OCing video card

I read awhile back that 999 or 1000 core was all the higher it'd go as one of the forceware updates prevented going any higher and would actually drop the settings. I know 1025 core was a common max people could get before the cap, so figured I'd shoot for 998. The voltage is capped close to where I was on the last try, so wasn't to eager to set it there. I'll try some software benches when some down time is available and see what's up. Thanks.

OK, I spent some time OCing with a current MSI Afterburner, then with EVGA Precision X and running benchmarks with EVGA OC Scanner X, while monitoring with gpu-z and here's the results which can be applied to other cards as well.

If you go over 999mhz (with the GTX 550ti and possibly other similar, older cards with newer forceware drivers) AND it crashes an application, it'll either LOCK the speeds to the 2D clock rates (or default them if K-Boost is on) and only way I could undo this was a PC reset. Even opening an application that would use 3D clock speeds, it'd stay at 2D. I needed gpu-z to even see this. I was able however to get to 1025, but it needs more voltage to be stable. Neither Afterburner nor Precision X will allow more than 1.150v. Both applications DO allow going higher than "recommended" voltages, BUT neither allowed it with this particular card (unless I'm missing something). Using the EVGA OC Scanner X which has a multitude of rather simple and quick benchmarking tests, I was able to pass those w/o issues at 1025 gpu clock and a 2065 memory clock (1026/2052 memory is stock) using the 1.15v. Also, the voltages were just actual presets like with the older Afterburner. Both the recent Afterburner and EVGA P X had the same presets, so no fine tuning with the voltages. Running BF4 with 999 clock and 2065 memory ended up with a crash. Eventually rolled back the 999 to 980 and all was good. Temps never went past 68c.

EVGA OC Scanner X:
Running the 4M partical test in 1280x720 windowed form, AA off and artifact scanner off (seems to be the defaults if you open 1st time and just press "Benchmark" button saw score of 5941 with min. FPS of 88 and max of 111 fps which was the lowest of my tests with stock settings. 94 min. Fps was more common, but the test had to start out strong without a hiccup to get that.

Running same test with 999 and 2065 (Memory was very touchy. 2066 or higher saw a decrease in score) and 1.137v saw the score jump to 6729 with a minimum of 100 fps and max of 125 fps which is pretty significant, though the benchmarking at the very beginning of the test would stutter when transitioning after the load screen, so gotta watch that or it'll artificially lower the score, especially if the gpu is running at 2D clocks and needs to jump to 3D clocks (cause of the stutter I believe). The K-Boost button on helped it start the benchmark test smoother.

1011 core with 1032/2065 memory, 1.15v and it broke the 6818 mark with 102 min. fps and 127 max fps.

Was able to get a 1027 core with 2060 memory, same 1.15v. and a 6900, 102 min fps, 129 max fps but it isn't stable enough and crashes the driver right after it completed the test. Last night I got an insanely high 7200 when I first selected the K-Boost, but haven't even gotten close to repeating that. It did however start the benchmark very smoothly and initially a high fps, so I'm thinking that's why. The K-Boost alone is worth it's weight as it forces the card to be ready to go.

Well, it seems BF4 is more demanding than some of the benchmark tests. Doesn't really explain why it was at playable settings one night, then crashing the next, unless it was dropping the clocks to 2D or something silly. Real shame they locked out the higher voltages as I was getting pretty impressive performance increases, more impressive than I'd ever expect otherwise. Going from a min. FPS of 88 to 102-103, and a max FPS of 111 to 129 is pretty neat with this older card. I also found out that more voltage when it wasn't needed dropped the scores, as did the memory which plateaued very quickly. If you are reading this and want to OC your card, be sure to run gpu-Z or similar in the background and watch your clocks after a crash in case they lock-in at 2D rates. Leaving the K-Boost on seemed to prevent the 2D reversal and I was able to reset the rates w/o resetting the PC. Also, both Afterburner and EVGA P X have a profiles tab in the settings. If you save an OC setting, you can select that setting as a 2D and 3D setting, though I found with Afterburner it'd eventually ignore this and revert, and wasn't until K-Boost was selected with EVGA that it seems to stick and not revert to 2D or less speeds which can cause a stutter if a high load is quickly placed on the card and it's sleeping at a slow "energy saving" speed. Using nvidia's "prefer maximum performance" under the control center's settings didn't seem to help either. So, even if you don't OC and have a newer card, try turning on the K-boost under the voltage settings and force the fans on to max and try your favorite game. I promise you'll see an improvement with response as it'll actually use all of the video card's muscle instead of only a small percent. IF you can't get over 60 fps with a benchmark or similar, be sure to turn v-sync OFF in the driver's settings i.e. nvidia control panel for nvidia cards. Also, be SURE to blow the dust out of the video card every 1-3 months. Winter time with a furnace running can cause more dust build up as it gets pulled in. I like to put my hose on the outlet of my shop vac and carefully blow all the dust out of the PC. You'd be amazed at how much gets built up in rather quick time, and just how much it hurts your efficiency and long term life of components.