Solana Hardware Compatibility List

A comprehensive list of compatible hardware for Solana validators

View on GitHub

Solana Hardware Compatibility List

Last updated: e68f7901f7326b69bee2ac28e445da18df6b2508

This is the Solana Hardware Compatibility List for running a mainnet validator.

Baseline Hardware

Source: https://docs.anza.xyz/operations/requirements

Component Validator Requirements Additional RPC Node Requirements
CPU - 2.8GHz base clock speed, or faster
- SHA extensions instruction support
- AMD Gen 3 or newer
- Intel Ice Lake or newer
- Higher clock speed is preferable over more cores
- AVX2 instruction support (to use official release binaries, self-compile otherwise)
- Support for AVX512f is helpful
 
  12 cores / 24 threads, or more 16 cores / 32 threads, or more
RAM Error Correction Code (ECC) memory is suggested
Motherboard with 512GB capacity suggested
 
  256GB or more 512 GB or more for all account indexes
Disk PCIe Gen3 x4 NVME SSD, or better, on each of:
- Accounts: 500GB, or larger. High TBW (Total Bytes Written)
- Ledger: 1TB or larger. High TBW suggested
- Snapshots: 250GB or larger. High TBW suggested
- OS: (Optional) 500GB, or larger. SATA OK

The OS may be installed on the ledger disk, though testing has shown better performance with the ledger on its own disk

Accounts and ledger can be stored on the same disk, however due to high IOPS, this is not recommended

The Samsung 970 and 980 Pro series SSDs are popular with the validator community
Consider a larger ledger disk if longer transaction history is required

Accounts and ledger should not be stored on the same disk
GPUs Not necessary at this time
Operators in the validator community do not use GPUs currently
 

CPU

Agave CPU

AMD is the CPU maker of choice when it comes to running Solana validators. Running Intel processors for a new validator is strongly discouraged. Intel Xeon Gold/Platinum (6[45]xx) are the only ones known to be able to possibly keep up with the chain.

NOTE: This list is not exhaustive, just a selection of CPUs different folks on the discord have run or are currently running.

The recommendations also apply to RPC nodes, but RPC nodes require more cores and RAM than validators to perform sufficiently under load. RPC node operators should therefore aim for a slightly higher core count than validator operators.

Recommended Manufacturer Model Base Clock Max Boost Clock Cores Threads Default TDP PoH speed *
Yes AMD Ryzen Threadripper PRO 7965WX 4.20 GHz Up to 5.30 GHz 24 48 350 W 22.2M, 20.4M
Yes AMD EPYC 9274F 4.05 GHz Up to 4.30 GHz 24 48 320 W 18.1M
Yes AMD Ryzen Threadripper PRO 7975WX 4.00 GHz Up to 5.30 GHz 32 64 350 W  
Yes AMD Ryzen Threadripper PRO 7985WX 3.20 GHz Up to 5.10 GHz 64 128 350 W  
Yes AMD EPYC 9374F 3.85 GHz Up to 4.30 GHz 32 64 320 W 18.2M
Yes AMD EPYC 9375F 3.80 GHz Up to 4.80 GHz 32 64 320 W  
Yes AMD EPYC 9275F 4.10 GHz Up to 4.80 GHz 24 48 320 W 19.3M
Yes AMD Ryzen Threadripper 7960X 4.20 GHz Up to 5.30 GHz 24 48 350 W 20.6M, 19.9M
Yes AMD Ryzen 9 7950X 4.50 GHz Up to 5.70 GHz 16 32 170 W 22.4M
Yes AMD Ryzen 9 9950X 4.30 GHz Up to 5.70 GHz 16 32 170 W 23M
Yes AMD EPYC 9254 2.90 GHz Up to 4.15 GHz 24 48 200 W 17.5M
Yes AMD EPYC 9354P 3.25 GHz Up to 3.80 GHz 32 64 280 W 16.1M, 14.4M

* The CPU should be capable of at least 10M hashes/sec. Results based on governor set at performance mode.

Source:

Frankendancer CPU

For Frankendancer its beneficial to have Simultaneous Multi-Threading (SMT) enabled. With each CPU it was tested if its possible to catch up with a non-voting node, while SMT was enabled.

Recommended Manufacturer Model Base Clock Max Boost Clock Cores Threads Default TDP Caught up in Testnet Caught up in Mainnet
Yes AMD EPYC 9474F 3.60 GHz Up to 4.10 GHz 48 96 360 W Yes, default layout Yes, default layout
Yes AMD EPYC 9374F 3.85 GHz Up to 4.30 GHz 32 64 320 W Yes, default layout Yes, default layout
Yes AMD EPYC 9354P 3.25 GHz Up to 3.80 GHz 32 64 280 W Yes, default layout Yes, default layout
Yes AMD EPYC 9354 3.25 GHz Up to 3.80 GHz 32 64 280 W Yes, default layout Not tested yet
Yes AMD EPYC 9274F 4.05 GHz Up to 4.30 GHz 24 48 320 W Yes, default layout Yes, default layout
Yes AMD EPYC 9254 2.90 GHz Up to 4.15 GHz 24 48 200 W Yes, custom layout* Yes, default layout
Yes AMD EPYC 7742 2.25 GHz Up to 3.40 GHz 64 128 225 W Yes, default layout Yes, default layout
Yes AMD EPYC 7513 2.60 GHz Up to 3.65 GHz 32 64 200 W Yes, default layout Not tested yet
Yes AMD EPYC 74F3 3.20 GHz Up to 4.00 GHz 24 48 240 W Yes, default layout Yes, default layout
Yes AMD Ryzen Threadripper Pro 7965WX 4.20 GHz Up to 5.30 GHz 24 48 350 W Yes, default layout Yes, default layout
No AMD Ryzen 9 5900X 3.70 GHz Up to 4.80 GHz 12 24 105 W Yes, custom layout* No
No AMD Ryzen 9 5950X 3.40 GHz Up to 4.90 GHz 16 32 105 W Yes, default layout -
No AMD Ryzen 9 7950X 4.50 GHz Up to 5.70 GHz 16 32 170 W Yes, default layout -

* Tested Testnet layout for Ryzen 9 5900X: one verify tile, one bank tile and agave affinity “auto”
* Per Bored King 9254 with custom layout and disable accounts index patch

Storage

Drives are mainly recommended based on performance figures from the manufacturer. All tables are sorted by the claimed random write performance (IOPS), as this is an essential property for Solana validators.

Enterprise gen 5 (PCIe 5.0)

Manufacturer Model Size Sequential Read Sequential Write Random Read Random Write
Phison X200E 6.40 TB 14,800 MB/s 8,700 MB/s 3,200K IOPS 880K IOPS
Solidigm D7-PS1030 6.40 TB 14,500 MB/s 10,000 MB/s 3,000K IOPS 800K IOPS
SanDisk SN861 6.40 TB 13,700 MB/s 7,500 MB/s 3,300K IOPS 800K IOPS
Phison X200E 3.20 TB 14,800 MB/s 8,600 MB/s 3,300K IOPS 790K IOPS
Solidigm D7-PS1030 3.20 TB 14,500 MB/s 8,200 MB/s 3,100K IOPS 716K IOPS
SanDisk SN861 3.20 TB 13,700 MB/s 7,200 MB/s 3,300K IOPS 665K IOPS
Micron 9550 MAX 6.40 TB 14,000 MB/s 10,000 MB/s 3,300K IOPS 640K IOPS
Kioxia CM7-V 3.20 TB 14,000 MB/s 6,750 MB/s 2,700K IOPS 600K IOPS
Kioxia CM7-V 6.40 TB 14,000 MB/s 6,750 MB/s 2,450K IOPS 550K IOPS
Micron 9550 MAX 3.20 TB 14,000 MB/s 6,000 MB/s 3,000K IOPS 540K IOPS
SanDisk SN861 7.68 TB 13,700 MB/s 7,500 MB/s 3,300K IOPS 430K IOPS
Solidigm D7-PS1010 7.68 TB 14,500 MB/s 10,000 MB/s 3,000K IOPS 400K IOPS
Kioxia CD8P-V 3.20 TB 12,000 MB/s 5,500 MB/s 1,900K IOPS 400K IOPS
Kioxia CD8P-V 6.40 TB 12,000 MB/s 5,500 MB/s 2,000K IOPS 400K IOPS
Phison X200P 7.68 TB 14,800 MB/s 8,700 MB/s 3,200K IOPS 390K IOPS
Micron 9550 PRO 7.68 TB 14,000 MB/s 10,000 MB/s 3,300K IOPS 380K IOPS
SanDisk SN861 3.84 TB 13,700 MB/s 7,200 MB/s 3,300K IOPS 330K IOPS
Phison X200P 3.84 TB 14,800 MB/s 8,600 MB/s 3,300K IOPS 320K IOPS
Solidigm D7-PS1010 3.84 TB 14,500 MB/s 8,200 MB/s 3,100K IOPS 315K IOPS
Kioxia CM7-R 3.84 TB 14,000 MB/s 6,750 MB/s 2,700K IOPS 310K IOPS
Micron 9550 PRO 3.84 TB 14,000 MB/s 6,000 MB/s 3,000K IOPS 300K IOPS
Kioxia CM7-R 7.68 TB 14,000 MB/s 6,750 MB/s 2,450K IOPS 300K IOPS
Samsung PM1743 7.68 TB 14,000 MB/s 6,000 MB/s 2,500K IOPS 300K IOPS
Samsung PM1743 3.84 TB 14,000 MB/s 6,000 MB/s 2,500K IOPS 280K IOPS

Enterprise gen 4 (PCIe 4.0)

Manufacturer Model Size Sequential Read Sequential Write Random Read Random Write
Samsung PM9A1 2.00 TB 7,000 MB/s 5,200 MB/s 1,000K IOPS 850K IOPS
Micron 7450 3.84 TB 6,800 MB/s 5,300 MB/s 1,500K IOPS 550K IOPS
Samsung PM9A3 3.84 TB 6,900 MB/s 4,100 MB/s 1,000K IOPS 180K IOPS
Samsung PM9A3 1.92 TB 6,800 MB/s 2,700 MB/s 850K IOPS 130K IOPS

NOTE: There have been a few reports of Kioxia’s being unreliable on the discord, so proceed with caution.

Consumer

Consumer drives are less suitable for Solana validators, because they are designed for lighter workloads and have lower write endurance compared to enterprise drives. Minimum 2 TB per drive is recommended. Larger drives offer increased write endurance, which is crucial for handling the intensive read/write operations of Solana validators.

Manufacturer Model Size Generation Sequential Read Sequential Write Random Read Random Write
Samsung 9100 PRO > 2 TB Gen 5 14,700 MB/s 13,400 MB/s 1,850K IOPS 2,600K IOPS
SanDisk WD_BLACK SN8100 > 2 TB Gen 5 14,900 MB/s 14,000 MB/s 2,300K IOPS 2,400K IOPS
Kingston FURY Renegade G5 > 2 TB Gen 5 14,700 MB/s 14,000 MB/s 2,200K IOPS 2,200K IOPS
Micron 4600 > 2 TB Gen 5 14,500 MB/s 12,000 MB/s 2,100K IOPS 2,100K IOPS
Crucial T705 > 2 TB Gen 5 14,500 MB/s 12,700 MB/s 1,550K IOPS 1,800K IOPS
Seagate FireCuda 540 > 2 TB Gen 5 10,000 MB/s 10,000 MB/s 1,490K IOPS 1,500K IOPS
SanDisk WD_BLACK SN850X > 2 TB Gen 4 7,300 MB/s 6,600 MB/s 1,200K IOPS 1,000K IOPS
Samsung 980 Pro > 2 TB Gen 4 7,000 MB/s 5,100 MB/s 1,000K IOPS 1,000K IOPS

NOTE: The Samsung 990 PRO has been reported to have severe endurance issues, and is therefore not recommended.

Swap

Running swap files is highly discouraged since it masks problems with your setup and will lead to negative consequences as chain activity increases.

Datacenter Providers

Provider Link
Latitude Latitude
Edgevana Edgevana
Teraswitch Teraswitch
Vultr Vultr
Cherry Cherry
DedicatedNodes DedicatedNodes
Allnodes Allnodes

More here: Marinade ISP list

Notes to consider for renting:

Free to test for beginners

Some providers offer a test phase before you commit to renting with them, often as a playground to explore running a Solana node. You can check the list, which we update frequently with new providers.

Notes

Feel free to reach out if you need further assistance or more detailed specifications!

FAQ

Q: My validator doesn’t catch up, catches up slowly or does not keep up with the tip. Why?

A: If you’ve followed the setup instructions correctly, make sure to: (via ax on discord)

i would like to suggest using the acronym “core” when beginners are asking for help, because its always the same question with same answers. it helps to break down the steps and make the guidance easier to follow:

Enable performance mode

sudo apt install cpufrequtils

sudo bash -c 'echo performance > /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor'

Resources

Contributors

Inspired by