[SOLVED] My brand new i7-10700k isn't performing as well as it should. Need help

[klct555]

Hi all,
Last Sunday, I built a new machine and bought an i7-10700k and an NZXT Kraken M22 to cool it.
I wanted to run some benchmarks and I couldn't help but notice I am underperforming.
My first run on Cinebench R23 Multi core was a 12070 and my second was a 12323. Sources online say it's supposed to be a 13302. I also tested my entire system on UserBenchmark and my CPU scored 97.3%, which is in the 38th percentile.
XMP is on and I didn't touch anything in the BIOS related to my CPU, and I am achieving the base turbo speed of 4.7. For reference, I idle at roughly 36-41C and during the Cinebench benchmarks I am at about 80C. COD Warzone nets me low to mid 60's.
My CPU usages before running the benchmarks is nothing abnormal, with just a few programs running.
Any help would be appreciated, thanks

 

[Phaaze88]

A breakdown of sorts:
1)Your sources need to be more specific.

2)Userbenchmark doesn't even count because the programmers never set a baseline for performance comparisons.
Example: Instead of comparing your 10700K to an Intel stock 10700K, it's being compared to ALL 10700Ks, overclocked or not. You can't make an accurate comparison like that.

3)Intel's 10th gen is a double edged sword.
Performance is good, thermals are also good.
But the Intel power limits are a bit conservative. If you want to get the most out of them - if you care about that min-maxing stuff - then you'll have to raise/remove the power limits... only to then see the cpu thermals literally go to hell.

4)Kraken M22.
Just forget about even trying to tinker with power limits with this cooler. It's too small, plus that unit in particular has a high failure rate.
The pump in radiator design didn't seem to work too well for NZXT.

5)Be happy with the performance you're getting now - the tradeoff in higher power and thermals practically isn't worth the gains afterwards.
The diminishing returns from overclocking/removing power limits has gotten worse over the generations(6th - 10th).

6)Ram frequency matters some.
It's a common occurrence to forget to enable XMP.

 

[klct555]

A breakdown of sorts:
1)Your sources need to be more specific.

2)Userbenchmark doesn't even count because the programmers never set a baseline for performance comparisons.
Example: Instead of comparing your 10700K to an Intel stock 10700K, it's being compared to ALL 10700Ks, overclocked or not. You can't make an accurate comparison like that.

3)Intel's 10th gen is a double edged sword.
Performance is good, thermals are also good.
But the Intel power limits are a bit conservative. If you want to get the most out of them - if you care about that min-maxing stuff - then you'll have to raise/remove the power limits... only to then see the cpu thermals literally go to hell.

4)Kraken M22.
Just forget about even trying to tinker with power limits with this cooler. It's too small, plus that unit in particular has a high failure rate.
The pump in radiator design didn't seem to work too well for NZXT.

5)Be happy with the performance you're getting now - the tradeoff in higher power and thermals practically isn't worth the gains afterwards.
The diminishing returns from overclocking/removing power limits has gotten worse over the generations(6th - 10th).

6)Ram frequency matters some.
It's a common occurrence to forget to enable XMP.

1. Sources? Can you clarify?

2. A buddy of mine has the same exact processor as me with a similar cooler and he performed "as expected" on UserBenchmark, a 103%.

3. I haven't touched anything like that on my CPU, and I don't plan on it. I would just like it to perform as expected while stock.

4. Agreed. I should have mentioned I have no plans on overclocking. I know 120mm AIO's aren't good for it, but I can't lie, I'm a sucker for good-looking AIO's and I didn't want to spend a lot of money so I bought it.

5. Again, buddy of mine has his 10700k performing as expected on UserBenchmark and Cinebench with 120mm AIO.

6. XMP is on and my RAM is running at 3600Mhz as it should.

 

[Phaaze88]

1)My first run on Cinebench R23 Multi core was a 12070 and my second was a 12323. Sources online say it's supposed to be a 13302.
This part needs clarification. Under what settings(bios, Windows, drivers, background applications, etc) did you achieve your score Vs the settings your sources used?
Have everything set up like they and your buddy did, and things may change.

2)Your bud also could be running different drivers, bios settings, power plan, ram speeds and so on. It's unfortunately not the same as just having the same cpu.

3)It's a brand new system. If you've already updated all the necessary drivers for the mobo/Windows/other devices, enabled XMP for the ram... then you're already getting stock performance.

4/5/6)Ok.

 

[klct555]

1)My first run on Cinebench R23 Multi core was a 12070 and my second was a 12323. Sources online say it's supposed to be a 13302.
This part needs clarification. Under what settings(bios, Windows, drivers, background applications, etc) did you achieve your score Vs the settings your sources used?
Have everything set up like they and your buddy did, and things may change.

2)Your bud also could be running different drivers, bios settings, power plan, ram speeds and so on. It's unfortunately not the same as just having the same cpu.

3)It's a brand new system. If you've already updated all the necessary drivers for the mobo/Windows/other devices, enabled XMP for the ram... then you're already getting stock performance.

4/5/6)Ok.

1. Got it

2. We have the same mobo, except mine is the slightly higher spec with wifi and such, so we have the same BIOS and he also didn't touch anything in his BIOS related to CPU. I also have faster ram. 3600 MHz vs. 3200 MHz. We have different PSUs but same wattage. Also, I never thought about the power plan, I totally forgot to update that when I got the new build. Will change it and see how it is. Are there any drivers that could possibly affect CPU performance?

 

[Phaaze88]

We have the same mobo, except mine is the slightly higher spec

Well, there's one variable right there. Some models aren't shipped with Intel's stock bios settings and are tweaked for a little extra performance - but no one would know it unless they looked for it.

I also have faster ram. 3600 MHz vs. 3200 MHz.

That's another variable. Frequency isn't everything though; the timings are another variable - lower timings are better.

Also, I never thought about the power plan, I totally forgot to update that when I got the new build.

It should be on the balanced plan by default.

Are there any drivers that could possibly affect CPU performance?

Yes.
Bios: some updates contain bug, security, and compatibility fixes, as well as performance enhancements.
Windows: Windows will be Windows... don't fall behind on updates and clean install it like once a year, because the mess that is Windows can't be relied upon to flush itself out.
Motherboard drivers: Chipset, lan, and audio. Be aware of the file versions; there may be more than one chipset driver to install, for example.

 

[klct555]

Well, there's one variable right there. Some models aren't shipped with Intel's stock bios settings and are tweaked for a little extra performance - but no one would know it unless they looked for it.


That's another variable. Frequency isn't everything though; the timings are another variable - lower timings are better.


It should be on the balanced plan by default.


Yes.
Bios: some updates contain bug, security, and compatibility fixes, as well as performance enhancements.
Windows: Windows will be Windows... don't fall behind on updates and clean install it like once a year, because the mess that is Windows can't be relied upon to flush itself out.
Motherboard drivers: Chipset, lan, and audio. Be aware of the file versions; there may be more than one chipset driver to install, for example.

So I changed my power plan from balanced to Ultimate performance, and my CPU began performing as expected on UserBenchmark, around 102%, but my Cinebench multi core score decreased by about 1000.....my temps actually decreased by literally 10 degrees during it tho...

 

[uWebb429]

The performance of the 10700K can vary significantly depending on how you have adjusted the turbo ratios and the turbo power limits. Do you know what values you are using? Is everything in the BIOS set to AUTO?

What does CPU-Z report for MHz while Cinebench is running? Does it decrease significantly during the benchmark?

You can also run HWiNFO to check for reasons for throttling while the CPU is loaded.

 

[klct555]

The performance of the 10700K can vary significantly depending on how you have adjusted the turbo ratios and the turbo power limits. Do you know what values you are using? Is everything in the BIOS set to AUTO?

What does CPU-Z report for MHz while Cinebench is running? Does it decrease significantly during the benchmark?

You can also run HWiNFO to check for reasons for throttling while the CPU is loaded.

Everything is set to Auto

I can verify it is running at its stock turbo clock, 4.7Ghz

Using AIDA64 I can see throttling stays at 0% for the entire benchmark.

 

[uWebb429]

Most people that buy unlocked K series CPUs buy these so they can overclock them. 4.7 GHz might be the default turbo speed but if no one else is running their 10700K at that speed, it is impossible to make a fair comparison.

5.10 GHz is the default max turbo speed. Lots of people immediately go into the BIOS and set an all core overclock so all cores run at this speed. Your benchmark results are going to be significantly behind their results.

For a point of reference, I ran Cinebench R23 with my Comet Lake CPU set to 8 cores 16 threads and 5.10 GHz across all 8 cores.

I will go do another run at 4.7 GHz for a better comparison to the speed that you are running at. Both are with DDR 4000 memory.

Try running ThrottleStop.

ThrottleStop 9.2.9

ThrottleStop 9.2.9 https://www.techpowerup.com/download/techpowerup-throttlestop/ New Features - added 10850K / 10900K support including a new Turbo Group access window. - updated the TS Bench and the C State window for the 10 core CPUs. - enabled Limit Reasons support for Comet Lake CPUs. -...

www.techpowerup.com

Make sure your CPU multiplier is steady for the entire run. Some types of throttling are ignored by some monitoring software.

Also check the amount of time your CPU is spending in the C0 state when idle. Some people have too much stuff running in the background that will interfere with CPU performance. My screenshot shows that 0.1% is normal for an 8 core - 16 thread Comet Lake CPU. If your computer is spending 5% of the time in the C0 state when idle processing background tasks, your R23 results will be 5% less.

 

[klct555]

Most people that buy unlocked K series CPUs buy these so they can overclock them. 4.7 GHz might be the default turbo speed but if no one else is running their 10700K at that speed, it is impossible to make a fair comparison.

5.10 GHz is the default max turbo speed. Lots of people immediately go into the BIOS and set an all core overclock so all cores run at this speed. Your benchmark results are going to be significantly behind their results.

For a point of reference, I ran Cinebench R23 with my Comet Lake CPU set to 8 cores 16 threads and 5.10 GHz across all 8 cores.

I will go do another run at 4.7 GHz for a better comparison to the speed that you are running at. Both are with DDR 4000 memory.

Try running ThrottleStop.

ThrottleStop 9.2.9

ThrottleStop 9.2.9 https://www.techpowerup.com/download/techpowerup-throttlestop/ New Features - added 10850K / 10900K support including a new Turbo Group access window. - updated the TS Bench and the C State window for the 10 core CPUs. - enabled Limit Reasons support for Comet Lake CPUs. -...

www.techpowerup.com

Make sure your CPU multiplier is steady for the entire run. Some types of throttling are ignored by some monitoring software.

Also check the amount of time your CPU is spending in the C0 state when idle. Some people have too much stuff running in the background that will interfere with CPU performance. My screenshot shows that 0.1% is normal for an 8 core - 16 thread Comet Lake CPU. If your computer is spending 5% of the time in the C0 state when idle processing background tasks, your R23 results will be 5% less.

Thanks, I'll make sure to check that out.

But a buddy of mine has the same processor with similar cooling and mobo from the same manufacturer except mine is slightly higher spec. He isn't really too tech savvy, hell, I'm the one who built his machine, hence he didn't touch any OC settings in his BIOS or downloaded any additional drivers that I may not have.

During his benchmarks, he is obtaining results that are actually average for the processor, and I had him check his clock speed and it stayed at 4.7 GHz.

 

[Phaaze88]

During his benchmarks, he is obtaining results that are actually average for the processor, and I had him check his clock speed and it stayed at 4.7 GHz.

See, this is actually a critical point - it's not supposed to do that at stock operation.
This would mean yours is running at Intel stock, and his isn't.

 

[Deleted member 2838871]

Most people that buy unlocked K series CPUs buy these so they can overclock them. 4.7 GHz might be the default turbo speed but if no one else is running their 10700K at that speed, it is impossible to make a fair comparison.

5.10 GHz is the default max turbo speed. Lots of people immediately go into the BIOS and set an all core overclock so all cores run at this speed. Your benchmark results are going to be significantly behind their results.

For a point of reference, I ran Cinebench R23 with my Comet Lake CPU set to 8 cores 16 threads and 5.10 GHz across all 8 cores.

I will go do another run at 4.7 GHz for a better comparison to the speed that you are running at. Both are with DDR 4000 memory.

... and for the sake of comparison my 10900k multicore score on my new build @ 5.1 (simple BIOS one click OC, haven't gotten around to manual OC yet) with DDR 4000 was 16,928.

3000 pt gap between 10850k and 10900k seems a little large, I expected it to be a lot closer but I'm not an expert.

 

[uWebb429]

3000 pt gap between 10850k and 10900k seems a little large

That is a little large. If you looked at the testing I was doing, I had my 10850K set to 8 cores - 16 threads for a fair comparison to a 10700K.

If you look at the Cinebench R23 results I posted, in third place is my 10850K. When all 10 cores and 20 threads were active, it scored 17073. Very similar performance to your 10900K but the 10850K sells for $100 less. Only $429 this week on Amazon.

 

[Deleted member 2838871]

That is a little large. If you looked at the testing I was doing, I had my 10850K set to 8 cores - 16 threads for a fair comparison to a 10700K.

If you look at the Cinebench R23 results I posted, in third place is my 10850K. When all 10 cores and 20 threads were active, it scored 17073. Very similar performance to your 10900K but the 10850K sells for $100 less. Only $429 this week on Amazon.

Just goes to show you how much I know about the 10850k... didn't do much research into it. Did read your post though, was just thinking it was 8/16 like you had posted. I did say in my first post I wasn't an expert... Some people are great at remembering the specs on dozens of different processors... but I'm not... and only looked at the score at the top of the pics. My point being I was expecting less of a gap between 8 cores and 10. Obviously I wouldn't expect a 3000 pt gap between processors with an identical number of cores... I may not be an expert but I'm not stupid either.

Anyway, I just looked it up and now I remember reading about it... "similar performance to the 10900k at 100mhz less clock speed." Grats on the $100.

 

[uWebb429]

The 10850K and 10900K are identical. When a CPU comes down the assembly line, if it runs well without needing too much voltage, Intel puts it in a 10900K box. If it is kind of crappy and needs lots of voltage to run fast, Intel puts it in a 10850K box. Intel's only goal is to sell as many CPUs as possible. They created the 10850K line to help move some of their excess inventory of poor running 10900K CPUs.

100mhz less clock speed

It is an unlocked K series CPU. You can set the CPU multiplier to whatever you like. Add voltage until stable. Just like you, I went with a 51 all core multiplier to get 17073. The 52 multiplier during Cinebench creates too much heat for my AIO.

 

[Deleted member 2838871]

The 10850K and 10900K are identical. When a CPU comes down the assembly line, if it runs well without needing too much voltage, Intel puts it in a 10900K box. If it is kind of crappy and needs lots of voltage to run fast, Intel puts it in a 10850K box. Intel's only goal is to sell as many CPUs as possible. They created the 10850K line to help move some of their excess inventory of poor running 10900K CPUs.

It is an unlocked K series CPU. You can set the CPU multiplier to whatever you like. Add voltage until stable. Just like you, I went with a 51 all core multiplier to get 17073. The 52 multiplier during Cinebench creates too much heat for my AIO.

Thanks for the info.... I'm familar with most of what you mention here... but the part about the ""similar performance to the 10900k at 100mhz less clock speed" was just a copy/paste from the top result in my Google search.

I actually just dialed in my 3090 OC the other day and ran all my tests and it's doing well. I started fooling around with the 10900k OC last night... and first go also created too much heat with Cinebench. Plan on tweaking it more this weekend to dial it in. At the moment I'm just running the one click OC in the BIOS as a placeholder until I get time to manually do it... which I haven't due to end/beginning of the year craziness at work.

Anyway... should be fun. Shooting for a stable 5.3 all core with good temps.

 

[Phaaze88]

You guys/gals have an easy alternative: Intel Performance Maximizer
If nothing else, it'll give ya'll something to work around.

 

[Deleted member 2838871]

You guys/gals have an easy alternative: Intel Performance Maximizer
If nothing else, it'll give ya'll something to work around.

I'm at work so can't download it and check it out... is that anything like Intel Extreme Tuning Utility? I've seen that mentioned before... good and bad.

 

[Phaaze88]

I'm at work so can't download it and check it out... is that anything like Intel Extreme Tuning Utility? I've seen that mentioned before... good and bad.

It's more like Msi Afterburner OC Scanner, if you're familiar with that.

It's software, so I'm not expecting it to be perfect - I'm sure better can be done manually, but it gives users somewhere to start.

 

[Deleted member 2838871]

It's more like Msi Afterburner OC Scanner, if you're familiar with that.

It's software, so I'm not expecting it to be perfect - I'm sure better can be done manually, but it gives users somewhere to start.

Excellent. Thanks for the link. Will check it out at home later. I'm of the "set it and forget it" mindset and am ready to OC the 10900k to stability and call it good. I've done it manually before but it's been a while... so like you said this will be a good place to start. Definitely not an OC junkie... just want to get the performance that's available.

 

[Deleted member 2838871]

It's software, so I'm not expecting it to be perfect - I'm sure better can be done manually, but it gives users somewhere to start.

It was a good starting point... got a solid OC last night with maybe even room to wiggle... benchmarks improved... gonna post my own thread in the OC forum and not hijack this one.

Thanks again!