Hi All,
After reading thru the other 2 threads concerning this issue I ran a little test w/ my EVGA GTX 470 SC to see if I could make it stutter while flying Aces High II.
I set up Afterburner to log the card's vitals (GPU temp, GPU load %, voltage, fan speed %, clocks: GPU, CC's & mem, mem amount used) to a file.
Then I intentionally set the voltage lower than stock (.987v>.937v) w/ Afterburner then ran the game for several minutes to see what the drivers would do to control the card according to the ever changing load demands placed on these cards. I also set up Task Manager in Performance tab to monitor the CPU core usage as well to see how the C2D (E8600 OC'd to 4 GHz) was performing. I chose to go this route as I was suspecting that the stutters were coming from a low voltage condition (& the odds of not harming the card were well in my favor
).
I tried to post the results but this BBS (as most others) doesn't allow posting files except .jpg, .png & .gif files so I'll paraphrase:
The game ran & was playable but stuttered. The vid card started out as it always had--59-60 FPS regardless & stutter-free (GPU at full clocks 625/1250/1701 @ .937v, 70*C, fan speed @ 89%, 63%-65% GPU usage, mem usage at 444Mb w/ Aero turned off w/ drivers in Adaptive Power Mode) & CPU running at 12%-13% core utilization across both cores. But within 5 minutes of flying the drivers (258.96 WQHL's) downclocked the vid card clocks as follows: 405/810/1701 @ .937v, 71*C, fanspeed @ 89%, 96%-99% GPU usage, mem usage at 445Mb. CPU held steady. FPS started to fluctuate from 47-60 FPS. After another 10 mins the drivers downclocked the following: 405/810/324 @ .887v, 63*C, fanspeed at 72%, 96%-99% GPU usage, mem usage at 445Mb. FPS now jumping from 36-57 w/ CPU still running steady as before. The stuttering got no worse & the game was still somewhat playable. Shut down the game, went into Afterburner & reset the voltage back to stock .987v setting then restarted AB to reset the card. Went back into the game & went up but the drivers would not reset off a restart of the game's .exe & held onto the last settings no matter what I did. The results being the same. I then shut down the game, went in AB & shut down (unchecked) the unlock voltage control in AB & left the rest as set.
Then I performed a cold boot on my box to clear all out all mem & have the box start up fresh. Once all was back up I checked AB to ensure that settings had taken--all was fine. Knowing that the vid card's voltage is back at stock settings & AB can't touch them I then went into Aces High II & flew for approx 35 mins & got the following:
The game ran stutter-free as it did before the above test w/ FPS locked at 59-60 regardless (GPU at full clocks 625/1250/1701 @ .987v, 71*C, fan speed @ 90%, 63%-67% GPU usage, mem usage at 445Mb) & CPU running steady at 12%-13% utilization across both cores. The vid card drivers made no attempt to change or adjust anything during the entire 2nd test time. Just like I like it!
I make note that I use AB for the auto fan speed control & the vitals monitoring only. I usually leave the voltage control shut down in AB as I can't see where/why I would need to be messing w/ this as I don't think that my C2D--as good as it was showing to perform--could keep up w/ this 470 SC running flat out at stock clocks, much less OC'd. I also run using the NVCP to set all graphics rendering options-not the game. The drivers run w/ card v-synched & triple buffered to LCD's RR (I can't STAND tearing!). I use the game settings to set the level of eye candy (which is MAXED OUT!) as I simply LOVE beautiful graphical eye candy in the games I play along w/ smooth clean performance. I want it all!
From this exercise I determined that the drivers do have some part to play in the stuttering by awarkwardly downclocking the card "thinking" that the card doesn't need to run at full speed. What I saw from the data was the drivers were shutting down the GPU/Cuda Core clocks (linked) then later the voltage & mem speed. Was interesting that the voltage setting & mem speed were showing to be related/linked. Even when the GPU usage went to full (this is from reduced cuda core performance--shader stuff being front loaded back onto GPU at reduced clocks) this did not trigger the drivers to reup the clocks & voltage back to full power so there is 1 thing to consider. This "could" also be triggered by a rippling PSU or a card OC that hasn't enough voltage available. I starting to believe that there is some built in voltage protection scheme in event of a power dip onboard (thus why I intentionally set the power low--.937v is almost back to 470 vanilla range of .925v w/ card clocks still at SC clocks--representative of a OC'd vid card on stock voltage setting) but can't/won't prove either way. The 2nd test showed that the drivers were now back to good clean control of the hardware & all is good--as it was before I started this test.
This could also be from--I'm only thinking out loud so this is a THEORY, not fact--a 3rd party program (like AB, RT, NST, GPU-Z, etc) changing driver "defaults" (best I could come up with to explain) in relation to the vid card's BIOS image to achieve a desired result but somehow, some way in a few totally unrelated cases has "dirtied" the driver & created this issue. But the same thing can happen from a bad driver install/flakey vid card BIOS image as well so 'tis why I said this thought is a THEORY & not fact but could be a possibility--however rare it may actually be to have occurred.
YMMV & take this post for what it's worth. A post w/ some information to share.
The reality is that I reproduced the stutters on my GTX 470 SC & then cleaned them up on MY BOX.
