possibilities are not adding up, need to talk to a true Bitcoiner ???

[Feras]

So for a solo miner to solve a block -as I understand it- the process includes taking the Merkle tree of a suggested block, but other information like time and then iterate the 2^32 possible Nounce until finding a result that match the less than criteria , and if didn’t then fetch another suggested block and repeat.
My problem is with the numbers, 2^32 is about 4 Giga possible Nonce, and looking at blockchain.info there is an average of 6 transactions a second.
So looking at S9 for example, it does 14Tera hash per second.
So an S9 will check about (14 000 G/ 4 G ) = about 3000 suggested block a second! (if tera hash means tera tested full iteration).
But there are only 6 transactions per second, so a single S9 should be able to test the suggested 9 blocks happening every second!
Unless the suggested blocks list is not as simple as I think, any documentation on that will be great.
Thanks all

[leowonderful]

No.

The S9 indeed does 14TH/S but factoring in difficulty and global hashrate one miner by itself is extremely insignificant and you'll be lucky to find one block in a year, much less 9 blocks a second. Even if this were to happen, difficulty adjustments would bring the block time to 1 per 10 minutes again.

For reference, the total network hashpower is near 8000PH/S and difficulty is near 900 billion.

[Feras]

I did more reading, and I think I'm getting closer to a technical mathematical answer, I'd like to share my understanding so far:

first , yes a single S9 will be able to try all the 2^32 Nonce for about 3000 suggested blocs in one second, but most probably none of these blocks will have a hash output that starts with the defined famous numbers of leading zeros, even though you tried all 2^32 variations for each of the 3000 suggested blocks, so the miner would probably go back and put together a different set of transactions in a new suggested block, hash that block and then iterate thru all possible Nonce for that block and so on, notice 2 things here:

1) there are a decent amount of unconfirmed transactions in a memory pool, and even thu only an average of 6 is added a second the number of variations in both order and size(number of included transaction) is humongous.
2) part of the Message to be hashed is a especial transaction that will reward the miner with his 12.5(for now) bitcoin to his wallet(called coin base transaction), which makes the hashing trial for that miner unique than other miners! this is GENIUS...
it will make what works for one miner not working for the others, distributing the possibility of solving block even more.     

so in other words, the race is not only to find a Nonce for a certain suggested block, it's more about arranging a suggested block, with all of it's ordering, length, time stamp...and coinbase transaction, such that there is a Nonce for it that will make it's hash meet the difficulty needed, only a rare percentage of these block would actually have the ability to give you leading zeros when hashed!
 
reading more now on to actually obtain the unconfirmed transactions and the broadcast of a solution,
 
just LOVE IT

any references to find information about that?!
thanks again