I see his 5600Х cpu-z mt score is a bit on a low side.Except we all know that it takes more than 18.1% power to get 13.3% higher clocks/freq, as shown in the bottom spreadsheet where 5600x takes 83w (5w more than 12400) in AIDA 64 FPU at the freq needed to match 12400 in CPU-Z benchmark.
Is this stock or tweaked? Not that it matters, reviews should settle this sometime in the near future.I see his 5600Х cpu-z mt score is a bit on a low side.
76W is stock PPT limit, so yes.Is this stock or tweaked?
Stock, yes, but are you using Curve Optimizer or applying some type of undervolting? Because if you are, your result is not exactly stock.
So 12400 has the same efficiency or a bit better than 5600x at least If It's at 4GHz, right? How did you come to the conclusion that ADL is hugely more efficient than Tiger Lake? Zen3 is not that much more efficient at higher power draw than Tiger Lake.
Techspot
At 75W Zen3 is about 10-13%. I can't say that is a huge increase in efficiency.
My graph also shows that Tiger Lake is under Zen 3, but with increased TDP the difference becomes smaller.Tigerlake is clearly below Zen 3 in efficiency, that's why. This is based on several reviews or other tests like this. The Tigerlake based NUC didn't look great either. The i5-12400 seems to have a 5-10% efficiency advantage over a comparable Zen 3, this is overall a huge improvement to me.
The CPUZ benchmark, especially version 17, takes less power than the AIDA64 FPU stress test. Since the power consumption of the 12400 was not reported for the CPUZ bench, the example had to first determine the clock speed of the 5600X needed to match the 12400 in CPUZ (4.5GHz) then run the AIDA64 stress test at that clock speed to come up with the closest equivalent power figure.I see his 5600Х cpu-z mt score is a bit on a low side.
I don't get what relation cpu clock has to power consumed when comparing different architectures. What was the goal then? Can it be approximated this way?the example had to first determine the clock speed of the 5600X needed to match the 12400 in CPUZ
My graph also shows that Tiger Lake is under Zen 3, but with increased TDP the difference becomes smaller.
Considering how low Alder Lake needs to be clocked to be within 125W(P: ~3.6GHz and E: ~2.8GHz), based on the R20 leak, I seriously doubt It is more efficient than Zen 3.
What I posted should be an official Lenovo presentation If I understood It correctly.I'm not talking about 125 watts nor E-cores, I'm talking about Golden Cove and mobile. Once again the 12400 (it's a 6+0 model) looks very promising and therefore it's promising for mobile as well. 78W in AIDA64 FPU could result in 70W Cinebench MT running at 4.0 Ghz. 5600x requires the same if not more power in Cinebench and runs slower. I haven't seen any reliable 125W ADL-S test by the way. The leak you are probably talking about isn't reliable, the 125W score is too low a tester hinted.
What are you guys talking about? How can cpu consume less or more than what is deifined by the power limit preset in these tests?78W in AIDA64 FPU could result in 70W Cinebench MT running at 4.0 Ghz. 5600x requires the same if not more power in Cinebench and runs slower
You are right, that 7492 points for 12900K looks pretty low, If 12400 really manages 4784 points. It would mean only 56% better performance, and that is a very low score for 8C+8c If an 8 core at 4GHz should manage 6377 points(4784 * 1.333). It would mean 8 Gracemont cores provide only ~18% better performance, that's worse than what SMT provides. I would expect ~50% higher performance in Cinebench.
What are you guys talking about? How can cpu consume less or more than what is deifined by the power limit preset in these tests?
It's a rough approximation but it's something for the curious and enthusiasts to do while waiting for the review embargo to lift.I don't get what relation cpu clock has to power consumed when comparing different architectures. What was the goal then? Can it be approximated this way?
Based on the behavior on Tiger Lake-H that I see. On an Inspiron 16 Plus laptop, at 8x3.4GHz (undervolted) results in 56W maximum package power in AIDA64 FPU, 50W in Cinebench R23 and just slightly more than 36W in CPUZ v17 benchmark.PS: If I get what was on screenshot right, during the AIDA FPU test it consumed 78W ?
What makes you think then, that in CPU-Z it would consume less, I mean what this is based on?
Oh really...? Is it mobile parts specific or it applies to all 11 series? Didn't know that.Based on the behavior on Tiger Lake-H that I see
Ah, yes, I see. Didn't know that.imited by the multithread turbo clock speed in the first place
It would mean 8 Gracemont cores provide only ~18% better performance, that's worse than what SMT provides. I would expect ~50% higher performance in Cinebench.
I was only discussing the low score from supposedly official Lenovo marketing slides for 125W power limit.I swear, Gracemont goes from being a Bulldozer to a Conroe and back to a Bulldozer every leak.
In Cinebench R23 there is only a ~15% difference between 125W and unlimited (260W) on a 12900k.
It means that at 125W the CPU will do the bench at 180W on average, doing half the bench at 240W before temperature limiting the CPU at 125W for the second half of the bench.
But if PL1=PL2=125W then that's not the case.
The difference would be bigger than 15%, from brick wall limit at 125W to 240W unlimited the perf difference should be around 30%.
On this topic...Guess we'll see in a few days if reviewers are smart enough to test at fixed power levels.
Will be interesting to watch how transparent all the reviewers will be stating the exact version of all those moving targets. Add the inherent difficulty to do "correct" performance tests on hybrid designs and we are set for crazy town indeed.we're all set for crazy town.