Cinebench R23 efficiency race

[glnn_23]

12900k 8 (P) cores at 50W.

15276 pts

 

[usiname]

This is about the expected result for a 5600 at 50W; OP's 10000 for a 5500 seems really fishy though...

Did you missed the HWBot url, the result cannot be manipulated

 

[openbox1980]

i9 12900h laptop processor at 50w. 6 p-cores, 8 e-cores





This is at 65w rated 6p 8c-cores


This one is for $hit and giggles, rated at the full 127w 6p, 8-cores

 

[freeagent]

When I limit my CPU to 50w my score is pitiful..

Eco mode probably performs better lol.

That is both CCD's, 12/24, the only thing I changed with my PPT limit from 235 to 50w.

It scores like an R5 3600 I think.

 

[openbox1980]

I redid it again with the BM app
at 50w i9 12900h laptop processor 6p, 8e-cores






at 65w



at 127w, the full power rated of the processor.


This is at 80w, the max wattage allowed when using the graphics card

 

[Psychoholic]

This seems like a bizarre topic to argue about, but here we are and I'm enjoying the read.. still too early for popcorn though

 

[Taraquin]

I get 11850 running 5600X -30CO, 0 PBO at 76W limit. 50W limit gives me over 85% of that score. Not bad. Effective clocks in hwmon is 3974-4049MHz.

I run Ram tuned at 3800cl15. SOC 1.06v, IOD 0.98v, CCD 0.8v, VDDP 0.8v, VDD18 1.6v. This reduces I\O-die consumption quite a bit vs stock SOC 1.2v, IOD\CCD 1v, VDDP 0.9v and VDD18 1.8v. I see benchmate makes one mistake and thinks my MB has 4dimms, it only has 2.

Currently my I\O-die uses about 18-19W during load. If I run ram at 2133 I could probably run SOC at 0.9v, IOD at 0.8v and mayby shave off 5W more for 2-300MHz higher clockspeed and about 500 more points on CB23, but in gaming etc my PC will be waaaay slower then.

This is about the expected result for a 5600 at 50W; OP's 10000 for a 5500 seems really fishy though...

If SOC, IOD, CCD, VDDP and VDD18 is run as low as they can and you use curve optimizer for undervolt then 10k on 5500 is doable. My 5600X do 10350 at 50W Running stock on everything it gets around 8500.

 

[Voluman]

Lenovo Legion 5 notebook @ default, very basic bios, can't set anything.

 

[agent_x007]

So I'm lazy and just used 65W 8C/16T limit for my 10-Core/20-Thread CPU.

Pretty good score for 6 year old tech I think
Also, more transisors with less volts = better score as this benchmark scales to any number of threads.
I wonder how low Vcore a Tr Pro 5975X would need, to get 65W...

PS. Benchmate 10.10.0 or later should be mandatory for good CPU Package Power reading.

 

[HenrySomeone]

This seems like a bizarre topic to argue about, but here we are and I'm enjoying the read.. still too early for popcorn though

I don't know about bizarre, but it's more and more obvious that core-for-core Alder Lake isn't just (way) more powerful, but also more efficient, just like fevgatos has been saying for months (and Randallflagg before him) and got nothing but dismissals and ridicule from ardent team red members...

 

[Det0x]

I finally remembered that i had to use the amd agesa part of the bios to set lower than 0.8 vcore with a static OC.
So i have done some more runs, mainly with all 16 cores enabled.

Few words before we dive into data:

• There have been no telemetry trickery with these numbers
• SOC powerusage is ~2w together with 13w usage for rest of CPU. Anything over this constant ~15w draw is going to the cores/caches on the CCD.
• For all the new runs a very wimpy LLC together with set 0.75vcore was used to reach close to 0.7v under load on the lowest power runs

Below we have results for a simulated 8core Zen3 5800x:

8 Zen3 cores @ 33 watt = 11149 points in Cinebench r23 (~lowest powerlevel i could run with my current setup)
3200mhz @ 0.719 vcore under load -> ~2.25watt per core under load
11149points/33watt = 337 points per watt


8 Zen3 cores @ 41 watt = 12603 points in Cinebench r23
3625mhz @ 0.8 vcore under load -> ~3.25watt per core under load
12603points/41watt = 307 points per watt


8 Zen3 cores @ 50 watt = 13698 points in Cinebench r23
3950mhz @ 0.88 vcore under load -> ~4.375watt per core under load
13698points/50watt = ~274 points per watt


8 Zen3 cores @ 64 watt = 15099 points in Cinebench r23
4350mhz @ 0.98 vcore under load -> ~6.125watt per core under load
15099points/64watt = ~235 points per watt


Results for a real 16core Zen3 5950x:

16 Zen3 cores @ 49 watt = 20441 points in Cinebench r23 (~lowest powerlevel i could run with my current setup)
3100mhz / 2950mhz @ 0.7 vcore under load -> ~2.125watt per core under load
20441points/49watt = 417 points per watt


16 Zen3 cores @ 64 watt = 23701 points in Cinebench r23
3575mhz / 3425mhz @ 0.781 vcore under load -> ~3.063watt per core under load
23701points/64watt = 370 points per watt


16 Zen3 cores @ 88 watt = 27037 points in Cinebench r23
4075mhz / 3925mhz @ 0.887 vcore under load -> ~4.5625watt per core under load
27037points/88watt = 307 points per watt


So what have we learned by this comparison ?
A underclocked 12900k can be more efficient then a underclocked 5800x when you handicap Zen3 with its size advantage and only compare core for core with GC, but at the same time it cant touch a underclocked 5950x in energy efficiency as the numbers show.

Like i said earlier, in the end it all boils down to GC physical size, they are so big that intel could only put 8(10) of those on a consumer cpu(die) and keep the price in check at the same time. (10P cores would score lower than 8p+8E in full multithreaded benchmarks)

But that is no reason to handicap desktop Zen3 with a artificial limit for 8 cores maximum in this efficiency comparison when we both have to 5900x and 5950x as normal desktop consumer cpus for sale today

A too expensive to produce hypothetical 16core GC could properly match and beat the 5950x in MT efficiency with ease, but it would meet the same fate against a 32core Threadripper. There is a reason why Intel dont have any HEDT class products for sale atm.

If you think i'm wrong, please share how many points Alder lake can put on the board in Cinebenchr23 at the given powerlevels 50w, 65w and 88w as the OP asked for

 

[Psychoholic]

I don't know about bizarre, but it's more and more obvious that core-for-core Alder Lake isn't just (way) more powerful, but also more efficient, just like fevgatos has been saying for months (and Randallflagg before him) and got nothing but dismissals and ridicule from ardent team red members...

Yep, I noticed day one (around launch day) when i bought my 12900K that i could take it from the default 241W down to around 170-180ish watts and not really lose anything performance wise.

it would have been better received had it launched with a TDP less than 200W, while keeping pretty much the same performance.

 

[HenrySomeone]

So what have we learned by this comparison ?
A underclocked 12900k can be more efficient then a underclocked 5800x when you handicap Zen3 with its size advantage and only compare core for core with GC, but at the same time it cant touch a underclocked 5950x in energy efficiency as the numbers show.

A too expensive to produce hypothetical 16core GC could properly match and beat the 5950x in MT efficiency with ease, but it would meet the same fate against a 32core Threadripper. There is a reason why Intel dont have any HEDT class products for sale atm.

If you think i'm wrong, please share how many points Alder lake can put on the board in Cinebenchr23 at the given powerlevels 50w, 65w and 88w as the OP asked for

Handicap Zen3? Core-for-core comparison is the sole purpose of this thread! But because that doesn't suit your narrative, you keep on butting in with your 5950x with all cores enabled (which is something that we had before, many times). Oh, and recent 32 core Threadripper costs sweltering $3300 (AMD Ryzen Threadripper PRO 5975WX - Ryzen Threadripper PRO Chagall PRO (Zen 3) 32-Core 3.6 GHz Socket sWRX8 280W Desktop Processor - 100-100000445WOF - Newegg.com), so hardly a worthy comparison, even with the theoretical 16P core Alder Lake, wouldn't you say? Actually, just like Steve from HU says (usually an AMD fan favorite site), AMD is moving away from HEDT as well (likely abandoning the category altogether), as the 5000 pro series is a purely workstation line. And finally, because you seem to have forgotten, here is the OP's opening post in which they are specifically naming 6/12 and 8/16 chips as the primary focus of the thread:

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.

 

[Taraquin]

Perhaps a bench in for instance SOTTR would be interesting to see scaling there? CB23 scales with frequency, but cares bot for ram, cache etc. Some apps/games are much more dependent on ram and/or cache. With the TPU powerscaling of 12900K in mind the picture can vary greatly between apps/games. In most games I bet the 5800X3D would be the king of efficiency at low Watt limit, while on most apps 12900K is the king.

Nah, the difference is around 20%, I'm just too bored to go into manual tuning right now

But sure, let's go with 10% for now until I decide to take this more seriously, cause right now im just letting the motherboard decide wtf it's doing

A bit off topic:

You are more familiar with Alder lake than me, my 12400F has very limited viltage control due to locked FIVR, only UV on combined core/cache available. On my 6700HQ (unlocked bios) I could regulate:
SOC
Core
Cache
System agent
I/O

On ADL we have VDDQ aswell, are there more?

On Zen 3 we have
SOC
IOD
CCD
VDDP
VDD18
And a few more which I'm unfamilar with.

Tuned these voltages seems key to get good efficiency at limited powerbudget.

 

[openbox1980]

I don't know about bizarre, but it's more and more obvious that core-for-core Alder Lake isn't just (way) more powerful, but also more efficient, just like fevgatos has been saying for months (and Randallflagg before him) and got nothing but dismissals and ridicule from ardent team red members...

On the mobile side, it seems alder lake and AMD 6000 series are about the same speed clock per clock upto about 60w. But since alder lake can ramp up much more in wattage, it starts to lose the efficiency race at around 60w.

I saw a youtube video on it, I dont remember which youtuber it was.

 

[fevgatos]

A underclocked 12900k can be more efficient then a underclocked 5800x when you handicap Zen3 with its size advantage and only compare core for core with GC, but at the same time it cant touch a underclocked 5950x in energy efficiency as the numbers show.

A too expensive to produce hypothetical 16core GC could properly match and beat the 5950x in MT efficiency with ease, but it would meet the same fate against a 32core Threadripper. There is a reason why Intel dont have any HEDT class products for sale atm.

If you think i'm wrong, please share how many points Alder lake can put on the board in Cinebenchr23 at the given powerlevels 50w, 65w and 88w as the OP asked for

But you do realize that's the whole point right? There are people saying that a 16P core would be too hot and power hungry to compete, which is the whole reason the topic even exists in the first place. Of course the current 8+8 configuration cannot compete in efficiency (at heavy MT tasks that is) with the 5950x, that's a given, since basically it uses 8 cores to match 16, e cores just stop scaling all together at 4ghz. Although even in that scenario the difference isn't as big as you think, can probably hit 23k at 90w with the full configuration. That's just with AUTO motherboard settings btw, no manual tuning whatsoever.

 

[Det0x]

So with the release of Zen4 and raptor lake which both seem to be overvolted and running kinda hot, this could be a interesting topic again
Lets see who can make their hardware run the most efficient.. Ill start with my handtuned 7950x

• 4800MT/s memory with auto timings were used for all the runs.
• SOC draw alone is around ~10w (important for the lower PPT runs)
• PBO was used to limit the powerdraw for all runs, at the high-end (~above 120w) i'm pretty sure i could do a better job with static OC.

PPT 250w, 160A TDC and 225A EDC limit = 40360 points (OBS: CPU only pulled 230w maximum)


PPT 210w, 140A TDC and 200A EDC limit = 40085 points (OBS: CPU only pulled 204w maximum)


PPT 160w, 115A TDC and 175A EDC limit = 39068 points


PPT 130w, 100A TDC and 160A EDC limit = 37344 points


PPT 100w, 80A TDC and 150A EDC limit = 34574 points


PPT 65w, 70A TDC and 130A EDC limit = 28387 points


PPT 50w, 65A TDC and 125A EDC limit = 22832 points


PPT 35w, 60A TDC and 120A EDC limit = 144667 points


PPT 25w, 40A TDC and 80A EDC limit = 4180 points (cores were running at a wooping 500mhz each while rendering cinebench lol)


At the lower end of the PPT scale the 7950x is limited by the chiplet design with SOC on a separate die drawing ~10w alone
At PPT limit 65w the cores had ~3.44w each
At PPT limit 50w the cores had ~2.5w each
At PPT limit 35w the cores had ~1.56w each
At PPT limit 25w the cores had ~0.93w each

A monolithic APU should be much better suited here.. Lets see them Raptor Lake results

 

[Dr. Dro]

So with the release of Zen4 and raptor lake which both seem to be overvolted and running kinda hot, this could be a interesting topic again
Lets see who can make their hardware run the most efficient.. Ill start with my handtuned 7950x

• 4800MT/s memory with auto timings were used for all the runs.
• SOC draw alone is around ~10w (important for the lower PPT runs)
• PBO was used to limit the powerdraw for all runs, at the high-end (~above 120w) i'm pretty sure i could do a better job with static OC.

PPT 250w, 160A TDC and 225A EDC limit = 40360 points (OBS: CPU only pulled 230w maximum)
View attachment 266702

PPT 210w, 140A TDC and 200A EDC limit = 40085 points (OBS: CPU only pulled 204w maximum)
View attachment 266703

PPT 160w, 115A TDC and 175A EDC limit = 39068 points
View attachment 266704

PPT 130w, 100A TDC and 160A EDC limit = 37344 points
View attachment 266705

PPT 100w, 80A TDC and 150A EDC limit = 34574 points
View attachment 266706

PPT 65w, 70A TDC and 130A EDC limit = 28387 points
View attachment 266707

PPT 50w, 65A TDC and 125A EDC limit = 22832 points
View attachment 266708

PPT 35w, 60A TDC and 120A EDC limit = 144667 points
View attachment 266709

PPT 25w, 40A TDC and 80A EDC limit = 4180 points (cores were running at a wooping 500mhz each while rendering cinebench lol)
View attachment 266710

At the lower end of the PPT scale the 7950x is limited by the chiplet design with SOC on a separate die drawing ~10w alone
At PPT limit 65w the cores had ~3.44w each
At PPT limit 50w the cores had ~2.5w each
At PPT limit 35w the cores had ~1.56w each
At PPT limit 25w the cores had ~0.93w each

A monolithic APU should be much better suited here.. Lets see them Raptor Lake results

At 65W you have roughly the same CB scores my 5950X with a very quick n dirty -2 all core Curve Optimizer pulls. Mine is around 28,8-29k usually. Motherboard and chip just don't cooperate.

Not bad. Just a shame the EDC bug on AM4 isn't fixed and probably never will be. Thanks a lot, AMD. I'll keep that in mind when I buy my next (Intel) CPU.

 

[fevgatos]

So with the release of Zen4 and raptor lake which both seem to be overvolted and running kinda hot, this could be a interesting topic again
Lets see who can make their hardware run the most efficient.. Ill start with my handtuned 7950x

• 4800MT/s memory with auto timings were used for all the runs.
• SOC draw alone is around ~10w (important for the lower PPT runs)
• PBO was used to limit the powerdraw for all runs, at the high-end (~above 120w) i'm pretty sure i could do a better job with static OC.

PPT 250w, 160A TDC and 225A EDC limit = 40360 points (OBS: CPU only pulled 230w maximum)
View attachment 266702

PPT 210w, 140A TDC and 200A EDC limit = 40085 points (OBS: CPU only pulled 204w maximum)
View attachment 266703

PPT 160w, 115A TDC and 175A EDC limit = 39068 points
View attachment 266704

PPT 130w, 100A TDC and 160A EDC limit = 37344 points
View attachment 266705

PPT 100w, 80A TDC and 150A EDC limit = 34574 points
View attachment 266706

PPT 65w, 70A TDC and 130A EDC limit = 28387 points
View attachment 266707

PPT 50w, 65A TDC and 125A EDC limit = 22832 points
View attachment 266708

PPT 35w, 60A TDC and 120A EDC limit = 144667 points
View attachment 266709

PPT 25w, 40A TDC and 80A EDC limit = 4180 points (cores were running at a wooping 500mhz each while rendering cinebench lol)
View attachment 266710

At the lower end of the PPT scale the 7950x is limited by the chiplet design with SOC on a separate die drawing ~10w alone
At PPT limit 65w the cores had ~3.44w each
At PPT limit 50w the cores had ~2.5w each
At PPT limit 35w the cores had ~1.56w each
At PPT limit 25w the cores had ~0.93w each

A monolithic APU should be much better suited here.. Lets see them Raptor Lake results

I think the 12900k won this already, only the 13900k might be able to touch it. 35w = over 15k points.

 

[HD64G]

TSMC 5nm is providing great efficiency combined with Zen4 arch. AMD chose to push for top-spot in benchmarks since Intel is doing so for 4 years now with over 200W power draw for their top CPUs. I would prefer they had 3 modes for their Ryzen 9 CPUs with 100W, 150W and the maxed out 230W limits that reviews would test. They have the most efficient products both in CPU and GPU by some distance for at least 2-3 years now but that was sadly overlooked.

 

[mama]

So this is my 7900x stock (10 minute run). I will post efficiency results shortly.