What good is upgrading without real numbers to compare the systems? Well, actually, it’s just as good. After all, it is still the same upgrade with or without benchmarks. But I did some anyway. I even overclocked the CPU and GPU and did some benchmarking with those updates as well. This resulted in a surprisingly small increase in performance.
|OLD PC||NEW PC||Δ%:||New PC OC||Δ% over non-OC:|
|8GB RAM||16GB RAM|
|GTX 550ti||GTX 960|
|PC Mark 8 Home||3392||4450||31%|
|PC Mark 8 Work||2998||3817||27%|
|PC Mark 8 Creative||2633||4585||74%|
|3D Mark Ice Storm||71504||170795||139%||176984||4%|
|3D Mark Cloud Gate||11786||24219||105%||24943||3%|
|3D Mark Sky Diver||6359||21602||240%||21872||1%|
|3D Mark Firestrike||1834||6899||276%||6991||1%|
|3D Mark Firstrike Extreme||514||3511||583%||3544||1%|
So, as you can see, the old computer was not terrible, but the new computer smokes it. Something interesting came out of the PC Mark tests. PC Mark tests things like audio compression, video scaling, spreadsheets, gaming, web browsing, etc… The new computer did slightly worse times on the web browsing (I blame no windows 8 drivers for my wireless network adapter) and, here is the big surprise, it was a ton slower on the gaming tests. Even though, as you can see by the graphics benchmarks, that gaming performance should be its staple measure. The gaming tests on PC Mark use Direct X 9. The tests for 3D Mark use Direct X 11 (IIRC). So is the new card really that bad at outdated tech or is something else messing with its chi?
BTW, every other test in the PC Mark tests was crazy fast. To the point that total test time dropped from 64 minutes to 44 minutes on the longest one. By crazy, I am talking things like the “video to go part 2” test dropped from 68 seconds to just 14. That’s an 80% drop.
So I think the overall numbers, even though they increased by a good margin, I think are severely being skewed by the poor gaming test performance.
After doing all the benchmarks, I decided to play with overclocking. The Gigabyte motherboard comes with some software that will automatically overclock you CPU and run it through some stability tests until it breaks it then it turns you back a few notches for your final setting. It tested my 4.0GHz system all the way up to 5.1 GHz before it failed. And the software settled on 4.6GHz for my final setting.
The Asus STRIX GTX 960 OC2 is made for overclocking and came pre-overclocked a little. I kicked it all the way up to 1420MHz and it ran fine. So I retested GPU benchmarks. And as you can see, the improvements were essentially statistical anomalies. In other words, the numbers vary that much each time you test anyway.
At this point, I did some load testing on the CPU and found that at 4.6GHz, it was getting up to 91°C which is a little higher than I wanted to see. With the CPU running at 4.0GHz, it never got over 75°C. So I turned it down to 4.4GHz and load tested again. At 4.4GHz, it never got over 75°C, just like the stock setting. So I left it at 4.4GHz. Looks like I found the limitation of that CPU cooler. And due to the insignificant performance gains the GPU overclocking showed, I turned it back down to the factory overclock setting.
Now, the website I used to help decide on CPU and video card choice is maintained by Passmark. They have their own benchmarking software for the full PC. And after running those tests, Iam thinking they may be more reliable than the PCMark tests.
Here are my Passmark scores:
|OLD PC||NEW PC||Δ%:|
|G3D Mark (video card)||2072||6681||222%|
According to the Passmark site, my CPU should score 11245. I am overclocked a bit, so I beat it. And their site says a standard GTX 960 video card is rated at 5987 where mine got 6681. I attribute that to the Asus card being overclocked. As for the hard drive score, have I said lately how much I love SSD’s? And mine score better than a bunch of Crucial and Kingston that seem to be gamer favorites.
Also one of the baseline reference systems Passmark compares my results to is a i7-5820K with a gigabyte X99 MB, 8GB RAM, same size Crucial SSD, and the vastly superior GTX 980 video card. And that system is rated at 4995 compared to my 5251. So yeah, I feel pretty good about this new build.