I'm playing dead island with fraps running, I noticed I was getting around 120 fps the it would bog down to 30 fps then it would go back up to 120 fps then back down to 30 fps. It just stuck with this cycle continuously changing every 3-5 seconds.
I looked at my GPU usage and it was like a high peak and then a low peak and it just did that the entire usage... I'm not sure if my video card is dying or what...
Your GPU is throttling, that is what is going on with your setup. For some reason the video card is giving you max output, then throttling down to near-zero Mhz, right? What about the compatibility of your PSU, the CPU, and the overclock you are undoubtedly running? Is your PSU strong enough to cover all of it the way you are presently running it?
You need to list your specs completely, from A to Z. Doesn't this place have a template for posting here, like we have over at EVGA.com Forums? There if you ask a tech question you HAVE TO post your system's specs before anyone seriously will even deal with the answers part of it, you dig?
What is your OC, if any? On the CPU, type of cooler for the CPU, if any; general specs about voltages you are running would help also. Anything significant would be in your best interests, I assure you.
I am personally interested in the OC, the settings for the OC on the CPU, ie volts at the key registers would help immensely, type of PSU and ratings of same, number of fans, type of case and any mods to it etc.
I run twin GTX-560 Ti SC's at 980-1960-2170 but I've increased the voltage at the video card using MSI Afterburner, which is superior to EVGA's controller by a long bunch of gulls, Precision it is like you are using...I am running just a slight voltage increase to 1.08750V, up from the stock 1.06750 in my particular Ti SC card. And it works a treat, for example...! That's in a Gulftown 12- Core server, with 24GB RAM, OC'd to 4215Mhz 24/7, Megahalems CPU Cooler, 9 internal Scythe fans, plus the stock 230mm front LED fan, and a 140mm Scythe exhaust fan.
You might add what temps you typically run under load, and at idle also, that would help also. Especially the video card's readings!
That helps a bunch. Personally I'd like to see at least a 750W PSU with an overclocked 2500k and a 560 Ti video card, but if push comes to shove that Corsair 650W "enthusiast" PSU should have the guts to pull the ship out of the harbor without much sweat. I've personally never used anything smaller than 750W and on the Pro end of Corsair's models, which have some built-in headroom for their power ratings from what I've seen and tested myself, along with others in my computer groups.
But you really don't have that much going on to warrant GPU throttling rearing its ugly little head and screwing up your user experience. Could you download and test the GPU with OCCT v.3.1.0 (not the new Beta 4.x, the older file v3.1), and see if you get the throttling going on with that utility also? Check both the GPU's CPU itself, and then run the Memory exclusive test, and for no less than 1 hour each test, so your situation will reveal itself at some point therein. OCCT has some fantastic graphs which will get produced if you check the "write to file" option, and it will place a bunch of great graphics and graphs in your Documents folder in a folder called "OCCT", which it will make for you, and you'll have a good record of the things that happened during the test. DO NOT abort the test, let it go full speed ahead for an hour at least, or the graphics will cease to exist in some cases, other times they just get corrupted and don't display well.
Throttling in OCCT is an obvious graphic line for "CPU Usage" that will waver drastically as the session persists, you can't miss it! If the CPU keeps going from "On" to almost "Off" the graph will tell you that, and then you begin to investigate the causes which are possibly making the problem happen.
Do not rule out a faulty PSU, a bad cable, important defect within the video card itself, but the OCCT graphics should tell you a LOT about what is going on...great test!
That's where we start this movie out, and if you get Throttling during the OCCT sessions then you know there's a cause somewhere. OCCT will also stress test the PSU very, very nicely, which you can do after the graphix tests are finished, if there's throttling going on. Do you have another PSU of any type, that is adequate for the job, just in case the PSU turns out to be faulty? If the PSU turns out faulty, not saying it will or won't just speculating here...then you're facing an RMA with Corsair, which takes 10-15 days for a PSU if everything kicks into gear the right way, longer if there are hangups. I've RMA'd three PSU's of the Corsair brand since I started building computers about 10 years ago, and they are generally very quick about getting the new product back to the customer, knowing the machine is most likely down for the count in the interim.
That's my best advice for you tonight...run the OCCT tests and see what happens there...you can also run Furmark's test, but the graphics are not so sweet as OCCT's. It will tell you if the video card is generating ERRORS, that it will do. I use Furmark for long-term tests of video cards that are suspect as it really can "toast" a card thoroughly to the maxx, but it's best to literally watch the testing with Furmark constantly....it's not a test to "set and forget" like you can with OCCT.
Good luck, and hope something above makes sense for you...