Gaming Performance: 4K

Last, we have our 4K gaming results.

Civilization VI

(a-5) Civilization VI - 4K Min - Average FPS(a-6) Civilization VI - 4K Min - 95th Percentile

World of Tanks

(b-7) World of Tanks - 4K Max - Average FPS(b-8) World of Tanks - 4K Max - 95th Percentile

Borderlands 3

(c-5) Borderlands 3 - 4K VLow - Average FPS(c-6) Borderlands 3 - 4K VLow - 95th Percentile

Grand Theft Auto V

(e-5) Grand Theft Auto V - 4K Low - Average FPS(e-6) Grand Theft Auto V - 4K Low - 95th Percentile

Red Dead Redemption 2

(f-5) Red Dead 2 - 4K Min - Average FPS(f-6) Red Dead 2 - 4K Min - 95th Percentile

F1 2022

(g-7) F1 2022 - 4K High - Average FPS(g-8) F1 2022 - 4K High - 95th Percentile

Hitman 3

(h-7) Hitman 3 - 4K High - Average FPS(h-8) Hitman 3 - 4K High - 95th Percentile

Total War: Warhammer 3

(i-4) Total War Warhammer 3 - 4K High - Average FPS

We noticed some discrepancies in our Cyberpunk 2077 testing at 1440p and 4K; we will publish these results once we identify the issue.

As we've seen throughout our game testing, things are quite competitive between the top contenders, including the Intel Core i9-13900K, the Core i9-12900K/KS, and the AMD Ryzen 9 7950X chips. In our 4K testing, however, where the Ryzen 7 5800X3D and loads of 3D V-Cache can't be utilized, then the Core i9-13900K and Core i5-13600K perform very well. 

Gaming Performance: 1440p Closing Thoughts
POST A COMMENT

169 Comments

View All Comments

  • Pjotr - Thursday, October 20, 2022 - link

    Closing thoughts typos: Ryzen 580X3D and Ryzen 700. Reply
  • Ryan Smith - Thursday, October 20, 2022 - link

    Thanks! Reply
  • mode_13h - Thursday, October 20, 2022 - link

    Thanks for the review!

    Could you please add the aggregates, in the SPEC 2017 scores? There's usually a summary chart that has an average of the individual benchmarks, and then it often has the equivalent scores from more CPUs/configurations than the individual test graphs contain. For example, see the Alder Lake review:

    https://www.anandtech.com/show/17047/the-intel-12t...
    Reply
  • Arbie - Thursday, October 20, 2022 - link

    TechSpot / Hardware Unboxed show that to complete a Blender job the 13900K takes 50% more total system energy than does the 7950X. Intel completing a Cinebench job takes 70% more energy. Meaning heat in the room. And that's with the Intel chip thermal throttling instantly on even the best cooling.

    Looking at AT's "Power" charts here, which list the Intel chip as "125W" and AMD as "170W", many readers will get EXACTLY THE OPPOSITE impression.

    Sure, you mention the difficulties in comparing TDPs etc, and compare this gen Intel to last gen etc but none of that "un-obscures" the totally erroneous Intel vs AMD picture you've conveyed.

    ESPECIALLY when your conclusion says they're "very close in performance" !! BAD JOB, AT. The worst I've seen here in a very long time. Incomprehensibly bad.
    Reply
  • gezafisch - Thursday, October 20, 2022 - link

    Cope harder - watch Der8auer's video showing that the 13900k can beat any chip at efficiency with the right settings - https://youtu.be/H4Bm0Wr6OEQ Reply
  • Ryan Smith - Thursday, October 20, 2022 - link

    We go into the subject of power consumption at multiple points and with multiple graphs, including outlining the 13900K's high peak power consumption in the conclusion.

    https://images.anandtech.com/graphs/graph17601/130...

    Otherwise, the only place you see 125W and 170W are in the specification tables. And those values are the official specifications for those chips.
    Reply
  • boeush - Thursday, October 20, 2022 - link

    Not true. You have those insanely misleading "TDP" labels on every CPU in the legend of every performance comparison chart. This paints a very misleading picture of "competitive" performance, whereas performance at iso-power (e.g. normalized per watt, based on total system power consumption measured at the outlet) would be much more enlightening. Reply
  • boeush - Thursday, October 20, 2022 - link

    *per watt-hour (not per watt)

    [summed over the duration of the benchmark run]
    Reply
  • dgingeri - Thursday, October 20, 2022 - link

    Is it just me, or does the L1 cache arrangement seem a bit odd? 48k data and 32k instruction for the P cores and 32k data and 64k instruction on the e-cores. Seems a bit odd to me. Reply
  • Otritus - Thursday, October 20, 2022 - link

    Golden/Raptor Cove has a micro-op cache for instructions. 4096 micro-ops is about equal to 16Kb of instruction cache, which is effectively 48Kb-D + 48Kb-I. I don’t remember whether Gracemont has a micro-op cache. However, it doesn’t have hyperthreading, so maybe it just needs less data cache per core. Reply

Log in

Don't have an account? Sign up now