You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This is the _headliner_ of the Fusaka fork, the main feature added in this upgrade. Layer 2s currently post their data to Ethereum in blobs, the ephemeral data type created specifically for layer 2s. Pre-Fusaka, every full node has to store every blob to ensure that the data exists. As blob throughput rises, having to download all of this data becomes untenably resource-intensive.
28
24
29
25
With [data availability sampling](https://notes.ethereum.org/@fradamt/das-fork-choice) , instead of having to store all of the blob data, each node will be responsible for a subset of the blob data. Blobs are uniformly randomly distributed across nodes in the network with each full node holding only 1/8th of the data, therefore enabling theoretical scale up to 8x. To ensure availability of the data, any portion of the data can be reconstructed from any existing 50% of the whole with methods that drive down the probability of wrong or missing data to a cryptographically negligible level (~one in 10²⁰ to one in 10²⁴).
@@ -37,9 +33,7 @@ This keeps hardware and bandwidth requirements for nodes tenable while enabling
37
33
-[DappLion on PeerDAS: Scaling Ethereum Today | ETHSofia 2024](https://youtu.be/bONWd1x2TjQ?t=328)
38
34
-[Academic: A Documentation of Ethereum’s PeerDAS (PDF)](https://eprint.iacr.org/2024/1362.pdf)
39
35
40
-
#### Blob parameter only forks {#blob-parameter-only-forks}
Layer 2s scale Ethereum - as their networks grow, they need to post more data to Ethereum. This means that Ethereum will need to increase the number of blobs available to them as time goes on. Although PeerDAS enables scaling blob data, it needs to be done gradually and safely.
45
39
@@ -49,11 +43,15 @@ These coordinated upgrades generally include a lot of changes, require a lot of
49
43
50
44
Blob parameter only forks can be set by clients, similarly to other configuration like gas limit. Between major Ethereum upgrades, clients can agree to increase the `target` and `max` blobs to e.g. 9 and 12 and then node operators will update to take part in that tiny fork. These blob parameter only forks can be configured at any time.
51
45
52
-
#### Blob base-fee bounded by execution costs {#blob-base-fee-bounded-by-execution-costs}
46
+
When blobs were first added to the network in the Dencun upgrade, the target was 3. That was increased to 6 in Pectra and, after Fusaka, that can now be increased at a sustainable rate independently of these major network upgrades.
47
+
48
+
#### Blob base-fee bounded by execution costs {#blob-base-fee-bounded-by-execution-costs}
57
55
58
56
Layer 2s pay two bills when they post data: the blob fee and the execution gas needed to verify those blobs. If execution gas dominates, the blob fee auction can spiral down to 1 wei and stop being a price signal.
59
57
@@ -63,89 +61,129 @@ EIP-7918 pins a proportional reserve price under every blob. When the reserve is
63
61
- layer 2s pay at least a meaningful slice of the compute they force on nodes
64
62
- base-fee spikes on the EL can no longer strand the blob fee at 1 wei
65
63
66
-
### Gas limits, fees & DoS hardening {#gas-limits-fees-and-dos-hardening}
#### History expiry and simpler receipts {#history-expiry}
72
+
73
+
In July 2025, Ethereum execution clients [began to support partial history expiry](https://blog.ethereum.org/2025/07/08/partial-history-exp). This dropped history older than [the Merge](https://ethereum.org/roadmap/merge/) in order to reduce the disk space required by node operators as Ethereum continues to grow.
74
+
75
+
This EIP is in a section apart from the "Core EIPs" because the fork doesn't actually implement any changes - it's a notice that client teams must support history expiry by the Fusaka upgrade. Practically, clients can implement this any time but adding it to the upgrade concretely put it on their to-do list and enabled them to test Fusaka changes in conjunction with this feature.
#### Set upper bounds for MODEXP {#set-upper-bounds-for-modexp}
71
80
72
81
Until now, the MODEXP precompile accepted numbers of virtually any size. That made it hard to test, easy to abuse, and risky for client stability. EIP-7823 puts a clear limit in place: each input number can be at most 8192 bits (1024 bytes) long. Anything bigger is rejected, the transaction’s gas is burned, and no state changes occur. It very comfortably covers real-world needs while removing the extreme cases that complicated gas limit planning and security reviews. This change provides more security and DoS protection without affecting user or developer experience.
73
82
74
-
#### Transaction Gas Limit Cap {#transaction-gas-limit-cap}
#### Transaction Gas Limit Cap {#transaction-gas-limit-cap}
77
86
78
87
EIP-[7825](https://eips.ethereum.org/EIPS/eip-7825) adds a cap of 16,777,216 (2^24) gas per transaction. It’s proactive DoS hardening by bounding the worst-case cost of any single transaction as we raise the block gas limit. It makes validation and propagation easier to model to allow us to tackle scaling via raising the gas limit.
79
88
80
89
Why exactly 2^24 gas? It’s comfortably smaller than today’s gas limit, is large enough for real contract deployments & heavy precompiles, and a power of 2 makes it easy to implement across clients. This new maximum transaction size is a similar to pre-Pectra average block size, making it a reasonable limit for any operation on Ethereum.
81
90
82
-
#### MODEXP Gas Cost Increase {#modexp-gas-cost-increase}
#### `MODEXP` gas cost increase {#modexp-gas-cost-increase}
85
94
86
95
MODEXP is a precompile built‑in function that calculates modular exponentiation, a type of large‑number math used in RSA signature verification and proof systems. It allows contracts to run these calculations directly without having to implement them themselves.
87
96
88
97
Devs and client teams identified MODEXP as a major obstacle to increasing the block gas limit because the current gas pricing often underestimates how much computing power certain inputs require. This means one transaction using MODEXP could take up most of the time needed to process an entire block, slowing down the network.
89
98
90
-
EIP‑7883 changes the pricing to match real computational costs by:
99
+
This EIP changes the pricing to match real computational costs by:
91
100
92
101
- raising the minimum charge from 200 to 500 gas and removing the one‑third discount from EIP-2565 on the general cost calculation
93
102
- increasing the cost more sharply when the exponent input is very long. if the exponent (the “power” number you pass as the second argument) is longer than 32 bytes / 256 bits, the gas charge climbs much faster for each extra byte
94
103
- charging large base or modulus extra as well. The other two numbers (the base and the modulus) are assumed to be at least 32 bytes - if either one is bigger, the cost rises in proportion to its size
95
104
96
105
By better matching costs to actual processing time, MODEXP can no longer cause a block to take too long to validate. This change is one of several aimed at making it safe to increase Ethereum’s block gas limit in the future.
This creates a ceiling on how big a block is allowed to be - this is a limit on what's *sent* over the network and is separate from the gas limit, which limits the *work* inside a block. The block size cap is 10 MiB, with a small allowance (2 MiB) reserved for consensus data so that everything fits and propagates cleanly. If a block shows up bigger than that, the clients reject it.
112
+
This is needed because very large blocks take longer to spread and verify across the network and can create consensus issues or be abused as a DoS vector. Also, the consensus layer’s gossip already won’t forward blocks over ~10 MiB, so aligning the execution layer to that limit avoids weird “seen by some, dropped by others” situations.
101
113
102
-
Ethereum adds a hard cap on the [RLP](/developers/docs/data-structures-and-encoding/rlp/)-encoded execution block size: 10 MiB total, with a 2 MiB safety margin reserved for beacon-block framing. Practically, clients define`MAX_BLOCK_SIZE = 10,485,760` bytes and `SAFETY_MARGIN = 2,097,152` bytes, and reject any execution block whose RLP payload exceeds `MAX_RLP_BLOCK_SIZE = MAX_BLOCK_SIZE − SAFETY_MARGIN`. The goal is to bound worst-case propagation/validation time and align with CL gossip behavior (blocks over ~10 MiB aren’t propagated), reducing reorg/DoS risk without changing gas accounting.
114
+
The nitty-gritty: this is a cap on the [RLP](/developers/docs/data-structures-and-encoding/rlp/)-encoded execution block size. 10 MiB total, with a 2 MiB safety margin reserved for beacon-block framing. Practically, clients define
103
115
104
-
#### Set default gas limit to XX million {#set-default-gas-limit-to-xx-million}
116
+
`MAX_BLOCK_SIZE = 10,485,760` bytes and
117
+
118
+
`SAFETY_MARGIN = 2,097,152` bytes,
119
+
120
+
and reject any execution block whose RLP payload exceeds
The goal is to bound worst-case propagation/validation time and align with consensus layer gossip behavior, reducing reorg/DoS risk without changing gas accounting.
#### Set default gas limit to XX million {#set-default-gas-limit-to-xx-million}
107
129
108
130
Prior to raising the gas limit from 30M to 36M in February 2025 (and subsequently to 45M), this value hadn’t changed since the Merge (September 2022). This EIP aims to make consistent scaling a priority.
109
131
110
132
EIP-7935 coordinates EL client teams to raise the default gas-limit above today’s 45M for Fusaka. It’s an Informational EIP, but it explicitly asks clients to test higher limits on devnets, converge on a safe value, and ship that number in their Fusaka releases.
111
133
112
134
Devnet planning targets ~60M stress (full blocks with synthetic load) and iterative bumps; research says worst-case block-size pathologies shouldn’t bind below ~150M. Rollout should be paired with the transaction gas-limit cap (EIP-7825) so no single transaction can dominate as limits rise.
113
135
114
-
### Preconfirmation support {#preconfirmation-support}
With EIP-7917, Beacon Chain will become aware of upcoming block proposers for the next epoch. Having a deterministic view on which validators will be proposing future blocks can enable [preconfirmations](https://ethresear.ch/t/based-preconfirmations/17353) - a commitment with the upcoming proposer that guarantees the user transaction will be included in their block without waiting for the actual block.
121
143
122
144
This feature benefits client implementations and security of the network as it prevents edge cases where validators could manipulate the proposer schedule. The lookahead also allows for less complexity of the implementation.
This feature adds a small EVM instruction, **count leading zeroes (CLZ)**. Most everything in the EVM is represented as a 256-bit value—this new opcode returns how many zero bits are at the front. This is a common feature in many instruction set architectures as it enables more efficient arithmetic operations. In practice this collapses today’s hand-rolled bit scans into one step, so finding the first set bit, scanning bytes, or parsing bitfields becomes simpler and cheaper. The opcode is low, fixed-cost and has been benchmarked to be on par with a basic add, which trims bytecode and saves gas for the same work.
129
151
130
-
EIP-7939 adds a small EVM instruction, CLZ (“count leading zeros”). Given a 256-bit value, it returns how many zero bits are at the front — and returns 256 if the value is entirely zero. This is a common feature in many instruction set architectures as it enables more efficient arithmetic operations. In practice this collapses today’s hand-rolled bit scans into one step, so finding the first set bit, scanning bytes, or parsing bitfields becomes simpler and cheaper. The opcode is low, fixed-cost and has been benchmarked to be on par with a basic add, which trims bytecode and saves gas for the same work.
#### Precompile for secp256r1 Curve Support {#secp256r1-precompile}
135
155
136
156
Introduces a built-in, passkey-style secp256r1 (P-256) signature checker at the fixed address `0x100` using the same call format already adopted by many L2s and fixing edge cases, so contracts written for those environments work on L1 without changes.
137
157
138
158
UX upgrade! For users, this unlocks device-native signing and passkeys. Wallets can tap into Apple Secure Enclave, Android Keystore, hardware security modules (HSMs), and FIDO2/WebAuthn directly - no seed phrase, smoother onboarding, and multi-factor flows that feel like modern apps. This results in better UX, easier recovery, and account abstraction patterns that match what billions of devices do already.
139
159
140
160
For developers, it takes a 160-byte input and returns a 32-byte output, making it easy to port existing libraries and L2 contracts. Under the hood, it includes point-at-infinity and modular-comparison checks to eliminate tricky edge cases without breaking valid callers.
141
161
142
-
[More about RIP-7212](https://www.alchemy.com/blog/what-is-rip-7212)_(Note that EIP-7951 superseded RIP-7212)_
-[More about RIP-7212](https://www.alchemy.com/blog/what-is-rip-7212)_(Note that EIP-7951 superseded RIP-7212)_
165
+
166
+
### Meta {#meta}
167
+
168
+
#### `eth_config` JSON-RPC method {#eth-config}
169
+
170
+
This is a JSON-RPC call that allows you to ask your node what fork settings you're running. It returns three snapshots: `current`, `next`, & `last` so that validators and monitoring tools can verify that clients are lined up for an upcoming fork.
171
+
172
+
Practically speaking, this is to address a shortcoming discovered when the Pectra fork went live on the Holesky testnet in early 2025 with minor misconfigurations which resulted in a non-finalizing state. This helps testing teams and developers ensure that major forks will behave as expected when moving from devnets to testnets, and from testnets to Mainnet.
173
+
174
+
Snapshots include: `chainId`, `forkId`, planned fork activation time, which precompiles are active, precompile addresses, system contract dependencies, and the fork's blob schedule.
175
+
176
+
This EIP is in a section apart from the "Core EIPs" because the fork doesn't actually implement any changes - it's a notice that client teams must implement this JSON-RPC method by the Fusaka upgrade.
## Does this upgrade affect all Ethereum nodes and validators? {#does-this-upgrade-affect-all-ethereum-nodes-and-validators}
182
+
###Does this upgrade affect all Ethereum nodes and validators? {#does-this-upgrade-affect-all-ethereum-nodes-and-validators}
145
183
146
184
Yes, the Fusaka upgrade requires updates to both [execution clients and consensus clients](/developers/docs/nodes-and-clients/). All main Ethereum clients will release versions supporting the hard fork marked as high priority. You can keep up with when these releases will be available in client Github repos, their [Discord channels](https://ethstaker.org/support), the [EthStaker Discord](https://dsc.gg/ethstaker), or by subscribing to the Ethereum blog for protocol updates. To maintain synchronization with the Ethereum network post-upgrade, node operators must ensure they are running a supported client version. Note that the information about client releases is time-sensitive, and users should refer to the latest updates for the most current details.
147
185
148
-
## How can ETH be converted after the hard fork? {#how-can-eth-be-converted-after-the-hardfork}
186
+
###How can ETH be converted after the hard fork? {#how-can-eth-be-converted-after-the-hardfork}
149
187
150
188
-**No Action Required for Your ETH**: Following the Ethereum Fusaka upgrade, there is no need to convert or upgrade your ETH. Your account balances will remain the same, and the ETH you currently hold will remain accessible in its existing form after the hard fork.
151
189
-**Beware of Scams!** <Emojitext="⚠️" /> **anyone instructing you to "upgrade" your ETH is trying to scam you.** There is nothing you need to do in relation to this upgrade. Your assets will stay completely unaffected. Remember, staying informed is the best defense against scams.
0 commit comments