Why is the optimization space for ETH's EVM limited?

1. Architectural design determines single-threaded sequential execution

When the EVM was initially designed (in 2015), parallel transaction processing was not considered at all. It assumes:

Each transaction changes the global state (State Trie)

The execution order of transactions must be strictly determined (no out-of-order execution)

There may be dependencies between transactions (for example, A transfers to B, and B immediately transfers to C)

→ Therefore, transactions can only be executed one by one, in a single-threaded manner.

Issue:

Even if your node machine is faster, single-threaded speed will not increase exponentially, so TPS improvement is very limited.

2. Increasing block speed and Gas Limit encounters physical limits

If the blocks are too large (for example, several hundred M gas per block), it would lead to:

Longer synchronization time (new nodes may take several hours or even days to join)

Increased verification pressure (ordinary nodes cannot keep up)

If blocks are produced too quickly (for example, one block every 0.5 seconds), it would lead to:

Network propagation delay becoming a bottleneck (new blocks are generated before the P2P network has finished propagating)

More chain forks (high fork rate)

Thus, it is not just about unlimited improvements; a conflict between decentralization and performance will quickly arise.

3. The compatibility burden of Ethereum's EVM

If you completely overhaul the EVM (for example, by supporting parallel execution), the old smart contracts (such as Uniswap, Aave, OpenSea) cannot be migrated directly.

→ Compatibility is broken, and the cost of ecosystem migration is enormous.

Therefore, Ethereum itself is quite conservative, preferring to do Rollups and not touch the EVM core.

Current industry status:

Now, if extremely high TPS (for example, tens of thousands of TPS) is needed, new chains basically have abandoned the EVM.

Aptos, Sui: Abandon the EVM and design their own Move VM.

Solana: Starts from scratch with a completely different VM.

Those who still insist on EVM usually expand through Layer2 or Rollups, not expecting explosive TPS from a single chain.

In summary:

The optimization space for the EVM itself is not large anymore; at most, it can achieve several thousand TPS. To go higher, it must either significantly change the EVM or simply switch to a new generation virtual machine (such as Move VM, SVM).