Overclock.net banner

1 - 12 of 12 Posts

·
Registered
Joined
·
2,591 Posts
Discussion Starter #1
Hello, I just got my new CPU back from Intel as my last 7700K was defective. I didn't overclock my other CPU because I didn't have it long enough to do so. My question is, I have an Asus Z170 and it has auto overclock options. Well, my new CPU is overclocked to 4.58ghz would it be any benefit gaming to overclock it anymore or no? Would I see any usable gains say going 4.8 from 4.58? It's liquid cooled btw.
 

·
Registered
Joined
·
3,004 Posts
I would first start off with a manual overclock. Auto over clocks tend to use a lot more voltage then is needed resulting in unneeded high temps. After that overclocking is gonna be limited to temps or your cpu voltage hitting a wall.
And it depends on game if you would benefit.
If the game is not gpu bound then you would see gains.
Some games like more cores and others like higher clock speeds.
Games that like clock speeds us my also like higher ram speeds.
 

·
Registered
Joined
·
5,374 Posts
I'd reset to defaults and OC from scratch. It would bother me to know that auto is always giving my CPU way more voltage than it needs. Especially to the whole socket.
Plus if you can get 5ghz off the voltage already being sent to your CPU then why not run 5ghz? Or keep the current clock and back off the voltage for a cooler chip.
Pick a voltage you're comfortable with (manual for now) then sync your cores and bump the multiplier. Keep bumping the multiplier till Windows crashes then back down 100mhz. Now you need to see if it's stable, if it's not I'd raise the voltage by 0.02 (assuming thermals allow). If adding a few hundredths doesn't help just back off the multiplier. Now you've got a quick and dirty OC ready to stress.
I prefer to stress test with real world loads so I use handbrake for my final stress test but before I do that I'll run a synthetic like Aida64 so hopefully it'll crash faster.
Your voltage is still manual so you'll need to switch to adaptive and set the voltage you used manually. If HWInfo is reading something way too high you'll need a negative offset. Say you set 1.25v and only need that 1.25v but HWinfo is showing 1.30, you'd want to use an offset of -0.05. After you get that synced up I'd run Handbrake again just not for as long.
After you're as stable as can be at your desired speed you can bump up cache and ram.

A much better and more thorough explanation was written by Darkwizzle and if you're even thinking of overclocking (even using the auto OC crap) you should read his guide:
http://www.overclock.net/forum/5-intel-cpus/1621347-kaby-lake-overclocking-guide-statistics.html

Not only are there directions for the OC but there's many explanations pertaining to stuff like C-States, stress testing, what should be disabled/enabled, etc.
If you read it a few times you'll have a much better understanding of what you're doing and why.


My personal thing on overclocking is to take it just under the point where you start needing big jumps in vcore to get stability. Many of us are using vcore very close to stock and getting better temps to boot (delid).

And yes, overclocking your CPU will help games/performance almost always especially in CPU bound games and when you're going for high frame rates.
You might not get more FPS but faster cores, cache and ram can significantly increase your MINIMUM FPS which IMO is the most important one.

Good luck!
 

·
Registered
Joined
·
3,747 Posts
For games with the i7 7700k I have been doing the math from the reviews on the net and you will see 0 to 4% at 4.8GHz.
 

·
Registered
Joined
·
5,374 Posts
For games with the i7 7700k I have been doing the math from the reviews on the net and you will see 0 to 4% at 4.8GHz.
Was that on minimums, average or maximums?
High frame rate 1080p where the CPU needs to draw frames as fast as possible or 4k where the CPU barely matters?
Or in massive open world games with lots of players and locations to calculate?
Whether it'll help or not depends on so many things, it's hard to just say yes or no without knowing the use case.
Some console ports are never coded to take advantage of CPU power so it's likely you won't see an advantage in those. GTA really loves CPU power but if it's the GPU that's holding you back then an OC will likely only bring up minimums which again is the most important thing for me.
 

·
⤷ αC
Joined
·
11,240 Posts
It's not so much about clockspeeds as direct control over the voltage and power.

If you use manual voltage and leave C-states on , your CPU will actually be more efficient than a stock CPU provided you don't overvolt it much from stock volts.

The way I see it is :
1) take your stock voltage and clocks (no automatic multicore overclocks , so disable MCE or auto boosting all cores to turbo from motherboard vendor)
2) see how high you can clock on stock voltage
3) see how much of an overclock you can get with 0.1V or <10% more voltage (say 1.3 from 1.2V or about 1.35 from 1.25V)
4) If cooling allows, see how much you can get with safe voltages (so about 1.4V typically provided your LLC isn't ridiculous) at the socket <--- don't do this if you aren't measuring at socket or at least on the board
5) Do a clocks vs power utilization test with a normal application that uses no AVX instructions (Cinebench R15 for example or prime95 26.6) and then one that uses AVX (such as x265 encoding, Linpack, OCCT AVX, or Prime95 AVX)
6) Do additional testing with looping CPU+GPU at the same time (for example x265 encode + Unigine Heaven on loop)
7) Start BOINC on both CPU+GPU at the same time with a quorum 2 project (validation by another PC). If you error out WUs (not from poorly made driver crashes , but outright wrong results) then it's not 100% stable for all purposes. Folding works as well, but I don't believe they provide you that level of information on workunits. It simply provides checkpointing and digital signature verification.

Do a work/power (performance per watt) curve. Find the most performance per watt , ideally within the power envelope of the CPU design (so for a LGA1151 CPU it would be ~91W) and your CPU cooler (if it's a H100i that's not going to be a limiting factor so long as you keep fan speeds reasonably high).
 

·
Registered
Joined
·
5,374 Posts
It's not so much about clockspeeds as direct control over the voltage and power.

If you use manual voltage and leave C-states on , your CPU will actually be more efficient than a stock CPU provided you don't overvolt it much from stock volts.

The way I see it is :
1) take your stock voltage and clocks (no automatic multicore overclocks , so disable MCE or auto boosting all cores to turbo from motherboard vendor)
2) see how high you can clock on stock voltage
3) see how much of an overclock you can get with 0.1V or <10% more voltage (say 1.3 from 1.2V or about 1.35 from 1.25V)
4) If cooling allows, see how much you can get with safe voltages (so about 1.4V typically provided your LLC isn't ridiculous) at the socket <--- don't do this if you aren't measuring at socket or at least on the board
5) Do a clocks vs power utilization test with a normal application that uses no AVX instructions (Cinebench R15 for example or prime95 26.6) and then one that uses AVX (such as x265 encoding, Linpack, OCCT AVX, or Prime95 AVX)
6) Do additional testing with looping CPU+GPU at the same time (for example x265 encode + Unigine Heaven on loop)
7) Start BOINC on both CPU+GPU at the same time with a quorum 2 project (validation by another PC). If you error out WUs (not from poorly made driver crashes , but outright wrong results) then it's not 100% stable for all purposes. Folding works as well, but I don't believe they provide you that level of information on workunits. It simply provides checkpointing and digital signature verification.

Do a work/power (performance per watt) curve. Find the most performance per watt , ideally within the power envelope of the CPU design (so for a LGA1151 CPU it would be ~91W) and your CPU cooler (if it's a H100i that's not going to be a limiting factor so long as you keep fan speeds reasonably high).
I tend to agree with this method as well.
That's why I posted the guide as a lot of this is explained there.
I'd like to point out that after switching to Windows 10, using manual voltage and C states no longer worked and I had to use adaptive. 4690k on an Asus Sabertooth MKII.
I haven't retested after the BIOS update but without a GPU yet it doesn't really mater yet.
 

·
Registered
Joined
·
3,747 Posts
For games with the i7 7700k I have been doing the math from the reviews on the net and you will see 0 to 4% at 4.8GHz.
Was that on minimums, average or maximums?
High frame rate 1080p where the CPU needs to draw frames as fast as possible or 4k where the CPU barely matters?
Or in massive open world games with lots of players and locations to calculate?
Whether it'll help or not depends on so many things, it's hard to just say yes or no without knowing the use case.
Some console ports are never coded to take advantage of CPU power so it's likely you won't see an advantage in those. GTA really loves CPU power but if it's the GPU that's holding you back then an OC will likely only bring up minimums which again is the most important thing for me.
It depends on the game if you see 0 to 4% maximum gain at 4.8GHz.
 

·
Registered
Joined
·
147 Posts
Everyone...

So i have the following...

Gigabyte G1 Gaming Z-170 mobo and paired it with 7700k, i brought the CPU just recently as i heard the horror story about temps...

So i though lets overclock, so i set CPU 1.26 or 1.28 volts and upped to 4.8Ghz and turned turbo off..

Too my surprise im idling at 42c with Corsair H100i and load was about 75c at max...

Is this good ? as i had a 6700k before on the same mobo and that thing was a beast at 4.8Ghz with same setup..

I know before you say why bother go up to the 7700k and not get the 8700k, i got the 7700k for a steal and i hate intel not allowing us Z170 owners since its really the same chipset to use the 8700k... Intel reall F ed up there...

I was thinking temps should be higher, is it worth going to the magic 5.0ghz range...

One thing is noticed that the memory worked perfectly with the XMP profile of 3200Mhz unlike before with the 6700k it would not, maybe it could be the bios update i did to allow the 7700k to work with my board that helped...

Anyways after reading this, am i good at 4.8ghz or should i go for the elusive 5.0ghz range...

Thanks all...
 
1 - 12 of 12 Posts
Top