Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Bootcamp gamers on 650m and 750m ?

B

baris3

macrumors member
Original poster
Nov 21, 2012
60
1
Hi,

I just received a rMBP and was gaming on w10 via bootcamp. 750m driver is up to date. Problem is the frame rate was not stable and was dropping so low for a moment which renders the game unplayable. Decided to watch gpu temps and clocks while gaming and I've found 750m gets too hot and clocks down and up which causes freezing/lags because it constantly changes it clocks.

Since 750m was running at lowest clocks possible and starts to fluctuate when it's hot, I tried manually maxing out fans before starting the game. Yeah with pre-maxed out fans maybe clocks were a little bit higher than before at first 5 minutes, but the problem still persisted as clocks are always fluctuate and rendered the game unplayable.


My first thought of the reason was the insufficient 85w power adapter. But I don't understand that because back in 2012 I could play games on 2012 rMBP with 650m without any issues. Frame rates on 650m was always stable as rock. Fans were kicking out of course, but everything was stable. I did not check the clocks of 650m back then but I'm sure it was not changing itself constantly.

Anyone experiencing the same issue with 750m or 650m ? My rMBP is a mid-2014 with 2.5 ghz. I've tried both power plan setting balanced and high performance, high performance forces cpu to run at turbo boost speed and temps get hot too quickly. Balanced power plan adjusts cpu speed so it is used only when needed.

Can I force the 750m to run at stock speeds constantly ? It supposed to do run at stock speeds, not fluctuate. I do not even want boost feature, I just want stable stock clocks when gaming so games would be playable.

There is the readings of gpu-z program showing the clocks when running Dota 2 :

capture-png.575433


As we see it fluctuates constantly in game, only when I press alt+tab and go desktop it locks to stock clocks. It does not pass 80 celsius and stays around 70-75 in-game.

Can 650m users share their gpu-z readings (clocks) while gaming on bootcamp ?
 
80C should still be save and not be a problem. The GPU should start throtteling at 90C+ not sooner.
Personally I never had that problem and I frequently run mine at 1066 Mhz.

I cannot say what is exactly the problem but one thing you can do to deliver more cooling power to the GPU is to limit the CPU. Go into the power plan setting and the advanced ones. Set the max CPU power to 90%. That effectively kills the Turbo mode and you will have a max clock of 2Ghz and something. It will still leave you with enough cpu performance to run most games well. Only something like Total War needs full CPU speed but first person shooter, car sims actually need very little. But CPU will be limited to some 20W actual TDP.
Just make one power plan with those settings, you can then quickly switch power plans to get full or limited cpu performance. @2Ghz a quad haswell needs about 20W @3Ghz it is 35W @3.3Ghz it is 40W.
 
Last edited:
dusk007 said:
80C should still be save and not be a problem. The GPU should start throtteling at 90C+ not sooner.
Personally I never had that problem and I frequently run mine at 1066 Mhz.

I cannot say what is exactly the problem but one thing you can do to deliver more cooling power to the GPU is to limit the CPU. Go into the power plan setting and the advanced ones. Set the max CPU power to 90%. That effectively kills the Turbo mode and you will have a max clock of 2Ghz and something. It will still leave you with enough cpu performance to run most games well. Only something like Total War needs full CPU speed but first person shooter, car sims actually need very little. But CPU will be limited to some 20W actual TDP.
Just make one power plan with those settings, you can then quickly switch power plans to get full or limited cpu performance. @2Ghz a quad haswell needs about 20W @3Ghz it is 35W @3.3Ghz it is 40W.
Click to expand...


Thank you for your answer. Do you kill cpu speed for overclock ? Was there any problems at stock clocks for you ? Which drivers and OS you are using ? Can you be more specific ?

I don't think it is caused by CPU's excessive powerdraw, because I already set my max cpu state to 90 or 95. I will try to set it lower and see what happens too.

This morning I booted into windows directly and ran a steam game called nosgoth. I finished one round and gpu-z readings were stable, 1.0250v VVDC constantly, therefore clocks were stable at stock clocks. No problem at all at first. But after another round, lags started and gpu-z showed like 0.0950v or 0.0850v VVDC. So GPU basically does not receives enough power, and beacuse of the downvolt clocks fluctutate, game starts to lag.

I realized the power adapter was cold while gpu was operating at 1.0250v. After two rounds of gaming, (it should be like 10-15 min) power adapter was hot. It seems it is a heat issue, but it still does not makes sense to me because I remember back in 2012 I was running games stable with 650m, even with a very hot power adapter (I was using it to warm my feet because it was winter ;)).

I noticed when I launch the game Nosgoth, or alt-tab from windows to it, the background of the menu turns green and pixelates. It backs to normal after 10-15 secs. You can see the background below, it shows an area with flames, buildings and trees. Does that mean why card is faulty ? I am upset because of this problem, and i am starting to think my card is faulty.

https://www.nosgoth.com/system/rich/rich_files/rich_files/000/000/295/blog/noscam-203.jpg


Any helps would be appreciated.
 
Last edited:
I was running Win 8.1 my MBP has been in repair for the last month so I have not gotten around to testing Win 10 yet.
I often overclock to the max 1066Mhz clock speed because I never had any problems with it. I also just underclock for noise reasons. The truth is CPU performance is just not needed, it just makes noise and extra heat that is better spent on the GPU.
I don't think going past 95/90 is necessary. As the disabling Turbo is what cuts the most power, more shouldn't be necessary except for noise reduction.

Try overclocking the CPU and see if the problem occurs sooner. I suspect there is some power delivery issue. Heat really shouldn't be the issue if it does only reach 80C. That is perfectly fine temps.
 
dusk007 said:
I was running Win 8.1 my MBP has been in repair for the last month so I have not gotten around to testing Win 10 yet.
I often overclock to the max 1066Mhz clock speed because I never had any problems with it. I also just underclock for noise reasons. The truth is CPU performance is just not needed, it just makes noise and extra heat that is better spent on the GPU.
I don't think going past 95/90 is necessary. As the disabling Turbo is what cuts the most power, more shouldn't be necessary except for noise reduction.

Try overclocking the CPU and see if the problem occurs sooner. I suspect there is some power delivery issue. Heat really shouldn't be the issue if it does only reach 80C. That is perfectly fine temps.
Click to expand...

If you can tell me which windows 8.1 driver you are using, I want to uninstall w10 and install w8. Or I can wait until you test out with w10.

I performed a test while the power adapter sitting on a ice bag, and played game like that. It still downvolts and downclocks. It is not about power adapter/heat.

Then I performed a second test as you said : cpu power state set to %100, it was forcing cpu at 3.6 ghz turboboost constantly. Played game like that, nothing changes. Same downvolt, same downclocks. The performance has not get worse or better or had not occur sooner.

When I was playing at evening, after some fluctuations card locked itself to the downvolt and clocks as 765/1000. As you can see below, the line is stable in-game before going into desktop (going into desktop makes clocks back to 925/1254 instantly).

clocks new.gif
 
Last edited:
I will test it in Win 10 but it will take at least another week before I get mine. My notebook is being repaired by complete uagggh. * Language not permitted on this forum. *
 
Thank you for your concern. What is your problem with notebook ? I bought this rMBP a week ago and I was thinking returning it because of this issue, and it has a very small green pixel at the centre of the display, only noticeable on black backgrounds. If this overpriced machine won't satisfy my casual gaming needs, I'd like to return it because my old 2012 rMBP could satisfy my casual gaming needs without any hassle.
 
After one and a half years it started having RAM errors. First it crashed without cause after a couple hours. Then Windows wouldn't even start anymore without crashing, OSX never ran more than 1 or 2 hours without crashing. I ran memtest and found a couple RAM errors.
They exchanged the logic board but than damaged the AR coating on the display. Sent the repaired unit with the damaged display to my local shop and thought I will take it in that state. Later refused to repair the display that they clearly damaged. Sure it costs a lot, you have to exchange the whole panel, but the AR coating matters (I often use it outdoors or in daylight lit rooms) and it is their fault if they don't know how to treat those displays or properly educate their personal in how to treat the stuff. Not that you need to even open or ever put pressure on the display while exchanging a logic board. There was no need to clean it, and there was definitely no need to clean it so vigorously as to rub the AR coating off. That doesn't happen by it self you actually have to rub quite a bit or use aggressive detergents, which you really should.
Now they said they will repair it on Apple's goodwill which I still think is a bad joke, but at least it seems I finally get it back repaired. The incompetence is astounding I think.

Honestly if it is just a week old. Return it. The two weeks return period are usually the least troublesome way to treat with problems. It always becomes more difficult later. I also doubt it is a driver issue. It is probably hardware. If it can still be returned do it quickly.
The 650M and the 750M are the same GPU. Nvidia just renamed and added the Turbo mode which overclocks it by 15% on average. The thing is Apple disabled the Turbo mode and reduced clocks a little. So the Apple 750M is exactly the same as the Apple 650M. For gaming there really is no reason to upgrade. There is literally zero difference well 25Mhz but that is nothing.
Windows 750Ms run with Turbo up to 1300Mhz. Besides if you want to upgrade what you need is Maxwell the 850m/860M or 960M those are twice as good and even significantly better than that M370X Apple uses.
 
dusk007 said:
After one and a half years it started having RAM errors. First it crashed without cause after a couple hours. Then Windows wouldn't even start anymore without crashing, OSX never ran more than 1 or 2 hours without crashing. I ran memtest and found a couple RAM errors.
They exchanged the logic board but than damaged the AR coating on the display. Sent the repaired unit with the damaged display to my local shop and thought I will take it in that state. Later refused to repair the display that they clearly damaged. Sure it costs a lot, you have to exchange the whole panel, but the AR coating matters (I often use it outdoors or in daylight lit rooms) and it is their fault if they don't know how to treat those displays or properly educate their personal in how to treat the stuff. Not that you need to even open or ever put pressure on the display while exchanging a logic board. There was no need to clean it, and there was definitely no need to clean it so vigorously as to rub the AR coating off. That doesn't happen by it self you actually have to rub quite a bit or use aggressive detergents, which you really should.
Now they said they will repair it on Apple's goodwill which I still think is a bad joke, but at least it seems I finally get it back repaired. The incompetence is astounding I think.

Honestly if it is just a week old. Return it. The two weeks return period are usually the least troublesome way to treat with problems. It always becomes more difficult later. I also doubt it is a driver issue. It is probably hardware. If it can still be returned do it quickly.
The 650M and the 750M are the same GPU. Nvidia just renamed and added the Turbo mode which overclocks it by 15% on average. The thing is Apple disabled the Turbo mode and reduced clocks a little. So the Apple 750M is exactly the same as the Apple 650M. For gaming there really is no reason to upgrade. There is literally zero difference well 25Mhz but that is nothing.
Windows 750Ms run with Turbo up to 1300Mhz. Besides if you want to upgrade what you need is Maxwell the 850m/860M or 960M those are twice as good and even significantly better than that M370X Apple uses.
Click to expand...

I only bought this machine because of it's display, I study photography&video currently, and other windows laptop displays are crap, at least in my country. This is a special display, no need for any calibration and colors look same at all angles. Other IPS displays are not even close to that. If I would not travel a lot, I'd build a desktop pc with a solid external screen, in my country it is hard to find a solid monitor though. After my bad experience with 2012 LG displays this one has a samsung with a little green pixel on black backgrounds on it. I don't notice it if I don't zoom my head into screen, with a dark background.

Do you have any issue like that, yellow tint maybe ? At bottom left of the screen, mine has a leaking of yellow.

I don't even want to play high-demanding games. I know even 650m at stock clocks could drive DOTA 2 at ultra settings, full hd. I will install w8.1 and let you know if problem still persist.

To be honest I don't want to return this machine if I can solve this problem. Because it cost me 6k, and a decent Asus 17" with 960m has a price of 5k. There is also asus n550jx /w 950m priced at 3.5k but it does not even have ssd, and only 8 gb ram. And of course, those asus's come with a crap display. Those are the prices in my country, when buying electronic, i need to make it count.
 
Last edited:
Yeah mine got a bit of yellow tint. But it could be described as somewhat too warm colors 5600k or so. I just changed the color profile to one that sets it to 7000+k which ends up around 6300k. I like my screens a bit warmer than standard.
Not really a calibration free screen. If you want it accurate you still have to calibrate and in theory you can calibrate other screens too. Personally I don't care as long as its not annoying. I am a programmer for my work text just has to be readable. I usually prefer my color temps to fit the lighting and that means at night 6500k is too cold. That is something that I don't like about some accuracy obsessed people. I get it about the average color range but color temp should not be simply 6500k, especially not on smartphones.
I get the need for photographers but I mean consumers should feel that accurate color temp is dependent on lighting.
 
dusk007 said:
Yeah mine got a bit of yellow tint. But it could be described as somewhat too warm colors 5600k or so. I just changed the color profile to one that sets it to 7000+k which ends up around 6300k. I like my screens a bit warmer than standard.
Not really a calibration free screen. If you want it accurate you still have to calibrate and in theory you can calibrate other screens too. Personally I don't care as long as its not annoying. I am a programmer for my work text just has to be readable. I usually prefer my color temps to fit the lighting and that means at night 6500k is too cold. That is something that I don't like about some accuracy obsessed people. I get it about the average color range but color temp should not be simply 6500k, especially not on smartphones.
I get the need for photographers but I mean consumers should feel that accurate color temp is dependent on lighting.
Click to expand...

You should try flux, I was using it in the windows and I remember it is available for mac. It minimizes the blue light from screen gradually while it's becoming night. For calibration, I don't want to spend any money on calibration tool (i mentioned the high prices in my country above) and this display is accurate out of the box.
 
I am on w8.1 right now, using default drivers and ran a Furmark GPU stress test :

aSasasas.gif



Downvolts started to happen when card reached 80 C. I have not tried in game, but it is the same, Furmark just accelarates the process of downvolting. And lowest voltage card reaches is 0.9000V, which equals 660/390.

And here is the readings when playing Nosgoth. Gaming is worse than W10 :

gaming.gif
 
Last edited:
I even set max cpu state at %50, and the problem still persist.

Do you think I should try with w7 too before going to service ? It is too hard to describe this problem with this machine, though. I am sure it would do the same downvolts in OS X in intense tasks and gaming.
 
It will always downclock in furmark. That is done by the driver just because AMD/Nvidia despise Furmark. Use 3dmark or games for load testing. Furmark is unrealistic and really not good for anything these days. It is a stress testing utility and since modern gpu's aren't built to actually handle using all their execution units at the same time at high clocks, it isn't even any good for stress testing. Forget furmark.

Honestly I doubt it is a problem with which Windows you have. Because then more people would have the problem.
 
  • Like
Reactions: baris3
dusk007 said:
It will always downclock in furmark. That is done by the driver just because AMD/Nvidia despise Furmark. Use 3dmark or games for load testing. Furmark is unrealistic and really not good for anything these days. It is a stress testing utility and since modern gpu's aren't built to actually handle using all their execution units at the same time at high clocks, it isn't even any good for stress testing. Forget furmark.

Honestly I doubt it is a problem with which Windows you have. Because then more people would have the problem.
Click to expand...

Do you think i have a problem with motherboard? If so, how can i prove it? I dont think service tests will reveal anything. I need to prove it is faulty because you are saying you have no problems with clocks while gaming. Can you tell which games you play at what fps range, so i can test with same settings?

I hate to play games with lower clocks because it gives low performance with crappy graphics. It will hinder the performance in gpu accelarated video Photo editing programs too.

Its like card or mobo trying to keep it cool with giving low voltage. Yeah as we can see card is 66 C but with awful performance.
 
Last edited:
Well you just tell them and send the gpu-z screenshots as proof via Mail. They should be reasonable. Because those downclocks just should not be. It is not normal unless you run it in an unreasonably hot environment, which you don't because heat would be the issue then.
 
  • Like
Reactions: baris3
dusk007 said:
Well you just tell them and send the gpu-z screenshots as proof via Mail. They should be reasonable. Because those downclocks just should not be. It is not normal unless you run it in an unreasonably hot environment, which you don't because heat would be the issue then.
Click to expand...

I was playing last night and clocks were 925/1254. After 30-40 min or so, core clock became 845, memory stayed same. And then started fluctuate, after awhile back to 925/1254 stable for 5 min. And after that fluctuated again.. What I realized is from a cold boot I can play with stable stock clocks up to 30-60 min. And then it will fluctuate. I've tried cutting CPU power and nvidia control panel adaptive and high performance settings. No matter what I am doing, it's going to fluctuate after awhile and reboot does not fix it too.

I think I can't prove them because at first everything seems normal. GPU-Z sees that performance cap as "Util" which means in my case card utilizes itself beacause of voltage reasons, or preventing energy draw from battery. Battery is draining though, after hours it won't stay at %100.

Do you remember any power drain from battery when gaming ?
 
Could be it is a heat problem on some of the VRMs aka power delivery. And after they heat up they cannot keep the power stable enough or something. Either way its an issue that is Apple's to fix.
I never really checked my battery for power drain. I never noticed any but I know it happens. Nothing to worry about. I never noticed the battery going low and I played for multiple hours with a friend.
 
dusk007 said:
Could be it is a heat problem on some of the VRMs aka power delivery. And after they heat up they cannot keep the power stable enough or something. Either way its an issue that is Apple's to fix.
I never really checked my battery for power drain. I never noticed any but I know it happens. Nothing to worry about. I never noticed the battery going low and I played for multiple hours with a friend.
Click to expand...

Today I was at the service, service guy ran a test for gpu for like 10 min and it passed. The test does not even load the gpu, no fan noise at all. I said apple's tests won't fail because it is a heat issue about the motherboard. He refuses the gpu-z readings as proof, because it is windows, and there is no obvious defect. He says he can change motherboard or give refund only if apple tests fail.

He said he will run the test for prolonged times, so I dropped the machine at the service.
 
I got my notebook back with its new logicboard. I tried some car driving game. Not very demanding but I ran it at 2880x1800. CPU full out and GPU clocked at 1060Mhz.
I found the same fluctuations in clocks. It seems more widespread. I never noticed it before but on this new board it frequently clocks down to 843Mhz without heat (never gone about 75C) or anything ever becoming an issue. It might be a GPU driver problem and not a hardware problem after all. Maybe it has to do with Nvidia powermizer or how its called. That is relatively new and reduces GPU clocks to provide only just enough performance.
I only ran the test for about 2 minutes or so. No real warm up but it fluctuated definitely already. The game is really of the graphical fidelity kind you find on smartphones, absolutely not demanding. I didn't check the fps but I think just about any GPU could handle it. Still it clocks up to 1060Mhz and fluctuates.
I have to test a few more games.
 
a new more extensive test on a game that actually can push it reaching only 20 fps or so it clocked all the way down to 765Mhz and mostly stayed there. Reducing the max processor state changed nothing. There is something seriously off. I am sure I didn't have that problem. With 1066Mhz it stayed on max all the time and reached some 80C but now it barely reaches 75C and clocks lower even in a game that could use those extra 40% performance.
It is also not powermizer that has been around longer and those power saving states are lower at 400mhz.
 
I guess we can use nvidia inspector to lock things on windows...
I did it in my PC desktop with a gtx660, but I no time to test it on bootcamp (macbook pro) yet.
here:

http://www.overclock.net/t/1267918/...-disable-boost-fixed-clock-speed-undervolting

Look at the last post (from 3dwin4tor):

http://www.tomshardware.com/forum/372385-33-wanna-overclock-650m-660m

the thing is:

1- Make two .bat files to execute: bat 1 - normal speeds , bat 2: fixed or overclock speeds:


like this (made for his laptop):

C:\Users\xxx\Desktop\OC\nvidiaInspector\nvidiaInspector.exe -forcepstate:0,0 -setGpuClock:0,2,1220 -setMemoryClock:0,2,2580

we only need to modify for our macs

I'll keep looking for better understanding...
 
Last edited:
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back