Custom 3.62 GHz CPU and DDR-1900
Here is where things get interesting. This profile was tested with Hyper Threading off, and then benchmarked with Sandra with Hyper Threading on. The temperatures represent the off setting. Notice the massive difference in CPU benchmark between the two modes. Does this correlate to real-world computing? It might, but I have to save something for later articles, right? To even be able to test that difference, you would have to encode something like 4 movies at the same time. That’s the only real-world performance test I could think of. I don’t classify everything as ‘real-world’, so please don’t bring up Folding@home, SETI, or other distributed computing projects.
Regardless, neither Hyper Threading mode passed Prime95. Increasing voltage here did not solve that. This suggests that memory speed/timings need work. We did learn a good lesson here, though. Turning Hyper Threading off realistically saved us 20°C here. That is an absolutely huge margin in CPU temperatures. The temperatures experienced under this 3.62 GHz setting are the same as the lowly 3.07 GHz setting that had Hyper Threading turned on.
Overall, I love the speeds in this benchmark, but it is a bit too hot and not stable. For our final benchmark, we will build off of these settings and leave Hyper Threading turned off. Before checking that out, take a final note of the Sandra CPU scores below. The Hyper Threading results in drastically different scores. Researching the outcome suggests that Sandra is the only place you will ever see this difference and it has no bearing on real-world performance.
|Hyper Threading||ON OR OFF|
|Core Voltage||1.160 V|
|IMC/VTT Voltage||1.2125 V|
|DRAM Voltage||1.400 V|
|Sandra Bandwidth||23.33 GB/s|
|Sandra CPU Math||87.70 GOPS|
|Sandra CPU Math HT:OFF||65.80 GOPS|