6700k - Is this the ideal mix of core and cache?
6700k - Is this the ideal mix of core and cache?
I've been searching for the ideal OC for my I7 6700k running on an MSI M9 ACK motherboard, water-cooled with a Kraken x61 system, paired with two 8GB DDR4 Corsair Dominator Platinum RAM sticks at 3333Mhz, all operating on a Samsung M.2 SSD with 256GB storage. Additionally, I have a Titan XP graphics card for my 4K gaming needs (mainly Battlefield 1).
As the title indicates, I'm currently running at 4.6Ghz core speed, with a cache ratio of 46, and operating at 1.35V.
I've conducted several stability tests using XTU (Intel Extreme Tuning Utility), and I'm pleased to report that this configuration remains stable, achieving impressive temperatures as well. (Refer to the pictures I shared.)
My question is: does matching the cache ratio with your core speed provide better performance compared to increasing the core speed to 4.7Ghz and reducing the cache ratio to 44 or 43?
In CPU-Z, when benchmarking with an I7 6700k as a reference, I'm seeing about a 10% improvement in single-core performance and better results in multi-core tests. Should I be satisfied with these outcomes or continue optimizing further?
Thank you to the community for your support and insights! I'm just looking for some feedback here.
Because the majority of games are designed for standard systems and provide only small improvements via overclocking, I believe your time would be more effectively used by fine-tuning your CPU/GPU settings and memory timings to get the best performance. If this doesn’t meet your expectations, feel free to start overclocking—but make sure your components are compatible and your memory doesn’t slow things down.
The cache ratio won't match the effect of the core clocks. If your clock speed reaches 4.7 and your cache exceeds 4ghz, you're likely to achieve better performance than 4.6 on the core and 4.6 on the cache. I’d recommend ensuring your cache frequency is at least 4.0.
However, you can always run benchmarks to verify. Execute a few Cinebench tests using 4.6/4.6 and average the results. Try a few with 4.7/4.0 (or any stable setting) to compare score differences.
Moreover, the XTU stress test tends to be quite cautious. What remains stable in XTU may become inconsistent in other tests. Still, I’ve noticed that if you can complete a full Cinebench run without significant temperature spikes, your system is likely very reliable—so long benchmarks are usually enough to confirm stability.
If your application demands absolute reliability, Prime95 serves as the benchmark of choice. However, it simulates extreme conditions not typical in everyday use, so pushing your system this hard may lead to excessively high temperatures.
In mission-critical scenarios where uptime matters most, Prime95 is the preferred option. But it’s not a practical representation of real-world workloads, as it will stress your hardware more than any other software I know. Therefore, unless stability is non-negotiable, you might want to avoid aggressive overclocking.
In summary, for gaming at 4K resolution, the GPU becomes the main bottleneck rather than the CPU. Raising the clock speed slightly from 4.6 to 4.7 (or even 5GHz) could marginally improve performance, but it’s unlikely to noticeably boost frame rates. This is due to the CPU-GPU handoff at higher resolutions. It ultimately depends on how much effort you’re willing to invest to achieve a small gain without affecting your gaming experience.
Generally, increasing the CPU cache leads to improved memory bandwidth measurements, which can slightly enhance R15 or GB scores.
Some instructions suggest increasing the cache size when using higher CPU speeds, but it's generally advised to avoid this as it may cause instability. Confirms?
Rodrigodrt :
A bit off topic but, I suppose this could be useful for both. Some guides suggest that with higher CPU multis, increasing the cache is also necessary, isn't it? I believe you'd usually say that boosting your cache speed might cause instability. Does that match what you know?
CPU cache refers to the "uncore" section of the CPU, including the memory controller and other components. Overclocking it requires sufficient voltage, and certain boards allow you to adjust load-line calibration for stability.
You'll see impressive AIDA64 memory benchmarks, but gaming and applications probably won't notice much improvement—maybe just a few percent at most.
It's interesting how your CPU-Z benchmark result compares—especially with your vcore target. It seems the score you achieved is quite high, but balancing it with your vcore goal might be tricky.
I'm curious about the benchmark score you received. The 10% boost across single and multi-core is impressive, but if you're aiming for a stable vcore of 1.35v, it's challenging to push the core speed to 4.7Ghz with 47 cache cores. You'd likely need to increase the voltage for the cache core to handle the higher multiplier, which raises the overall temperature and affects performance. For regular use, it's better to keep the settings at a more balanced level, like max 40 min Auto, and let both CPU and cache operate in adaptive mode.
What if I play a lot and my girlfriend uses a video editor that sometimes exports huge files? I've heard increasing the cache core could help with those large sizes. If it really works, why wouldn't I keep my cache at 46, just like my main speed? The performance shouldn't be affected because of the Kraken X61.
Here’s a revised version of your text:
What if I frequently play games and my girlfriend uses a video editor that sometimes exports massive files, around 60GB? I’ve heard increasing the cache core can provide an extra boost in those areas. If that’s true, why wouldn’t I keep my cache core at 46, matching my core speed? The performance issues shouldn’t be a problem thanks to the Kraken X61.
Well, yes—if you do that, just leave them where they are, as long as your system is stable. Generally, boosting your core clock by 0.2GHz will offer a much stronger improvement than increasing the uncore or cache by another 0.2GHz. I’d suggest adjusting the uncore and running an AIDA64 memory benchmark to find the optimal settings for your setup.