Hello my friends. It has been quite a while since I got my GTX 970 and since I had some spare hardware laying around, I thought that it would be interesting to test how it performs paired with various cpus. Well for me certainly it is. So what I did, is benchmark this card in many games (custom gameplays and built in benchmarks), demos and stand alone benchmarks so we can see what to expect with different configurations. To tell you the truth, the initial goal was to create a benchmark database, with a highly clocked i5-2500k and a GTX 970 @1.5Ghz. I wanted to have benchmark results without cpu limits, so I can compare them against future cards I get. Under this scope, you must understand that this test is based primarily on a suite that was supposed to do GPU benchmarking, but since I was at it, the thought occur to see how older cpus would perform with real life gaming settings. Also consider this test, a tribute to older cpus, with which we had a lot of fun. Of course since I decided to go on and test my other cpus, the benchmark suite deviated somewhat, that's why you will see framerate capped games in this suite, since after a point I wanted to see how the cpus would fair. All in all then, it is a mixed test, with focus on the gpu but has the very important goal to inform people owning cpus with a varying degree of performance, what kind of gpu could their system handle and what to expect from it. The good thing is that having this database ready, I can benchmark new graphics cards against the 970, only redoing benchmarks that are greatly affected by newer game versions or drivers. I could test similarly powered AMD graphics cards versus the 970, my 7950 CFX vs the 970 or I could just test AMD cards in cpu limited games to see how the AMD driver fairs. The possibilities are endless. But these are stories for another time. I still have this one to finish, lol. To that goal, many game engines have been benchmarked, implemented on even more, very well or less known games. Cry Engine, Frostbite, Dunia, Unreal, Anvil, Unigine, Asura, IW6, Hammer, Chrome, EGO, you name it, all have been benchmarked! The test suite contains 70+ benchmark tests (dunno how many exactly but I guess they are more than 70 ), that have produced 108 benchmark results. My rigs are the following: The 2500k system Case: Coolermaster HAF 932 Mobo: Asrock P67 Extreme 6 RAM : 2X8GB TEAMGROUP Vulcan DDR3-2400 PC3 19200 CL 11-13-13-35 CPU : Intel Core i5-2500k @4.8Ghz CPU cooler : Noctua NH D14 GPU : Gigabyte G1 GTX 970 @1.5Ghz PSU : Coolermaster Silent Pro 1000W The i7-860 system Case: Thermaltake Armor Mobo: MSI Big Bang Trinergy RAM : 2X4GB G.Skill Ripjaws DDR3-2000 PC3 16000 CL 9-11-9-29 1.60v CPU : Intel Core i7-860 @4Ghz CPU cooler : Noctua NH U12P SE2 GPU : Gigabyte G1 GTX 970 @1.5Ghz PSU : Enermax Galaxy 850W The Q9550 system Case: Coolermaster Elite 430 black Mobo: Asus P5Q Deluxe RAM : 4X2GB OCZ PC2-8000 Platinum DDR2 CL 5-5-5-18 2.10v CPU : Intel Core 2 Quad Q9550 E0 @4GHz CPU cooler : Thermalright Ultra 120 GPU : Gigabyte G1 GTX 970 @1.5GHz PSU : Nexus Silent Modular 850W PSU (RX-8500) For all the tests, same settings and same game versions have been used. There is only one discrepancy in Far Cry 4, where I have used 1.8.0 for the Q9550 benchmark, 1.7.0 for the 2500k and 1.9.0 for the 860. According to the 1.8.0 changelog, this patch contains no performance changes, so we are OK. http://www.guru3d.com/news-story/far-cry-4-pc-patch-1-8.html and afaik 1.9.0 only added the Valley of the Yeti DLC. Now the semi ugly part comes to the driver versions used, because due to the sheer size of this benchmark suite, plus my Youtube uploads that took too long (will explain later), the benchmark suite has spanned four driver releases. 347.25, 347.52, 347.74 and 347.88. 350.12 has been used for all systems in GTA V, so that doesn't count. The ugly part comes to the fact that I was using the new driver as I went, so the older drivers are used on the 2500k for which I did the benchmarks first, then newer drivers for the Q9550, then went back to add some benchmarks to the 2500k and now I am still doing the 860 benchmarks which will have the newer 350.12 drivers for half its benchmarks. Just to come clean, I provide the release highlights, so you know what changed and how it affects this benchmark suite. 347.52 http://www.nvidia.com/download/driverResults.aspx/81877/en-us only Alien Isolation, Dragon Age Inquisition and Grid Autosport are affected with low percentage performance increases 347.71 http://www.guru3d.com/files-details/geforce-347-71-driver-download.html no performance changes 347.88 http://www.guru3d.com/files-details/geforce-347-88-whql-driver-download.html Battlefield Hardline driver and some new SLI profiles, irrelevant with the validity of this test 350.12 http://www.nvidia.com/download/driverResults.aspx/83819/en-us GTA V driver, which has been used for all GTA V tests in this suite anyway So all in all, as you can see, we are OK. The performance improvements of 347.52 have gone to waste anyway, since the Q9550 was very cpu limited in these games. Maybe these improvements will help the 860 a bit, but we will see that in an update. Now finally the time has come for the good part. The results! I will provide the results in various ways. First there will be a text based vbulleting table, then some screenies from my excel file that contains the sum up in alphabetical order, sorted screenies by highest performance difference and of course graphs. Lots of them. For now only the i5-2500k vs Q9550 results will be published, but I am halfway through for the i7-860. On the 860 I will only test the most cpu limited games/benchmarks. Lets go and see them one by one. First are the text based vbulleting results. What is so special about these results, is that each score is a hyperlink which leads to my Youtube channel and shows how this benchmark result came to be. If you want to check it out that is (warning for spicy wallpapers being used). As you can understand, we are talking serious data uploading here, patiently done with 1mbit upload speed. Especially considering that I upload huge 20mbit files :S All videos are recorded and uploaded raw at 1080p/60fps. I use an external recorder (Elgato HD 60) so what you will see in all videos is pure performance. I spend a lot of time in every video, showing the system and settings. You may want to skip that. At the end of each video, I show the graphs from Task Manager and MSI Afterburner, with framerate, frametimes, cpu usage, etc. In some videos you will hear something that resembles coil whine. It is not. In my infinite wisdom I had passed my microphone cable underneath the case of one of my systems, right under the PSU. So every electrical disturbance the sytem created, the microphone captured it. My 970 is dead silent. The quality/versus column, apart from giving you a basic info of the settings I used (every detail is recorded in the respective video links), also contains hyperlinks that will take you to youtube doubler, where you can see side by side comparisons of the runs. The difference of the hyperlinks in the columns, is that the ones with the numeric framerate results, lead to the video from the beginning, but the "versus" one takes you to the part where the benchmark started. In videos that contain more than one result of the table (3dmark, AMD Demos, Lost Planet 2, Mafia, Call of Pripyat, Hitman Absolution, Just Cause 2, Rome II, UE4 demos) , the hyperlink of the result will take you to the beginning of the video, but the versus hyperlink will take you to the part that each respective benchmark starts. For the side by side hyperlinks, you can use ctrl+mouse wheel up to enlarge the windows, use pause to resync the videos when they go out of sync etc. Ok I have followed the exact same path in my custom gameplay benchmarks, but I am only human guys. Hoping this isn't too confusing, here is the vbulletin text based table with the benchmarks.