltage until at ~80C max during prolonged, (high) stress testing or ~1.4v vcore, whichever comes first (assuming other temps are OK). Then I'd see how high the multiplier can be pushed.

That link was a good read schmuckley. That article makes me wonder again about how BLCK overclocking effects these chips, and if it's safe or useful to tinker with. Guess my CPU is pretty average, with 4.8ghz @ 1.4v and 5ghz @ ~1.5V.

 

[mrbunny]

Thanks for the great help, It's a busy week here. I still have to go through those articles and links, and play with the settings more. I'll see what I can get to without pushing it too hard.
For the mean time, I put my CPU set to ECO : )
Thank You!

 

[mrbunny]

Thank you Bebop460 and Schmuckley
I should of read the sticky and your links more.

http://www.tomshardware.com/forum/265056-29-2600k-2500k-overclocking-guide
OP - mrface

"For my setup, the following equipment was used:
Motherboard: MSI P67A-GD65 Intel Motherboard.
Processor (CPU): Intel Core i7-2600K Sandy Bridge 3.4GHz, 4 x 256KB L2 Cache, 8MB L3 Cache, LGA 1155 Quad-Core Desktop
RAM: G.SKILL Ripjaws X Series 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 PC3 12800"

~

Voltage and their extremities:
Refer to the following to understand your limits before setting. While most can be increased with no problems, you can fry stuff and void your warranty at higher voltages.
VCore: Stock is 1.3v and should not be raised higher than 1.5v
VCCSA (System Agent): Stock is 0.925v and should not be raised higher than 0.971v
VDDQ (RAM VOLTS): Stock is 1.5v and should not be raised above 1.65v (this depends on the manufacturer of the RAM, so please note when you buy it.)
VCCPLL (Phased-Locked Loop): the stock is 1.8v and should not be raised over 1.89v
VCCIO (I/O Supplied voltage): The stock is 1.05v and should not be raised over 1.0815v"



Cheers to MSI and Intel.


The above settings are for P-67. Mine is z77.
That's the only setting I over-volted, I think, I can't remember If I played with S/A
Did I get lucky with my previous settings; Vdroop at 100%, and i/o at 1.15v?
The first time I had i/o set to 1.25v based on a guide I read about overclocking memory!

VRM temperature,
Is there a probe for my mobo?

EVGA 560 ti 448 U C x 2 OC VRM TEMP
I have to check it manually for my video cards, but I haven't. Stock 1.025v - 1.150v OC is a little much? Its stable and the chips never read over 70c, rarely 65C.
http://www.tomshardware.com/answers/id-1654145/vrm-temp-detection-gtx-560ti-448.html
http://www.tomshardware.com/forum/392416-33-evga-ultra-classified-thermal-paste
The coolers on the cards are pretty good, my only gripe is how LOUD they are when both fans are spinning at 5,200 rpms. They are effective at dumping hot air out of the box.


I'm going for 4.7.... after I read more.

 

[bebop460]

1.0815v VCCIO is the Intel stated maximum (1.05v +/- 3%), and I wouldn't exceed it unless necessary, but its not uncommon to see it pushed as high as 1.15v.

Some say VCCSA shouldn't be changed and results in instability and possible damage. I wouldn't bother with it unless you've tried everything else and want to see if it helps. I found no useful change in stability across the min/max range (0.879-0.971v) at 5ghz, but YMMV.

I'd highly recommend overclocking the CPU before the RAM, and that you try undervolting VCCIO and PLL to see if it helps. Finding the sweet spot helps a lot in stability beyond 4.5ghz, and once you've found it, overclocking pretty much becomes just a matter of adjusting vcore (and LLC) to the appropriate level for the speed.

If there's a VRM probe for your mobo or GPU, it should show up in HWMonitor. If not, you could perhaps get a rough idea by measuring it with an IR thermometer with the GPU case off, though I guess that may require some careful fan placement to ensure it doesnt overheat. Did your GPU come with a lifetime warranty? If so, I wouldn't even worry about it unless your pushing very high voltage.