This page presents the RFC 2544 benchmark test results for the Brainboxes SW-595 5-port Gigabit Ethernet switch with SFP. RFC 2544 defines a standardised methodology for benchmarking network interconnect devices, covering throughput, latency, frame loss rate, and back-to-back frame handling.
Testing was performed using the Xena2544 v2.97 automated test suite on 13 May 2026, with a total test duration of 12 hours 27 minutes. Tests used custom frame sizes including IEEE defaults and jumbo frames (64, 128, 256, 512, 1024, 1280, 1518, 2048, 4096, 8192, and 9000 bytes) in a full-mesh, bidirectional topology across 5 ports at 1 Gbps (5 Gbps aggregate, 20 bidirectional flows).
Result: All Tests Passed
The SW-595 passed all RFC 2544 tests. It sustained full 4.98 Gbps aggregate line rate with zero frame loss across all 20 bidirectional flows for every IEEE-standard frame size (64–1518 bytes) and jumbo frame sizes up to 4096 bytes. Forwarding latency was exceptionally low at 1.6–43 µs at 50% load.
Test
Result
Throughput
Pass — 99.6% line rate, zero loss for all frame sizes 64–4096 bytes
Latency & Jitter
Pass — at 50% and 100% rates
Frame Loss Rate
Pass — 0% loss at 50% rate for 64–4096 bytes
Back-to-Back
Pass — at 50% and 100% rates
Jumbo frame note (8192 / 9000 bytes)
At 8192- and 9000-byte frames the Xena test instrument's traffic generator could only offer 25–35% of line rate; the switch forwarded those frames with zero loss at the achievable offered rate. Under 100% offered rate, elevated loss was observed at 8192 / 9000 bytes — this is expected behaviour when the aggregate jumbo-frame load exceeds the switch's internal forwarding budget. See the per-test sections below for full numbers.
The throughput test determines the maximum rate at which none of the offered frames are dropped by the device. Each frame size was tested for 60 seconds over 3 iterations, with a binary search from 0–100% line rate and 0% acceptable loss.
The SW-595 achieved 99.609% throughput (4.98 Gbps of the 5 Gbps aggregate line rate) with zero frame loss for all frame sizes from 64 through 4096 bytes. For 8192- and 9000-byte jumbo frames the Xena instrument's generator was limited to 27.2% and 25.4% of line rate respectively; the switch forwarded all offered frames with zero loss at those reduced rates.
Frame Size (bytes)
Offered Rate
Tx Rate (L1)
Tx Rate (L2)
Tx (Frames)
Rx (Frames)
Loss
Result
64
99.609%
4.980 Gbps
3.795 Gbps
444,684,000
444,684,000
0
Pass
128
99.609%
4.980 Gbps
4.307 Gbps
252,387,600
252,387,600
0
Pass
256
99.609%
4.980 Gbps
4.620 Gbps
135,338,400
135,338,400
0
Pass
512
99.609%
4.980 Gbps
4.793 Gbps
70,213,200
70,213,200
0
Pass
1024
99.609%
4.980 Gbps
4.885 Gbps
35,779,200
35,779,200
0
Pass
1280
99.609%
4.980 Gbps
4.904 Gbps
28,732,800
28,732,800
0
Pass
1518
99.609%
4.980 Gbps
4.916 Gbps
24,286,800
24,286,800
0
Pass
2048
99.609%
4.980 Gbps
4.932 Gbps
18,062,400
18,062,400
0
Pass
4096
99.609%
4.980 Gbps
4.956 Gbps
9,074,400
9,074,400
0
Pass
8192
27.214%
1.360 Gbps
1.357 Gbps
1,242,000
1,242,000
0
Pass
9000
25.391%
1.269 Gbps
1.266 Gbps
1,055,200
1,055,200
0
Pass
All 20 bidirectional flows in the full mesh showed identical throughput with zero loss at every frame size.
Per-Port Throughput Results
All five ports produced identical throughput in the full-mesh topology. Each port transmits 1/5 of the aggregate traffic, and receives the same.
Port P-0-2-0 (representative — all five ports identical)
Frame Size (bytes)
Tx (Frames)
Tx Rate (L1)
Tx Rate (L2)
Tx Rate (fps)
Rx (Frames)
64
88,936,800
996.09 Mbps
758.93 Mbps
1,482,280
88,936,800
128
50,477,520
996.09 Mbps
861.48 Mbps
841,292
50,477,520
256
27,067,680
996.09 Mbps
923.91 Mbps
451,128
27,067,680
512
14,042,640
996.09 Mbps
958.64 Mbps
234,044
14,042,640
1024
7,155,840
996.09 Mbps
977.01 Mbps
119,264
7,155,840
1280
5,746,560
996.07 Mbps
980.75 Mbps
95,776
5,746,560
1518
4,857,360
996.08 Mbps
983.13 Mbps
80,956
4,857,360
2048
3,612,480
996.08 Mbps
986.45 Mbps
60,208
3,612,480
4096
1,814,880
996.01 Mbps
991.17 Mbps
30,248
1,814,880
8192
248,400
271.98 Mbps
271.32 Mbps
4,140
248,400
9000
211,040
253.81 Mbps
253.25 Mbps
3,517
211,040
Ports P-0-2-1 through P-0-2-4 — identical results
Ports P-0-2-1, P-0-2-2, P-0-2-3, and P-0-2-4 each produced the same per-port frame counts, line rates, and zero loss as P-0-2-0 above.
The latency and jitter test measures forwarding delay and its variation across frame sizes. Tests were run for 60 seconds over 3 iterations using last-to-last latency mode, relative to the throughput test results. Testing was performed at both 50% and 100% of line rate.
Summary tables below show the mean of the five per-port averages, with min/max latency taken from the worst case across all ports.
The frame loss rate test measures the percentage of frames that are not forwarded by the device under steady-state conditions. Each frame size was tested for 60 seconds over 3 iterations. No pass/fail criteria were configured.
At 50% rate the SW-595 achieved zero frame loss for all frame sizes from 64 through 4096 bytes. For 8192 and 9000 the offered rate exceeded what one path could sustain in the mesh, producing 6.33% and 19.54% loss respectively. At 100% rate small per-flow losses (0.01–0.1%) were observed for 64- through 4096-byte frames, and substantially higher loss for 8192/9000-byte jumbo frames.
All five ports transmitted identical frame counts at 50% rate. Loss occurred only at 8192- and 9000-byte frame sizes, distributed across the mesh flows.
Port P-0-2-0 (representative — all five ports identical)
Frame Size (bytes)
Tx (Frames)
Tx Rate (L1)
Tx Rate (fps)
Rx (Frames)
64
44,642,640
500 Mbps
744,044
44,642,640
128
25,337,760
500 Mbps
422,296
25,337,760
256
13,586,880
500 Mbps
226,448
13,586,880
512
7,048,800
499.99 Mbps
117,480
7,048,800
1024
3,591,840
499.98 Mbps
59,864
3,591,840
1280
2,884,560
499.99 Mbps
48,076
2,884,560
1518
2,438,160
499.99 Mbps
40,636
2,438,160
2048
1,813,200
499.96 Mbps
30,220
1,813,200
4096
911,040
499.98 Mbps
15,184
911,040
8192
456,480
499.82 Mbps
7,608
388,010
9000
415,680
499.92 Mbps
6,928
281,830
Ports P-0-2-1 through P-0-2-4 — identical Tx, similar Rx with loss at jumbo sizes
All other ports transmitted the same per-port frame counts. Receive counts at 8192 and 9000 bytes varied per flow; see the source PDF for per-flow detail.
At 100% rate, each port offered 1 Gbps and received 1 Gbps in the full mesh. The very small losses observed at 64- through 4096-byte frame sizes (well under 0.1%) are characteristic of mesh saturation transients during port-sync. At 8192- and 9000-byte jumbo frames the mesh aggregate exceeds the switch's jumbo-frame forwarding budget, producing 30–34% loss.
The back-to-back test measures the maximum number of frames that can be sent at a given rate without any frame loss. Each frame size was tested with a 60-second burst over 3 iterations.
At 100% rate (full 5 Gbps aggregate) the back-to-back burst sizes are smaller per frame size because the rate divisor is doubled, but every burst completed with zero frame loss.
Frame Size (bytes)
Tx Burst (Frames)
Tx Burst (Bytes)
Max Offered Rate (fps)
Max Offered Rate (Gbps)
Result
64
5,917,600
378,726,400
7,440,476
5.000
Pass
128
5,862,187
750,359,893
4,222,973
5.000
Pass
256
5,825,960
1,491,445,760
2,264,493
5.000
Pass
512
2,722,180
1,393,756,160
1,174,812
5.000
Pass
1024
959,780
982,814,720
598,659
5.000
Pass
1280
537,720
688,281,600
480,769
5.000
Pass
1518
396,960
602,585,280
406,372
5.000
Pass
2048
145,627
298,243,413
302,224
5.000
Pass
4096
50,960
208,732,160
151,846
5.000
Pass
8192
727
5,952,853
76,108
5.000
Pass
9000
673
6,060,000
69,290
5.000
Pass
Per-Port Back-to-Back at 100% Rate
Port P-0-2-0 (representative — all five ports identical)
RFC 2544 defines a standardised benchmarking methodology for network interconnect devices. The tests cover:
Throughput — maximum rate at which no frames are lost
Latency — forwarding delay at various frame sizes and rates
Frame Loss Rate — percentage of frames lost under load
Back-to-Back — maximum burst length the device can handle without loss
Testing was performed using the Xena2544 automated test suite. Each test was run for 3 iterations and results averaged to minimise measurement variance. This report extends the standard IEEE frame sizes with jumbo frames (2048, 4096, 8192, and 9000 bytes) to validate jumbo frame support.