Benchmarks
Test System:
5960X ES @ 3.5GHz
Asus Rampage V Extreme
Thermaltake Water 3.0 Extreme
Kingston HyperX Predator DDR4 3000 # HX430C15PB2K4/16
Cooler Master V1200 PSU
MSI 290X Lightning
Mushkin Chronos 240GB SSD
AIDA64 Memory Test
In this test the results show that the XMP profile is not as fine tuned as it could be, as looking at the much lower write results at lower cache speeds. This could be due to looser sub-timings in the first XMP profiled. The write speeds do recover nicely with the increase of the cache speeds. These results fall in line with what was expected at these speeds and timings. XMP profile 2 seems to have been tweaked much better than the XMP profile 1. The results are in line with the previous test increases with each jump in cache. The increases with the uncore speed really prove why it is necessary to try and get the cache speeds as high as possible. It is possible to see that the switch to 1T really helps the modules gain some traction with the copy and read speeds that were achieved. It was fairly easy to go to 1T with this kit and it would be advised to run the kit at 1T since it is a very easy way to pick up some free performance boost. Some DDR4 kits have proven to be a bit harder to run with 1T enabled, so being able to run it at 3000 was a welcome event.
SiSoft Sandra Bandwidth Test
These bandwidth tests are very impressive and shows a great deal of improvement with the jumps in the cache speeds. This shows that paying close attention to cache speeds can definitely help out applications that rely on a lot of bandwidth. The gains in bandwidth are helped by the speeds at which the memory operates, as at higher speeds of RAM can help boost the bandwidth a lot. The change from 2T to 1T shows a very nice bump to the bandwidth results. The change from XMP profile 1’s CAS 15 to the 3000 CAS 14 does not make a huge impact on the results. With DDR4 the main things that will help improve performance are the sub-timings, of which tightening up can make drastic improvements to the results. All the results are in line with what was expected from the tests. It was interesting to see that the jump from 3500 cache to 4000 cache didn’t give as big of a jump as the others did.
Super Pi mod 1.5 XS
Here we see that the XMP profiles 1 and 2 are very close to each other with the results and the speeds are not really making that much of a difference. In Super Pi having the memory tweaked and having sub-timings tight give the best results. This seems to say that the XMP profiles are not that fine tuned in the subs which may be done for compatibility sake and making the profiles easier to run. Having something more buttoned down will run the risk of being more finicky to run in systems. The switch to 1T does show a nice gain right off the bat and shows that the speed increase across the board with just that one change.
