"And what impact does it might have on overall performance of both switches ?"
It would most impact the switch the command is applied to.
As to potential impact, there may be none, or it makes things worse or better. Hopefully, the latter, but insufficient information to actually say (i.e. guarantee) what the result will be.
If the issue is drops due to transient burst congestion, and if you have buffers not being used only because you're bumping into logical queue size limits, this change might yield a dramatic decrease in drops. (This is often the case.)
If the issue is drops due to sustained congestion, and if you have buffers not being used only because you're bumping into logical queue size limits, there may be no change or perhaps even more drops.
If for either of the prior two issues, if there are no free buffers, i.e. logical and physical limits are the same, no change.
Of if for the same two prior issues, if there are free buffers, but other interfaces are occasionally making use of them, those buffers might no longer be available for other interfaces, so the above results for the currently congested interface might be the same as noted above but additionally other interfaces will also start to drop packets.
In my first hand experience, and suggesting similar adjustments, to others on these forums, going back to the 3750 series, so far, have not seen or heard anything beyond a positive result without any negative side effects. Again, even Cisco recommends this.
Highly recommend at least trying it, but do monitor the results. If overall impact is "worse", I would NOT expect it to be a major/big worse, just a slight worse.
Again, if there's no positive result, or not as positive as hoped for, there's still more that might be done to "tune" buffers or other QoS "tuning" that may decrease drops on those interfaces. And/or, (often the first "fix" by many) is to provide additional bandwidth might be considered too, e.g. 10g or Etherchannel.
