Understand RDS PIOPS, EBS IO and EBS BYTE Balance (%) ?

I have a Posgrest RDS instance using r5.xlarge type. 500GB SSD gp2 type. As I understand, with 500GB gp2, i got a baseline with (3 x 500) = 1500 IOPS.

Now I need to increase it to 2500 IOPS, what should I do ? As documents said, I have 2 option (please correct me if I'm wrong) :

1. I can increase the size of DB to ~ 850 ( 3x850 ~ 2500 IOPS)
2. Change the disk type to IO1 and set PIOPS = 2500.

500GB Gp2 cost 115$ per month

With option 1, I has to pay 195$ per month. With option 2, I has to pay 115$ + 500$ ( 0.2 * 2500) = 615$ per month.

I know that Gp1 Provide more throughput and SLA level, but do I really need to use io1 + PIOPS? Which case should I use it (assume that I just need 99% SLA) ?

And one more question, assume that I have RDS with 1000 GB Gp2, so the baseline is 3000 IOPS, what happen if I change it to io1 and set PIOPS to 1000? Now what is the baseline IO of my RDS? 3000 or 1000 or 3000 + 1000 ?

I saw EBS IO Balance (%) and EBS Byte Balance (%) in CloudWatch metric, as I understand, it's my reserved balance of (IO and Throughput), but how do I know the absolute value of it ? (So I can count how many IO balance remaining)

Let say I have RDS with 1000GB Gp2, as I understand from documents, I got 3000 IOPS, if my RDS used < 3000 IOPS, it will reserved the IO credits to my balance, but what is the maximum balance that I can reserved? I couldn't find the documents said about that.

Is there anyway to monitor how RDS consume my IO ( independently of AWS ) ?

Thank you so much.