Skip to main content

Intel's Enthusiast-Grade Core i9-11900KB Tiger Lake CPU Benchmarked

Intel
(Image credit: Intel)

Benchmarks for Intel's mysterious Core i9-11900KB "Tiger Lake" desktop CPU surfaced today. Testing of Team Blue's upcoming NUC 11 Extreme "Beast Canyon" show the CPU performing similarly to Intel's socketed eight-core Core i9-11900K "Rocket Lake." 

The i9-11900KB looks to compete against the best CPUs with an octa-core chip with Hyper-Threading, clock speeds of 3.3 / 4.9 / 5.3 GHz, 24MB of cache and a 65W TDP. The NUC tested also uses an Nvidia GeForce RTX 3060. Someone with access to the system posted its results in the 3DMark TimeSpy and Fire Strike benchmarks. While 3DMark does not represent performance in real-world games and applications, it still demonstrates overall capabilities of a PC. 

Considering specifications of the Core i9-11900KB, it makes sense to compare it to a similarly configured system powered by the i9-11900K (8C/16T, 2.5 / 5.0/ 5.2 GHz, 16MB, 65W).

i9-11900KB 'Tiger Lake' vs. i9-11900K 'Rocket Lake'

Core i9-11900KB + GeForce RTX 3060Core i9-11900 + GeForce RTX 3060
Time Spy8,4209,089
Time Spy | CPU10,87211,064
Time Spy | GPU8,0988,812
Fire Strike19,02019,898
Fire Strike | GPU20,52322,002
Fire Strike | Physics25,57125,260
Fire Strike | Combined9,8379,777

While the Core i9-11900KB has a higher base frequency, rather high burst clocks and a larger cache, it still couldn't beat the Core i9-11900K in the 3DMark Time Spy CPU test. But that's likely because a better power supply and cooling allowed the latter to work at higher clocks for longer time. 

Meanwhile, the Tiger Lake chip slightly outperformed the Rocket Lake in the 3DMark Fire Strike Physics test. 

Intel's NUC 11 Extreme "Beast Canyon" doesn't have a price or release date yet but promises to be considerably smaller than standard desktop PCs, so still being able to offer CPU performance on par with bigger systems would be a big deal. Yet, graphics cards will unlikely hit their maximum boost clocks in a small form factor system, due to cooling constraints.

  • JamesJones44
    The benchmarks are interesting, but not surprising given the clocks. One would expect them to be slightly less given they are clocked lower with a similar architecture on a shrunk node.

    I'm much more interested in the thermals and power draw. It might give some insight as to what to expect with Alder Lake. Hopefully that information comes out soon.
    Reply
  • littlechipsbigchips
    It is a good chip but the problem with intel is that it will be more expensive than a standard i9 11900 in a Mini ITX case and ITX motherboard. and I dont think that the size difference between the NUC and ITX is that big to justify the price difference.
    Reply
  • funguseater
    I guess Intel didn't want AMD CPU's on the chart?
    Reply
  • Joseph_138
    funguseater said:
    I guess Intel didn't want AMD CPU's on the chart?

    This wasn't a comparison against AMD, it was a comparison against Intel's own product to highlight the differences between the two. Not everything is about competition. Leave your fanboyism at the door. Your comment sounds like it was written by a 12 year old.
    Reply
  • thGe17
    Overall quiet impressive results considering the fact that this is a 65 W mobile part.
    Still, you should correct the article, because in the text and over the table you state a comparison against the 11900K ... I'll guess you compared against the 65 W desktop part 11900 w/o "K"?

    "to Intel's socketed eight-core Core i9-11900K "Rocket Lake.""
    "system powered by the i9-11900K (8C/16T, 2.5 / 5.0/ 5.2 GHz, 16MB, 65W)."
    "i9-11900KB 'Tiger Lake' vs. i9-11900K 'Rocket Lake'"
    "it still couldn't beat the Core i9-11900K in the 3DMark"If the "K" is correct, and it was the 125 W part, the technical details above are incorrect, so it is a little bit confusing (and the results would be even more impressive if you really compared against a 125 W desktop part ;)).
    Reply
  • Amelia321
    thGe17 said:
    Overall quiet impressive results considering the fact that this is a 65 W mobile part.
    Still, you should correct the article, because in the text and over the table you state a comparison against the 11900K ... I'll guess you compared against the 65 W desktop part 11900 w/o "K"?

    "to Intel's socketed eight-core Core i9-11900K "Rocket Lake.""
    "system powered by the i9-11900K (8C/16T, 2.5 / 5.0/ 5.2 GHz, 16MB, 65W)."
    "i9-11900KB 'Tiger Lake' vs. i9-11900K 'Rocket Lake'"
    "it still couldn't beat the Core i9-11900K in the 3DMark"If the "K" is correct, and it was the 125 W part, the technical details above are incorrect, so it is a little bit confusing (and the results would be even more impressive if you really compared against a 125 W desktop part ;)).

    Thanks for pointing this out!
    Reply
  • usiname
    thGe17 said:
    Overall quiet impressive results considering the fact that this is a 65 W mobile part.
    Still, you should correct the article, because in the text and over the table you state a comparison against the 11900K ... I'll guess you compared against the 65 W desktop part 11900 w/o "K"?

    "to Intel's socketed eight-core Core i9-11900K "Rocket Lake.""
    "system powered by the i9-11900K (8C/16T, 2.5 / 5.0/ 5.2 GHz, 16MB, 65W)."
    "i9-11900KB 'Tiger Lake' vs. i9-11900K 'Rocket Lake'"
    "it still couldn't beat the Core i9-11900K in the 3DMark"If the "K" is correct, and it was the 125 W part, the technical details above are incorrect, so it is a little bit confusing (and the results would be even more impressive if you really compared against a 125 W desktop part ;)).
    What is impressive? 65W 10nm lose to backported 65W 14nm?
    Reply
  • passivecool
    Joseph_138 said:
    Leave your fanboyism at the door. Your comment sounds like it was written by a 12 year old.

    Ditto, mister 138.
    Reply
  • watzupken
    thGe17 said:
    Overall quiet impressive results considering the fact that this is a 65 W mobile part.
    Still, you should correct the article, because in the text and over the table you state a comparison against the 11900K ... I'll guess you compared against the 65 W desktop part 11900 w/o "K"?

    "to Intel's socketed eight-core Core i9-11900K "Rocket Lake.""
    "system powered by the i9-11900K (8C/16T, 2.5 / 5.0/ 5.2 GHz, 16MB, 65W)."
    "i9-11900KB 'Tiger Lake' vs. i9-11900K 'Rocket Lake'"
    "it still couldn't beat the Core i9-11900K in the 3DMark"If the "K" is correct, and it was the 125 W part, the technical details above are incorrect, so it is a little bit confusing (and the results would be even more impressive if you really compared against a 125 W desktop part ;)).
    I suggest that you leave the TDP out in the comparison here because you know its not staying at 65W. Considering the Tiger Lake H "45W TDP" is pulling in 90W or more easily, I am not surprise the actual power requirement is much higher in this case. It may not be pulling in as much as Rocket Lake, which makes sense because of the form factor and limited cooling, but its possible that it can draw double the supposed TDP. The lower power limit and cooling solution are probably the reason why it can't keep up with Rocket Lake despite the core being faster clock for clock. .
    Reply
  • thGe17
    That doesn't make sense, because there still exists a major difference. If the 11900 was the comapred target, this still is a regular Desktop-CPU with a different socket and much more room for a more powerfull cooling solution, therefore even if both CPUs would have been exactly the same, it might be possible, that the desktop pendant had an advantage.
    If the 11900K was the compared target, it gets even more skewed, because this is a desktop CPU with already a regular 125 W TDP and it easily can use 150+ W w/o being overclocked, so this makes a huge difference.

    And additionally turbo duration is a factor, because the Tiger Lake with its more limited cooling solution can keep up its highest frequency only for a few seconds, whereas the desktop chip might run at this frequency e. g. 10 times longer.

    A full-blown test would have supported power consumption during a test run, because it would be interesting to compare 14nm+++ with 10nm++.
    An even more in-depth test would lock both CPUs to e.g. 4,0 GHz and check power consumption with the same workload. How much better can 10nm++ perform compared to the old process node?
    Reply