Hi everyone,
I've recently bought a 1440p 144hz GSync monitor, since I thought that my P650 (specs in signature) could comfortably push it in most of the games.
I started tweaking to reach at least 120 FPS all the time in most of my games, but I've noticed, that some games performance does not scale well. By this, I mean that I started tweaking from highest graphics, and lowered some settings to reach the desired level. But after a while, I've noticed, that my FPS does not really increase, only my GPU usage gets lower.
Laptop is in dedicated mode, high performance windows profile, power connected, latest nvidia driver, nvidia control panel vsync is application controlled, vsync in games is turned off, nothing is overclocked.
The problem is there in both automatic an overclock fan profile, with and without CCC, though in overclock fan mode, with a +150Mhz GPU overclock, and a bit of CPU undervolt, the issue seems to be a bit less apparent, at least in Overwatch.
Some examples:
- in Overwatch, I average at around 130FPS with my settings, while the GPU usage is between 70% and 90%.
- in League of legends training mode, I average at about 150 FPS on the highest settings possible. If I select the lowest possible settings, I get around 200FPS. If I also reduce the resolution to the lowest, the FPS does not change. This case might be a single threaded CPU bottleneck, though it still only hangs around 85-95%, and not capped at 99.
- in Heroes of the Storm, I also average around 110 FPS, with dips into the 90s.
- in Battlerite, this problem is even more apparent, since on full resolution, high graphics, I get around 110-120 FPS. If I lower the res and the settings to the minimum, I have barely 150 FPS, with single core usage around 70-80%, GPU usage 20%, CPU total usage 60-70, temps under 70C
In short, it seems like something is capping my systems graphic performance, the more I lower the settings, the less gain I got from it. It could be single core performance, but the last example more or less proves this theory wrong. (This still could be the problem, I'll try to run some extended tests)
I would also cross out overheating, because in some cases, I do know that the GPU throttles, but in other cases, like with League of Legends, Battlerite, and lower settings OW, even though the temps might not drop a lot (they are still mostly in the 60-80s for both), but the fans audibly spin at a lower frequency, and the GPU/CPU usage is lower by a lot.
I'm quite baffled by this problem, as I have no more concrete ideas what could be the cause of this, only wild guesses, like a buggy nvidia driver, or some Clevo bios ********.
Has anybody experienced something like this before? Does anybody has an idea what I could do?
Also, if someone could provide some comparison FPS values from a game, just to check whether I'm simply expecting more from my laptop than I should, would be really helpful.
-
yrekabakery Notebook Virtuoso
Also keep in mind when monitoring CPU usage that Hyper-Threading changes the way utilization is reported. -
I do believe that the p650 series cannot use both the 6700hq and the GPU at max performance simultaneously.
Generally, if its a CPU bottleneck, turning down the graphics doesnt make such a large difference in FPS. -
I find with my P650RP (1060) that overclocking tends to do nothing (as it bounces off the hard power limiter) and undervolting is far more successful. Clocks are much more stable which is hugely important since I use this as a VR demo rig.
The only downside is undervolting is SUPER annoying to do in the MSI Afterburner curve editor. You have to drag nearly every damn point to a fixed clockspeed and you can't have any variations. -
Stooj: Just as you wrote your reply, I've found some information about GPU undervolting, and right now tweaking a profile to test. My only problem is, as Prmt said, the system seems to be quite underpowered, and even with some undervolt, the GPU frequency seems to jump around quite a bit, though I'm around 140-150 FPS now in OW, so a 120hz display based vsync might solve the jumping.
What especially bothers me, is the case with Battlerite. With that game, I do not see a single core CPU bottleneck, as all cores show well below 80%. Could this be an unprecise reading, caused by hyperthreading? -
You're better off graphing each core usage and keep an eye on cores 0/2/4/6 which are the physical cores. If they're at 100% then that's all you've got. -
yrekabakery Notebook Virtuoso
-
Stooj: The CPU percentage I've seen is from the task manager, with the "show all logical cores" enabled, so I think I'm reporting not a cumulative, but single core load. Also, with a bit of CPU undervolt, and normal GPU usage, I see EDP throttling quite much in general, and other than that, mostly temp throttling. The Clevo fan profiles usually let the GPU get to 90+C, where it starts to throttle a bit.
yrekabakery: Wouldn't setting the core affinity to the physical cores in Task manager have almost the same effect as disabling hyperthreading? -
How do you manage 90C for the GPU? I had the p650 with the 1070 and at worst it was in the 70-73c region. Even when I installed an unlocked bios and overclocked it, it was still in the mid 80s.
-
You can't overclock a pascal. When you " overclock" a pascal card all you actually do is undervolt the card. Press ctrl + F while moving the core sliders, you'll see how the card gets undervolted.
90c on pascal is really bad. They are WAY more annoyed by high temps than previous cards. -
OP, I suggest you go back to stock settings and get those temps in check and then check your FPS again. Lowering the settings may have ad an impact on your FPS simply because it wasnt working as hard when you lowered the graphics and therefore didnt throttle as much.
-
About the high GPU temp, I've always had some temperature problems. After some asking around, I did a repaste with NT-H1 paste, and as far as i can remember, the temps quite much stabilized.
I've dug up some older forum posts and emails about the older issue, and at that time, as far as I can see, the main problem was that my CPU was hitting the temp limit, which normalized a bit after repasting, and went into the 70s after undervolt.
Now that I've looked into it, I think there might be a whole other problem going on.
Does anybody have any guesses what could be increasing my temps that much? I periodically remove dust form the fans, so I don't think it is dust buildup.
Edit:
Prmt: Will do that as soon as possible, do you have any guidelines what should I test? Shall I reinstall nvidia drivers, or anything, or just roll back all undervolt / OC settings? -
-
Though there is a bit of clutter on my desk, so I will make as much room as I can around it, and test it when I get home. Any more tips? -
Ok, follow up, did a test run with Overwatch. All graphis ridiculusly up, laptop on cooler, desk clutter removed, overclock fan profile, and after that, maximum fan profile, no undervolts / overclocks.
The CPU was around 60-70C° the whole time, the GPU rose to 90C° in just under 3 minutes, with 99% usage.
Going to pop off the back now, and take a look at the fans and the heatpipes.
Update:
I thoroughly blew out the dust from the laptop, and now my temps look a lot friendlier.
GPU normalized around 85C°, and on a good 150-200 MHz higher than before.
The earlier problem still holds up though, I will do some more testing, but the problem seems to be single core CPU performance.Last edited: Dec 15, 2017 -
Prostar Computer Company Representative
-
I might have been a little hasty to judge, looks like the temps are still rising until 90C° with or without undervolt, just a bit slower.
-
Prostar Computer Company Representative
You don't need replacement thermal pads, but I hear good things about the ones from Fujipoly. They may help a bit as well. -
Ok I've done some further testing. Most of the games I tried maxed out my GPU temps at 90C° in 3 minutes, even with overclock or maximum fan profile, and with CPU and GPU undervolt. Meanwhile the CPU sat at around a comfortable 70C°. The good thing is, I've managed to get a GPU undervolt, that manages around 1700Mhz at these temps. The airflow is fine, and I've removed the dust as much as I could, so neither of these are causing this, I think.
Now I would like to ask for advice, whether I should:
a. accept that my laptop runs at these temps, and not worry about it too much, since these are within the safe operational temperatures.
b. try repasting, hoping that my repaste from half a year ago is somehow causing these issues.
c. contact a repair shop or my vendor (even though I don't think they could help, as I'm from another country) for further inspection or RMA or anything.
The performance problem is also still there, in some games I get ~ 40% GPU usage, with no CPU cores above 75%, and my FPS hangs around 100-110. -
Meaker@Sager Company Representative
Some game engines do have frame limits built in, what titles are you seeing the 100-110 fps?
Clevo p650 Low performance / performance does not scale
Discussion in 'Sager and Clevo' started by Brabulla, Dec 14, 2017.