Intel Turbo Boost VS Manual Overclocking ?

[BLUE-ANGEL]

Hi fellow carb members,

So I watched this thread LINK from Hardware Unboxed when I noticed that I can actually look at the intel i9 9900K running stock vs 5.0GHz overclocked to see the performance difference, because of HU running the benchmarks to compare the AMD vs INTEL performance they obviously show both for both cpu's. I have noticed that there is nothing really great about the performance boost gains when the i9 is running stock compared to overclocked to 5.0GHz. It seems that the Intel Turbo Boost performance is actually pretty impressive by the looks of things. So why overclock and run at insane temps to just say I am running at 5.0GHz, is there really a big difference in performance? Is there maybe a bigger difference at 2k or maybe even 4K?

I would like to know if anyone else have some good feedback on this and to share their experience.
What is your thought's on the Intel Turbo Boost when you run stock vs OC on your cpu and compare the performance difference when bench marking the cpu stock vs overclocked? Is it maybe just the i9 9900K that shows this little improvement? I might be very wrong to think that there is after all not such huge difference in performance and for me having a 5-7 fps loss is not the end of the world, but 15-20 is a different story yes.

 

[pXius]

For gaming the difference is meh. Remember that Turbo Boost is also time limited, so if you run a workload that'll pin the cores at 100% it will eventually drop off quite a bit. These are also the workloads that will benefit from a manual OC the most. The manual OC will keep all the cores at 5Ghz+ indefinitely or until the workload finishes. Photoshop renders for instance scale very well with all core OC's on the 9900k.

Games generally don't load up the CPU in such a way where you'll really see a huge benefit.

Why overclock? Free performance, for the love of benchmarking, some of us find 3DMark more entertaining than Witcher 3, enthusiast mindset, I donno.
If your temps aren't out of hand and you're not being loadshed, then why not. It's fun. xD

The 9900K is a weird chip that really doesn't fit anywhere in the market right now. The 9700K can match it's gaming performance for cheaper and most of the AMD lineup wrecks it in multi-threaded workloads. It was a king when it launched, but I can't see it as viable in any new system right now.
That said, the second hand market has a completely different value scale when it comes to system building and it'll have it's place at the right price point for years to come. Think 4790k.

 

[dainluke]

Some turbo voltages are absolutely ginormous. If I let my CPU run at stock it'll turbo to 4.9 at like 1.35V or something ridiculous. My chip doesn't do 4.9 even at like 1.325. So I set it at 1.25 with a few different LLC and AC/DC settings and it runs absolutely perfectly at 4.8 all core. My CPU runs at 38 idle and around 55-60 when gaming as I have an AIO GPU now. My bedroom is a furnace and there is no air con. I also can't open a window because there's a wasp nest outside.

 

[BLUE-ANGEL]

Yeah I would think for someone like myself who primarily just use my pc for gaming and audio production ( Not video editing or any streaming ) it might just be best to keep my settings at stock. I was a bit skeptic and contemplating about intel vs AMD and whether it might be better to change from i9 to 3900 chip and looking at the performance in gaming between the two I might just rather stay with the Intel for a few years and and upgrade at a much later stage.

 

[BLUE-ANGEL]

Some turbo voltages are absolutely ginormous. If I let my CPU run at stock it'll turbo to 4.9 at like 1.35V or something ridiculous. My chip doesn't do 4.9 even at like 1.325. So I set it at 1.25 with a few different LLC and AC/DC settings and it runs absolutely perfectly at 4.8 all core. My CPU runs at 38 idle and around 55-60 when gaming as I have an AIO GPU now. My bedroom is a furnace and there is no air con. I also can't open a window because there's a wasp nest outside.

I can relate, I am on air cooler and I run at 4.8GHz @ 1.3V and I keep my LLC and AVX at stock so I have a continuous 4.8GHz regardless the load or AVX demand on the cpu. I also idle at about 36-40 degrees and run on good day when gaming on ULTRA between 60 - 65 degrees and on hot days I do 70-75 degrees. I tried to do more but an additional 100Hz wants too much volts and shoots the temps up.

 

[dainluke]

I can relate, I am on air cooler and I run at 4.8GHz @ 1.3V and I keep my LLC and AVX at stock so I have a continuous 4.8GHz regardless the load or AVX demand on the cpu. I also idle at about 36-40 degrees and run on good day when gaming on ULTRA between 60 - 65 degrees and on hot days I do 70-75 degrees. I tried to do more but an additional 100Hz wants too much volts and shoots the temps up.

That's so weird how we both got similar binned parts for the same components XD. LLC at stock is actually usually either conservative or too high depending on the MB manufacturer. With my current board I run LLC on Turbo and it actually keeps my voltage perfectly level at 1.25 with some spikes up to 1.26 ish when running a stress test or any CPU benchmark. I actually find it helps stability and allows me to run a lower idle voltage. AVX offset I also have at 0 though.

 

[BLUE-ANGEL]

That's so weird how we both got similar binned parts for the same components XD. LLC at stock is actually usually either conservative or too high depending on the MB manufacturer. With my current board I run LLC on Turbo and it actually keeps my voltage perfectly level at 1.25 with some spikes up to 1.26 ish when running a stress test or any CPU benchmark. I actually find it helps stability and allows me to run a lower idle voltage. AVX offset I also have at 0 though.

Apologies, I just had a look and I see I have my LLC at mode 4 (MSI mobo) which is flat-line for load and volts to be equal regardless. Also I am running at 1.29volts at 4.8GHz OC and not 1.3volts. I just did both test on Aida64, CPU and FPU separate tests, and I will do the same for stock settings now to see what my difference will be and I will post it in the thread to compare Intel Turbo Boost and manual OC. With you having an AIO that must be much more beneficial for overclocking.

 

[dainluke]

Apologies, I just had a look and I see I have my LLC at mode 4 (MSI mobo) which is flat-line for load and volts to be equal regardless. Also I am running at 1.29volts at 4.8GHz OC and not 1.3volts. I just did both test on Aida64, CPU and FPU separate tests, and I will do the same for stock settings now to see what my difference will be and I will post it in the thread to compare Intel Turbo Boost and manual OC. With you having an AIO that must be much more beneficial for overclocking.

Air cooling is fine provided it's a decent air cooler. My AIO itself is not the most groundbreaking in terms of performance. I'd say 1.29 for 4.8 might be a tad bit high but it could just be binning. That or perhaps my CPU isn't stable. I'm fairly confident it is though.

 

[BLUE-ANGEL]

I have a pretty decent air cooler, but I am also kinda soaking my tower with hot air...room ambient temp is about 28 degrees average hence I don't get the greatest temps. I know if I could drop a little bit more on the temps I could probably settle down by another 0.2 volts, but for now the guy wants 1.28 - 1.29 volts. I found that games spike the volts much more than any benchmark programs so I can pass all the tests on 1.28 volts but I still get a random game crash, but not on 1.29 volts.

So here are the findings I just got for the different results. On manual OC the clocks ran stable all the way for the CPU test at 4.8GHz. The FPU test adjusted both manual OC and stock clocks down to about 4.5GHz hence the difference in the temps and volts. Not huge difference though, so I assume that Intel Turbo Boost is NOT bad at all. I guess when you run 5.0GHz manual OC there will definitely be a bigger difference also having a decent AIO to keep the 5.0GHz OC. I will test this in gameplay to see how the clock speed and temps are effected.

For the average gamer I think Intel Turbo Boost will do great if you don't have the urge to manual OC or don't have a decent cooler or even just don't want to bother with overclocking, but for the guys who do want to get every bit of juice out of their lemons, it definitely seems worth it.

All 4 tests are done in Aida64 for about 7min. I know there is a difference between for ex. Prime95 vs Aida64. but this for me is fine to see the difference.

Tests done in following order:
1.) Stock settings CPU test
2.) Stock settings FPU test
3.) Manual OC settings CPU test
4.) Manual OC settings FPU test


Images

 

[BLUE-ANGEL]

Just to add, I did a quick in-game test to see what Intel boost does to the cpu regarding performance. It actually took the cpu speed down to a stable 4.7GHz in the game. I barely noticed a real difference in fps between 4.7 and 4.8GHz. So I assume the AVX load from the game itself was a decent demand for the cpu to auto set the AVX by -3 with Intel boost and the auto settings going from the 5.0 to 4.7GHz. So I guess this is a sign that yes the clock speeds are not impressively weaker than what the average joe like myself can achieve with overclocking to 4.8GHz for stable clocks on air cooler.

I know that you can actually achieve 5.0GHz on a air cooler, but then your room ambient temp needs to be about 18 degrees with a decent airflow through your case.

Also, I found about a 200-300 score difference in Timespy for CPU score as well. To me it looks like the real gains is really only happening when you actually do overclock to 5.0GHz or more to see an impressive gain compared to Intel Turbo Boost with Auto settings.