Question i7-9700K beats i9-9900KF on a single core bench?

There must be some differences between your systems other than the CPU that are affecting scores.

With a bunch of programs and a browser with lots of tabs open, my stock i9 9900K scores 1356 single core. With some basic tweaks it could probably score a little bit more.

i9 9900K @ 1356 single core

I think your freind needs to look at his setup, for example if the CPU is thermal- or power throttling.

EDIT :

He should go through his BIOS settings. I see he has an ASUS board.

ASUS boards stick very closely to the Intel rated 95W TDP at stock BIOS settings for a 9900K/F.

My i9 power throttles quite quickly, all the way down to 4.1 or 4.2GHz (I don't remember exactly which one it was) and stays there during load, unless I setup BIOS to allow power draw in excess of the 95W TPD.

I had an MSI MPG Z390 Gaming Pro Carbon which ignored the Intel TDP limit and my i9 ran the advertised speeds all day long, at BIOS default settings.

But my ASUS board at "optimized defaults" throttled the CPU all the time, to stay around 95W.

Allowing the ASUS motherboard to draw extra power, means that my i9 never throttles at all.

I'm sure there is performance to gain from this for your friend - unless the BIOS has already been setup correctly.
 
Last edited:

Tex-Twil

Distinguished
Dec 13, 2007
171
2
18,695
yeah I'm also suspecting some kind of throttling. I go through the BIOS when I'll have physically access to the computer.

The OS itself is clean and there were no open tabs or running programs that could have impacted the bench.

thanks for your tips. I'll report back
 
The menu to look for is called Digi+ Power Control, on my board it is located under Extreme Tweaker. Set CPU Current Capability to maximum (140%).

But it is best to check if it is temperature throttling or power throttling, or if it is throttling at all before you do this. If everything is running normally, just ignore my comment.

If it is throttling to stay around 95W TPD, and you allow it to draw more power to stay at boost clocks, temperatures could rise a bit, since it will be running at higher frequencies continously. It doesn't affect vCore and it is not changing standard core clock, it just means it can consume more watt and stay at boost speed without downclocking itself.


EDIT


Cooling is a NOCTUA NH-U12S

Although Noctua air coolers are some of the best in my opinion, I think he should have gone for a beefier model with a higher watt rating -for instance the NH-D14 or NH-D15, I'm not sure this cooler is sufficient - I might be wrong, someone else will know for sure, but it is something to look into.
 
Last edited:
The setting to look for is called Digi+ Power Control, on my board it is located under Extreme Tweaker. Set it to maximum (140%)

But it is best to check if it is temperature throttling or power throttling, or if it is throttling at all before you do this. If everything is running normally, just ignore my comment.

If it is throttling to stay around 95W TPD, and you allow it to draw more power to stay at boost clocks, temperatures could rise a bit, since it will be running at higher frequencies continously. It doesn't affect vCore and it is not changing standard core clock, it just means it can consume more watt and stay at boost speed without downclocking itself.


EDIT




Although Noctua air coolers are some of the best in my opinion, I think he should have gone for a beefier model with a higher watt rating -for instance the NH-D14 or NH-D15, I'm not sure this cooler is sufficient - I might be wrong, someone else will know for sure, but it is something to look into.
Cooler is more than enough. Wattage rating is sufficient, and higher wattage rating on the same CPU doesn't necessarily mean lower temps. NH D14/15 would be better if overclocking, but not necessary on stock settings.
 
  • Like
Reactions: MadsModsat
My older Asus board (Z270/7700K) had MCE enabled, and did not hesitate to go to max turbo on all cores...

A new build (Z390) I did for a friend using a 9700K did strangely seem to adhere to a 'semi- economy' mode, seemingly only boosting in all core loadings to about 3.8 GHz or so, even with MCE enabled... (used XTU to set it up the way I considered 'normal')

As normal max boost clock speeds will only be 100 MHz apart between a 9700K and a 9900K, the scores could be quite close, or, reverse, depending on what clock speeds are hit, and might easily be largely dependent on BIOS and/or XTU settings...
 
Last edited:
You can't really compare a highly optimized overclocked system running at 5GHz (I suspect @ all core) directly to an 9700K, and say that it is strange a bone stock 9700K gets a lower single core score, when it only has a single core boost clock of 4.9 GHz, and lower all core clocks at 4.6GHz.

The stock i7 9700K OP has, still scores higher multi-core score than your OC'ed 9600K, with its additional cores and no OC.

But it does illustrate however, that the i9 9900K score mentioned in OP's original post is way off.

Unless OP returns and let us know what the actual problem was, I'd still say it is power throttling since it is an ASUS motherboard, and on some of their boards, they enforce Intel TDP specs strictly, unless you specifically let the 9900K draw more power than its 95Watt TDP.

The single-core score I posted earlier was with a lot of stuff happening at the same time, it was never meant to be a max performance run, and it was also an earlier version of Geekbanch 5, my actual single core score when I don't have a lot of apps running is 1418

Geekbench 5 9900K Single Core 1418
 
Last edited:

Tex-Twil

Distinguished
Dec 13, 2007
171
2
18,695
Hey
so I'm back with a physical access to the computer.

I ran Geekbanch again while watching the frequencies and they never went higher that 3.6 Ghz

The menu to look for is called Digi+ Power Control, on my board it is located under Extreme Tweaker. Set CPU Current Capability to maximum (140%).

So I changed this setting in the BIOS (it turns out it's accessible also via the Asus AI Suite) as suggested the frequency is still max at 3.6Ghz during the bench.

I tried running the Aida64 stability test for a while and again the frequency is at 3.6 Ghz mac. Temps do not go above 60C during the 3.6 load.

I also made sure the Windows 10 Power Setting is set to Performance.

The Geekbench score did not change. Should the CPU go to 5Ghz under load (without overclocking) ?

Update:
I also set the Short and Long Duration Power Limit to its max (4095) in the BIOS as described here
and now the Geekbench 5 score is around 1250. It is still lower than other builds with the same CPU but I think it's acceptable
 
Last edited:
This certainly explains the low score.

Is Turbo Boost enabled in BIOS/UEFI?

The i9 9900K should boost to these frequencies depending on active cores :

1-2 active cores : 5.0 GHz (x50 multiplier) I always thought it was single core 5Ghz, but I've seen theese numbers different places, I've added links to two sources.

3 active cores : 4.9 GHz (x49 multiplier)

4-5 active cores : 4.8 GHz (x48 multiplier)

6-8 active cores : 4.7 GHz (x47 multiplier)

It is not often i see 5 GHz, unless I close down any unneccessary background processes when benching, or it is a single core benchmark.

4.8 - 4.7 is what my CPU is at most of the time on the desktop under light load. 4.7GHz always at full load.

If Turbo Boost is disabled, the CPU will stick to its max base clock of 3.6GHz

Turbo boost is not an overclocking option, the frequecies posted above are default specifications for the 9900K.

--------------------------------------------------
Second source for two-core x50 multiplier
 
Last edited:

Karadjgne

Titan
Ambassador
I7-9700k and i9-9900k are the same cpu basically. Few internal changes like enabling HT. So single thread should be pretty much the same, if power limits are the same. There's always a margin for error because of differences in setup, not so much the cpu. 9900k advantage is mostly in multi-thread.
 
The silicon in the cores are essentially the same. The 9900K just has hyperthreading, giving it more threads.
A single thread benchmark will depend on the clock rate of the single thread in question.
If there is overclocking involved. the upper limit will be determined by the silicon lottery.