Packet Buffer Reference

Packet Buffer Reference

➕ Add a New Switch
This site is an evolution of Jim Warner's extensive work researching and making available data regarding port buffers on network hardware. It is community maintained, and has no guarantee of correctness. If you have information to contribute, please submit a pull request or open an issue. https://github.com/buraglio/port-buffers. This is intended as a community resource, in keeping with the spirit of the original site. There will be inevitable mistakes, and we welcome corrections; please validate any results before relying on them for production use.
Data quality note: Information here is by rumor, innuendo, and extrapolation. Manufacturers rarely publish packet buffer information in their data sheets. However, all data presented here is the result of data mining public resources, including vendor data sheets, white papers, and community discussions. See the summary page for context on why buffers matter, or read about microbursts.

Shared memory means hardware permits buffers to be used by any port that needs them. In a shared memory design, it is not possible to let ALL memory go to queued packets (no room for new arrivals → head-of-line blocking). Some switches have multiple switch ICs, each managing their own memory pool — memory from one IC cannot be loaned to ports on another chip. Max TX buffer is the most a single queue/port can claim from the shared pool.

Trident+ — Broadcom BCM56840, Shared Memory, 9 MB on-chip
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Accton 5652 48 SFP+ and 4 QSFP 8Q — 9 MB — 5? MB
Edge-corE 5600-52X 48 SFP+ and 4 QSFP 8Q — 9 MB — 5? MB
Juniper QFX3500 48 SFP+ and 4 QSFP 12Q — 9 MB — 5? MB
Arista 7050S-64 48 SFP+ and 4 QSFP 8Q — 9 MB/switch — 5 MB
Dell 8132F & 4032F 24 SFP+ and 2x QSFP 8Q — 9 MB — —
Dell 8164F & 4064F 48 SFP+ and 4x QSFP 8Q — 9 MB — —
Pica8 P-3920 48 SFP+ and 4x QSFP — — 9 MB — —
Penguin 4804x 48 SFP+ and 4x QSFP — — 9 MB — —
Cisco Nexus 3064X 48 SFP+ and 4 QSFP 12Q — 9 MB — 5 MB
IBM G8264 48 SFP+ and 4 QSFP 8Q — 9 MB/switch — not on data sheet See Tolly report for microburst testing
Force10 S4810 48 SFP+ and 4 QSFP 4Q — 9 MB — —
Allied Telesis DC2552 48 SFP+ and 4 QSFP 8Q — 9 MB — 5 MB
NEC PF5820 48 SFP+ and 4 QSFP — — — — — No buffer data available
Accton 5652 48 SFP+ and 4 QSFP 8Q — 9 MB — 5? MB
Edge-corE 5600-52X 48 SFP+ and 4 QSFP 8Q — 9 MB — 5? MB
Juniper QFX3500 48 SFP+ and 4 QSFP 12Q — 9 MB — 5? MB
Arista 7050S-64 48 SFP+ and 4 QSFP 8Q — 9 MB/switch — 5 MB
Dell 8132F & 4032F 24 SFP+ and 2x QSFP 8Q — 9 MB — —
Dell 8164F & 4064F 48 SFP+ and 4x QSFP 8Q — 9 MB — —
Pica8 P-3920 48 SFP+ and 4x QSFP — — 9 MB — —
Penguin 4804x 48 SFP+ and 4x QSFP — — 9 MB — —
Cisco Nexus 3064X 48 SFP+ and 4 QSFP 12Q — 9 MB — 5 MB
IBM G8264 48 SFP+ and 4 QSFP 8Q — 9 MB/switch — not on data sheet See Tolly report for microburst testing
Force10 S4810 48 SFP+ and 4 QSFP 4Q — 9 MB — —
Allied Telesis DC2552 48 SFP+ and 4 QSFP 8Q — 9 MB — 5 MB
NEC PF5820 48 SFP+ and 4 QSFP — — — — — No buffer data available
Trident II — Broadcom BCM56850/56854, Shared Memory, 12 MB on-chip
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Arista 7050X 32 QSFP 8Q — 12 MB — Dynamic up to 8 MB
Arista 7250X 64 QSFP 8Q — 12 MB/ASIC, 48 MB total — Dynamic up to 8 MB 4-chip design
Brocade 7750-26Q 26 QSFP + expansion slot 8Q — 12.2 MB — Dynamic up to 8 MB
Brocade 7750-48F 48 SFP+ and 6 QSFP and expansion 8Q — 12.2 MB — Dynamic up to 8 MB
Extreme Summit 770 32 QSFP 8Q — No info — No info
Cisco Nexus 3100 32 QSFP or 31 QSFP and 4 SFP+ 8Q — 12.2 MB — Dynamic up to 8 MB
Dell S6000 32 QSFP 8Q — 12 MB — Dynamic up to 8 MB
Juniper QFX5100-24Q 24 QSFP + expansion slots 8Q — 12 MB — Dynamic up to 8 MB
HP 5930 32 QSFP 8Q — 12.2 MB — Dynamic up to 8 MB
Aruba CX 8320 48x 10G and 6x 40G — — 16 MB shared shared Broadcom Trident II/Maverick; 16MB packet buffer
FS N8000-32Q 32x 40G QSFP+ 8Q — 12 MB shared Dynamic up to 8 MB BCM56850 (Trident 2); 2.56 Tbps
Arista 7050X 32 QSFP 8Q — 12 MB — Dynamic up to 8 MB
Arista 7250X 64 QSFP 8Q — 12 MB/ASIC, 48 MB total — Dynamic up to 8 MB 4-chip design
Brocade 7750-26Q 26 QSFP + expansion slot 8Q — 12.2 MB — Dynamic up to 8 MB
Brocade 7750-48F 48 SFP+ and 6 QSFP and expansion 8Q — 12.2 MB — Dynamic up to 8 MB
Extreme Summit 770 32 QSFP 8Q — No info — No info
Cisco Nexus 3100 32 QSFP or 31 QSFP and 4 SFP+ 8Q — 12.2 MB — Dynamic up to 8 MB
Dell S6000 32 QSFP 8Q — 12 MB — Dynamic up to 8 MB
Juniper QFX5100-24Q 24 QSFP + expansion slots 8Q — 12 MB — Dynamic up to 8 MB
HP 5930 32 QSFP 8Q — 12.2 MB — Dynamic up to 8 MB
Aruba CX 8320 48x 10G and 6x 40G — — 16 MB shared shared Broadcom Trident II/Maverick; 16MB packet buffer
FS N8000-32Q 32x 40G QSFP+ 8Q — 12 MB shared Dynamic up to 8 MB BCM56850 (Trident 2); 2.56 Tbps
Mellanox — Mellanox Spectrum ASIC, Shared Memory
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Mellanox SX1024 48 SFP+ and 12 QSFP ?? — 4.6 MB — 64 KB to 1 port Spectrum ASIC
Mellanox SX1036 36 QSFP ?? — 4.6 MB — 128 KB to 1 port Spectrum ASIC
Mellanox SX1024 48 SFP+ and 12 QSFP ?? — 4.6 MB — 64 KB to 1 port Spectrum ASIC
Mellanox SX1036 36 QSFP ?? — 4.6 MB — 128 KB to 1 port Spectrum ASIC
Other Shared Memory — Various ASICs
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Brocade FCX624S 24 GigE with 4 SFP+ 8Q — 2 MB — 1.04 MB to 1 port
Juniper EX4500 40 SFP+ plus 8 SFP+ 8Q — — — 230 KB
Brocade ICX6610-24 24 GigE with 8 SFP+ 8Q — 4 MB — 1 MB to 1 port
HP 3800 24 GigE with 4 10Gb/s 8Q or 4Q or 2Q — — — 0.23 MB
HP 2920-24 24 twisted-pair GigE with 4 optional 10Gb/s 8Q — 11.25 MB 4.5 MB shared 6.75 MB shared
HP A5800-24G 24 GigE with 4 10Gb/s 8Q — 4 MB — 4 MB
HP E6600-24G-4XG 24 GigE with 2 10Gb/s 8Q — 18 MB for GE — 18 MB
Arista 7124SX 24x SFP+ 8Q — 2 MB/switch-chip shared 1.238 + 0.02 MB
Arista 7148SX 48 SFP+ 8Q — 2 MB/switch-chip, 8 MB total — 1.238 + 0.02 MB
Arista DCS-7150S-24 24 SFP+ 8Q — 9.5 MB — —
Arista DCS-7150S-52 52 SFP+ 8Q — 9.5 MB — —
Arista DCS-7150S-64 48 SFP+ and 4 QSFP 8Q — 9.5 MB — —
Dell 8024 24 SFP+ 2Q — 2 MB — —
Dell 6248 48 GigE, 4x SFP+ 8Q — 6.1 MB — 98 KB per port
Dell 7024 24 GigE, 4x SFP+ 8Q — 32 MB — 4 MB
IBM BNT G8052 48 GE + 4 SFP+ 8Q — 4 MB/switch — 4 MB
Cisco Catalyst 3750G-48TS 6 ASICs, 8 GE ports each 8Q 1P3Q3T 576 KB per ASIC 192 KB per ASIC 384 KB per ASIC Multiple switch ICs – memory not shared across chips
Cisco Catalyst 3750E / 3560E / 3750X / 3560X ASIC supports 24 GE ports or 2x 10G 2Q 1P3Q3T 2750 KB per ASIC 750 KB per ASIC 2 MB per ASIC
Cisco Catalyst 4948E 48 GigE + 4 SFP+ 4Q — 17.5 MB — Max 16 MB to 1 port
Cisco Nexus 3548 48 SFP+ 4Q — 18 MB — 5.8 MB Algorithmic buffer; see detail page
Brocade FCX624S 24 GigE with 4 SFP+ 8Q — 2 MB — 1.04 MB to 1 port
Juniper EX4500 40 SFP+ plus 8 SFP+ 8Q — — — 230 KB
Brocade ICX6610-24 24 GigE with 8 SFP+ 8Q — 4 MB — 1 MB to 1 port
HP 3800 24 GigE with 4 10Gb/s 8Q or 4Q or 2Q — — — 0.23 MB
HP 2920-24 24 twisted-pair GigE with 4 optional 10Gb/s 8Q — 11.25 MB 4.5 MB shared 6.75 MB shared
HP A5800-24G 24 GigE with 4 10Gb/s 8Q — 4 MB — 4 MB
HP E6600-24G-4XG 24 GigE with 2 10Gb/s 8Q — 18 MB for GE — 18 MB
Arista 7124SX 24x SFP+ 8Q — 2 MB/switch-chip shared 1.238 + 0.02 MB
Arista 7148SX 48 SFP+ 8Q — 2 MB/switch-chip, 8 MB total — 1.238 + 0.02 MB
Arista DCS-7150S-24 24 SFP+ 8Q — 9.5 MB — —
Arista DCS-7150S-52 52 SFP+ 8Q — 9.5 MB — —
Arista DCS-7150S-64 48 SFP+ and 4 QSFP 8Q — 9.5 MB — —
Dell 8024 24 SFP+ 2Q — 2 MB — —
Dell 6248 48 GigE, 4x SFP+ 8Q — 6.1 MB — 98 KB per port
Dell 7024 24 GigE, 4x SFP+ 8Q — 32 MB — 4 MB
IBM BNT G8052 48 GE + 4 SFP+ 8Q — 4 MB/switch — 4 MB
Cisco Catalyst 3750G-48TS 6 ASICs, 8 GE ports each 8Q 1P3Q3T 576 KB per ASIC 192 KB per ASIC 384 KB per ASIC Multiple switch ICs – memory not shared across chips
Cisco Catalyst 3750E / 3560E / 3750X / 3560X ASIC supports 24 GE ports or 2x 10G 2Q 1P3Q3T 2750 KB per ASIC 750 KB per ASIC 2 MB per ASIC
Cisco Catalyst 4948E 48 GigE + 4 SFP+ 4Q — 17.5 MB — Max 16 MB to 1 port
Cisco Nexus 3548 48 SFP+ 4Q — 18 MB — 5.8 MB Algorithmic buffer; see detail page
Large Queue 1RU Switches — External DRAM, Very Large Buffers
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Force10 S60 48 GigE + 4 SFP+ 4Q — 1250 MB — enough External DRAM; designed for speed-step scenarios
Arista 7048 48 GigE + 4 SFP+ 8Q — 768 MB 16 MB/port? GOBBS External DRAM; claims 50ms buffers per port simultaneously
Force10 S60 48 GigE + 4 SFP+ 4Q — 1250 MB — enough External DRAM; designed for speed-step scenarios
Arista 7048 48 GigE + 4 SFP+ 8Q — 768 MB 16 MB/port? GOBBS External DRAM; claims 50ms buffers per port simultaneously
Virtual Output Port (VoQ) — Per-egress queues at ingress, eliminates head-of-line blocking
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Arista 7504E / 7508E 48-port SFP+ line card 8Q — 3 GB per packet processor 50 MB/port-queue VoQ
Brocade MLX 2-port 100 Gb/s line card 8Q — 3 GB 256 MB/port-queue VoQ
Brocade MLX 8-port 10 Gb/s line card 8Q — 3 GB 256 MB/port-queue VoQ
Brocade MLX 24-port 1 Gb/s line card 8Q — 1 GB 64 MB/port-queue VoQ
Cisco Nexus 5010 26 SFP+ 8Q 1 (uncertain) 480 KB per port 480 KB VoQ
Cisco Nexus 5548P 32 fixed SFP+ and 16-port expansion module 8 Unicast + 8 Multicast 1 (uncertain) 640 KB per port 640 KB VoQ
Cisco Nexus 6001 48 fixed SFP+ and 4 QSFP 8 Unicast + 8 Multicast 1 (uncertain) 25 MB per 12-port 10G ASIC 14 MB shared per ASIC VoQ
Cisco Nexus 7000 M1 48-port twisted-pair GE 2q4t 1p3q4t — 7.56 MB per port 6.15 MB per port
Cisco Nexus 7000 F3 6-port 100GE 4q 4q 144 MB shared — VoQ
Cisco Nexus 7000 F2 48 SFP+ 4q1t 1p3q1t 72 MB shared — VoQ
Cisco Nexus 7000 M1 32-port 10GE shared 4-port groups 8q4t 1p7q4t — 65 MB per group 80 MB per group
Cisco Catalyst 6500 Various line cards — — — — — Complex architecture — see dedicated detail page
Arista 7504E / 7508E 48-port SFP+ line card 8Q — 3 GB per packet processor 50 MB/port-queue VoQ
Brocade MLX 2-port 100 Gb/s line card 8Q — 3 GB 256 MB/port-queue VoQ
Brocade MLX 8-port 10 Gb/s line card 8Q — 3 GB 256 MB/port-queue VoQ
Brocade MLX 24-port 1 Gb/s line card 8Q — 1 GB 64 MB/port-queue VoQ
Cisco Nexus 5010 26 SFP+ 8Q 1 (uncertain) 480 KB per port 480 KB VoQ
Cisco Nexus 5548P 32 fixed SFP+ and 16-port expansion module 8 Unicast + 8 Multicast 1 (uncertain) 640 KB per port 640 KB VoQ
Cisco Nexus 6001 48 fixed SFP+ and 4 QSFP 8 Unicast + 8 Multicast 1 (uncertain) 25 MB per 12-port 10G ASIC 14 MB shared per ASIC VoQ
Cisco Nexus 7000 M1 48-port twisted-pair GE 2q4t 1p3q4t — 7.56 MB per port 6.15 MB per port
Cisco Nexus 7000 F3 6-port 100GE 4q 4q 144 MB shared — VoQ
Cisco Nexus 7000 F2 48 SFP+ 4q1t 1p3q1t 72 MB shared — VoQ
Cisco Nexus 7000 M1 32-port 10GE shared 4-port groups 8q4t 1p7q4t — 65 MB per group 80 MB per group
Cisco Catalyst 6500 Various line cards — — — — — Complex architecture — see dedicated detail page
Nokia SR (FP-based)
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Nokia 7750 SR (FP4) Various line cards (up to 400G) 128k ingress queues 128k egress queues ~64 GB per line card shared shared Fully shared memory pool; pre-buffer multi-million packets; ~3 Tbps class NPU; 8k queue realloc increments
Nokia 7750 SR (FP5) Various line cards (400G/800G) 128k ingress queues 128k egress queues 32–64 GB per line card shared shared Fully shared memory pool; pre-buffer 10.8M–21.6M packets; ~6 Tbps class NPU; 16k queue realloc increments
Nokia 7750 SR (FP4) Various line cards (up to 400G) 128k ingress queues 128k egress queues ~64 GB per line card shared shared Fully shared memory pool; pre-buffer multi-million packets; ~3 Tbps class NPU; 8k queue realloc increments
Nokia 7750 SR (FP5) Various line cards (400G/800G) 128k ingress queues 128k egress queues 32–64 GB per line card shared shared Fully shared memory pool; pre-buffer 10.8M–21.6M packets; ~6 Tbps class NPU; 16k queue realloc increments
Nokia Access / Aggregation
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Nokia 7705 SAR Various adapter cards — — ~200–300 MB per adapter card — — SAR ASIC; fixed 512B (chained) or 2304B buffer segments; MBS per queue; mobile backhaul / TDM transport
Nokia 7730 SXR Various (core/DCI) 256k queues — multi-GB (FP5-class) shared shared FP5 silicon; same family as 7750 SR FP5; fully shared buffer; deterministic QoS; core/DCI role
Nokia 7705 SAR Various adapter cards — — ~200–300 MB per adapter card — — SAR ASIC; fixed 512B (chained) or 2304B buffer segments; MBS per queue; mobile backhaul / TDM transport
Nokia 7730 SXR Various (core/DCI) 256k queues — multi-GB (FP5-class) shared shared FP5 silicon; same family as 7750 SR FP5; fully shared buffer; deterministic QoS; core/DCI role
Merchant Access Silicon
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Nokia 7210 SAS Various (access/demarcation/cell-site) — — 64–512 MB shared ingress/egress shared ingress/egress Merchant Ethernet silicon; ingress-dominant; hierarchical QoS; access edge role
Nokia 7210 SAS Various (access/demarcation/cell-site) — — 64–512 MB shared ingress/egress shared ingress/egress Merchant Ethernet silicon; ingress-dominant; hierarchical QoS; access edge role
Broadcom Qumran — Broadcom Access & Aggregation VOQ Deep Buffer (Qumran-MX, Qumran-AX)
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Nokia 7250 IXR 100G/400G (aggregation/DC edge) — — moderate (shared) shared + VOQ VoQ Broadcom Jericho/Qumran family; VOQ reduces head-of-line blocking; low-latency DC/aggregation
Cisco NCS 5000 Series Various (10G/100G aggregation) — — 16 MB shared shared Qumran-MX; On-chip buffer only; metro aggregation
Cisco NCS 540 Series Various (1G/10G/100G) — — Up to 3 GB VOQ VOQ Qumran family; Deep packet buffers for microbursts
Cisco NCS 560 Series Modular (up to 100G line cards) — — Line card dependent (Deep Buffer) VOQ VOQ Modular platform using Qumran/Jericho subsets in line cards
Edgecore AS5916-54XKS (AGR130) 48x 10G SFP+, 6x 100G QSFP28 VOQ VOQ 8 GB (Deep Buffer) VOQ VOQ Broadcom Qumran-MX (BCM88375); 16 MB on-chip + 8 GB external DRAM
Edgecore AS7316-26XB (CSR320) 16x 10G SFP+, 8x 25G SFP28, 2x 100G QSFP28 VOQ VOQ ~2 GB (Deep Buffer) VOQ VOQ Broadcom Qumran-AX (BCM88470); Deep packet buffer typically ~2GB via DRAM
UfiSpace S9510-28DC 24x 10G/25G SFP28, 2x 100G QSFP28, 2x 40/100G VOQ VOQ 2 GB (Deep Buffer) VOQ VOQ Broadcom Qumran-AX; 2GB external deep buffer for 5G backhaul
Nokia 7250 IXR 100G/400G (aggregation/DC edge) — — moderate (shared) shared + VOQ VoQ Broadcom Jericho/Qumran family; VOQ reduces head-of-line blocking; low-latency DC/aggregation
Cisco NCS 5000 Series Various (10G/100G aggregation) — — 16 MB shared shared Qumran-MX; On-chip buffer only; metro aggregation
Cisco NCS 540 Series Various (1G/10G/100G) — — Up to 3 GB VOQ VOQ Qumran family; Deep packet buffers for microbursts
Cisco NCS 560 Series Modular (up to 100G line cards) — — Line card dependent (Deep Buffer) VOQ VOQ Modular platform using Qumran/Jericho subsets in line cards
Edgecore AS5916-54XKS (AGR130) 48x 10G SFP+, 6x 100G QSFP28 VOQ VOQ 8 GB (Deep Buffer) VOQ VOQ Broadcom Qumran-MX (BCM88375); 16 MB on-chip + 8 GB external DRAM
Edgecore AS7316-26XB (CSR320) 16x 10G SFP+, 8x 25G SFP28, 2x 100G QSFP28 VOQ VOQ ~2 GB (Deep Buffer) VOQ VOQ Broadcom Qumran-AX (BCM88470); Deep packet buffer typically ~2GB via DRAM
UfiSpace S9510-28DC 24x 10G/25G SFP28, 2x 100G QSFP28, 2x 40/100G VOQ VOQ 2 GB (Deep Buffer) VOQ VOQ Broadcom Qumran-AX; 2GB external deep buffer for 5G backhaul
Broadcom Jericho — Broadcom BCM88670/BCM88675, VOQ + GDDR5 deep buffer, ~16 MB OCB
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Cisco NCS 5500 Series Various (up to 400G line cards) VOQ VOQ 16 MB on-chip + 4-8 GB off-chip per ASIC VOQ VOQ Jericho/Jericho+ ASIC; 16MB fast buffer, large GDDR5 deep buffer
Cisco NCS 5500 Series Various (up to 400G line cards) VOQ VOQ 16 MB on-chip + 4-8 GB off-chip per ASIC VOQ VOQ Jericho/Jericho+ ASIC; 16MB fast buffer, large GDDR5 deep buffer
Broadcom Jericho2 — Broadcom BCM88690/BCM88800, VOQ + 8 GB HBM, ~32 MB OCB
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Cisco NCS 5700 Series Various (up to 400G line cards) VOQ VOQ 32 MB on-chip + 8 GB off-chip HBM per ASIC VOQ VOQ Jericho2 / J2c ASIC; HBM (High Bandwidth Memory) integration
Arista 7280R3 / 7800R3 Series Various (10G/25G to 400G line cards/edge routers) VOQ VOQ Ultra-deep buffer (up to 24 GB / 32 GB via HBM) VOQ VOQ Broadcom Jericho2; HBM integration for deep queueing
Cisco NCS 5700 Series Various (up to 400G line cards) VOQ VOQ 32 MB on-chip + 8 GB off-chip HBM per ASIC VOQ VOQ Jericho2 / J2c ASIC; HBM (High Bandwidth Memory) integration
Arista 7280R3 / 7800R3 Series Various (10G/25G to 400G line cards/edge routers) VOQ VOQ Ultra-deep buffer (up to 24 GB / 32 GB via HBM) VOQ VOQ Broadcom Jericho2; HBM integration for deep queueing
Broadcom Qumran2c — Broadcom BCM88820, VOQ + 4 GB HBM integrated deep buffer
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
UfiSpace S9600-72XC 64x 25G SFP28, 8x 100G QSFP28 VOQ VOQ 4 GB (Deep Buffer via HBM) VOQ VOQ Broadcom Qumran2c (BCM88820); 4GB deep packet buffer via integrated High Bandwidth Memory
UfiSpace S9600-72XC 64x 25G SFP28, 8x 100G QSFP28 VOQ VOQ 4 GB (Deep Buffer via HBM) VOQ VOQ Broadcom Qumran2c (BCM88820); 4GB deep packet buffer via integrated High Bandwidth Memory
Trident 3 — Broadcom BCM56870/BCM56771, Unified Shared Memory, 32 MB on-chip
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Arista 7050X3 Various (10G/25G/100G edge) — — 32 MB shared shared Broadcom Trident 3; flexible packet processor
Aruba CX 8325 32x 100G or 48x 25G — — 32 MB shared shared Broadcom Trident 3; 3 configurable lossless QoS pools
Aruba CX 8360 Various (32x 100G, 48x 25G aggregation/edge) — — 32 MB shared shared Broadcom Trident 3; Single fixed QoS pool
Juniper QFX5120 Various (32x 100G, 48x 25G) — — 32 MB shared shared Broadcom Trident 3; 25% dedicated buffer, remaining shared
FS N5570-48S6C 48x 10G SFP+, 6x 100G QSFP28 8Q — 32 MB shared Dynamic (shared) BCM56771 (Trident 3 X5); 1.08 Tbps
FS N5850-48X6C 48x 10G RJ45, 6x 100G QSFP28 8Q — 32 MB shared Dynamic (shared) BCM56771 (Trident 3 X5); 1.08 Tbps; copper ToR
FS N8550-32C 32x 100G QSFP28, 2x 10G SFP+ 8Q — 32 MB shared Dynamic (shared) BCM56870 (Trident 3 X7); 3.2 Tbps
FS N8550-48B8C 48x 25G SFP28, 8x 100G QSFP28, 2x 10G SFP+ 8Q — 32 MB shared Dynamic (shared) BCM56873 (Trident 3); 2.0 Tbps
FS N8560-32C 32x 100G QSFP28, 2x 10G SFP+ 8Q — 32 MB shared Dynamic (shared) BCM56870 (Trident 3 X7); 3.2 Tbps; 2nd-gen platform
FS N8560-48BC 48x 25G SFP28, 8x 100G QSFP28, 2x 10G SFP+ 8Q — 32 MB shared Dynamic (shared) BCM56873 (Trident 3); 2.0 Tbps; 2nd-gen platform
Edgecore AS7726-32X (DCS204) 32x 100G QSFP28 — — 32 MB shared shared Broadcom Trident 3 (BCM56870); Fully shared 32MB on-chip buffer
Arista 7050X3 Various (10G/25G/100G edge) — — 32 MB shared shared Broadcom Trident 3; flexible packet processor
Aruba CX 8325 32x 100G or 48x 25G — — 32 MB shared shared Broadcom Trident 3; 3 configurable lossless QoS pools
Aruba CX 8360 Various (32x 100G, 48x 25G aggregation/edge) — — 32 MB shared shared Broadcom Trident 3; Single fixed QoS pool
Juniper QFX5120 Various (32x 100G, 48x 25G) — — 32 MB shared shared Broadcom Trident 3; 25% dedicated buffer, remaining shared
FS N5570-48S6C 48x 10G SFP+, 6x 100G QSFP28 8Q — 32 MB shared Dynamic (shared) BCM56771 (Trident 3 X5); 1.08 Tbps
FS N5850-48X6C 48x 10G RJ45, 6x 100G QSFP28 8Q — 32 MB shared Dynamic (shared) BCM56771 (Trident 3 X5); 1.08 Tbps; copper ToR
FS N8550-32C 32x 100G QSFP28, 2x 10G SFP+ 8Q — 32 MB shared Dynamic (shared) BCM56870 (Trident 3 X7); 3.2 Tbps
FS N8550-48B8C 48x 25G SFP28, 8x 100G QSFP28, 2x 10G SFP+ 8Q — 32 MB shared Dynamic (shared) BCM56873 (Trident 3); 2.0 Tbps
FS N8560-32C 32x 100G QSFP28, 2x 10G SFP+ 8Q — 32 MB shared Dynamic (shared) BCM56870 (Trident 3 X7); 3.2 Tbps; 2nd-gen platform
FS N8560-48BC 48x 25G SFP28, 8x 100G QSFP28, 2x 10G SFP+ 8Q — 32 MB shared Dynamic (shared) BCM56873 (Trident 3); 2.0 Tbps; 2nd-gen platform
Edgecore AS7726-32X (DCS204) 32x 100G QSFP28 — — 32 MB shared shared Broadcom Trident 3 (BCM56870); Fully shared 32MB on-chip buffer
Tomahawk — Broadcom BCM56960/BCM56967, Shared Memory, 16 MB on-chip (4 pipes × 4 MB each; not unified)
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Juniper QFX5200 32x 100G — — 16 MB shared per pipe shared per pipe Broadcom Tomahawk; 4 pipes, 4 MB per pipe serving 8 ports
FS N5850-48S6Q 48x 10G SFP+, 6x 40G QSFP+ 8Q — 16 MB shared Dynamic (shared) BCM56864 (Trident 2+); 720 Gbps; unified buffer
FS N8500-32C 32x 100G QSFP28 8Q — 16 MB shared per pipe shared per pipe BCM56960 (Tomahawk); 6.4 Tbps; 4 pipes × 4 MB (not unified)
FS N8500-48B6C 48x 25G SFP28, 6x 100G QSFP28 8Q — 16 MB shared Dynamic (shared) BCM56967 (Tomahawk+); 3.6 Tbps
Juniper QFX5200 32x 100G — — 16 MB shared per pipe shared per pipe Broadcom Tomahawk; 4 pipes, 4 MB per pipe serving 8 ports
FS N5850-48S6Q 48x 10G SFP+, 6x 40G QSFP+ 8Q — 16 MB shared Dynamic (shared) BCM56864 (Trident 2+); 720 Gbps; unified buffer
FS N8500-32C 32x 100G QSFP28 8Q — 16 MB shared per pipe shared per pipe BCM56960 (Tomahawk); 6.4 Tbps; 4 pipes × 4 MB (not unified)
FS N8500-48B6C 48x 25G SFP28, 6x 100G QSFP28 8Q — 16 MB shared Dynamic (shared) BCM56967 (Tomahawk+); 3.6 Tbps
Juniper Q5
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Juniper QFX10002 Fixed 36x 100G or 72x 40G VOQ VOQ Up to 100 ms per port VOQ VOQ Juniper Q5 custom ASIC; high-efficiency HMC memory for deep VOQ buffering
Juniper QFX10002 Fixed 36x 100G or 72x 40G VOQ VOQ Up to 100 ms per port VOQ VOQ Juniper Q5 custom ASIC; high-efficiency HMC memory for deep VOQ buffering
Trident 4 — Broadcom BCM56880/BCM56780, Unified Shared Memory, up to 132 MB on-chip
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Arista 7050X4 Series 10G/25G/100G (Enterprise/DC edge) — — Up to 132 MB shared shared Broadcom Trident 4; Up to 132MB shared packet buffer
FS N8520-32D 32x 400G QSFP-DD 8Q — 132 MB shared Dynamic (shared) BCM56880 (Trident 4 X11); 12.8 Tbps
FS N8550-24CD8D 24x 200G QSFP56, 8x 400G QSFP-DD 8Q — 82 MB shared Dynamic (shared) BCM56780 (Trident 4 X9); ~8 Tbps; 200G-native design
Arista 7050X4 Series 10G/25G/100G (Enterprise/DC edge) — — Up to 132 MB shared shared Broadcom Trident 4; Up to 132MB shared packet buffer
FS N8520-32D 32x 400G QSFP-DD 8Q — 132 MB shared Dynamic (shared) BCM56880 (Trident 4 X11); 12.8 Tbps
FS N8550-24CD8D 24x 200G QSFP56, 8x 400G QSFP-DD 8Q — 82 MB shared Dynamic (shared) BCM56780 (Trident 4 X9); ~8 Tbps; 200G-native design
Tomahawk 3 — Broadcom BCM56980, Unified Shared Memory, 64 MB on-chip
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Arista 7060X4 Series 32x 400G or dense 100G — — 64 MB shared shared Broadcom Tomahawk 3; 64MB shared buffer (some models 32MB)
FS N9500-32D 32x 400G QSFP-DD 8Q — 64 MB shared Dynamic (shared) BCM56980 (Tomahawk 3); 12.8 Tbps
FS N9550-32D 32x 400G QSFP-DD, 2x 10G SFP+ 8Q — 64 MB shared Dynamic (shared) BCM56980 (Tomahawk 3); 12.8 Tbps
Arista 7060X4 Series 32x 400G or dense 100G — — 64 MB shared shared Broadcom Tomahawk 3; 64MB shared buffer (some models 32MB)
FS N9500-32D 32x 400G QSFP-DD 8Q — 64 MB shared Dynamic (shared) BCM56980 (Tomahawk 3); 12.8 Tbps
FS N9550-32D 32x 400G QSFP-DD, 2x 10G SFP+ 8Q — 64 MB shared Dynamic (shared) BCM56980 (Tomahawk 3); 12.8 Tbps
Tomahawk 4 — Broadcom BCM56990/BCM56993, Unified Shared Memory, 56.83–113.66 MB on-chip
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Arista 7060X5 Series 32x 400G / 800G — — Up to 114 MB shared shared Broadcom Tomahawk 4; Up to 114MB shared buffer
FS N9510-64D 64x 400G QSFP-DD 8Q — 113.66 MB shared Dynamic (shared) BCM56990 (Tomahawk 4); 25.6 Tbps; confirmed from official spec page
FS N9550-64D 64x 400G QSFP-DD, 2x 10G SFP+ 8Q — 113.66 MB shared Dynamic (shared) BCM56990 (Tomahawk 4); 25.6 Tbps
FS N8610-64D 64x 400G QSFP-DD 8Q — 113.66 MB shared Dynamic (shared) BCM56990 (Tomahawk 4); 25.6 Tbps; 2U platform
FS N8610-32D 32x 400G QSFP-DD, 2x 10G SFP+ 8Q — 56.83 MB shared Dynamic (shared) BCM56993 (Tomahawk 4 half-capacity); 12.8 Tbps
Arista 7060X5 Series 32x 400G / 800G — — Up to 114 MB shared shared Broadcom Tomahawk 4; Up to 114MB shared buffer
FS N9510-64D 64x 400G QSFP-DD 8Q — 113.66 MB shared Dynamic (shared) BCM56990 (Tomahawk 4); 25.6 Tbps; confirmed from official spec page
FS N9550-64D 64x 400G QSFP-DD, 2x 10G SFP+ 8Q — 113.66 MB shared Dynamic (shared) BCM56990 (Tomahawk 4); 25.6 Tbps
FS N8610-64D 64x 400G QSFP-DD 8Q — 113.66 MB shared Dynamic (shared) BCM56990 (Tomahawk 4); 25.6 Tbps; 2U platform
FS N8610-32D 32x 400G QSFP-DD, 2x 10G SFP+ 8Q — 56.83 MB shared Dynamic (shared) BCM56993 (Tomahawk 4 half-capacity); 12.8 Tbps
Tomahawk 2 — Broadcom BCM56970, Unified Shared Memory, 42 MB on-chip
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
FS N8550-64C 64x 100G QSFP28 8Q — 42 MB shared Dynamic (shared) BCM56970 (Tomahawk 2); 6.4 Tbps
FS N8560-64C 64x 100G QSFP28 8Q — 42 MB shared Dynamic (shared) BCM56970 (Tomahawk 2); 6.4 Tbps; 2nd-gen platform
FS N8550-64C 64x 100G QSFP28 8Q — 42 MB shared Dynamic (shared) BCM56970 (Tomahawk 2); 6.4 Tbps
FS N8560-64C 64x 100G QSFP28 8Q — 42 MB shared Dynamic (shared) BCM56970 (Tomahawk 2); 6.4 Tbps; 2nd-gen platform
Tomahawk 5 — Broadcom BCM78900/BCM78902, Unified Shared Memory, up to 165.2 MB on-chip
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
FS N9600-64OD 64x 800G OSFP, 2x 25G SFP28 8Q — 165.2 MB shared Dynamic (shared) BCM78900 (Tomahawk 5); 51.2 Tbps; OSFP form factor
FS N8650-32OD 32x 800G OSFP, 2x 25G SFP28 8Q — ~83 MB shared Dynamic (shared) BCM78902 (Tomahawk 5, half-capacity); 25.6 Tbps; buffer inferred — not confirmed from official source
FS N9600-64OD 64x 800G OSFP, 2x 25G SFP28 8Q — 165.2 MB shared Dynamic (shared) BCM78900 (Tomahawk 5); 51.2 Tbps; OSFP form factor
FS N8650-32OD 32x 800G OSFP, 2x 25G SFP28 8Q — ~83 MB shared Dynamic (shared) BCM78902 (Tomahawk 5, half-capacity); 25.6 Tbps; buffer inferred — not confirmed from official source
Marvell Prestera (MikroTik) — Marvell 98DX3236/98DX8216/98DX8525 (Aldrin2), Shared SRAM, ~1.5–8 MB estimated
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
MikroTik CRS317-1G-16S+RM 16x 10G SFP+ — — ~4 MB (Approximation) shared shared Marvell 98DX8216; highly approximated via RouterOS CLI total-byte-cap
MikroTik CRS326-24G-2S+RM 24x 1G, 2x 10G SFP+ — — ~1.5 MB (Approximation) shared shared Marvell 98DX3236; highly approximated via RouterOS CLI total-byte-cap
MikroTik CCR2216-1G-12XS-2XQ 12x 25G SFP28, 2x 100G QSFP28 — — Small/Undisclosed (Approximation) dynamic dynamic Marvell Aldrin2 (98DX8525) HW offload; CPU RAM handles main routing queues
MikroTik CRS317-1G-16S+RM 16x 10G SFP+ — — ~4 MB (Approximation) shared shared Marvell 98DX8216; highly approximated via RouterOS CLI total-byte-cap
MikroTik CRS326-24G-2S+RM 24x 1G, 2x 10G SFP+ — — ~1.5 MB (Approximation) shared shared Marvell 98DX3236; highly approximated via RouterOS CLI total-byte-cap
MikroTik CCR2216-1G-12XS-2XQ 12x 25G SFP28, 2x 100G QSFP28 — — Small/Undisclosed (Approximation) dynamic dynamic Marvell Aldrin2 (98DX8525) HW offload; CPU RAM handles main routing queues
Marvell OCTEON 10 (CN10K) — Programmable NP/DPU; DDR5-backed packet buffer (16–48 GB per NP), no fixed on-chip SRAM pool
Model Port Type RX Queue TX Queue Total Buffer RX Buffer TX Buffer Notes
Asterfusion ET3608 2x 100G QSFP28, 2x 10G SFP+ Software-defined (DPDK/SONiC-VPP) — 16 GB DDR5 (expandable to 48 GB) DRAM pool (NPA) DRAM pool (NPA) CN103XX NP/DPU; buffer is external DDR5 DRAM, not on-chip SRAM; no public on-chip packet buffer figure from Marvell
Asterfusion ET3600 4x 100G QSFP28, 4x 10G SFP+ Software-defined (DPDK/SONiC-VPP) — ~32 GB DDR5 (2x CN103XX, 16 GB each) DRAM pool (NPA) DRAM pool (NPA) Dual CN103XX NP/DPU; 200 Gbps aggregate; buffer is external DDR5 DRAM per NP; no public on-chip packet buffer figure from Marvell
Asterfusion ET3608 2x 100G QSFP28, 2x 10G SFP+ Software-defined (DPDK/SONiC-VPP) — 16 GB DDR5 (expandable to 48 GB) DRAM pool (NPA) DRAM pool (NPA) CN103XX NP/DPU; buffer is external DDR5 DRAM, not on-chip SRAM; no public on-chip packet buffer figure from Marvell
Asterfusion ET3600 4x 100G QSFP28, 4x 10G SFP+ Software-defined (DPDK/SONiC-VPP) — ~32 GB DDR5 (2x CN103XX, 16 GB each) DRAM pool (NPA) DRAM pool (NPA) Dual CN103XX NP/DPU; 200 Gbps aggregate; buffer is external DDR5 DRAM per NP; no public on-chip packet buffer figure from Marvell

To add or update a switch, edit _data/switches.yaml. Each switch is one YAML block. Add a detail_page slug to link to a page in content/switches/.