Orius International Investment LLC retweetet
Is it possible to scale to millions of TPS with 80ms finality on a decentralized network?
Most people would say no.
The physics don't work. The latency alone makes it impossible.
Yesterday's Supernova demo proved them wrong.
And here's the part nobody's talking about yet. ⬇️
———
MultiversX's Supernova upgrade just had its first live demo yesterday and hit numbers that even surprised the team.
With a goal of taking block time from 6s down to 600ms with instant finality, they showed that 600ms is just the starting point.
Reaching 400ms and even 200ms is very possible.
But wait, I thought you said 80ms?
Let me break down what this means.
———
With the full release of Supernova dropping in the next couple months, MultiversX's block time will be lowered to 600ms.
Currently 6 seconds.
Soon 600 milliseconds.
That means a new block with transactions will be formed and finalized every 600ms.
Those transactions are truly final too. Since the last update, Andromeda, MultiversX released single slot finality.
No more waiting for confirmations like is required for most blockchains.
Confirmations? Huh?
Ok let's take a step back.
———
When you submit a transaction, it is seen and confirmed by a group of validators.
In order for a chain to fully verify and confirm a tx is legit, it often takes multiple rounds or blocks.
Here's how MultiversX compares to many major L1:
📊 Ethereum
Block time: ~12s
Finality: 13 minutes (64 confirmations)
⚡ Solana
Block time: ~400ms
Finality: ~13 seconds (32 confirmations)
🔷 Cardano
Block time: ~20s
Finality: 1-2 minutes
🔴 Polkadot
Block time: ~6s
Finality: ~30 seconds
🟢 Algorand (one of the few with instant finality)
Block time: ~2.8s
Finality: Instant, no confirmations
⭕ Tron
Block time: ~3s
Finality: ~1 minute (19-20 confirmations)
⚫ MultiversX (with Supernova)
Block time: 600ms → 200ms possible
Finality: Instant, same shard
Cross-shard: ~1.5s average
———
See the problem?
Everyone else is measuring finality in seconds or minutes.
MultiversX is measuring in milliseconds.
And they're not even pushing the limits yet.
———
But that's not even the really impressive part.
Currently we are witnessing over 500ms of idle time on each block. But what does that mean?
When a round begins, a leader is selected with VRF (provable randomness), leader bundles the txs, 2/3rd +1 of validators sign, and the block is now irreversible, aka final.
This is taking on average 82ms right now:
⏱️ 1-5ms → Leader Selection
⏱️ 10-40ms → Voting
⏱️ 10-40ms → Signature aggregation
Total: ~82ms to FULL finality
That leaves 500ms+ of idle time.
Translation? The network can go even faster.
———
Let's touch on a few things here.
First, the signature aggregation and broadcast is where some magic happens. Any honest node can aggregate signatures, verify >2/3 voting, and submit.
The Andromeda release introduced ECP which allows any validator to finalize the block, NOT just the selected leader. This removes single leader failure which is far bigger than most realize.
Read more here: x.com/DBCrypt0/statu…
Second, with Supernova, execution now happens in parallel across nodes POST-FINALITY.
This is where typical network latency comes into play which currently is around 120ms for global distribution.
———
So what's the big takeaway here?
The team set a goal of reducing block time from 6 seconds to 600ms and far outperformed even their own expectations.
MultiversX will soon see block times of 400ms or even as low as 200ms.
That is on top of instant finality so we are talking fully final intra-shard txs of possibly 200ms and cross-shard of ~500ms.
All on top of the first fully sharded chain that is capable of scaling to millions of TPS on minimal hardware, not server-grade machines.
A Samsung smartphone is currently validating blocks on the network to give you an idea of what's possible.
No other network can come close to touching that that I am aware of.
This is what is needed for a truly decentralized and scalable global network.
This is MultiversX.
English
