Full Speed 9980HK With No Throttling! New M15 R2

Posted this in another section and thought I'd share here as well. Bought an M15 R2 that was delivered in September but setup a return since I could never get the thermals consistent. Dell persuaded me to try another system and I'm glad I did. In the other sub forum the effectiveness of the 9980HK was in question and called out by Notebookcheck. They specifically stated the 9980H beat all 9980HK's that have come across for reviews, until now...

Results of 30x runs of Cinebench R15 ran automatically in a loop - took me awhile to figure out but I was successful:

Score average - 1842.7 (can go much higher on single run but throttling will occur on later tests leading to lower average)
Average Core Frequency - 4.45 GHz
Average Package TDP - 121w
Max temp - 89c
Liquid Metal - absolutely
# of times thermal throttled = 0

From Notebookcheck article:
"We should note that our MSI in this example is the fastest laptop in our database with the i9-9880H processor. Other laptops with the i9-9880H like the HP Omen 15 may perform slower than the i9-9980HK. Still, the fact that almost no i9-9980HK laptop thus far can outperform the MSI is a slap in the face to Intel's top-of-the-line mobile Core i9 option. We're still waiting for that one Core i9-9980HK laptop to hopefully break this trend."

The MSI has been beaten, losing by 14% while weighing 23% more...

 

Playing the devils advocate, those systems tested all run stock paste with no undervolt.

 

Can you test with
CB20 and post Hwinfo CPU Package power numbers? Cb15 is much less demanding, and TDP is the limit, not the consumed power.

 

15x Cinebench R20, 30x would take a long time since it's over a minute a run.

CPU-Z, HWInfo, XTU, and notepad output below.

Haven't calculated the score average, but the final run was 4324.

Little warmer today here in California, so laptop was on a cooling pad with fans running.

 

I see the clocks dropped to 4.19 with less Cpu Package Power vs. your previous test with CBR-15. Any reason for this? Can't be Thermal throttling if temp stay in the 80th's. Is it AVX negative offset?

 

Wondered the same thing and ran a R15 run right after the R20 test. Went right to 4.45 ghz. No idea why there is a difference, was pleased with the temps though!

 

From what I can see... You have put 50x on all cores but the clocks stay at 4.45GHz in CBR-15. It seems the chips is unable to utlize the whole PL1 and PL2 power headroom. Because the chips should be able to take what it need to hold highest possible clock speed as long it's below thermal throttling temp. Is it a hided Power Cap?

 

It might be the 140A setting. Volts x Amps = Watts.

I briefly played with the amps initially and found decreased performance past 140A, but I’ll take a look at it a little more tomorrow.

 

Respect to Notebookcheck staff, they did an update to their 9980HK vs 9980H article and referenced my testing.

https://www.notebookcheck.net/Appar...H-under-the-right-circumstances.437052.0.html

 

@littlezipp I'm very impressed with your cpu, quite frankly I never thought an alienware laptop THAT THIN could pull of these amazing results, kudos to you (+rep)

For comparison here is my 11th Cinebench R15 run out of 20 that I did consecutively on the 9880H CPU: (This benchmark is run with the lowest scoring laptop on the chart, HP Omen 17 2019, RTX 2080 )




I did manage to pull off 1741cb as a max score out of all runs taken, but that's not the average...

Temps were reasonable as I spent 3 hours lapping the damn heatsink before applying liquid metal, fans 100% throughout the tests.

Damn I can get close, but with good cooling your laptop cpu could frankly clock higher as it's unlocked, but would the vrms handle it that's the question

 

I'm sure this isn't the max, but an additional 200 points over that 9880H looks pretty good. 1949 is all I can edge out tonight. Temps aren't even getting to 90c so temps aren't my problem. Single run speed runs to get to this, pretty sure I couldn't loop stably this high. Some very odd behaviors at play when you get to these speeds trying to get the chip happy and stable. Setting the core multiplier too high lowers performance even when the core does not go above 4.5 / 4.6 ghz. I'm determined to break 2000 on R15, just need to figure out how to keep it at 4.9 turbo which it will happily do for short periods on all cores.

 

Probably going to be limited by power delivery on motherboard..

 

I think you are right. I believe 140w to be the max +/- 5w.

2000 R15 surpassed with a 2010 score. I got to 2016 but then it crashed. These settings appear to be somewhat stable, but moving the 8 core multiplier to 48 leads to a lower score and -.145 undervolt is unstable. Believe I may be close to the max, and hit this with a max temp of 83c.

 

What Cashe ratio you run? Tried lower it and put 8 cores @48?

 

Cinebench 20 --- I7 9750H --- 3110CB

 

It's the all-silicon lottery if you ask me. Sure, out of 1000 processors, this just might be the one to hit those higher clocks without eating 150W of power.
Ever since the 8000 series from Intel, it was evident that there are gross power consumption / thermal dissipation inconsistencies across processors of the same model.

Gone are the days when a 45W or 55W TDP was giving you exactly 55W TDP @ Max Turbo.

 

CB R20 --> i9-8950HK --> 3.600 = Power Limits and K matters

 

This! basically we all kept paying money for an old architecture with no real upgrades except for more cores of the same architecture since skylake. Power usage literally doubled and morewhile they keep smacking the 45watt tdp label in it and just state "only for base clocks" Yet they all advertise with max boost clocks.

 

It’s still up to the ODM/OEM manufacturer’s if they want cap long time Power at 45W and cap PL2 at 1.25 Intel recommendation for unlocked BGA from 8th gen is 100/125w. But as usual... Thin and light come always first before needed cooling capacity. Can’t fill up 1 liter plastic bag with 2 liter water.

 

Only for a short time boost. Intel isnt very up front of it because they still reference a cooling design for 45watt TDP which isnt true at all. 45watt TDP cooling solutions of 2 generations ago could keep a 7700HQ running at 24/7 boost clock at just 45watt power usage. The same size heatsink now even without GPU load can only cool the follow up 8750HQ at around base clocks in Prime95.

 

Haven’t the link in front of me. But Intel themselves recommends cooling as I mentioned in link above. Now it’s more up to the manufacturers where they want put the power limits (stated by Intel). Can’t create cooling for the lower 45w(1.25 pl2) if the Notebook manufactures prefer let 8750h stick at 55/58w for PL1 and run wild on short power in their flashy gamingbooks. This for locked down chips. Same for unlocked BGA Chips... They put limits way above 120W and give a damn in their cheaply made cooling.

This isn’t so difficult. If they can’t (due chassis limits) or can’t afford a cooling for etc 100W then cap power around own coolers capacity. But competition between the different brands make them lose their heads and prefer 100C and throttling mess.

Edit. See also https://www.notebookcheck.net/Asus-ZenBook-14-UM431DA-With-a-slowed-down-AMD-CPU.453726.0.html
“Time and again, laptop manufacturers equip their devices with up-to-date CPUs without adapting the system to the requirements of the new hardware”

https://www.notebookcheck.net/Lenov...eing-worse-than-its-predecessor.453861.0.html
“Additionally, the X1 Extreme Gen 2 has quieter fans than its predecessor, especially under load. However, Lenovo has stuck with the same cooling system, so performance suffers as a result”

Doesn’t matter what chips it is if they prefer go cheap or don’t bother make awesome!

 

Intel shares reference heatsink designs for what would be sufficient for cooling. Even their own reference design Mag15 cant hold the boost clocks for long. WIth just the 45watts recommend power they never reach the advertised speeds which Intel writes on their boxes. The advertised max boost clocks for all cores never can be met in most laptops out there fo rlonger than a couple of seconds even though they follow Intel's directions. WHen they do force it to go higher, they also will run more hot.

 

How much does the MAG-15 get for a 6-core boost after 10 minutes of full load? Intel's specifications state it should 4.00 GHz.

 

In prime95 it doesnt even touch the 3ghz. 6core boost in games tend to be around 3.6ghz from my experience. Same as the Blade 15.

 

Well yeah, that's prime95, I wouldn't expect it to boost too high. Even my old 6950X barely touches 4.0 GHz in a 10 minute 4K torture test and struggles at 95C.

 

The quadcore skylakes would just keep boosting. with a mild undervolt and would sit around 65~75c (depending if you had a bit of a decent laptop ofcourse). the 3.4ghz boostclocks of the 7700HQ would just sit around the 45watt limit.