This is promising, but it is important to keep in mind what is actually in the paper.
First, the average reductions in TFI in the active treatment and active washout phases are 7.51 and 6.71 points, respectively. These were both significantly different from 0, but they are both much less than 13 which is the cutoff for clinical significance. Additionally, the mean reduction during the sham phase is about 2 points, and the sham washout has a mean reduction of a bit less than 2. Thus, the effect of the treatment itself (active - sham) is about 5.5 to 4.7 points - not large. Either way, on average the treatment has a statistically significant but clinically insignificant effect on the TFI.
Second, 10 of the 20 subjects did have a clinically meaningful reduction of at least 13 points during the active treatment. This is good. However, we don't know what reductions these individuals had during the sham phase. In fact, 4 participants had clinically significant reductions during the sham phase, and 2 participants had clinically significant reductions during BOTH phases. The authors do say that the two who experienced reductions during both phases experienced larger reductions during the active phase, but they don't provide the information so that the reader can understand the differences. It would have been useful to see individual level data. This would be possible with only 20 subjects, and it would allow the reader to see the degree to which individuals who saw improvement in the active phase also saw improvement in the sham phase. (Ideally you would like to see responses during the active phase and no response during the sham phase for each person. Assuming they are correct that the sham is truly a sham, it acts like a placebo so if you see a reduction in TFI or loudness during the sham period this is due to the placebo effect. That's why you need to subtract out the effect of the placebo in order to determine the effect of the treatment.)
Third, assuming the treatment truly does have a clinically significant reduction for a significant proportion of people who are treated, it would be useful to know ahead of time who might benefit. In this case, there were no demographic differences between the people who improved significantly and those that didn't. This may be due to the very small sample. Assuming the results persist in a larger sample, perhaps observable differences will observe.
Fourth, there is a reduction in loudness during the treatment of about 8 dB. Unlike the reduction in TFI, this effect becomes smaller during the active washout. Thus, it appears that any sustained reduction in perceived loudness requires continued use of the device. However, the 8 dB reduction isn't all due to the treatment. The effect during sham treatment is a reduction in loudness of about 3.5 dB or so (hard to say exactly from the figure). Thus the net reduction due to the treatment is about 4.5 dB. (In Figure 6C, it's interesting that there is a similar relationship from sham to sham washout that there is between the active and active washout.)
Fifth, 2 people did report the elimination of tinnitus "toward the end of the active treatment period" though there is no information the durability of that result.
So there are promising results here, but they aren't earth-shattering. It's important to remember this is a small study with some promising results. We won't really know more until they do a larger study which is or will soon be underway.