Weird thermal throttling on 50ºC with T5750 cpu while gaming

I'm having trouble while gaming on an acer 6920g with a t5750 cpu. Turns out that, around 50ºC (measured by Hardware Monitor) or a bit higher the thermal throttling reduces cpu performance by 50% (as indicated by RMclock). Looking at other people's heat measures, I'm quite sure I shouldn't have throttle at this temp.

Doing some research, I found out that altering the thermal throttle treshold is impossible, or at least I could find a single thread or guide on how to do it.

Exploring other options, I recon I could try an external cooler, but this alternative is expensive (for a good one) and may not work. Also, I feel that having a cpu at this temp I shouldn't have to spend any more money on something I shouldn't need in the first place. Another option would be to undervolt. I've never done it and I'd rather not, if possible, but I'm seriously considering it, since it seems to have no risks involved.

That's when I read the undervolting guide of this site (very nice) and found out another odd thing. Running orthos to find out my cpu max load temp, after 31 minutes of stress the cores went only to 61\59ºC (when the guide assumed I would reach at least 70ºC). Most importantly, the cores reached this temp with no throttling happening at all.

To summarize:

- the cpu throttles at early 50c while gaming;
- it does not throttle running Orthos at higher temps;

I should also note that other temps in the notebook are good (GPU 65c at most) and I'm using it suspended on the sides (to free ventilation under it), which gave me a larger margin before the throttle happens. Also, my GPU is a 9500gs and it sometimes underclocks after 45-60 minutes of play, but some research leads me to believe it is a separate problem linked to the model ( for which I'll have to look for another solution). Also, I'm having 25-30ºC ambient temp while doing all this.

My questions:

What's going on?

Are there any other solutions I've overlooked, besides trying to lower even more the cpu temp?

Is it possible to change thermal throttle treshold for a higher temp?

Can any of the RMclock settings help me (besides undervolting)?

Sorry for the long post. Any help would be appreciated, since I've looked everywhere and haven't found similar cases with solutions.

Thanks in advance

 

Makryu, welcome to the forums!
Us lads over in the Acer forum have been dealing with this for a while now. If you used the search function, or visited the AS6920 Owners Lounge, you'd be able to see what we've figured out, and some solutions we've achieved. Here's what we've discovered.

1. HWMonitor and RMClock read your CPU temperature about 15 degrees lower than actual. So your 50 degrees Celsius is really 65. And yes, the thermal throttle temperature really IS that low. We have no idea why.

2. The GPU downclocks because of its PowerMizer technology - it powers down to lower speeds when it's not being used as much. So if the CPU can only perform physics calculations and process data for 15 frames a second in games (because it's being throttled), the GPU will only have those 15 frames a second to render; thus, it downclocks. It's perfectly normal behaviour, albeit undesirable.

3. Undervolting makes a HUGE impact - most of us who have encountered this issue with our 6920Gs have undervolted, and we haven't run into any issues since. My CPU maxes in the low 60s now, and idles in the low-to-mid 30s.

4. There is a less desirable alternative to undervolting - going into RMClock's Advanced settings page, and disabling the Thermal Monitors 1 and 2. However, doing so will remove the internal overheating protection methods of the Core 2 processor, and is extremely dangerous.

5. My personal theory about why ORTHOS doesn't result in a CPU throttling is b/c it doesn't use the GPU at all. Games use both, and since the CPU and GPU share a heatpipe, the overall system heat increases.

6. Don't be afraid to clean your vents every so often. Dust builds up like crazy in the 6920G, so unscrew the bottom panel and blow out the crud from the mesh filter.

Feel free to read up more in the Acer forum here - I'll be seeing if a CPU upgrade fixes the problem in a few days.

 

Hi, thanks for the answer. Actually, I did use the search function, but I didn't thought the owner's lounge thread had something on this. Correct me if I'm wrong, what you're saying is that the GPU underclocks because of CPU throttling, thus solving the throttling will also avoid the undeclock?
Forgive me if I'm asking the obvious, I just want to be sure, this is a subject I'm wasn't familiar with.

I made an initial undervolt of .100, it was stable and lowered cpu temp by 10c. What I'm wondering is if I have to follow the procedure with all the ticked boxes (multipliers). It seems very time consuming to do (7 multipliers with 45 minutes test for each .025 undervolt), so I will do it only if needed to deal with the throttle issue (I don't care at this time about the excessive energy or extended battery time).

I'm imagining I could alter only the 11x multiplier and leave the rest alone, since it is only this multiplier that is activated during gaming. The guide says you could let RMclock fill the gaps, but there's no guarantee the voltages would be stable. To avoid this, I would like to avoid tampering with the other multipliers, effectively only undervolting the 11x. Could anyone tell me if that's ok to do it, that is, will solve my problem and not cause any weird issues?

Thanks in advance for the answers, this is an annoying issue in a notebook that otherwise surprised me at how well it runs most games.

 

Makryu, no worries. Better to ask questions of us than to ask Acer Support where to send your broken unit

You've got the basic idea - fixing the CPU throttling will ensure that enough frames get sent to the GPU to be rendered to keep it busy, thus halting it from downclocking.
Trust me, this problem had me stumped for a good month or two. I kept thinking it had to do with audiodg.exe suddenly taking up so much CPU power - I never thought to check to see if my CPU was overheating!

Here's what I did for stability testing. I lowered the top multiplier (in my case 11x - you're on a T5750 so you should have 12x) while I was running ORTHOS in the background, letting it sit for about 5-10 minutes before dropping the voltage again. I stopped at about 0.200V below stock - I might've been able to go lower, but my system's been rock solid - and ran ORTHOS for about 4 hours. Then, I just did quick 5-10 minute tests for each of the other multipliers at the auto-adjusted intermediate states (no need to worry about 6x at 0.950V).

You certainly could just use the top multiplier - just change your Maximum Performance setting in RMClock to use only the top, then set that to be your default! If you wanted to just undervolt the top and use the rest, un-tick the box that says "Auto-adjust intermediate states" on the Profiles page.

It's our pleasure to help you out. If you have any other questions, feel free to bound over to the Acer forum and ask us Unofficial Acer Support Crew members.