1st time OCer * Done some reading * Have some questions.

[LeperMessiah]

I've been reading a lot of guides and watching videos and read many threads and I have some specific questions I would like to clarify if anyone is up for the task.

First off my specs:
Intel i5-3570K 3.4 Ghz
Gigabyte GA-Z77-D3H board
8 GB Ram
850w Power Supply
Asetek 510LC Liquid cooling
GeForce GTX650

Just to give you some sense of my goals: I'd like to clock it up to 45, and mainly I'm a casual gamer and photo editor, but mostly the computer will be doing simpler tasks.

Questions:

1) What is the downside of using the Intel TurboBoost? As I understand it, setting the cores to something like 45/45/44/43 is more power efficient and it will only run it up to 4.5ghz when there's a demand... the rest of the time (like for browsing) it will run closer to 1.6ghz. Also the voltage can be automatically adjusted, specified by my Offset Voltage (eg. .015). The alternative is having all of the cores running, say, at 4.5ghz all the time and the voltage being at a fixed rate (unless I change it). I'm a casual gamer, so which is better, using TurboBoost or using separate fixed profiles for gaming/photo editing and for more basic computing.
Subquestion: Should TurboBoost OCing be benchmarked in the same way as manual OCing? For example, would I start TurboBoost at 38/38/38/38, benchmark it, and then increase it incrementally from there?

2) I've heard that Prime95 is too aggressive for Ivy Bridge processors and a program called AIDA64 is better. Is that true? What about IntelBurnTest?
Subquestion: Some people say you only need to benchmark for 15/30 min, some say an hour, some say 12/24 hrs. What gives? lol

3) Is Gigabyte EasyTune useful at all, even if just for benchmarking different profiles? I haven't heard much anything good about it, but I like the idea of changing settings in real time without the need to reboot.

Sorry for being wordy, I get the sense that there is a shorthand for what I'm describing here but I don't know the jargon, lol. Thanks!

 

[I.M.O.G.]

Welcome to the forums.

I took a stab at getting you started on your questions...

1. The downside is initial Overclocking. When the frequencies and voltages are set statically, without fluctuating, it is easier to find what is stable and what is not. After you have dialed in your overclock and you know what settings work with static frequency and voltage, there is nothing wrong with using power savings features or dynamic frequencies. They may effect your OC, and most people recommend turning these off at least while testing.

I would find your max stable that you are comfortable with, based on voltage and temperature, then I'd consider setting up turbo profiles if I was interested in saving power.

2. Prime95 being too aggressive is phooey. I disregard those kinds of comments. I always use Prime95. Used it years ago, still use it today - it works for my OC routine. I suggest using what you like best. I like Prime95 because it has settings to stress primarily memory, primarily CPU, or both - that helps in my experience.

Benchmark/stability testing is up to you. I usually just do it for a short period to see if the overclock seems stable even when the chip gets warmed up - but it could very well crash later. The longer you stability test, the less likely you are to encounter unexpected issues - so this is mainly a question of what you are comfortable with.

3. I use EasyTune. It can be buggy, and I prefer using Gigabyte Tweak Launcher. Either one is helpful in finding settings and testing stability quickly, but for daily clocks once I find what works I always set it in BIOS.

 

[EarthDog]

+1 to IMOG. I can elaborate a bit more on the P95 thing too...

I had a discussion with....cant think of his name now - a reviewer from another site that posts here - and he said the same thing. Its what ASUS(?) told him. Being me, I pushed for some more information and proof and it was provided...Basically, P95 and what it tests (as in CPU instructions sets) hasnt been updated in years, while AIDA64 is updated and uses all the instruction sets like SSE3,4, etc. AIDA64 may not heat up the processor quite as much as P95, but it seemingly stresses the heck out of EVERYTHING. FWIW, since that information and the rare times I need to use such applications, AIDA64 would bomb out quicker than P95 at the same settings...(yes that is anecdotal).

 

[I.M.O.G.]

Thanks for that comment Earthdog, thats interesting, and I hadn't heard actual facts behind the differences before. Very informative.

 

[EarthDog]

NP.. and sorry I dont have the name of the reviewer, I have been scouring the site trying to find something more than just that, LOL.

I dont think P95 is overkill regardless (but IBT, OCCT, Linx, etc I feel are).

 

[LeperMessiah]

NP.. and sorry I dont have the name of the reviewer, I have been scouring the site trying to find something more than just that, LOL.

I dont think P95 is overkill regardless (but IBT, OCCT, Linx, etc I feel are).

I think I saw the same thing as you. It was a Newegg video review /overclocking tutorial of an Asus motherboard. It was the guy from Asus who said AIDA was better for testing the Intel 3rd gen's capabilities and that other stress tests, like P95, might damage the processor. The guy from Newegg was Paul and the ASUS guy was JJ.

 

[LeperMessiah]

Welcome to the forums.

I took a stab at getting you started on your questions...

That's very helpful, thank you. I actually don't even have the PC yet, it arrives on Friday... so I'll test things out then and post back with my experience.

It's funny; overclocking isn't one of those things where it's difficult to find information about how to do it, the problem is that there is SO MUCH info out there and so many different opinions, it's hard to know where to start.

Thanks for the replies, I really appreciate it.

 

[EarthDog]

We know JJ well here... and that is exactly who that reviewer heard it from as well.

I dont buy in to P95 damaging IB (tons of people still use it - direct report of chips degrading/dieing b/c of it) or any other CPU, but I do think AIDA64 is a more complete test for those reasons he listed.

 

[EarthDog]

Sorry to double post, but I found the information. It was from a member here, Eldonko and an email from JJ.

Hello Zac,

Yes my recommendation is to not exceed 1.35vid not so much from
degradation but this is a possibility. More so the CPUs and there
tighter gate and lithography do not like high voltage levels. That being
said does not mean when cool sufficiently you can not push the chip. All
things considered overclocking is overclocking and all aspects should be
taken into consideration. Currently I have not seen degradation on our
end with our samples but then again I am sensible about how I test. HQ
has of course pushed outside sensible usage but variance in CPU quality
and steppings also account for some sketchiness.

As for unvalidated test applications this is just a question of driving
high load and high voltage and therefore potential causing damage to the
CPU. Overall Aida is advised as it is the only validated stress test
application for this platform. As always the use of applications like
OCCT, Prime, LinX run without the insight behind all aspects of the CPU
architecture.

For high level stress testing internally we use Intel PTU program the
advantage is we can specifically define load therefore being sensible
about how we push the CPU voltage and thermal wise.

In regards to difference a lot of it is on the CPU specification
support.

Instruction set usage is poor in many applications like P95 or LinX (
unless you have an updated LinPack Library ).

Things like AvX or AES and latest SSE instructions and more do not get
tested. While you may get a pretty solid result from testing it is not
testing all parts of the CPU.

Additionally P95 and LinX do not test subsystem interaction and this is
important as you can have issues with stability once interaction of data
occurs between the GPU and the storage array ( configuration ). As such
I prefer to use Aida as it allows for testing of parameters

CPU, Memory, GPU, HD

Additionally while I can understand the rational of doing "extreme
testing with applications like P95" it is un realistic as the max draw
is different as such efficiency is compromised to a farther degree
because you are defining oc variables based on conditions you will not
use.


Examples are below. ( on sandy bridge e )

P95 current build and unvalidated for Core Series processor approx
130watt pull
P95 new build with cpu support for Core Series is approx 145watt pull
plus it increase IA core power consumption
A standard 3D Game -approx 60-70watt pull
Normal desktop usage ( Multiple tabs in a browser, streaming, media
player, document editor, photo viewer etc ) 35 to 65 watt pull
Aida64 approx 110-115wat pull
As such to me it does not make sense to potentially validate with an
application 2x the delta of max usage of a system

No if someone is doing a more consistent high performance workload (
Audio, Video Editing, 3D Render , Folding, ) then you watt pull can be
similar in some cases even exceed P95 in those cases though due to the
specific workload patterns I advise validating with the application that
is actually being used.

Hope this provides some clarification.

Best Regards,
JJ

 

[eskimochaos]

How can a piece of software physically damage a piece of hardware? Temp/voltage independent of course. It makes no sense.

 

[ivanlabrie]

I think he meant it's overkill to test that kind of load if not using it which I agree.
But, there's one point I digress, and it's the extensions.
Prime95 ver 27.7 uses AVX extensions and the like and is optimized for Core series cpu's, so it is equally or more demanding than Aida64 and the like...I just like it more, it's more standarized, you can do custom tests for specific workloads and it's free.
Can't beat that.

Also, I have a 4.7ghz folding stable, bullet proof stable system, and found it using Prime95 custom tests for 30 min, and then running some blend for the heck of it.