6700k - The perfect Core / Cache balance ?

[raw_blitz]

Dear community,

I've been trying to find the perfect OC for my I7 6700k running in a MSI M9 ACK Motherboard, water-cooled by a Kraken x61 system, coupled with 2x8GB DDR4 Corsair Dominator Platinum RAM sticks at 3333Mhz, running Win10 which is installed on a Samsung M.2 mSSD 256gb Hard drive. Next, to this, I also have a Titan XP graphic card for my 4k gaming (Battlefield 1 mainly).

Right now, as the title says, I'm running at 4.6Ghz core speed, with a Cache ratio of 46, at a voltage of 1.35v.

Next, I've done quite some stability test's using XTU (Intel Extreme Tuning Utility), and I'm happy to say that this OC is stable. with some pretty damn good temps as well.

(See pictures I shared with you)







Here comes my question: I've read that the ideal OC is when you're cache ration matches (or is 300-500Mhz under) you're core speed.

Will this give me better performance than if I would OC my 6700k to 4.7Ghz, and lower the cache ratio to 44, or maybe 43?

Right now in CPU-Z, when I run the benchmark and put an I7 6700k as a reference, I'm getting 10% more performance in Single Core benchmarking and in Multi.

Should I be happy with these results or should I try to get more out of it?

Big thank's to the community for helping out others in need! I'm just seeking for some opinions here!

 

[feelinfroggy777]

The cache ratio wont have nearly the impact that the core clocks will. If you can get your clock speed to 4.7 and your cache is above 4ghz, then you will probably get better performance than 4.6 on the core and 4.6 on the cache. I would say just to make sure your cache frequency is at 4.0 or above.

But you can always just benchmark it to check. Run a couple of cinebench benchmarks with the 4.6/4.6 and take an average score. The run a couple of benches with 4.7/4.0 (or whatever you can keep stable) and see what the differences are in scores.

Additionally, the XTU stress test is fairly conservative. What will be stable in XTU will be unstable in other stress test. But I have found, that if you can run a complete bench in Cinebench and your temps are find during the bench, then your system is pretty much 99% stable and there really is not point in running long stress test to beat up your CPU.

Now if you have mission critical work where stability is a must, the Prime95 is the gold standard for stress test. But it is not really a real world example as it will push your system harder than any other software that I am aware of. So I am not a big fan and you will have very high temps. With that being said, if your work is that mission critical, you probably should not be overclocking in the first place.

Lastly, your are gaming at 4k where the stress is all on the GPU and not the CPU. Increasing the clock speed from 4.6 to 4.7 (or even 5ghz), might give you 1 fps more, it may not give you any more fps. This is because at higher resolution the bottleneck moves form the CPU to the GPU. So it is up to you to how much time and thought you want to put into getting the extra 100mhz OC that wont have any noticeable impact on gaming.

 

[larrycumming]

generaly if you up the cpu cache, you get higher memory bandwidth benchmarks, which results in marginally better R15 or GB scores.

 

[Rodrigodrt]

A bit off topic but, I guess this could be of mutual use.

So some guides says that at higher cpu multis you also need to up the cache a bit, is that correct? I think you'd usually say that upping your cache speed would lead to instability. Confirms?

 

[larrycumming]

A bit off topic but, I guess this could be of mutual use.

So some guides says that at higher cpu multis you also need to up the cache a bit, is that correct? I think you'd usually say that upping your cache speed would lead to instability. Confirms?

cpu cache is generally that "uncore" part of the cpu with memory controller and other goodies. overclocking it is the same as cpu core overclocking you need to apply adequate voltage and some boards will let you set its own load-line calibration levels for stability.

you'll get very sexy aida64 memory benchmarks but i think gaming and applications wont show much improvement atmost say 1-3%?

 

[raw_blitz]

I'd like to know if that CPU-Z benchmark score I got (the 10% performance extra on single core and multi) is a lot, or not. Funny thing is, if I want to keep my vcore at 1.35v, I can't seem to go for 4.7Ghz core speed and 47 cache core :/

 

[larrycumming]

I'd like to know if that CPU-Z benchmark score I got (the 10% performance extra on single core and multi) is a lot, or not. Funny thing is, if I want to keep my vcore at 1.35v, I can't seem to go for 4.7Ghz core speed and 47 cache core :/

you need to raise voltage to the cache core as you increase its multiplier, which results in higher overall temperature for the cpu package and cpu core cuz they're thermally connected. I would raise the cache core just to get a sexy benchmark scoring but for everyday use I'd dial it down to something like max 40 min Auto and leave both cpu/cache on adaptive mode.

 

[raw_blitz]

@Larrycummin (love the name btw)

What if I play a lot of games and my GF used a video editor on it and that she sometimes exports videos that are like 60gb big ? I heared that upping the cache core will give you some extra boost in those sectors, IF true, why wouldn't I leave my cache core at 46, same as my core speed ? Temps are no issues there thanks to the kraken x61.

 

[larrycumming]

@Larrycummin (love the name btw)

What if I play a lot of games and my GF used a video editor on it and that she sometimes exports videos that are like 60gb big ? I heared that upping the cache core will give you some extra boost in those sectors, IF true, why wouldn't I leave my cache core at 46, same as my core speed ? Temps are no issues there thanks to the kraken x61.

well, yea, in that case leave them where they at then, as long as you have a stable system, but usually raising your core clock a 0.2ghz will give you a much bigger boost than raising the uncore/cache another 0.2ghz

I would raise the uncore and run aida64 memory bench to see where the sweet spot is for your memory kit

 

[Skipping-Bear]

i have 6700k and my cache is at 4.5ghz with my cpu being at 4.6 with a 1.280 vcore but i dont see any performance gains in gaming which is what i made my rig for when compaird to 4.2 cache

 

[feelinfroggy777]

I'd like to know if that CPU-Z benchmark score I got (the 10% performance extra on single core and multi) is a lot, or not. Funny thing is, if I want to keep my vcore at 1.35v, I can't seem to go for 4.7Ghz core speed and 47 cache core :/

For me, I dont go over 1.35 volts for a daily driver. Some people will go higher, but I dont think that is wise. You may be able to adjust your LLC and it will be stable, but expect higher temps.

 

[larrycumming]

Well, I like to keep them below 1.25V and below 75C (stress test)

 

[feelinfroggy777]

Well, I like to keep them below 1.25V and below 75C (stress test)

I know guys that will rock their Intel CPUs to 1.45 volts (think that is the Intel max voltage recommendation). Just way to high for my liking. As for temps, it depends on the stress test. I want my average load temps below 80C. Peak temps above is okay for me.

But if you are running Prime95, then you can have some silly high temps running at stock speeds and voltages. Which is another reason I am not a fan of Prime.

 

[undoubling]

Since "most" games are "optimized" for "standard" systems, and offer only MINIMAL gains through OC'ing, I would "think" that your time is better spent just optimizing your CPU/GPU AND Memory timings to MAXIMIZE performance. If this doesn't "work" to YOUR satisfaction, then by all means start OC'ing your system---------but ensure that your CPU/GPU values are compatible with each other, and that your memory won't impede performance.