Can 11900K beats a LTX gold sample 10900K? Let's find out!

cagoblex

n00b
Joined
Aug 28, 2020
Messages
63
We just published the review for the 11700K. And here comes it’s bigger brother 11900k. Unlike the previous generations, the i9 flagship 11900K has exactly the same spec as the 11700K. Actually it even has a slightly lower base clock, 3.5GHz compares to 3.6GHz of the 11700K.


spec.jpg





However what makes it an i9 is Turbo Velocity boost. As long as you have adequate cooling, the i9 processor will boost its frequency for an extra 100MHz on top of the already high Turbo Boost 3 frequency. But apart from that, it is the same chip as the 11700K. Same core count, same threads, same cache, same PCI-E support and same iGPU. The reason this review is taking so long is because of the new Intel ABT, or adaptive boost technology. Basically what it does it that it will override the TB2 or TB3 multiplier limit and gives a higher all core boost given there is enough thermal and power budget. So long story short, it will give the 11900K an all core boost of 5.1GHz under ideal condition compares to the 4.8GHz default.




Intel-Adaptive-Boost-Technology.jpg







Asus just released the new BIOS update that supports the Intel ABT. In this review we will first take a look of this processor without the ABT enabled, and then compare the improvement ABT brings separately. And after that comes the important part, can it beat a LTT gold sample 10900K? Let’s find out!

Disclaimer on 5800X score. The AIO I am using does not have AM4 mount bracket, so these scores are tested with a cheap air cooler. With limit thermal budget, it is not achieving max PBO frequencies. A lot of readers have pointed that out. I don’t have the 5800X on hand right now so I cannot retest it. It is not that I intentionally lower the score for AMD to make Intel look better. Obviously Intel did not sponsor this review and myself is using an 5900X while writing this. So it is completely due to the limitation of testing equipment. The point of this review is not Intel vs AMD anyways so it shouldn’t be too big of a problem. If you really want to know how 11900K stands against AMD, I will include a little bit of 5900X score in the final part of the review so you can get an idea.


bios.PNG



Firstly let’s take a look at CPU-Z. Thanks to the higher boost clock and the presence of Thermal velocity boost, it is slightly faster than 11700K, but only by a tiny margin.



cpu-z.png










In Aida 64 Memory testings, it is pretty much the same as the 11700K.

aida64.png


Next is Cinebench.





In Cinebench R15, Again it’s almost the same as 11700K. It is technically higher but it’s within the margin of error.





r15.png






In cinebench R20, the difference is a little bigger than R15. It is about 100 points, which is about 2% faster than 11700K in multi core, and about 6% faster in single core.


r20.png





In V-ray, again, it has a higher average clock speed of 4.68Ghz compares to 4.55GHz of the 11700K. But the result is 200 points higher, which is about 1.2%.

vray.png




In Blender, we tested both classroom and BMW. It is 4 seconds faster in BMW and 28 seconds faster in Classroom.


blender.png



In 7 zip, it is about 3% faster in decompression while being almost within margin of error in compression.

7zip.png




In hand break, we are transcoding a 1:31 4K video into 1080P H.264. It is 3 seconds faster which is about 4%.


handbrake.png



In Geekbench 5, it is 5% faster in single core and 2% faster in multi core testings.

g5c.png


In Y-cruncher, it is only 1 second faster than 11700K in single thread testing, and 11 seconds faster in multi core testings.

ycruncher.png


Next let’s take a look at the games. All platforms are paired with EVGA RTX3080 FTW3 video card.




3dmark.png





First is 3D Mark, let’s just focus on the CPU part. Again they are identical, almost within the margin of error. It actually got slightly lower score than 11700K.


hitman 2.png




In hitman2, again it’s pretty boring. It is basically a 11700K with negligible gain.





In Horizon, it is exactly a 11700K.



horizon.png




In Shadow of the Tomb Raider, it’s even worse than 11700K. I actually retested just to make sure, but still I got the same result.



sot.png




In dirt 5, it’s 2 fps faster than 11700K. Yeah I exaggerated the difference so it looks less boring.

dirt 5.png


Ok this is kinda boring right, let’s see what Intel Adaptive Boost Technology brings. I just hate it because it made me run all the benchmarks again. I have included a stock 10900K here to give you an idea of where 11900K stands against 10900K. The reason I did it is simple. It is because ABT makes close to no difference to the performance. But since I spent all this time testing it with ABT, so I decide to keep it. Otherwise this whole section should not have existed if I don’t add something new to it. And here is the result.

The video card being used for the benchmarks below is the Asus TUF 6700XT video card, not the EVGA 3080 FTW3.





cpuz.png




11900K definitely a lot faster than 10900K in single threaded testings, but 10900K is still faster in multi threaded testings thanks to it’s two extra cores.



r15.png


Next it Cinebench R15. Again ABT makes almost no difference. The 11th gen is faster in single threaded tests and slower in multi threaded testings.


r20.png



And it’s the same with R20. There is nothing really worth to be talked about.

7zip.png




In 7 zip, 10900K is faster in decompression while tiny bit slower in compression.


blender.png



In Blender, 10th gen and 11th gen performs almost the same. Again ABT doesn’t help in any way.

ycruncher.png




Y Cruncher is where the 11th gen and the ABT shines because it uses AVX512 instruction which is lacking on the 10900K. However the single core score is lower when ABT is enabled, but multi core is a lot faster. The 10900K is a lot slower than 11900K in this case.

vray.png


In v-ray, ABT makes no difference, and 10900K is still faster thanks to its two extra cores, but not by a lot.

3dmark.png




In 3Dmark Timespy, the 11th gen is about 20% faster than 10900K in CPU score. ABT only gives it 1% gain on 11900K.





Ok that’s still boring right. Now it’s time to do something interesting, let’s overclock both of them and see how they competes. For the 10900K, I am using a LTT gold sample, which is a bundle of a cherry picked 10900K and a Cooler master ML360 Sub Zero cooler. Technically it should give excellent overclocking result. So let’s give it a try!

gs.PNG


gs1.PNG



ml360.PNG


Ok here we are in the BIOS, Asus gives it a 102 SP score, which is high but not the highest I have ever seen on a 10900K. It is predicting that it will need 1.398V at L4 LLC to run non-AVX workloads at 5.3GHz.So let’s start from that and see how it performs.

11900k bios.PNG


With the Cooler, we need to install the Cryo software and make it work in Cryo mode. So the cooler will communicate with the processor and adjust the cooling real time. We will make a separate video on this set if you want to know more details. Right now we can see, at idle, the cooler is pulling about 100W of power from the 8pin connector, and the CPU temp is around 22 degrees. Let’s open HWinfo64 monitor, and here we can see the processor is running at 5.3Ghz on all cores, the VID is actually a little higher at 1.44V. But VID only means what the CPU thinks it needs at a given frequency, not the actual voltage that’s being fed to it. And the temps are in the lower 20 degrees range, which is pretty cool. Now let’s run some R20 and see if it will get a pass. Keep in mind R20 actually uses AVX so it would normally require higher voltage. And we got a pass, and got 6252points.
cyro.PNG


cyro temp.PNG


r20 pass.PNG

vray error.PNG


If you take a closer look at the results, it is still slower than our 5.3GHz all core 11700K we tested last time. But that’s not the full potential of this gold sample. Let’s try something extreme. Let’s put 5.5GHz on four cores and 5.4GHz on six cores and see how CPU-Z score looks like. Now we got over 8000 points for multi threaded and 651 for single threaded! This is a very remarkable score under non LN2 cooling. Let’s try V-ray. And no it cannot pass V-ray. But will it pass Cinebench R20? Let’s see. And...it’s a freeze.


5.4 cpuz.png



Now we have restarted and let’s lower the frequency a little bit. Let’s do 5.4GHz on four cores and 5.3GHz on other six. And it still freezes for R20.



So after some tweaking, the most reasonable combination factoring temp, voltage and performance is 5.3GHz if we want to run AVX workloads.



And our 11900K is not as highly binned as the 10900K. It has a SP score of 71 in BIOS. And BIOS is predicting it needs 1.6V to reach 5.3GHz all core for non-AVX workloads. However when we apply 1.6V to it, it thermal throttles immediately even with the Cryo cooler. After some tweaking, the best combination for this particular chip running AVX workload is all core 5.2GHz at 1.52V. And here are the results.


sp11.PNG











In CPU-Z, the 10900K is still 1000 points faster than the 11900K thanks to its extra cores. But in single core performance, the 11900K is 10% faster than the 10900K.


cpuz.png



In R20, it’s the same situation. The 11900K is slightly slower in all core performance but definitely have a lead on single core.

r20.png


In V-ray, they are almost the same, with 10900K being 4 points faster.
vray.png



So conclusion. I have very mixed feelings about Rocket lake processor now. Yes, 11700 non K version, and the 11500 we are currently reviewing, are a competitive choice at their respective price point. But for 11900K, I’m not too sure. In Intel’s defense, it is able to achieve roughly same all core performance, with 2 less physical cores on the same manufacturer node, which is the biggest breakthrough for Intel in the past few years. However, being at the same price point, even an overclocked 11900K can’t see the taillight of the 5900X. Trying to be competitive at mid range, and completely give up on high end market. Does that sound familiar to you? Yes this was the position AMD has been in for almost a decade before Ryzen. So great job Intel, on Rocket lake, and on the other hand, you really need to do better than this.



You can watch a video version of this review here:

 
Its a shame the 10700K wasnt included.
The best value performance CPU for gaming (while pushing for high end framerate)...
And you can buy one.
 
Last edited:
Damn, thank you for posting this man. Made up my mind lol. My motif monument is getting an 11700K
 
Are these 11900K CPUs able to do per core overclocking like X series CPUs in Intel XTU?

(Would allow core 0 at 5.6ghz 1.6 vcore, core 1-2 5.4ghz 1.5vcore, and cores 3-4-5-6-7 at 5ghz 1.3vcore for example) Something like that to go for highest single threaded performance using the best cherry picked cores.
 
Last edited:
Its a shame the 10700K wasnt included.
The best value performance CPU for gaming (while pushing for high end framerate)...
And you can buy one.
IMO, the 10700f non-k set with unlimited turbo is the best value (assuming price separation at any given moment).
You can see in this review, that stock performance is not far behind. And if you open up the turbo, it will do 4.6 all core, all day, in most boards.
I've even seen some boards able to do 4.8 all core with some extra tweaks. But I think that requires a Z board anyway.
 
  • Like
Reactions: Nenu
like this
IMO, the 10700f non-k set with unlimited turbo is the best value (assuming price separation at any given moment).
You can see in this review, that stock performance is not far behind. And if you open up the turbo, it will do 4.6 all core, all day, in most boards.
I've even seen some boards able to do 4.8 all core with some extra tweaks. But I think that requires a Z board anyway.
I wasnt aware a none K will boost so high, thats sweet.
What cooling/voltage/temps did you see this with?
 
I wasnt aware a none K will boost so high, thats sweet.
What cooling/voltage/temps did you see this with?
With the turbo settings opened up, my non-K does 4.6 all core, all day. The program called "Core Temp" isn't the most accurate but, for its TDP/power usage estimation, the most I have seen is 140w. But its generally around 130w under full load (cinibench or encoding H.264 with staxrip). I need to do more testing on the power usage, though. Because mine seems lower than some reviews. I've seen some reviews actually track the power usage of the non-K a little bit higher than the K, because its a lesser bin and requires more power to keep up the speed.

I'm using a Scythe Mugen 5 Rev. B cooler with two aftermarket fans which were mostly chosen for being very quiet. In Cinibench or encoding with Staxrip, I've had 2 or 3 cores hit 72c when the room was pretty warm already. The other cores were 69-70. With faster/louder fans, I'm sure a couple of degress could be shaved off.

To get 4.8 all core, you either need a board which is really strict about matching the highest frequency to all cores (Maybe an Asus or MSI would do that. The non-K will technically do 4.8 on only two threads. A really strict board might be able to force that on the all core). You can also try to bump the b-clock up to 102mhz (I think you can only tweak b-clock on Z490/Z590 boards. I have an H570 from Asrock.)
 
Last edited:
  • Like
Reactions: Nenu
like this
Are these 11900K CPUs able to do per core overclocking like X series CPUs in Intel XTU?

(Would allow core 0 at 5.6ghz 1.6 vcore, core 1-2 5.4ghz 1.5vcore, and cores 3-4-5-6-7 at 5ghz 1.3vcore for example) Something like that to go for highest single threaded performance using the best cherry picked cores.
Yes you can do it. I just haven't got enough time to play with it. Was trying to post it before the embargo. I will have updates on this later.
 
It looks like you could in an ASUS motherboard, I see the settings in my Z390 and Z490 that you can, but I usually use the ASUS AI Suite 3 and overclock all 8 cores, but I'm using a 10700KF.


cagoblex, Microcenter has the 10700K for $249.99
I know... But still there is no point to buy a 10700K just for the review. I have way too many CPUs lying around.
 
IMO, the 10700f non-k set with unlimited turbo is the best value (assuming price separation at any given moment).
You can see in this review, that stock performance is not far behind. And if you open up the turbo, it will do 4.6 all core, all day, in most boards.
I've even seen some boards able to do 4.8 all core with some extra tweaks. But I think that requires a Z board anyway.
Yes. Z board with the extra 2 bin boost enabled will make it 4.8GHz.
 
Thanks for your review, I had a beer when I wrote my comment, it was a bit dead pan.
Who is "We" btw? interested to know.
We? Oh that's just me and bunch of my friends. We all work in the industry and usually exchange samples and information on hardwares.
 
  • Like
Reactions: Nenu
like this
Yes. Z board with the extra 2 bin boost enabled will make it 4.8GHz.
That's probably where reviewers are getting over 200w on the non-k 10700. its gotta push the extra power to force 6 more cores to run at 4.8.; over the stock behavior of 4.8 for two cores, 4.7 for 4 cores, and 4.6 for all cores.
 
I will say, you should fix that Dirt 5 graph. Same exact format as the 11700k review by making the differences look way larger than they really are.
 
  • Like
Reactions: noko
like this
Adored TV is showing that bios updates for some boards are increasing performance ( 1 key area being lowered memory latency) and also fixing booting bugs and some crashes. If you have a Rocket Lake CPU and/or a 5 series board, you should probably watch your motherboard's bios updates closely. I have an H570 from Asrock and they posted a new bios on launch day and another beta bios on April 1st.
https://adoredtv.com/houston-we-have-bios-problems-the-rocketlake-preview/
 
Back
Top