GPU throttling and FPS issues

Hey there! I am pretty much having the exact same issues as in this thread: http://forum.square-enix.com/ffxiv/t...GPU-throttling

I have the exact same situation, same computer (rmbp late 2013 with 750m card and running boot camp), same GPU throttling, jumping between the same mhz and voltage levels. I don't speak german so I'm unable to tell HOW this issue was resolved, but it's really impacting my ability to play the game, and I didn't used to have this problem. Rolling back my drivers haven't helped and switching from Direct X 11 to 9 hasn't solved anything either (although at least now I'm not crashing when exiting the game... "FFXIV has stopped working" ugh).

This is really starting to drive me bananas! I've done CT and second coil on this computer and now I can't even handle being in a low level dungeon roulette without seeing serious FPS loss.

i am having a serious fps loss when using dx11 when reverted to dx9 it all went ok

750m nvadia here

Going to copy/paste from Yokami here:

Quote:

---

My GPU clock is acting very strange. I can't really say when this problem started, but it was working perfectly fine at a time before.
My specs:
Macbook Pro Late 2013 15" with geforce gt 750m
Playing on bootcamp (windows 7 home premium) with latest graphics driver.
bought it Dec 2013

I already tried:

- Changing energy plan (changing processor state, high performance, etc.)
- Changing settings in Nvidia system control (changing energy setting for gpu to highest performance and some other settings)
- Forcing p-states (and forcing specific clock values) with Nvidia Inspector ---> helped at first, but after some time ingame it starts fluctuating again between 270 and 660 mhz
- Over- and Underclocking in conjunction with forcing p-states
- Force constant voltage with MSI afterburner
- Using ThrottleStop to control CPU usage
- clean uninstall of graphics driver and fresh install
- completely reinstall windows (also tried with windows 8)
- using older driver version (the version that bootcamp provides)
- Doing SMC and PRAM reset

Some of those of course weren't supposed to help in the first place but i was kind of desperate, so I tried anyways... also i'm not the extreme technician, I found all of those solutions via. google and really searched to death, but nothing helped.

I really don't think it's a temperature problem, since it stays at about 70°C, max 80°C with forcing p-states (so that it works for some minutes / seconds).

---

I'm running Windows 8.1 and my rMBP was purchased in April 2014, although it's a late 2013 model, with 750m GPU. My GPU's max core clock is 926mhz, and can be played without any FPS drops at 845mhz and 765mhz. It is only when it throttles down to 660mhz that my performance takes a hit, but because it's cycling between these power states, the performance is constantly choppy.

http://i.imgur.com/mewsmLj.png
Throttling begins immediately upon logging in and is constantly switching between 765mhz and 660mhz (which causes a FPS drop). The switch is every 1.5 seconds going from 765mhz and every 2.5 seconds coming back from 660mhz.

http://i.imgur.com/taHvoGz.png
Even while idle, with reasonable temps, it is throttling me for absolutely no reason.

I was able to play perfectly fine with no issues, even doing CT and SCOB without any throttling or FPS loss up through the end of the last year. I stopped playing for several months and now I am experiencing these GPU throttling issues.

That is a thermal throttle and you'd be better off throttling down than up if you can get it to just sit at that rate first. You need to remember that mobile devices can NOT have their thermal throttle reduced otherwise there will be severe damage to the motherboard. In a desktop system, the GPU is nearly always on a separate PCIe card and the cooling headroom is much much higher.

Also, in quad-core systems (especially in hyperthreaded mode) when all 4 cores are active, the top clock rate goes down. So for example my i4770 goes from 3.9 to 3.7ghz when all 4 physical cores are in operation. That is by design.

If there have been any firmware updates, or bootcamp updates since the last time you played, there may be cooling parameter changes that weren't present the last time you played. That's all anyone can really offer any insight into.

HI! I'm having trouble accepting that my GPU issues are by design, because they didn't exist until recently. Can you elaborate on some things?

Quote:

---

That is a thermal throttle and you'd be better off throttling down than up if you can get it to just sit at that rate first.

---

As far as I can tell, I don't seem to be overheating at all. The highest I get during constant play is mid 70s celsius, which should be fairly benign for a rMBP. Are you telling me that the GPU is now throttling on purpose at fairly safe temperatures? Why is it only NOW doing this and ONLY with FFXIV? I do not have these problems with any other games.

Quote:

---

You need to remember that mobile devices can NOT have their thermal throttle reduced otherwise there will be severe damage to the motherboard.

---

Until this problem began, I never touched the GPU, so these issues are not a result of me futzing with the settings. I'm only doing these things to test possible solutions. I also never attempted to raise my GPU's clock settings above what it can naturally do. I just really want it to stop cycling down every 2 seconds.

Quote:

---

Also, in quad-core systems (especially in hyperthreaded mode) when all 4 cores are active, the top clock rate goes down. So for example my i4770 goes from 3.9 to 3.7ghz when all 4 physical cores are in operation. That is by design.

---

According to my task manager, my CPU load never goes above 40% while playing FFXIV. I also installed MSI Afterburner to track my CPU cores and they all seems to be doing well, performance and temperature wise. It seems to really just be the GPU rapidly cycling between my P1 (765mhz) and P5 (660mhz) power states.

Quote:

---

If there have been any firmware updates, or bootcamp updates since the last time you played, there may be cooling parameter changes that weren't present the last time you played. That's all anyone can really offer any insight into.

---

I have not updated my firmware or bootcamp prior to these issues starting. I did update some bootcamp stuff in an attempt to fix the problem but nothing changed.

I am definitely willing to try whatever you can recommend! Can you explain to me how I am supposed to throttle down and how that will help with performance? Thanks!

This is a known issue with almost all models of the MBP, my work provides us with these models and luckily enough - We're allowed to use them for personal as well when off the clock. I've had the last three models, this issue is rampant with these laptops. Your best bet is to underclock and undervolt your GPU with afterburner or some similar software.

Mobile GPUs are designed to be run at 70c or under, this is their thermal throttling point unlike desktop GPUs where the thermal throttle is at 81c. It is possible to overwrite the throttle but I do not advise it.

I noticed when I switch to DirectX 9, it throttles even further, going from 660mhz to 270mhz just like Yokami reported in their thread.

I am sad to hear that this is a prevalent problem with these laptops. Mine was working perfectly in CT and SCOB last year, my woes only began when I decided to start playing again this week! I am very sad about that.

I've never actually futzed with my clock or voltage before (and the voltage settings are grayed out in both nvidiaInspector and MSI Afterburner) so I have no clue how to do this, haha. I've just used the two programs to watch my power states and attempt to force the computer to stick to a power state. Can someone explain how to underclock/undervolt for me?

If you open after burner and go thru the settings there is a check box to enable voltage controls. This will ungray the voltage controls on -most- models. Some models are hard locked in the bios to not allow voltage changes, even to lower. A part of the problem though could well be... Did you play during the summer last year? :P Ambient temps can make quite a difference for a laptop. My mbp is pretty much useless during the summer for gaming ;/

Yes I did! I played very heavily all through patch 2.3, which was from July until October.

Unfortunately my model seems to be hard locked, because the check box didn't enable voltage controls for me. Would underclocking the GPU (and maybe the CPUs) potentially help?

Underclocking will help to an extent, the main source of heat with your components is from the voltage though so you won't see nearly as big of a change. Using a laptop cooling fan or something is likely your best bet.

Quote:

---

Originally Posted by LappelDuvide

Yes I did! I played very heavily all through patch 2.3, which was from July until October.

Unfortunately my model seems to be hard locked, because the check box didn't enable voltage controls for me. Would underclocking the GPU (and maybe the CPUs) potentially help?

---

Have you blown the dust out with compressed air (carefully?)

If you want to try that, make sure the laptops been turned off for at least 30 minutes otherwise you'll crack anything warm in the laptop. I'm only suggesting this because on my ancient laptop, i had to do that every 3 months otherwise the dust would gum up the cooling. But that laptop since got taken in to be recycled because the video ultimately quit working.

Just based on that experience, I'd likely never buy a laptop again.

Quote:

---

Originally Posted by KisaiTenshi

Have you blown the dust out with compressed air (carefully?)

If you want to try that, make sure the laptops been turned off for at least 30 minutes otherwise you'll crack anything warm in the laptop. I'm only suggesting this because on my ancient laptop, i had to do that every 3 months otherwise the dust would gum up the cooling. But that laptop since got taken in to be recycled because the video ultimately quit working.

Just based on that experience, I'd likely never buy a laptop again.

---

I'm not even sure how to take my rMBP apart or if that would void the AppleCare. Maybe I should take it in and have someone look at it? Dust could very well be a culprit, it is very sneaky.