Second graphics card not working properly (slow speeds)

I am running two GTX 1070s in SLI, and I am noticed that I am having trouble with rendering with Blender (rendering a single simple object was running incredibly slow), so I ran benchmarking. I had previously used the Unigine Heaven Benchmark 4.0 when I first built my computer, and even though I was running two monitors, I was getting some pretty high scores (FPS: 171.5; Score 4321; Min FPS: 36.2; Max FPS: 300.7). When I was running the test then, I noticed that both the graphics and memory were running more or less top speed (Graphics: 1999 MHz; Memory: 4004 MHz). Now, when I run the same test under the same conditions, the first card still runs at that speed, but the second card has now dropped (Graphics 607 Mhz; Memory: 405 Mhz). The benchmark results are abysmal by comparison (FPS: 24.1; Score 607; Min FPS: 17.6; Max FPS: 35.7). The temperature on the first card goes up to the mid 70s, but the second card stay more or less in the upper 40s (not even enough to kick in the fan). I've seen it a couple times during the benchmark run that the second graphics card gets up to the speed of the other one, but that lasts only for a few seconds before dropping back down. I am not sure what is going on here.

There have been a couple changes since I built the computer. For one, it's approaching two years since I built it. I've also added a USB 3.0 card, a third monitor, and an Oculus Rift, in addition to drivers and BIOS updates (including the latest GeForce driver), software installs, etc. I also had to turn down my RAM speed to 2100 MHz (from 2933 MHz) in order to get the Oculus Rift to work (it was really laggy in tracking and kept jumping around in virtual space until I did that).

Here are the specks on my computer:
MB: Asus Z170-A
RAM: 32 GB (rated at 3200 MHz, running at 2100 MHz)
CPU: Intel i7 6700K running at 4.6 GHz (liquid cooled)
GPU: GTX 1070 x2 in SLI (24.21.13.9836; 4095 MB)
OS: Windows 10 64 bit

Any ideas about what is going on here?
 
PSU is an RM750 from Corsair.

I'll try pulling the PCIe card and see what happens, but there is currently nothing connected to it and it's not making a difference (plus I need it to run the Oculus, as for some reason it wouldn't work properly if I was just running on the USB ports on the MB).

I also found some benchmark tests for Blender, and I ran the BW one. Here are the results:
bmw27_cpu: 7:47.62 (Mem: 140.47M, Peak: 140.77M)
bmw27_gpu: 10:27.79 (Mem: 140.47M, Peak: 157.36M)
 
Hi Robert, thanks for your suggestion, but unfortunately it didn't solve the problem. I tried pulling the PCIe card and running the benchmark again, with no real difference. I also tried setting the computer back to more or less the state it was in after I built it (i.e. down to two monitors, disconnected the Oculus, no unnecessary USB devices, etc.), but saw only minor improvements. The second graphics card also tended to sit around the previous values with brief fluctuations (i.e. Graphics 607 Mhz; Memory: 405 Mhz).

Summary
Current test results:
-FPS: 24.1
-Score: 607
-Min FPS: 17.6
-Max FPS: 35.7

PCIe removed
-FPS: 24.7
-Score: 622
-Min FPS: 15.7
-Max FPS: 49.5

Stripped down to more or less original
-FPS: 34.4
-Score: 867
-Min FPS: 17.8
-Max FPS: 66.8

Original testing (2016)
-FPS: 171.5
-Score: 4321
-Min FPS: 36.2
-Max FPS: 300.7
 
I took a closer look at my previous benchmark results from 2016, and noticed that I was running on only one monitor. I ran the benchmark again with only the one monitor, and the results went back up (with even better results than two years ago in fact; score is now 4386 ). So I guess I should change my question to why the FPS and rendering drop so much when running multiple monitors, and why my render benchmarks in Blender are so low. (Or are they fast? I'm having trouble finding benchmark results for Blender that would be comparable to my set-up, but even one 1080 will render that image in under two minutes, and an i7-5820K will render it in under 5 min.)