Discussion in 'Sager and Clevo' started by 4W4K3, Jun 12, 2019.
Yes, always start with the easy fixes first
Hey there. @4W4K3 I Like to see that your trying to get the best out of the laptop and happy to see your good results with that.
I got the Clevo PB71EF aswell and currently I'm trying to tweak it.
16GB RAM, Crucial 2666mhz, Dual-Channel
Samsung 970 Evo Plus, 500GB NVME SSD
repasted CPU/GPU with "Kryonaut" thermal paste
Testing done with: discrete mode (no g-sync), performance mode and high performance powerplan
However, i'm not very successfull with that and the laptop performs quite poor after tweaking it with undervolting the cpu with -0.125v (core+cache) and MSI Afterburner (I followed @B0B 's guide basically) but even tried to copy 4W4K3's settings lately without getting even any near to his results.
The opposite is actually the case with any undervolting: for example not only was the native CineBenchR20 score very low with ~2250pts. but even went down from ~2250 pts. to ~1800 pts.!
The same with other benchmarks. I will upload my results and settings later on for a better picture, because I really think there's something off.
But before I return the laptop I want to try my best to figure out and fix what causes this bad performance, especially with tweaking (Undervolting, where you're supposed to increase it)
Hopefully you guys can help me witht that.
To start off, here are some benchmark results without any tweaking like undervolting:
PC Mark 10: https://www.3dmark.com/pcm10/38225414
In contrast to that, the undervolting results follow later on.[/spoiler]
@lappyluke Have you replaced the thermal paste in the laptop? This is going to be one of the main reasons you aren't getting similar benchmarks or similar temps.
When you run Cinebench, what does the clock speed do? If there is any deviation from 3.9GHz for even a second you are limited either by power or thermals. I'm running back to back benchmarks with no deviation in clock speed mainly due to the repaste with (most recently) Phobya NanoGrease Extreme.
In the grand scheme of things, a benchmark difference of a few hundred or even a few thousand points does not mean much. If you can game consistently for hours on end and you are happy I would forget an obscure benchmark score. I imply this thinking when I want to enable extra visual features and things that "hurt" performance. When it comes to reality, and my gaming experience, and my visual enjoyment, I'd rather sacrifice performance for enjoyment. So, if it works well and you can enjoy it consider keeping it!
Everyone does like a different t balance which is why PCs do have an advantage
Yes! The CPU/GPU are repasted with "Kryonaut" thermal paste (so a pretty good one aswell). The temps of the different cores seem to stay in a difference-margin of about 2-3 degrees. So as you said, that seems to indicate the repaste is good.
All testing done with discrete mode (no g-sync), performance mode and high performance powerplan. So I shouldn't get any power or thermal limiting, right?
As I said, the results are even worse after undervolting the cpu with -.125v. And we're undervolting to eliminate power/thermal throttling so I shouldn't get any of that and thus also better benchmark results, huh? Not the case for me here
Alright, so I finally got to upload the other results (I'll edit the post above aswell).
1st pic no UV | 2nd with UV
Here's my ThrottleStop:
I see that it constantly shows these yellow/red "EDP other or "PL1/PL2" but don't really know what it means or how to get rid of, if thats of any importance.
Edit1: My Idle temps are about ~50 degrees on both CPU cores and GPU (read that your idle temps are at 30°, wtf?)
Edit2: Did you OC and UV your GPU? Your Curve Editor looks a bit emphasized before being set to a fixed mhz/voltage (see below).
Even though you wrote that you think your CPU is a good one while your GPU is average, it seems that your current highest FireStrike Score (17800) has a pretty high graphics score (21 223) with a quite "normal" (?) physics CPU score (16 341)
Referring to your post here:
Notice the extreme physics CPU score drop at my FireStrike bench with UV: going from 16.247 without tweaks over waay below to 12.156 points (!)
Also, I just don't get how I could even get such an inferior result like 1800 at CinebenchR20 for the processor.
Oh jeez, this is so frustrating
There are a few differences between our TS profiles. Namely;
Non-Turbo Rato - 0
IccMax Amperage - Mine are all maxed out
Disable and Lock Turbo Limits - Checked/Enabled
Min/Max Cache Ratio - I limit mine to 36-38
SpeedStep - Disable
My CPU Core and Cache are undervolted significantly lower
Not TS related but also the DDR4 2666MHz in my machine is overclocked to 3000MHz with tighter timings
I also imagine our TPL windows settings are quite different. It's not a good idea to simply copy settings from another machine because your hardware might react differently. However, this was just to highlight the number of changes in this one software alone. Add to that I've made a number of Windows power/registry tweaks, I've repasted with 3 different thermal paste for testing, and I probably have 20-30 hours of "tweaking" into this specific machine alone; that is mostly why my signature score is higher.
If I make a few changes to improve visuals or lower fan noise I can't score as high as my signature scores. Those are all nose-bleed, sacrifice everything settings for the ultimate points. They are not the same settings I game on, or even browse on. Going back to what I said earlier; sometimes a score is just badge people like to show off. When it comes to real-world scenarios the highest scoring PC out there might actually be incredibly terrible to daily drive and use.
I believe the CPU performance comments were because my Cinebench scores were very high and when I uploaded them to HWBot I believe I was placing #3 or #2 in the world for 8750H. It's a very narrow criteria but I felt like I had won the lottery when I uploaded that score.
In comparison, 3DMark scores are much more popular and when I uploaded I saw something like 80-90% better than most results which is good but it's not 99% like the CPU seemed to be in HWBot.
Even now, go YouTube laptops reviews and notice MANY 8750H scoring in the low-mid 2000's. After repasting, undervolting, tinkering most people can get the score into the high 2000's. Even a lot of 9750H processors in 2019 laptops CANNOT break 3000 points. I see 2500-2800 points regularly from brand new machines.
With minimal effort I felt like I was breaking 3000 points straight out of the box with the PowerSpec. I can repeatedly hit 3100 points run after run after run. To me, this is an indication of a really well designed cooling system. Silicon lottery might have some influence too...
Yeah, sure there will be a few differences in the settings as you seem to have put a lot of time into tweaking and fiddling around with the settings.
However, there also seem to be many other around here getting way better results out of the box without any tweaking at all, and after tweaking shooting even higher. While my machine seems to perform even worse directly with just undervolting in a way lower margin.
So I think there could be more to it if I'm not completely overseeing smth here.
What are your temps while gaming something like e.g. Overwatch with 144 fps?
I'm currently playing WWZ, Wolfenstein YoungBlood, and some FarCry 5.
Load temps tend to be anywhere from 70-85C depending on game. Very occasionally I will hit 90C on a few cores especially when WWZ is throwing hundreds of zombies on-screen. If I don't limit the framerate and let it hit 200-300FPS the CPU really works very hard. Limiting the framerate to 144FPS brings dow nthe temps quite a bit in some games.
Ambient temp is currently 23-24C. With Chrome open (several tabs) and my 3D printing slicer software open I'm idling right now at 42C GPU and 42C CPU. If I close everything it might drop to 38-40C at best.
I've adjusted my fan profile in CCC to be more aggressive and ramp up about 20% sooner than default. By 65C the GPU is at 100% fan speed. At 75C the CPU is at 100% fan speed. This helps keep the load temps lower because it gives you maximum RPM at a lower temperature.
EDIT - Another thing to consider. I have enabled core-parking which lets some cores sit idle and consumer no/low power. I've also adjusted my Windows Power Plan CPU settings a bit. These will make a big difference in idle temperature. 6 active cores are much hotter than 1.
Hey lappyluke. I actually had a similar experience as you.
At first I followed a Throttlestop tutorial and would only get ~2300 on Cinebench R20 Multicore. What was happening was that the CPU was throttling down as soon as the benchmark started and not running at full turbo speeds. Check your Throttlestop TPL window. Make sure Turbo Boost Power Max and Turbo Boost Short Power Max are both checked. Clamp are both unchecked. I'm not in front of my laptop, but I believe I have my numbers set to 55, 70, and slider to 48. I'm not 100% sure what all the settings do, but the 2 numbers are related to the PL1 and PL2 throttling. It's the low and high Power Limit throttling of your CPU. You're not getting enough power to the CPU so it's down clocking. Once you get those settings set correctly you should be able to run an all turbo of ~3.9GHz throughout the whole Cinebench R20 and get a score >3000.
What's funny is by having the wrong settings I would've had a better score just running the laptop stock without any Throttlestop settings.
Separate names with a comma.