Discussion Cinebench 2024 Released

[Det0x]

Direct die 7950X @ large custom watercooling
2209mhz FCLK
8152MT/s @ 32-45-37-45-84

CPU Score = 2531

 

[Timur Born]

Power usage and noise to reach that score?

 

[Shmee]

Score from my 5800X3D. Pretty close to the other Zen 3 8 cores CPUs, both the 5800X reference and the 5700X tested by @DAPUNISHER. I am running 3200 MT/s on my RAM, curious if going higher would make a difference...of course you also have to consider different cooling solutions among systems.

 

[Det0x]

Power usage and noise to reach that score?

About same powerdraw as in Cinebench R23..

So not too shabby @ 318w peak when running delidded on the mycro direct die block

Cooling

 

[Timur Born]

About same powerdraw as in Cinebench R23..

Cinebench 24, multi and single core on undervolted 13900K (60x 1-core, 55x all-core):



CB23 hits my 253 W power-limit, so CB24 uses less power here. Your CB23 score is 7.5% higher for 25% higher max wattage, but I assume your average power is lower!? If not then your CB24 score is 8.5% higher for about 30% higher wattage.

 

[Det0x]

Cinebench 24, multi and single core on undervolted 13900K (60x 1-core, 55x all-core):

View attachment 91149View attachment 91150

CB23 hits my 253 W power-limit, so CB24 uses less power here. Your CB23 score is 7.5% higher for 25% higher max wattage, but I assume your average power is lower!? If not then your CB24 score is 8.5% higher for about 30% higher wattage.

So i'm not quite sure what your trying to prove/achieve here ?

But since your so interested in power draw and undervolting i did a new run with hwinfo open, at the same performance target as you. (2333points)
Turns out i need 181w peak / 176w average package power for this.

Screenshot when finished with the run:


Are we done ?

 

[Timur Born]

Yes, the Ryzen should be more efficient than the Intel part for Cinebench type of load, thanks for providing the numbers for everyone to compare. Overclocking these CPUs further eats a lot of power, though, because they come long past the point of diminishing returns out of the box.

 

[MarkPost]

I have tested my 9950X with different configurations:

9950X @Stock (200W PPT) DDR5 5600

9950X @Stock (200W PPT) DDR5 6400

9950X @PBO tuned 180W PPT DDR5 6400


9950X @PBO tuned 200W PPT DDR5 6400


9950X @PBO tuned 230W PPT DDR5 6400


Graph with the scores:



First conclusion (already known anyways): memory frequency has a clear measurable impact in CB 2024 (~6% from 5600 to 6400, in my testing)
Second conclusion: yeah, Zen 5 is really efficient

 

[Det0x]

9950X (ES) maxed out on custom watercooling

Cinebench R23 MT = 50k
(ST = 2394)


Cinebench R24 MT = 2810
(ST = 149 i think it was)

 

[Det0x]

And a more power efficient Cinebench R23 MT run

49k @ 250w PPT

 

[MarkPost]

wow

Mine "only" reaches a score of ~47300 @250W in CB R23



My custom WC is a puppy compared to yours though😅

 

[Seba]

I sold my Ryzen 5 3600 (for 50 euro) and bought a new Ryzen 7 5700X3D (for 200 euro).


Ryzen 7 5700X3D (with 32GB DDR4 3200)
CPU Multi Core: 817
CPU Single Core: 87

Radeon RX 7600 8GB
GPU: 4853


The system is the same, with the exception of the CPU and the CPU cooler.

The 3600 had the stock Wraith Stealth cooler, which was not quite adequate for it (certainly not for benchmarks).

For 5700X3D I am using a Noctua NH-D15 (first generation) (using both fans) which I had for quite some time, but never bothered to use it on the 3600.


I did not expect such difference for the GPU score.



The results with Ryzen 5 3600:

Ryzen 5 3600 (with 32GB DDR4 3200)
CPU Multi Core: 538
CPU Single Core: 74

Radeon RX 7600 8GB
GPU: 3842


View attachment 86904

 

[igor_kavinski]

So even Cinebench GPU rendering can get CPU limited!

 

[Det0x]

9950X (ES) maxed out on custom watercooling

Cinebench R23 MT = 50k
(ST = 2394)
View attachment 110565

Cinebench R24 MT = 2810
(ST = 149 i think it was)
View attachment 110566

Updated Cinebench R24 score with the ASUS ROG CROSSHAIR X870E APEX, and like i said earlier in that thread, APEX is showing really good efficiency together with excellent memory performance (more information and runs can be found in linked thread)

• Delidded 9950X ES running PBO for all benchmarks
• ROG CROSSHAIR X870E APEX
• 2x16GB G.Skill Trident Z5 Royal Neo A-DIE @ 1.85 VDD / 1.55 VDDQ running ~8550MT/S CL32 GDM disable
• Very large custom cooling with mora

Cinebench R24 MT = 2868 points


Cinebench R23 MT = 50.2k points and ~19.6k in CPU-Z MT

 

[igor_kavinski]

Cinebench R24 MT = 2868 points

+24% over a typical 9950X. Hardcore OC scene is alive and well!

 

[Hail The Brain Slug]

+24% over a typical 9950X. Hardcore OC scene is alive and well!

I don't think it ever died. The only thing that changed is the proportion of effort to % gain is so extremely different from 10 years ago.

There is no longer a useful amount of low hanging fruit that practically anyone can dial in to overclock.

 

[igor_kavinski]

There is no longer a useful amount low hanging fruit that practically anyone can dial in to overclock.

I really wish AMD would sell an XS series chip with a small "direct die covering" copper block and define a new heatsink form factor so mere mortals can get in on the fun too. Maybe with a little engineering, it may even work with current AM5 heatsinks.