FPS drops to 15-20 for a few seconds very often

Syktris

Honorable
Oct 14, 2012
59
0
10,630
Ever since I upgraded my cpu and cooling system last week, i have been getting strange fps drops that last for a few seconds then the fps go back to 60
i've already uninstalled and reinstalled graphics drivers

specs:
AMD Fx 8350 4.0 ghz
H80i water cooler
msi R7950 3GB
8gb ram
Windows 8.1
 
Solution
The practice of disabling every power saving feature on the board is actually being carried over from a by-gone erra. Most of those settings can be left ENABLED without negatively effecting a custom performance tune, in fact, most of them will simply make a tune more efficient. I always keep C1E, C6, and Cool/Quiet enabled. When overclocking on boards that properly support it I use offset voltage or custom P-States to take advantage of reduced power states in conjunction with the overclock.

Turbo and APM are the only settings that really need be disabled if you want to take manual control of voltage and clocks, as those features can override your handiwork. (In most cases, disabling APM will have the effect of disabling turbo as turbo...
There may just be a sudden stress in graphics, although it is unlikely for the fps to drop that much. Maybe the CPU is not quite compatible with the rest of the PC, or is damaged in some way. Try updating your computer from the control panel. Is it just that game, or does this happen universally. Sorry, I have no definitive answers to your question.
 
What temperatures is your processor getting up to while gaming? If it's hitting the thermal limits it could be throttling. Is it overclocked?

You may need to make a few changes in the bios and see about disabling cool and quiet or throttling settings, power saving settings and try enabling hpc if you have that option so the cores can't throttle back.

http://forums.anandtech.com/showthread.php?t=2345194

http://www.tomshardware.com/answers/id-1848790/amd-8350-bios-settings-advice.html

Just note that this will probably push the cores to full turbo and may result in slightly higher temps since it won't allow the cpu to throttle down to a percentage of it's total power for any energy savings. The switching back and forth as it throttles between 20% max to 100% max (or whatever the exact percentages are when throttling) can cause fps spikes.
 
Install and run HWinfo http://www.hwinfo.com/download.php. Start logging from the sensors screen, start your game and let it run for awhile until your problem has shown up a couple of times.
You can then open it in a spreadsheet program and see whether it's your CPU or GPU throttling and also check your temps.
 


i havent overclocked it and the cpu stays under 50c while playing and the gpu is around 70c
 


i mostly play dota 2, which isnt that intense, but it happens when im playing other games too like hitman absolution
and my motherboard is am3+ compatible
 


i did this but i dont really know how to read it can u help me with it
http://www46.zippyshare.com/v/97903289/file.html
 
Did you run a game during that logging session from HWinfo? Did you experience any FPS drops at all while it was logging?

If the answer is yes to both of the above then there is certainly nothing wrong with your temperatures or throttling. In fact they are hardly tasked at all.
 

ok maybe i recorded that one wrong, but here i retried with hitman absolution, when i start the game runs at 120 fps and im playing on the very populated level and it still stays above 60 and it shows my pc is able to play it well but then it suddenly starts to drop down to the 20's early in the log, and in the end it was low too
http://www52.zippyshare.com/v/60157500/file.html
and thank u for the help
 


What kind of motherboard do you have? A good one with beefy heatsink on the mosfets and 8+2 phase power? If not you may be getting thermal throttling and the accompanying vdroop. Actually it is very likely. Another thing, the H80i is a pretty weak cooler for an 8350. The radiator is just too small to dissipate the heat.
 


Wow, your CPU temps are fine at 60C but it's the auxiliary temps that spike from 26C to over 100C in a split second. That causes your CPU multiplier to drop from 26.5 to 7.

I have no idea what causes that I'm afraid (not even sure what auxiliary relates to). It's too sudden I'd have thought to be lack of cooling around the CPU area due to watercooling.

Anyone else like to step up for this one?
 


The auxiliary temps are probably chipset related.
 


ASRock 970 Extreme 4 AM3+
heres a pic of the inside of my pc, case is CM Elite 430
v3fdy0b.jpg

u notice the case is small and not much space for the mosfet heatsink
i have an extra cooler if needed, the CM Hyper 212 evo
 
After a bit of Googling and a think about those temps I might have hit upon a solution thanks to http://forums.tweaktown.com/asrock/47282-asrock-970-extreme3-fx-8120-keeps-throttling-im-not-overclocking-3.html.

It may just be the bios is reporting incorrect temps from the Aux sensor which forces the CPU to throttle.
Firstly check if you've got the most up to date bios on the MB which should be 2.60 dated 11/15/2013 - http://www.asrock.com/mb/AMD/970%20Extreme4/?cat=Download&os=BIOS

If it is then have a look at that tweaktown article and try some of their ideas - disabling load line calibration, disabling cool 'n' quite ect ect.

Only try one at a time and if it doesn't work, reset it and try another. It may take a combination to find out which setting ignores the Aux temp reading.


For anyone new to this thread the only thing that seems to change according to HWinfo is the CPU multiplier when Aux temp spikes from sub 30C to over 100C with a 2 second monitoring period for around 12 second segments at a time.

31
28.5
26
26.5
26.5
102
104
103.5
102.5
103.5
103.5

29
30.5
30.5
30
30
29.5
26
28.5
26.5
29.5
29.5
27
109.5
103
104.5
103
103.5
106

30
31
30.5
29


Edit: perhaps I should have added, if it wasn't clear, that those temps probably aren't real given that nothing else spikes.
 


thanks, i updated the drivers, but earlier i did kind of find a fix, i underclocked my cpu to 3.5ghz and turned off turbo core, and ive never seen the drops now, but do u think i should change it back to see if it works ?

edit: i changed cpu to 4.1ghz and turned off cool n quiet, but havnet played any games yet, also should i turn turbo core on ? or keep it off
 
The problem you are experiencing is extremely common in amateur system builds on the AM3+ platform that leverage CPU cooling solutions that REMOVE part of the VRM cooling that is provided by the OEM CPU cooler.

Simple answer here is that your motherboard is overheating.

Keep turbo disabled for best results, as turbo mode will almost always run higher voltage on the CPU (even when it's not at turbo speeds) than necessary.

Disabling cool/quiet is only advised as a band-aid for a software side problem... If your software configuration is not properly managing CPU power states, then disabling cool/quiet is an option that will remove access to those low power states.

---------

Power dissipation changes proportionally to clock speed changes and to the square of voltage changes on a CPU. If you under-clock, you should also under-volt to effect the greatest change in power dissipation. Most 8350's under-clocked to 3.5ghz would run with a full 0.1V LESS than stock VID just fine. (just keep in mind that as soon as you start doing custom clocks and voltage, you are in a realm of generating potential compute errors and file corruption. back up important data and always test a custom tuned system for stability to prevent system problems).
 


i updated the drivers and changed back to 4.1ghz and i think it fixed, and i kept those 2 things off
 
The statement "I upgraded my CPU cooling" almost unilaterally translates to "I downgraded my VRM cooling."

Glad to hear you got things working....

Teachable moment for people interested in AM3+. The platform often requires tinkering. Seems very common for people to strap CPU coolers that can handle 220W+ FX chips onto motherboards that basically fall on their face above 140W.
 


Although I'm no expert on AMD CPUs I'm pretty sure you've jumped in without reading the whole thread.

His Aux sensor is cycling between approx 24 seconds of sub 30C temps to over 100C for 12 seconds with out any intervening rises or drops. 70C to 80C rises don't occur in the blink of an eye with no intervening temps being recorded.

If you Google the problem it's either a faulty hardware sensor or faulty software reporting, NOT actual real temperatures.

All you are doing is confusing the OP.
 


Reset your bios to stock first, after that you'll have to play with the different options (using that Tweaktown thread as a guide) until you hit on the correct setting. It's probable (I'd have thought, but wouldn't be as noticeable) that it does exactly the same when browsing as playing a game if I'm right. Try logging with HWinfo while just browsing and then check the Aux temps to see if it's spiking.

Just remember to try one step at a time and to set it back to default before trying another.



Edit: Looking at that Tweaktown thread this part may apply for yourself -

Tip #1. CPU Power Saving Throttling (Bios related)

Disable
AMD Turbo Core Technology
APM (Application Power Management) *disabling the AMD Turbo Core Technology via AMD OverDrive utility will also disable APM)*
Cool ‘n’ Quiet *I actually keep this enabled so I don't waste power since high clocks aren't needed under low loads*
CPU Thermal Throttle *be careful keep an eye on your temps, especially core temps, leave this enabled if your not the cautious type* (FX-8120 < 61C, FX-8150 <61C, FX-6100 <70C, FX 4100 <71C)
Core C6 State
C1E
Enhanced Halt State
Any other Power Saving Features in your bios, usually under CPU Advanced, Google them first.


Enable
Load Line Calibration, play with this setting until you get stable voltages under load, remeber every bios is different, some might even be better with it disabled, but most best enabled and manually set.


Notes
*Ensure that CPU NB, Memory or HT Link values are running at default stable values.*
*CPU / Chasis Fan Control (set to full speed for best cooling, if too loud play with the manual speeds.)
 


I think it's equally likely that the boards method of self-preservation could be a simple on/off trigger that applies a multiplier to the output of that sensor in order to trigger throttling when a certain threshold is reached by a sensor that has been added as afterthought or revision to cut down on board failures. It's not uncommon at all to see these sort of band-aid solutions used on cheap motherboards.

The board design probably has a flawed or dysfunctional socket temp readout. Something that could not be fixed without re-running a the batch of boards. Someone probably came up with a bandaid solution that keeps the motherboard from catching fire but has the unfortunate side effect of generating an all-or-nothing throttling behavior.

I've seen this sort of all-or-nothing throttling on lots of boards. Nicer Asus boards are some of the only ones that actually "soft" throttle down the power state steps as they hit temp thresholds. LOTS of AM3+ boards don't report socket temps correctly.
 


Ahh, fair enough bud. Did I mention I'm no expert on AMD 😉