• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Cinebench R23 efficiency race

Joined
Dec 1, 2020
Messages
567 (0.33/day)
Processor Ryzen 5 7600X
Motherboard ASRock B650M PG Riptide
Cooling Noctua NH-D15
Memory DDR5 6000Mhz CL28 32GB
Video Card(s) Nvidia Geforce RTX 3070 Palit GamingPro OC
Storage Corsair MP600 Force Series Gen.4 1TB
Everyone with every cpu and architecture is welcome to join in our Cinebench R23 efficiency race!
We have two categories:
6/12 cores up to 50W
8/16 cores up to 65W
Disabling of cores is allowed.
The only requirement is screenshot with BenchMate and is recommended to share a link to the result uploaded in HWBot

Ryzen 5 5500 3.925Ghz 1.1V
10005pts
50W
1659871567759.png
 
Last edited:
It's important to note that Benchmate only runs on Windows machines which excludes Apple Macs running processors based on Apple Silicon.

Cinebench R23 does have a Mac-compatible binary but because of this Benchmate requirement you've excluded all Macs.

If the Benchmate requirement wasn't present, I'd expect Apple Silicon-powered devices to crush the competition in performance-per-watt in Cinebench R23 comparisons.

I happen to have an Intel-powered Mac mini 2018 but I don't fulfill the requirements of including a Benchmate screenshot.
 
It's important to note that Benchmate only runs on Windows machines which excludes Apple Macs running processors based on Apple Silicon.
Yes it was meant more as AMD vs Intel competition
Cinebench R23 does have a Mac-compatible binary but because of this Benchmate requirement you've excluded all Macs.
Benchmate is required to ensure that the CPU does not exceed the maximum power consumption. Of course, this is not the whole consumption, but all systems are different and that is why the consumption is limited only to CPU/SOC level.
If the Benchmate requirement wasn't present, I'd expect Apple Silicon-powered devices to crush the competition in performance-per-watt in Cinebench R23 comparisons.
It would be interesting to see also M1 family, but I don't know how to detect the exact consumption. Also M1 should be compared to mobile CPUs because that's the target it hits. For example 6900hs can score 13500+ with 35W and according Anandtech M1 max hit 12400 with 30w(34w - 4w idle) so "crush" is too strong statement
I happen to have an Intel-powered Mac mini 2018 but I don't fulfill the requirements of including a Benchmate screenshot.
We can include results without screenshots, but we won't be able to declare them as winners cause with such low consumption each watt can drastically changed the final result and is possible false result reporting - from a human and technical point of view
 
Last edited:
Here's my result with 65W Limit set in bios (Benchmate recorded 66W during the test)
I know..The 12900k doesnt fit into either 6/12 or 8/16 categories.

1659894124177.png
 
Here's my result with 65W Limit set in bios (Benchmate recorded 66W during the test)
I know..The 12900k doesnt fit into either 6/12 or 8/16 categories.

View attachment 257292

Why do you have your GSkill DDR5 6000 running at 5800?
I got a set of Corsair Dominator Platinum 6000 36-36-36-76.
 
Why do you have your GSkill DDR5 6000 running at 5800?
I got a set of Corsair Dominator Platinum 6000 36-36-36-76.

Ah, good catch.. I was messing around with it a while back trying to see if i could get CL30-32 stable so I had it set at 5800 while doing that.
Forgot to set it back when i set the timings back to auto/xmp.
 
Here's my result with 65W Limit set in bios (Benchmate recorded 66W during the test)
I know..The 12900k doesnt fit into either 6/12 or 8/16 categories.
12900k with disabled e-cores fit perfectly. Thanks for your result
 
12900k with disabled e-cores fit perfectly. Thanks for your result

Note: no e-cores were disabled in that run as it is against the rules...

Err, guess not.. i must have misread that, disabling of cores IS allowed.
 
Here you go, 5950x @ 51w PPT = 20k points :)
All 16core/32 threads enabled
Can test to how high i can reach @ 65watts later if there are a interest for that :)

This is the same cpu with unlimited amounts of power:
1659912323543.png
 
Last edited:
Note: no e-cores were disabled in that run as it is against the rules...

Err, guess not.. i must have misread that, disabling of cores IS allowed.
Sorry, yesterday I was on my phone and didn't saw very well your screen. The idea of this race was to sее how effective are the cores of AMD and Intel(the P-cores of AL) and what if12900k was 16p + 0e cores. Since 12900k has 8p cores, the max number of cores was limited to 8/16 and 6/12 with low power limits. Now I see that there is a desire for more categories with more different configuration of cores and consumption, but I am not sure how to proceed.

This question is to everyone - Do you want to have additional categories and what because the configurations of consumption and number of cores is to high and obviously CPUs with more cores always will have advantage if the power limit is same for everyone?

P.S. This 5950x running 3.125Ghz and 0.7v make me fell like the guy with the worst binned cpu in the world. Nice CPU!
 
This is going to come down to cooling...
 
Is the limit here 65~ watts?

OP should set up the weight classes :nutkick:

For AMD, good presets are Eco mode (45W, 54W PPT specification for 65W TDP CPUs), Eco mode 2 (65W, 84W PPT for 105W spec CPUs), the regular 65W (again with 84 PPT) and 105W (142W PPT) stock configurations :)

I'll try to get a contribution in shortly
 
The idea of this race was to sее how effective are the cores of AMD and Intel(the P-cores of AL) and what if12900k was 16p + 0e cores. Since 12900k has 8p cores, the max number of cores was limited to 8/16 and 6/12 with low power limits. Now I see that there is a desire for more categories with more different configuration of cores and consumption, but I am not sure how to proceed.

This question is to everyone - Do you want to have additional categories and what because the configurations of consumption and number of cores is to high and obviously CPUs with more cores always will have advantage if the power limit is same for everyone?
Why does the amount of cores matter ? i thought the goal was to get the highest amount performance/watt by any means possible ?

And no, the highest core count wont always win.. Go to high in that departement and the interconnects (IO-die) start using 50watt alone -> 64core epyc/xeon can not even compete in these 50watt and 65watt classlimits..

In my humble opinion, dont care about core amounts, only rank performance/power.
(Cinebench-points/watts)
 
Why does the amount of cores matter ? i thought the goal was to get the highest amount performance/watt by any means possible ?

And no, the highest core count wont always win.. Go to high in that departement and the interconnects (IO-die) start using 50watt alone -> 64core epyc/xeon can not even compete in these 50watt and 65watt classlimits..

In my humble opinion, dont care about core amounts, only rank performance/power.
(Cinebench-points/watts)
I am talking about mainstream platforms and focus on most effective cores that doesn't mean they won't be interesting to see 64 core cpu, but it turned out that 6 and 8 cores are not enought.
Single core is not good example for most efficient core so we need more cores to achieve real results in multi-threaded tasks while the power limit for specific number of cores will ensure that CPUs with 6-8 cores won't be smoked by CPUs with 12-16 cores, at least not because the CPUs with more cores are clocked much lower, but because of the better bin.
 
78W limit results in 11190pts at CB23. The result for the 50W limit is on the screenshot below. Also the load temp dropped to 53C from 64C when on the 78W limit. 64% of the 78W got to 74% of its performance.
CB23 CPU test@50W.jpg
 
78W limit results in 11190pts at CB23. The result for the 50W limit is on the screenshot below. Also the load temp dropped to 53C from 64C when on the 78W limit. 64% of the 78W got to 74% of its performance.
View attachment 257352
This is about the expected result for a 5600 at 50W; OP's 10000 for a 5500 seems really fishy though...
 
65w 8GC cores - E cores off

@Valantar
@mahirzukic2

We could continue the discussion here

If 8 zen 3 cores break this score at 65w ill admit I was wrong all along. Btw this is with just UV, wth static OC it can do better but it doesn't need to, it already absolutely scorches 8 zen 3 cores in efficiency.
16724.png

This is about the expected result for a 5600 at 50W; OP's 10000 for a 5500 seems really fishy though...
You know what's really fishy? Zen 3 are the best selling CPU's, but nobody is here with 8 zen 3 cores willing to test at 65w. Makes you wonder why, doesn't it? :roll:
 
Last edited:
Since my 5950x 20k score @ 50w dont seem to fit in this narrative that's being painted here, i have disabled half my cpu and ran it well outside the efficiency range with only 1 CCD.
Core for core GC is stronger than Zen3, but they are so much bigger that they could only fit 8 of those on the consumer-die before they had to start adding e-waste cores to make up for MT deficit :laugh:

8 Zen3 cores @ 41 watt = 12603 points
3625mhz @ 0.8vcore
12603points/41watt = 307 points per watt
1660226484156.png

8 Zen3 cores @ 50 watt = 13698 points
3950mhz @ 0.88vcore
13698points/50watt = ~274 points per watt
1660226549297.png

8 Zen3 cores @ 64 watt = 15099 points
4350mhz @ 0.98vcore
15099points/65watt = ~232 points per watt
1660226643908.png
Silicon quality for my 5950x

When i run whole cpu with all 16 cores enabled in the correct efficiency range i'm getting the following:
20681points/51watt= 405 points per watt

Compared to AL entries above:
16800points/65watt= 258 points per watt

As we see above, sweetspot for Zen3 is way below even 3625mhz, and that's why all higher core count ryzen CPU's absolutely scorches Alder Lake in efficiency :)
What numbers can you guys get with GC cores @ 50w max ?
 
Last edited:
Since my 5950x 20k score @ 50w dont seem to fit in this narrative that's being painted here, i have disabled half my cpu and ran with only 1 CCD.
Core for core GC is stronger than Zen3, but they are so much bigger that they could only fit 8 of those before they had to start adding e-waste cores :laugh:

8 Zen3 cores @ 41 watt = 12603 points
3625mhz @ 0.8vcore
View attachment 257685

8 Zen3 cores @ 50 watt = 13698 points
3950mhz @ 0.88vcore
View attachment 257686

8 Zen3 cores @ 64 watt = 15099 points
4350mhz @ 0.98vcore
View attachment 257687

Silicon quality for my 5950x

Sweetspot for Zen3 is way below even 3625mhz, and thats why higher core count CPU's absolutely scorches all Alder Lake cpus in efficiency :)
So GC is both faster and more efficient than zen 3 in CBR23. Thanks for posting

When i run whole cpu with all 16 cores enabled in the correct efficiency range i'm getting the following:
20681points/51watt= 405 points per watt

Compared to AL entries above:
16800points/65watt= 258 points per watt
Because you are comparing 16 cores to 8? LOL

When I run Alderlake with all 16 cores enabled I get
15215 points / 35watts = 434 points per watt.

So with half the cores, with E waste cores and with less threads alderlake > zen 3 ;)

It's clear by now that 16P cores at 125w would absolutely annihilate not just zen 3, but zen 4 as well
 
Last edited:
So GC is both faster and more efficient than zen 3 in CBR23 @65w. Thanks for posting
Fixed. ;)

Not trying to cause a flame war, but we will need to see various arbitrary power limits to get an idea of the efficiency curves of both architectures. @Det0x has posted 3 different power levels - to make a meaningful comparison you should post the same power levels and scores.

Also, is CBR23 sensitive to RAM speed?
 
Fixed. ;)

Not trying to cause a flame war, but we will need to see various arbitrary power limits to get an idea of the efficiency curves of both architectures. @Det0x has posted 3 different power levels - to make a meaningful comparison you should post the same power levels and scores.

Also, is CBR23 sensitive to RAM speed?
He can pick whatever power levels, 8 vs 8 it stands no chance. I'm not even manually tuning man, im just letting the motherboard figure out the core clocks, with static OC I would get higher numbers. But again, don't need to, it's already beating the zen 3 quite handily

No, it's not sensitive to ram, the difference between ram speeds is tiny, but thanks for reminding me, I can probably score better if I turn down the IMC voltages and run wtih stock ram :P
 
Because you are comparing 16 cores to 8? LOL

When I run Alderlake with all 16 cores enabled I get
15215 points / 35watts = 434 points per watt.

So with half the cores, with E waste cores and with less threads alderlake > zen 3 ;)

It's clear by now that 16P cores at 125w would absolutely annihilate not just zen 3, but zen 4 as well
Well, we are still comparing apples to oranges. ADL 434 points per watt @35W does not compare exactly to Z3 405 points per watt @51w.

If we could see 8x and 8x ADL P-cores Z3 at say 35w, 45w, 65w, 95w, 105w we could interpolate the efficiency curves. Right now, both sides are cherry-picking.
 
to make a meaningful comparison you should post the same power levels and scores.
Here you go50w.png

Well, we are still comparing apples to oranges. ADL 434 points per watt @35W does not compare exactly to Z3 405 points per watt @51w.

If we could see 8x and 8x ADL P-cores Z3 at say 35w, 45w, 65w, 95w, 105w we could interpolate the efficiency curves. Right now, both sides are cherry-picking.
Nope, only det0x is comparing apples to oranges, since he picked the score of 8 GC cores at 65w and compared it to 16 zen 3 cores at 50w. Of course afterwards I did the same, to point exactly what you just described, the cherrypicking.


I only wanted to see if core for core zen 3 or GC is more efficient, and the results speak for themselves I think, 8 GC cores slam the 8 zen 3 cores both in performance and efficiency. There is no wattage point that zen 3 do better than 8 GC cores. Heck, im not even sure Zen 4 can do better at this point.
 
Back
Top