Z77A-GD65 + i5 2500K Overclocking - How to start?

[Zach_]

Success!!! Finally, I was able to stabilize 4.6 GHz!

I updated my BIOS to 10.5
Set voltages:

Vcore to 1.43 (1.4 in CPU-Z)
I/O - 1.07
S/A - 0.97
PLL - 1.84

I also set the Short and Long Duration Power Limit values to 200 and 255 respectively and set the maintained time to 30 seconds before disabling EIST, then I disabled EIST so the values should be saved in the background.

Managed to run Prime95 blend for a full 12 hours without a single error and it finished a full cycle of tests from the smallest to the largest FFT, relief!
Also ran IBT for 10 passes on Maximum and it passed, finally I can call this config stable, this is enough for me.

I also did the Zip test, created a large 8GB archive with all sorts of files, MKVs, ISOs, data files and lots of executables, zipped them, checked for integrity and unzipped them, no errors.

Will move on to 4.8 soon but in the meantime it's time to start enjoying the PC.

Would like to hear your opinion about my current voltages and stability on 4.6, thanks:



More updates to follow :)

[profJim]

vcore and PLL are high but probably not a problem as long as temperatures are within reason.
Do you recall the difference in temperatures when running Prime95 with and without AVX?
Not to push your luck, run HCI Memtest overnight or while you're away for the day.

How about some screen-shots showing:
- temperatures at idle and under full load
- benchmark scores (including GFlops)

Keep in mind that stability is merely an illusion
Enjoy your pc, you've earned it!!

[Zach_]

Hey Jim,

The Vcore worries me a little, check out what this guy on the MSI forum said:

1.52v's is Intel's 'ABSOLUTE MAXIMUM', and not recommended for an everyday set value. In earlier datasheets, they were more defined as recommended limit and the absolute limit. 1.400v's is considered the everyday operating limit.

Variations in voltages, temps, and things are often different than actual or what is read by utilities in a Windows environment. Unless you use a digital multimeter at the V-Checkpoints, you really have no clue what the actual delivered voltages are.

Quote
many people have passed 1.4 and got their overclocks stable, without taking a performance hit or experiencing chip degregation

And how many 'have' experienced degradation that won't talk about it or admit it? It's your stuff, the experienced here will give you suggestions and recommendations, the rest is up to you. Out of 4 Sandy/Ivy rigs here, not one of them is set to run the CPU over 1.392v's for everyday. 3 out of the 4 are between 1.360v's and 1.384v's (this is when in a raw OC configuration).

Added: It's not only the CPU that is under more strain, it's also the higher voltages going to the VRM's/MOSFETS. Even Class II and III mainboard components can only take so much, and they should be taken into account as well.

2nd add: I actually lost two Sandy 2700k CPU's. One at exactly 1.472v's (verified by DMM), & the other I recall was lower than that when it went south. Just checked, the other one was at 1.432v's when it popped (verified by DMM).

I can lower the Vcore to below 1.4 and stabilize IBT at 4.6 GHz, I've managed to do that with 1.37, the question is, will it be really stable and can I rely on my PC based only on IBT. The guys there say that it should be more than enough to verify that your overclock is stable since nothing will stress the CPU so much.

The temps difference was about 5C lower with the regular version of Prime as compared to the 27.7 AVX version, not that big, GFlops scores were the same, 130 seconds for a loop and 120GFlops on average, I didn't take a screenshot. Max temp in both P95 and IBT reached about 70-74C.

Which benchmarks would you like me to post, and also, please can you explain why AVX is so important to use with these CPUs?

Here's what another guy at the MSI forum said about IBT vs P95 and my current voltages:

You are looking for the magic trick that gives you 4,6GHz. We don't have that wand. We can't change your CPU's power requirements.
If you left your machine at the current settings and used it for what you intended to use it, you will probably find that it may do everything you want to without ever having a hickup. You continue to add more and more voltage and then stress test for hours on end to find that elusive system crash, you may just find that your CPU performance deteriorates, especially when using 1,4V plus.
Settle on an OC that is stable at or below 1,4V core and get on with using your machine for its intended purpose.
You will test your machine to death the way you are continuing. You found this Prime95 AVX just to make triple sure ?
None of your everyday use programs will ever stress the system like that so the chances are your system will be as stable as it can be whilst overclocked.

[profJim]

I strongly disagree with their IBT (or LinX) only approach to stress testing.
Do these programs stress the memory sub-system the way Prime95/AVX does? I don't know.
Is stress testing with the non-AVX Prime95 good enough?
Maybe for today, but what about games or programs that are optimized to increase performance using AVX enhanced instructions?

Sorry, I don't have faith that using only one stress testing program will ever be adequate.

Is electromigration an issue with higher voltages? Near as I can tell, yes.
Will running your system with voltages near the high end have an impact on component life? Yes. How much? I don't know.

I like the shopping bag analogy.
If a bag is rated to carry 20 lbs, how long will it last if your frequently use it to carry 5 lbs? 10 lbs? 15 lbs? 20 lbs? 25 lbs?

From Advanced Vector Extensions (AVX) - Wikipedia:
Applications

• Suitable for floating point-intensive calculations in multimedia, scientific and financial applications (integer operations are expected in later extensions).
• Increases parallelism and throughput in floating point SIMD calculations.
• Reduces register load due to the non-destructive instructions.

Even though IBT or LinX cause the cpu to run hotter than Prime95 with AVX, it looks like Prime95 is a much more stressful test.

I agree with their recommendation to use a digital multi-meter (DMM) to measure actual voltages when stress testing and not to believe the reported voltages from the motherboard chip and software. This probably isn't required with a moderate overclock, but it is important when you're running with higher voltages.

If I had a higher end system, I'd run with a moderate overclock with relatively low voltages for everyday 24/7 operation.
If I had a very demanding application that required maximum performance, I'd boot with high overclock while running the game or program.

What happens with AVX2 (aka Haswell New Instructions)? Hmmm....

just my 3.14159 cents worth of opinion

[profJim]

Some good recent posts in your MSI forum thread, Z77A-GD65 + i5 2500K Overclocking - How to start?

Why are you ignoring the repeated advice to buy or borrow a DMM to verify the reported voltages?
You don't need to spend $50 - $100 for a DMM, you should be able to buy one for less than ~ $20.

For many years, enabling Spread Spectrum has been a cause of system instability.
Some mobo manuals recommend that it should only be enabled if you're experiencing specific problems.

With newer systems, it might now be possible to enable some or all power saving options in the bios and/or in win7 without causing instability.
I particularly like the post where HU16E says:

All I can say is give their suggestions a try. Trial and error has been an OC'ers time honored tradition for years. The information I gave you works for me, it may not work for everyone and their system configuration.

Endeavor to persevere! :-))

Nothing ventured, nothing gained.

** edit **
from your MSI thread:

Zach_:

Just for kicks, can you try stressing one of your Sandy rigs with Prime 95 27.7 AVX version with a blend test for about 8 hours and let me know what are your results with the same settings you gave me?

reply:

At my age, an 8 hour stress test would be an eternity. :rolleyes:

He can't do it while he's sleeping?
Sounds like he's over the hill and accelerating rapidly.....downhill

[Zach_]

About the DMM, just didn't have time to do that, but I might do it now just to verify things, do read:

I've decided to do the smart thing, I don't want to cross 1.4V in the BIOS so I decided I'm gonna try 4.5GHz and lowered the voltages as follows:

Vcore - 1.4 (1.384 in CPU-Z idle, 1.368 load)
I/O - 1.07
S/A - 0.95
PLL - 1.82
DRAM - 1.514

I then tested P95 AVX 27.7 Blend test which ran successfully for a full 12 hours! Check out the screenshot:



Which really surprised me, I was sure I'm gonna get a BSOD but it passed.
Then I ran IBT for 10 loops on Maximum to finish it off and it passed as well:



I'm very pleased with this, I was able to lower the Vcore by 0.3V, lower the PLL and S/A and still get it stable on 4.5GHz
I guess 4.6 is the crisis point for my particular CPU and it needs much more voltage in order to deal with this frequency, same thing that happened to me with the Q9550, I was able to stabilize 3.8GHz on 1.4V but for 4.0GHz I had to go north to 1.45, which is too high and risky for another 200MHz...

The difference in performance between 4.5 to 4.6 will be invisible, plus I've read several reviews which proved that there is hardly a difference if any at all in gaming or benchmarks so I'm fine with 4.5GHz at 1.4 Vcore.
I think I'm gonna settle down with 4.5 GHz it's a very nice 30% overclock to an already very powerful CPU, no need to push it to 5.0 and try to fry the CPU with high voltages :)



What do you think?

[profJim]

Ah, looking much better
Two screen-shots?? I'm having a heart attack
Maybe your updated bios contributed to system stability with your lower voltages.

Why not push your luck a bit:
Try enabling some of the bios power saving options and see if your system is still stable at 4.5GHz.
If it works, you'll have the best of both worlds.

I think that BF3 is currently one of the most demanding games.
You should run it for 6 continuous hours or longer as part of your "final" stress test.

You should "intrude" in the ***Official***Z77X-D3H UD3H UD5H Discussion\Owners Club*** thread and ask if any of the Gigaboys over there have tested their 4.xGHz overclocks using Prime95 AVX 27.7 Blend and Large FFT tests and if their tests were successful. Feel free to mention that this was my suggestion.

For the benefit of other Z77A-GD65 owners or wannabees you should post screen-shots of all of your overclocked settings.
When using the Insert Image option, each image will be much larger. You might have to un-tick the "Retrieve remote file and reference locally" option.

Haven't you learned by now that asking for my opinion is hazardous to your mental health?

[Zach_]

Hey Jim,

Your opinion always helps me :)

Anyway, I've posted in that thread, let's hear what they have to say, for now I think I'm done with this OC, will run some benchmarks for performance, never been much a benchmark user so if you could give me a few good ones (Aside from 3Dmark and Unigine Heaven) I'd be happy to post.

Will post my BIOS screenshots later on.

In regards to power saving features, EIST throttles the voltage too much, I'm seeing a bigger Vdroop when it's enabled, and anyway, I don't have my PC on 27/4 so I'm not sure it's needed, plus I don't have the energy for more stress testing atm.

Peace.

[profJim]

You've earned your o/c, now it's time to enjoy your system.

If any of your games have a built-in benchmark feature, others would be interested in your results.
If your video card needs overclocking, relax, video cards are much easier to overclock.

In Does DDR3 Memory Speed Really Matter with Ivy Bridge or Sandy Bridge?, I asked the question:

Assuming that the cpu is overclocked, do higher memory speeds improve real world performance when gaming, encoding or when running other intensive applications?

There weren't many responses that dealt with real world gaming or intensive applications when running with an Ivy or Sandy Bridge system.
Many owners are obsessed with running high memory speeds, but everything that I've seen indicates that 1600MHz is as good as it gets with real world programs.

When the o/c urge or itch strikes in the future, think about overclocking your memory to see if it improves the game play and/or gaming benchmarks.

I'm curious to see if they had any Prime95/AVX problems in the "Official" Z77X thread.
My final thoughts (for now) on Prime95/AVX testing:
1) Worker thread errors indicate system instability that should be addressed.
2) A Prime95 BSOD indicates a serious problem that must be addressed.

BTW, your TT & MSI forum signatures should be updated to show that you're using the 10.5 bios version.

[Zach_]

Updated my signature, will run some benchmarks and post my results and BIOS settings.

And also, notice that I updated my GPU in my signature, this is my new baby:



XFX Radeon 7950 Black Edition 3GB

Got a really good deal, traded my 6950 and added 150 USD, this GPU comes factory overclocked to 900MHz and reviews have proved that it can go as high as 1200 (!!!) MHz by overclocking and raising the voltage to 1.2 and still maintain reasonable temps.

Incredible :)