Discussion Cinebench 2026

[itsmydamnation]

bros is it time to hype 40% 1t

if its a qs/es and final is a few hunderd mhz higher right into ~35% T1.

 

[Joe NYC]

if its a qs/es and final is a few hunderd mhz higher right into ~35% T1.

What are you using for the base? Because the highest 9950x3d scores of 814 reflect some sort of tuning / overclocking.

Default 9950x3d may be in 775 range, which would make 1072 ~38% uplift.

 

[AMDK11]

Mike Clark

MC: It was actually dealing with two technologies [designing Zen 5 for both the 4nm and 3nm process technologies], especially a technology that the previous generation was in. And trying to do so much change, and therefore the unavoidable reality that in 4nm it's going to be [consume] more power than it's going to be in 3nm, no matter how smart we are.


But we need that flexibility in our roadmap, and it makes sense. But still that was really hard to try to control having the two technologies and the features, and a feature that looks great in 3nm not looking so great in 4nm because of the power impact of the not-as-efficient transistor and how it affects the floorplan. Normally, we do the architecture in one, and then we port on the next one, and then you have a lot of time to deal in the floor plan with the two technologies. [..] It was just really challenging. But that gives Zen 6 a lot of room to improve.

 

[eriksp92]

Here's my Core Ultra 5 125H (Zenbook 14 OLED) with HT on and off. The 255H in their results database must have the puniest cooling ever for me to beat it wil HT on despite it having two more P-cores and the Lion Cove/Skymont advantage. I also see why Intel wants to be rid of HT for consumers; the difference between it on and off in a laptop is miniscule.

 

[jpiniero]

bros is it time to hype 40% 1t

Single Core though.

You know what that means... SMT4!!!!

 

[Nothingness]

I also see why Intel wants to be rid of HT for consumers; the difference between it on and off in a laptop is miniscule.

HT brings 36% more performance in your ST vs SC results. Not that bad. The issue here is that the 125H has only 4 cores with HT.

Of course if all cores had HT, it's likely you wouldn't get 36% more MT performance due to other bottlenecks, but I'm not sure your result is enough to dismiss HT even on a laptop.

 

[eriksp92]

This is true, but if they're to keep to their strategy of P- and E-cores in a single chip where the e-cores are designed without HT (and are more numerous in all SKUs) in mind, HT won't add much more than 10 percent at best, and I imagine the scheduler complexity is a nightmare. Redwood Cove is also designed with HT in mind, so I would assume that Lion Cove has a front end designed to keep the core better fed. But I'm truly an amateur, so I claim no authority here!

 

[mikegg]

It was you who came up with the idea that AMD would have to improve efficiency by something like 720% to keep up with Apple.

"Apple Silicon is be 3-4x more efficient" = smoking the same crack pipe.

I think everyone will acknowledge that Apple silicon is more efficient than either Intel or AMD. But the gap is nowhere near either 300-400% or 700-800%

Can you quote me on the 700-800%?

The 300-400% is just data from Notebookcheck. How could I be wrong? It’s straight up facts.

 

[jdubs03]

Just doing some arithmetics, to try to fill in ST, assuming this is SC and further assuming that SMT ratio from Zen 5 processors holds.

1072 SC score would then translate to 800 ST score.

Given that highest M5 score that has been posted so far is 744 ST score, this would mean taking the ST lead from Apple.

"Big, if true"

Maybe it’s a LN2 cooled M5 Max running Parallels.

My other thought I’ve come around to is it is a Zen 6 ES/QS that is overclocked and LN2 cooled. That to me seems the most reasonable at this time.

 

[cherullo]

If this is Zen 5 it's over 7.3GHz. That is LN territory for sure but Detox could do it. Or he got a hold of some ES and got it up to like 6.4 or 6.5GHz.

Colorful Breaks Ryzen 7 9800X3D Frequency World OC Record with 7335.45 MHz

Colorful has claimed the world OC record for frequency with the AMD Ryzen 7 9800 X3D on its iGame X870E Vulcan OC motherboard. According to the HWBot listing submitted by Hero, the CPU reached a 7335.48 MHz peak clock under liquid nitrogen cooling, beating out the previous record—7313.05 MHz—by...

www.techpowerup.com

Coincidence?

 

[mvprod123]

Given that highest M5 score that has been posted so far is 744 ST score, this would mean taking the ST lead from Apple.

"Big, if true"

Unlikely, as the M6 is scheduled for release in November 2026.

 

[jdubs03]

Unlikely, as the M6 is scheduled for release in November 2026.

Indeed, the base M6 should get around 870 in ST.

 

[Joe NYC]

Can you quote me on the 700-800%?

Sure:

"Will Zen7 increase Strix Halo efficiency by 7.2x while also increasing ST performance by 2.2x?"

Page 18 - Discussion - Zen 7 speculation thread

Page 18 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[Joe NYC]

Unlikely, as the M6 is scheduled for release in November 2026.

Then, Zen 6 just needs to be released before November 2026

 

[Geddagod]

Just doing some arithmetics, to try to fill in ST, assuming this is SC and further assuming that SMT ratio from Zen 5 processors holds.

1072 SC score would then translate to 800 ST score.

Given that highest M5 score that has been posted so far is 744 ST score, this would mean taking the ST lead from Apple.

"Big, if true"

Also depends if Apple couldn't get a ~8% improvement with the M6 launching at a similar time then.
Let's see power too...

 

[poke01]

my i7 9750H 2019 16" Macbook pro

Windows:


macOS:

 

[Hulk]

my i7 9750H 2019 16" Macbook pro

Windows:
View attachment 136107

macOS:

View attachment 136108

Do you happen to know the average clock during those MT, SC, and ST runs? I'm curious as to how Coffee Lake (ie Skylake) does on this.

 

[poke01]

Do you happen to know the average clock during those MT, SC, and ST runs? I'm curious as to how Coffee Lake (ie Skylake) does on this.

I had a glance at task manager during the SC run, task manager reported 4.17GHz and it maintained that clock for the duration. I will check in depth later using HW info

 

[mikegg]

Sure:

"Will Zen7 increase Strix Halo efficiency by 7.2x while also increasing ST performance by 2.2x?"

Page 18 - Discussion - Zen 7 speculation thread

Page 18 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

How is that even remotely the same and why are people upvoting you? People are are upvoting misinformation just because it makes AMD look better.


This is what you said. Read the post again. Zen7 would need to increase efficiency by 7.2x over Strix Halo to match a theoretical M8 in Cinebench 2024.

I just calculated what Zen7 would actually need to do in order to beat an M8 in speed and efficiency, given that it's not an M4 that Zen7 would go up against. It'd be M8.

 

[poke01]

How is that even remotely the same and why are people upvoting you? People are are upvoting misinformation just because it makes AMD look better.

View attachment 136173
This is what you said. Read the post again. Zen7 would need to increase efficiency by 7.2x over Strix Halo to match a theoretical M8 in Cinebench 2024.

I just calculated what Zen7 would actually need to do in order to beat an M8 in speed and efficiency, given that it's not an M4 that Zen7 would go up against. It'd be M8.

IMO kinda pointless to talk about Zen7 and M8 efficiency this much when we haven’t even gotten their preceding architectures yet.

 

[mikegg]

IMO kinda pointless to talk about Zen7 and M8 efficiency this much when we haven’t even gotten their preceding architectures yet.

I agree. Hence, I wanted to show how silly it was for @Joe NYC to say Zen 7 was poised to take the performance and efficiency crown from Apple.

 

[DaaQ]

@Hulk more data lmk if you cant read out of SS.
HWinfo was only open for the ST test.
128GB 6000 34cas and 7950XTX w 550W bios but set at stock for test same with CPU 9800X3D ECO mode on default. Only EXPO turned on.
Tests used about 28GB ram, and GPU was at IIRC just over 16GB.


Below is GPU with undervolt OC and mem OC, gpu avg takes a hit on mhz from task manager compute image. Never hit 550w on RHS graph.
About 2k point increase.


EDIT:
So changing the TDP from eco to 105w, made little diff if not slightly lower, but made large impact on GPU score 61806 to 66352.

 

[Joe NYC]

How is that even remotely the same and why are people upvoting you? People are are upvoting misinformation just because it makes AMD look better.

It's probably the difference between sober analysis and a crack pipe that people were upvoting.

View attachment 136173
This is what you said. Read the post again. Zen7 would need to increase efficiency by 7.2x over Strix Halo to match a theoretical M8 in Cinebench 2024.

I just calculated what Zen7 would actually need to do in order to beat an M8 in speed and efficiency, given that it's not an M4 that Zen7 would go up against. It'd be M8.

And you made this projection, in the same post based on this assumption:

"M4 Pro is roughly 52% faster in Cinebench ST and 3.6x more efficient than Strix Halo."

Which is clearly a flawed assumption. Neither of the two (performance and efficiency) hold true when you do a "sanity check" by checking MT numbers.

So, you started with flawed assumption of +360% efficiency, doubled it and ended up twice as flawed +720% efficiency that AMD would have to make up.

It is a good idea to do a sanity check on any claim that looks insane, and MT is such a sanity check, because in MT test, the power consumption of the cores overwhelms all of the "noise" that exists in ST test. And that (including in current version of Cinebench) shows that your assumptions are off by an order of magnitude (10x).

Apple is leading, but by 1/10th of what you thought.

 

[Joe NYC]

I agree. Hence, I wanted to show how silly it was for @Joe NYC to say Zen 7 was poised to take the performance and efficiency crown from Apple.

I said "is determined to", which is an aspiration. Poised means more like inevitability, and that's not what I claimed.

 

[Hulk]

@Hulk more data lmk if you cant read out of SS.
HWinfo was only open for the ST test.
128GB 6000 34cas and 7950XTX w 550W bios but set at stock for test same with CPU 9800X3D ECO mode on default. Only EXPO turned on.
Tests used about 28GB ram, and GPU was at IIRC just over 16GB.

View attachment 136185
Below is GPU with undervolt OC and mem OC, gpu avg takes a hit on mhz from task manager compute image. Never hit 550w on RHS graph.
About 2k point increase.


EDIT:
So changing the TDP from eco to 105w, made little diff if not slightly lower, but made large impact on GPU score 61806 to 66352.

I can't extract your cpu average frequency from what you posted. You would need to watch in HWinfo during the 3 tests, or better/easier lock down max frequency in Power Plan.