【Maxon】Cinebench R20 Benchmark Thread

[Head1985]

Yeah I raised it to 1.475v and I'm pulling that when I use the profile and it seems fine.
I have the Noctua NH-D15, in the CoolerMaster H500P Mesh case.
I have some pictures of it here if you're interested.

1,475v???Your cpu will degrade very fast.Max safe voltage for 24/7 is 1.32v.

 

[burninatortech4]

3200mhz Flare X 14-14-14-34 1.35v

Can't seem to get 14-14-14-28 stable

 

[DrMrLordX]

2700x is no more "monolithic" than a 3900x. The two CCXs are separated by IF links. Look at intercore latency tests to see what I mean.

 

[nicalandia]

2700x is no more "monolithic" than a 3900x. The two CCXs are separated by IF links. Look at intercore latency tests to see what I mean.

True

would be nice to see more scores of 8C/8T Zen/Zen+ and Zen 2 on CB R20,

Strangely enough this 2x2 with SMT Off Zen processor(1200) shows nearly perfect scaling.

 

[Kenmitch]

True

would be nice to see more scores of 8C/8T Zen/Zen+ and Zen 2 on CB R20,

Strangely enough this 2x2 with SMT Off Zen processor(1200) shows nearly perfect scaling.

Here's my 3700x with PBO enabled and SMT turned off.



I didn't lock the clocks so the single core did boost higher. Does your cpu exhibit the same behavior?

 

[Hitman928]

Yeah, you can't compare the MP ratio between CPUs without taking the boost behavior into account. If you want to just see how well Cinebench scales from single to multiple cores between different CPUs, you'll need to make sure the frequency is static and not changing between the single and multi-core test.

 

[lopri]

^ So right. See also Amdahl's Law.

 

[lopri]

3200mhz Flare X 14-14-14-34 1.35v

Can't seem to get 14-14-14-28 stable

tRAS does little to nothing for performance. And this benchmark cares memory even less. I would not worry about either.

 

[nicalandia]

Here's my 3700x with PBO enabled and SMT turned off.

I didn't lock the clocks so the single core did boost higher. Does your cpu exhibit the same behavior?

Without Locking the Clocks it looks like 93% Core scaling, please take a moment to lock the clocks(anything from 3.8 to 4 Ghz would be good since we are after core scaling at lock speed) and see how much that improves.

 

[Kenmitch]

Without Locking the Clocks it looks like 93% Core scaling, please take a moment to lock the clocks(anything from 3.8 to 4 Ghz would be good since we are after core scaling at lock speed) and see how much that improves.

Locked at 4.0 GHz

SMT OFF



Same clocks SMT ON

 

[nicalandia]

Locked at 4.0 GHz

SMT OFF

Same clocks SMT ON

455 ST if 100% CPU scalling the number would have been 3640, but at 96% we get the number 3492(really good numbers) with SMT On we can see it gets a boost of 36% which is IMHO an amazing boost in performance. Thanks for posting this

 

[nicalandia]

It would be really interesting to know why user: lopri is getting worst CPU(94%) and SMT Scaling(28%),

It wasn't easy.

View attachment 8606

View attachment 8621

No meaningful difference in ST performance.

 

[Pandora's Box]

 

[DrMrLordX]

I tried running asynchronous speeds and lost a lot of performance doing so. Glad you were able to make it work.

 

[Pandora's Box]

I tried running asynchronous speeds and lost a lot of performance doing so. Glad you were able to make it work.

Yeah I also prefer ccx overclocking vs all-core as it allows the CPU to down-volt at idle.

 

[DrMrLordX]

Yeah I also prefer ccx overclocking vs all-core as it allows the CPU to down-volt at idle.

Hmm, mine can drop voltage even when going for all-core static. Maybe it won't do that if you set a hard voltage in the UEFI? I'm using Ryzen Master for everything these days.

 

[lopri]

It would be really interesting to know why user: lopri is getting worst CPU(94%) and SMT Scaling(28%),

Your guess is as good as mine. Don't forget that these CPUs have higher ST boost clocks than MT boost clocks.

 

[nicalandia]

Your guess is as good as mine. Don't forget that these CPUs have higher ST boost clocks than MT boost clocks.

It must be an algorithm issue since you don't have a Flat OC it's taking ST x MT = MP ratio, any chance that you do a Flat OC run? Even at 3.8 OC I believe we will see the true numbers

 

[SlowSpyder]

3509 - Ryzen 3600 under Coolermaster 240 (2x120mm) AIO. 16GB @ 2133MHz.

 

[nicalandia]

3509 - Ryzen 3600 under Coolermaster 240 (2x120mm) AIO. 16GB @ 2133MHz.

Thanks, please try to post info using SMT Off Single core and MT(8C/8T) with a flat OC.

 

[Markfw]

7025 and 497 3900x, 3600 memory stock, 1800 fclk

Bad...On my new x570 motherboard with the latest bios and all driver updates !

 

[DrMrLordX]

@Markfw

I've gotten a lot of results that low running stock on my system. With 1.0.0.3ABB and the latest chipset drivers, my boost clocks have been pretty low unless I start playing silly games with LLC.

 

[Markfw]

OK, EPYC 7742 (but says ES) running@2.6 ghz. 2nd run was 20,977 ! And max SYSTEM power draw from the wall was 275 watts ! And I have 2 NVME drives.

 

[EXCellR8]

showoff...

last run of my 3700X netted 4793cb ...

 

[EXCellR8]

this one slightly lower:

Happy to report my temps are better though, CPU never went over 58c during the multi-core test