Welcome to another PC Build thread. I have been on an i7-7700k desktop for a week and some change. In the last few builds I seem to be on a 2 to 3 year rotation, largely because I know people who could use computers and my custom builds would smoke a retail setup while also having good cooling layouts. Fancy means to say the equipment should run for a long time.
Current edition kit is:
- Processor / CPU:
Asus Hero IX Z270
32 GB Corsair DDR4 3000 MHz LED RAM
- Primary OS on a Samsung 860 Pro M.2 NVMe SSD
- Power Supply:
Corsair 850w PSU
Corsair H115i CPU Cooler
Corsair Obsidian 750D Airflow
- Video Card / GPU:
Asus Strix 980 GTX (Gen 1 – Non Ti)
I kept storage, graphics card, power supply from the previous build. I have so far put together some benchmarks from Final Fantasy XIV tools, 3D Mark products, and ran some GTA V and Watch Dogs 2 for comparison.
The short hand is that, each of the FFXIV Benchmarks gave about 1000 points higher of a score. Effects like elemental magic casts and screens with many extra objects on screen, ran much smoother. Checking benchmark details such as 3D Mark, the i7 does vastly better handling PhysX performance heavy content. Keeping in mind I am running the same video card as I did in the prior i5-4690k build.
For grins I also re-installed Bioshock Infinite and ran the Benchmark.exe in the install folder. Considering I recently switched to a 144 Hz display, I was able to see the benchmark ran that steadily. Looking back, I think I played the game on a 560 GTX and recalled some performance drops in some areas. Worth saying that would have also likely had been when I was on an i5-2500k setup.
3DMark Shootout: Please bear in mind most of the i5-4690k benchmarks were run on Windows 8.1. I believe that is some of where you see the higher FPS numbers from the i5 versus the i7 benchmarks.
Recapping this data, we see the PhysX scores are significantly higher, while base FPS are similar or a little slower than on the i5-4690k. Please keep in mind the only benchmark in this set run on Windows 10 with the i5-4690k was Time Spy, as it requires DirectX 12 / Windows 10. I added the compare links that also confirm this information and so you can add any benchmarks you may have run for comparison.
The last image and benchmark includes my i5-2500k build with a 570 GTX in it. I wanted to add that for more of a scaling over time and performance gain metric. In the short hand if you are asking does an i7 smooth up actual gameplay, I would say Yes to that. Watch Dogs 2 is a recent title I saw people mentioning benefits from extra threads and a higher CPU clock. I can confirm that to be the case as grass and tree heavy shadow environments are much smoother than they were on my i5-4690k. Similar gains can be seen in Final Fantasy XIV. Particularly as the Stormblood expansion pushed some higher usage textures and shading features to the game.
This rig is doing me well but I did have some issues running the RAM @ 3000MHz. I had a few crash application issues and some reboot problems as well. After seeing a few of those I fired up Memtest and let that run for 5 to 10 hours. Typically around the 5th hour, I started getting some errors when running the memory at the overclocked speed. Turns out I’m not the only person with issues running the RAM @ 3000MHz on an Asus motherboard. Stock is 2133MHz and that passed Memtest with flying colors and also does not have the reboot issue.
Just a heads up if you are doing a build, that you might see some issues if you hoped to set your XMP and it would just run. Increasing the voltage to the RAM made no long-term stability improvement either. I am running 32 GB (2x 16 GB) CMU32GX4M2C3000C15 Vengeance LED RAM for point of reference. I was happy to see the Memtest passed at stock timings, so I didn’t have to RMA them. From the product page the timings that should work are:
Tested Latency: 15-17-17-35
Set RAM to Auto
As I mentioned, simply using the XMP settings for 3000MHz will likely jam up your memory and system stability. Normally I would have let an exhaustive memtest run earlier, but you can get busy and I ended up building 3 machines that week. Granted my prior build was mostly a move to another case and burn in testing. Speaking of MemTest, I enjoy this version of MemTest.