gigagas L1

technical deep dive

t.me/ethproofs_community

gigagas L1

technical deep dive

less technical talk for EthCC

t.me/ethproofs_community

zkEL clients and validators

zkEL clients and validators

extreme decentralisation

extreme decentralisation

Raspberry Pi Pico 2 W

​

• cost—$7
• connectivity—WiFi
• CPU—dual Hazard3 RISC-V
• memory—520 KB onchip SRAM
• power—1W ($1/year electricity)

part 1—zero to one: latency

part 2—zero to one: safety

part 3—zero to one: liveness

part 4—one to gigagas

part 1—zero to one: latency

part 2—zero to one: safety

part 3—zero to one: liveness

part 4—one to gigagas

(same slot) delayed proving

block n

proposal

block n+1

proposal

(same slot) delayed proving

block n

proposal

block n+1

proposal

block n

proof

Δ

(same slot) delayed proving

block n

proposal

block n+1

proposal

block n

proof

Δ

EIP-7886

delayed execution

EIP-7732

enshrined PBS

zkEL pipeline

splitting

tracing

crunching

aggregation

fetching

prover

wrapping

zkEL pipeline

splitting

tracing

crunching

aggregation

fetching

gossip

k-of-n check

prover

verifier

wrapping

zkEL pipeline

splitting

tracing

crunching

aggregation

fetching

gossip

k-of-n check

prover

verifier

wrapping

O(1)

O(log(gas))

O(1)

O(1)

O(1)

O(1)

O(1)

O(1)

zkEL pipeline

splitting

tracing

crunching

aggregation

fetching

gossip

k-of-n check

prover

verifier

wrapping

O(1)

O(log(gas))

O(1)

O(1)

O(1)

O(1)

O(1)

O(1)

big enough slots!

zkEL breakthroughs

zkEL breakthroughs

zkEL breakthroughs

zkEL breakthroughs

parallel tracing

Fusaka (SFI)

30M cap

parallel tracing

Fusaka (SFI)

30M cap

2²⁴ cap

Glamsterdam?

proof wrapping

proof wrapping

proof wrapping

crazy idea?

64-bit ephemeral proofs

part 1—zero to one: latency

part 2—zero to one: safety

part 3—zero to one: liveness

part 4—one to gigagas

zkVM diversity

9 zkVMs

zkVM diversity

9 zkVMs

10 zkVMs

zkVM diversity

9 zkVMs

10 zkVMs

13 zkVMs

guest diversity

"host"

"guest"

zkVM

MIPS

LLVM

revm

rv32

bytecode

zevm

WASM

guest diversity

"host"

"guest"

zkVM

MIPS

LLVM

revm

evmone

gevm/eevm

rv32

bytecode

zevm

WASM

guest diversity

"host"

"guest"

zkVM

MIPS

LLVM

revm

evmone

gevm/eevm

"nevm"

"bevm"

rv32

rv64

bytecode

zevm

WASM

guest diversity

"host"

"guest"

zkVM

MIPS

LLVM

revm

evmone

gevm/eevm

"nevm"

"bevm"

Cairo

Keth

Linea EVM

Linea EVM

Lean 4

l4evm

rv32

rv64

bytecode

zevm

WASM

zkVM-guest blowup

​

EVM guest

count

​

×

quadratic blowup

zkVM

count

zkVM-guest blowup

​

EVM guest

count

​

×

quadratic blowup

zkVM

count

idea 1

slow-fast pairs

idea 2

multi-EVM proofs

slow-fast pairing

zkVM #1

zkVM #2

zkVM #3

zkVM #4

zkVM #5

slow-fast pairing

zkVM #1

zkVM #2

zkVM #3

zkVM #4

zkVM #5

EL client #5

EL client #4

EL client #3

EL client #2

EL client #1

slow-fast pairing

zkVM #1

zkVM #2

zkVM #3

zkVM #4

zkVM #5

EL client #5

EL client #4

EL client #3

EL client #2

EL client #1

k-of-n proof checking

n gossip

channels

...

k-of-n proof checking

n gossip

channels

attest if

k-of-n

valid & timely

...

k-of-n proof checking

n gossip

channels

attest if

k-of-n

valid & timely

...

(zk)Vouch

​

→ institutional client diversity

→ intra- (vs inter-)operator

k-of-n proof checking

(zk)Vouch

​

→ institutional client diversity

→ intra- (vs inter-)operator

n gossip

channels

attest if

k-of-n

valid & timely

...

Drake take™

​

→ k = 3 — 1MB of proofs

→ n = 5 — EL client count

k-of-n proof checking

(zk)Vouch

​

→ institutional client diversity

→ intra- (vs inter-)operator

n gossip

channels

attest if

k-of-n

valid & timely

...

Drake take™

​

→ k = 3 — 1MB of proofs

→ n = 5 — EL client count

Gold take™

​

→ k = 3

→ n = 3

formal verification

verified-zkevm.org

$20M

over 3y

formal verification

verified-zkevm.org

$20M

over 3y

github.com/Veridise/Picus

formal verification

verified-zkevm.org

$20M

over 3y

github.com/Veridise/Picus

bugs.zksecurity.xyz

86/89

under-constrained

part 1—zero to one: latency

part 2—zero to one: safety

part 3—zero to one: liveness

part 4—one to gigagas

mandatory proofs

*

mandatory proofs

→ insufficient fees (<1¢/Mgas)

→ prover killers

→ prover crashes

→ completeness bugs

liveness failures

*

mandatory proofs

→ insufficient fees (<1¢/Mgas)

→ prover killers

→ prover crashes

→ completeness bugs

liveness failures

FOCIL

DoS

*

*

cluster proving specs (for 1-of-n assumption)

$100K

10kW

cluster proving specs (for 1-of-n assumption)

$100K

10kW

bottleneck

power heuristic

1x 4090 for 16 slots

power heuristic

1x 4090 for 16 slots

16x 4090s for 1 slot

perfect parallelism

power heuristic

1x 4090 for 16 slots

16x 4090s for 1 slot

10x 5090s for 1 slot

perfect parallelism

Moore's

law

power heuristic

1x 4090 for 16 slots

16x 4090s for 1 slot

home cluster

10x 5090s for 1 slot

perfect parallelism

Moore's

law

GPU proving

realtime.succinct.xyz

bento

proofman

gpu_prover

GPU proving

realtime.succinct.xyz

open source

GPU prover

bento

proofman

gpu_prover

Ethereum L1 wishlist

Ethereum L1 wishlist

Ethereum L1 wishlist

blog.ethereum.org

gradual enshrinement

gradual enshrinement

gradual enshrinement

gradual enshrinement

gradual enshrinement

part 1—zero to one: latency

part 2—zero to one: safety

part 3—zero to one: liveness

part 4—one to gigagas

goal—3x per year

"300 Mgas limit"

goal—3x per year

"300 Mgas limit"

EIP-7938

goal—3x per year

"300 Mgas limit"

EIP-7938

"1 Ggas/s"

dynamic gas limit voting

dynamic gas limit voting

steady

predictable

ambitious

low-touch

primary bottleneck: solved™

real-time

proving parity

exponential

zkVM progress

=

+

3x/year

"easy"

power overhead

1.5GHz RISC-V

execution

15MHz RISC-V

GPU proving

15MHz RISC-V

GPU proving

...

100x

power overhead

1.5GHz RISC-V

execution

15MHz RISC-V

GPU proving

15MHz RISC-V

GPU proving

...

100x

100x

400 Watts

1x

4 Watts

10,000x

overhead

power opportunities

power opportunities

≥10x

software

≥10x

hardware

power opportunities

≥10x

software

≥10x

hardware

≥10x

multi-prem

≥10x

zkISA

node types

• validators
• wallets
• includers

• builders
• explorers
• RPCs

node

sophistication

node types

• validators
• wallets
• includers

• builders
• explorers
• RPCs

node

sophistication

primary bottleneck

secondary bottleneck

node types

• validators
• wallets
• includers

• builders
• explorers
• RPCs

node

sophistication

primary bottleneck

secondary bottleneck

solution:

→ SNARK engineering

solution:

→ traditional engineering

specialised databases

LevelDB

MDBX

used by Geth

used by Reth

specialised databases

github.com/base/triedb

github.com/LayerZero-Labs/qmdb

LevelDB

MDBX

used by Geth

used by Reth

secondary bottleneck: just do it™

secondary bottleneck: just do it™

• Geth
• LevelDB
• LSM-tree
• Reth
• MDBX
• B-tree

​

~50 disk ops

per traversal

​

secondary bottleneck: just do it™

github.com/base/triedb

• Geth
• LevelDB
• LSM-tree
• Reth
• MDBX
• B-tree

​

~50 disk ops

per traversal

​

​

4-8 disk ops

per traversal

​

perfnets

nethermind-ui.benaadams.vip

1M TPS

github.com/LayerZero-Labs/fafo

gigagas Ethereum

today

1.5Mgas/s

goal

1Ggas/s

delayed proving

real-time

proving

onchain

offchain

gigagas Ethereum

today

1.5Mgas/s

goal

1Ggas/s

zk gas

repricing

FOCIL

L1 txs

in blobs

zk-friendly

tree

multi-1559

delayed proving

real-time

proving

onchain

offchain

gigagas Ethereum

today

1.5Mgas/s

goal

1Ggas/s

zk gas

repricing

FOCIL

L1 txs

in blobs

zk-friendly

tree

multi-1559

delayed proving

better

sync

better

state DBs

history expiry

access

lists

p2p erasure

coding

real-time

proving

onchain

offchain

questions :)

personal goals

zkEL · stateless

zkEL · embedded

zkEL · proven

no NVMe

Pi Pico

20x 5090s

Christmas present

6x liquid-cooled 5090

Comino Grando

Ethproofs

"L2BEAT

for zkVMs"

EthCC Cannes

refreshed rollup-centric roadmap

+ thousands of L2s

1 gigagas/sec

1 teragas/sec

• shared assets
• shared minting
• shared deposits
• shared bridging

​

• shared settlement
• shared proofs
• shared blobs
• shared compression

hub

...

united chains of Ethereum

EF zkEVM

zkSTF

gradual enshrinement

• 2025 (Q4)
• no fork needed
• hybrid attesters
• 1% opted in?
• altruistic provers
• Ethproofs, grants
• delayed attesting
• smaller rewards
• alternative hack
• optimistic attesting

• 2027
• proof timeliness
• semi-enshrined
• 100% adoption
• fork choice
• rational builders
• fee incentives
• collateral
• incentive aligned & safe

phase 0

early adopters

phase 1

delayed execution

phase 2

mandatory proofs

• 2026 (Glamsterdam?)
• delayed execution
• hybrid attesters
• 10% opted in?
• altruistic provers
• Ethproofs, grants
• much less hacky
• aligned and safe
• except killers

phase 3

enshrined proofs

• 2028
• formal verification
• fully enshrined
• onchain proofs
• deterministic
• rational builders
• fee incentives
• collateral
• incentive aligned & safe
• new powers
• native validiums
• long-term cleaner

"no limit" native rollups

zk day

Berlin interop

t.me/ethproofs_community

L1 zkEVMs