Cinebench 2024 Out

D

Down8

2[H]4U
Joined
Sep 21, 2003
Messages
3,293
https://www.xda-developers.com/cinebench-2024-released/

It not only re-introduced GPU testing, but also compares Apple silicon & other ARM/Snapdragon stuffs. I'm sure other sites have discussion/reviews as well.

It did crash my PC the first time I opened it, then multiple times there after, as I re-ran my OC stability... only to wind up exactly where I started. Weird.

Quick run before my RX7800XT arrives: 4734/1079/116:
Screenshot 2023-09-06 185742.png


-bZj
 
Ryzen 7 3700X Stock : 714
RTX 2080Ti Stock : 8865

Doesn't run without AVX2, so I can't test it on my older PC.
 
Ryzen 7 7700X (TDP restricted to 80w): 993 multi core / 110 single core
RTX 3060 TI (undervolted & OCed): 9948
 
My CPU score seems a bit off (700-800, didn't let it finish 10 minutes)
RTX 4090: 33850
 
Reminder to myself to run this when i'm done with my current batch of Stable Diffusion images.
 
downloading now to run on my 6700k and to compare to my 7800X3D after I build it later this week :)
 
Done testing my Ryzen 7500F build (cooler using AIO Corsair H115i) and this is the result:

All at default settings:
Figure 3.JPG


This one below running at 5300mhz 1.235v all core OC:
Figure 5.JPG


Summary:
1. At default, ryzen 7500F beat ryzen 5600 (same MSRP at release) by 26% for both single and multi core*.
That's quite impressive gen to gen improvement.

2. At all core OC 5300mhz, ryzen 7500F beat ryzen 5800x by 3% for single core& 1% for multi score*.
So 6 core ryzen 7000 can match 8 core ryzen 5000.

*The cinebench 2024's ryzen 5000 score from internet.
 
  • Like
Reactions: TMCM
like this
managed to fix my CPU a bit . not much play with the 7800X3D
GPU stock (factory OC)

GPU 33835
CPU Multi Core 1014
CPU Single Core 109


1694447368682.png
 
Ryzen 5950x with a 3090 system #1
GPU 18406
Multi 1375
Single 98

Ryzen 5950x with a 3090 system #2
GPU 18462
Multi 1351
Single 97

Ryzen 5800x with a 3080 10gb
GPU 13478
Multi 866
Single 97

Ryzen 3600xt with a 3080 10gb
GPU 13505
Multi 589
Single 78

Ryzen 3600x with a 3080 10gb
GPU 13263
Multi 563
Single 80

ETA 3600x, 3600xt
 
Last edited:
applegrcoug said:
Ryzen 5950x with a 3090 system #1
GPU 18406
Multi 1375
Single 98

Ryzen 5950x with a 3090 system #2
GPU 18462
Multi 1351
Single 97

Ryzen 5800x with a 3080 10gb
GPU 13514
Multi 866
Single 97
Click to expand...
Out of curiosity, what are your overclock and ram settings? I'm gonna be comparing this to my 5950X system soon.
 
TheSlySyl said:
Out of curiosity, what are your overclock and ram settings? I'm gonna be comparing this to my 5950X system soon.
Click to expand...
Both are running 3600mhz ram. System #1 has PBO on along with per CCX offsets. However, system #2 needs some further investigation. It has one of those fancy boards that runs PBO on single core stuff and then should switch to an all core OC on multi thread, so it's multi-thread should have been markedly higher.
 
applegrcoug said:
Both are running 3600mhz ram. System #1 has PBO on along with per CCX offsets. However, system #2 needs some further investigation. It has one of those fancy boards that runs PBO on single core stuff and then should switch to an all core OC on multi thread, so it's multi-thread should have been markedly higher.
Click to expand...
How are the temps on the 5950X with the all-core tuning?

When I tried the "Active OC Tuner" feature on my Gigabyte X570S board with 5900X / 5950X I ran into some issues with really heavy all-core stress like Cinebench / P95 / Y-cruncher where the chips would get thermally overwhelmed and my scores were messed up because cores were momentarily dropping to 500Mhz during the run.

I wish these "PBO + all-core OC" features also had an upper limit that re-enables PBO under extreme load so the chip can float the clocks instead of hard throttling. The conundrum I ran into was that an all-core OC that's suitable for moderate multicore loads would catastrophically overheat under 100% load, and an OC that's thermally safe for 100% load ended up being worse than PBO under moderate load.
 
NattyKathy said:
How are the temps on the 5950X with the all-core tuning?

When I tried the "Active OC Tuner" feature on my Gigabyte X570S board with 5900X / 5950X I ran into some issues with really heavy all-core stress like Cinebench / P95 / Y-cruncher where the chips would get thermally overwhelmed and my scores were messed up because cores were momentarily dropping to 500Mhz during the run.

I wish these "PBO + all-core OC" features also had an upper limit that re-enables PBO under extreme load so the chip can float the clocks instead of hard throttling. The conundrum I ran into was that an all-core OC that's suitable for moderate multicore loads would catastrophically overheat under 100% load, and an OC that's thermally safe for 100% load ended up being worse than PBO under moderate load.
Click to expand...
So I tried running it again and the temps were in the mid-50s until it crashed. It is on a custom loop with a massive radiator, so temp shouldn't be an issue as long as the block can take it away. Makes me think while Prime95 stable, it isn't stable for cinebench24.
 
New AMD GPU drivers make a world of difference. Prior to update, my 7900XTX was getting around 5k, with the Sept update I'm hitting 14888. (y)
 
first time running this. excited to see what my new 7800X3D build will do after I build it later this week. no idea why it says 4.01Ghz?
 

Attachments

  • cinibench gpu score.jpg
    cinibench gpu score.jpg
    290.6 KB · Views: 0
  • cinibench single core score.jpg
    cinibench single core score.jpg
    237.5 KB · Views: 0
  • cinibench multicore score.jpg
    cinibench multicore score.jpg
    232.8 KB · Views: 0
Last edited:
Seems like I am right where I want to be! :) I use my 5950X different than most, with an all-core OC locked with no PBO and SMT turned off.... makes for an extremely smooth, stutter-free and seamless gaming experience. :cool:

1694529064590.png
 
III_Slyflyer_III said:
Seems like I am right where I want to be! :) I use my 5950X different than most, with an all-core OC locked with no PBO and SMT turned off.... makes for an extremely smooth, stutter-free and seamless gaming experience. :cool:
Click to expand...
So when you did that, what multiplier did you use and then, did you prime95 (or something) to test it? Also, what about voltage?

I just ditched my OC and went to the PBO and scored 20 more points. It also looks like my voltage went up during the run, which probably means I have more room to give it more juice and increase the clocks.
 
applegrcoug said:
So when you did that, what multiplier did you use and then, did you prime95 (or something) to test it? Also, what about voltage?

I just ditched my OC and went to the PBO and scored 20 more points. It also looks like my voltage went up during the run, which probably means I have more room to give it more juice and increase the clocks.
Click to expand...
On my specific setup, multiplier is 47x, BCLK is 100.50 and vCore is at 1.35V with LLC @ second most aggressive level. 100% Prime95 stable, I have had 0 crashes since going all core OC route, whereas even after a year+ of tweaking PBO, I'd still get a random crash now and then. System runs worlds better setup this way and also runs a lot cooler too. I don't even break 74C under full Prime95 torture testing and gaming it hovers in the mid 50C range tops.
 
Ryzen 5950X, 64GB DDR3200 RAM, RTX 3080 12GB
I still had background processes running, which is a decent list of stuff, (Plex, etc.) but I wasn't actively using the computer for things.
1694537249057.png

First: I have my entire setup power limited. My CPU is limited to 200W, my GPU is limited to 350W

Noticed some weird performance things, my Single Core score is weirdly low, my single CPU wasn't clocking right. Only seemed to be hitting about 4.5ghz, when normally it goes over 5.1ghz single core, my multi-core looks right. Roughly 4.4-4.5~ghz all core for the duration.
Both are slightly hindered by my slower ram speed.

My GPU score, while high, was interesting. My GPU rarely hit over 300W and stayed around the 1990mhz range, when I'm going all out on my GPU, its usually 350-400w and around 2050 sustained mhz.

Oh yeah, I use an Arctic LFII 420 on my 5950X and temps usually stay around 75c all core load.
 
Last edited:
Well, 4734 on the RTX2060, upgraded to just over 11000 on the fresh RX7800XT. Seems a bit low...? But only in comparison to others' scores. 2.33x is a fair bump. Oh, the RX7800XT was "auto-clocked" to 2703, I think.

Screenshot 2023-09-12 210743.png


-bZj
 
Dang my new 7800X3D build is quite the jump up from my 6700k. Went from GPU: 18461 to 19686 (same gpu), Multi core: 320 to 1070, Single core 73 to 111

cinebench 2024 sept 18 2023.jpg
 
applegrcoug said:
So I tried running it again and the temps were in the mid-50s until it crashed. It is on a custom loop with a massive radiator, so temp shouldn't be an issue as long as the block can take it away. Makes me think while Prime95 stable, it isn't stable for cinebench24.
Click to expand...
prime95 also has a setting to enable avx during the test. Running avx code is much more prone to crashing from overclocking, but when would you realistically run avx instructions for that long?
 
An interesting discovery...

Cinebench 2024 seems to like dual ranked memory...

All tests on x570S Aorus Master running a 5950x all core at 45

Test #1 using two single rank 16gb dimms at 3600mhz: multi 1355
Test #2 using two dual rank 16gb dimms at 3600mhz:multi 1619
Test #3 using four single rank 8gb dimms at 3600mhz: multi 1616

More tests later as I have time.
 
1696111344397.png


Edit: I just have PBO on for the 5950x nothing special - GPU is auto-OC via GeForce Experience (GPU still capped @ 200watts) nothing special there either
 
Last edited:
  • Like
Reactions: TMCM
like this
More test results looking at the impact of ram...

I updated the bios on both systems and it wiped out all my PBO curve settings. I've spent the last day+ running core optimizer to rebuild them.

This series of tests, were using PBO with the curves optimized. Because it was done with PBO, I decided to test single thread.

All tests are with 32gb 3600mhz memory in two dimms

Test #1 5950x with dual rank ram 1580 multi and 101 single
Test #2 5950x with single rank ram 1372 multi and 99 single

Next tests were on a different system running a 3600x on a B550 board with the same sticks of ram.

Test #3 3600x locked to 41.5 with single rank ram 586 multi
Test #4 3600x locked to 41.5 with dual rank ram 592 multi
Test #5 3600x PBO with single rank ram 571 multi
Test #6 3600x PBO with dual rank ram 590 multi
 
The only test I actually bothered with is the CPU multi core test, just to make sure the cooling on my 7800X3D was sufficient.
I got an 1100 score. I think that's pretty average. My ram is a bit tweaked, to Buildzoid's low effort Hynix timings, but I kind of doubt that changes much. I also had a bunch of crap running in the background.
1696229474330.png


The cooler I have is the Noctua NH-U14S, and it seems to be pretty much perfectly rightsized for this CPU. Temps were briefly at around 81, but then went down to 77-78 and just stayed there. I think the throttle point on this CPU is 87, so I guess this cooler is pretty much just perfectly right for this CPU. Though if I didn't get this open box at half price, I totally would have picked the Thermalright Peerless Assassin. More fans, more fins. Just better for the price, honestly. I think I'm starting to far prefer air coolers nowadays.
 
Last edited:
You must log in or register to reply here.
Back
Top