PROS:
- true 802.1q VLAN tagging support, at a very low price
- excellent wire-speed performance, tested @ 950+Mb/s in iperf on 6 ports at once, so long as Flow Control is disabled (see below). The main chip is a Broadcom BCM53128, which seems slightly faster and more efficient than the Realtek RTL8380 used on competing switches from TP-Link and Zyxel.
- very low power use, approx. 1W - 3W dependent on ports in use (~ 0.25W DC / 0.45W AC base plus 0.3W per connected 1000M port, or 0.2W per 100M port). Traffic levels seem not to affect this much, though cable length might, since it tries to use lower transmit power on short cables. The switch-matrix core runs on 1.2V, with 3.3V for I/O, both provided by efficient switchmode buck regulators (unlike some switches that use a hot-running, linear step-down from 3.3V to 1.2 or 1.8V, at < 50% efficiency).
- silent operation, with no fan or coil whine. Despite the lack of a fan, mine never even gets noticeably warm. There is a reasonably sized heatsink on the main chip, and a thermal pad under the PCB, allowing the case bottom to serve as an additional heatsink.
- good build quality. Just one electrolytic capacitor, which is a Nichicon (reputable Japanese brand) 220uF on the power input, over-rated at 25V when operating at 12V. Main switch chip is a Broadcom BCM53128. Case is solid metal, compromised slightly by a stick-on platic piece around the ports for labeling, which another reviewer complained about.
- happily runs on the variable 12V battery bus (10.5-14.4V) from an offgrid DC power system
- starts up quickly, from cold-powerdown to forwarding traffic in about 8 seconds, far faster than any "true" managed switch (Catalyst, Juniper, Procurve etc.)
CONS:
- management interface, whether via web or Windows app is limited and cumbersome, especially when configuring VLANs, though is most cases this is only a one-time annoyance, set-and-forget
- typical of switches in this class, there's no CLI management, nor SNMP, so tracking port activity with MRTG, etc. is not possible. Omitting these probably does reduce both cost and power use, though.
- The management controller, integrated within the BCM53128 is a weak CPU derived from the 1980s-vintage, 8-bit Intel 8051, which is easily overloaded. This explains the lack of HTTPS SSL support, occasional dropped HTTP requests, etc. It's actually impressive they managed to squeeze an IP stack and web interface onto such a small CPU at all.
- This limited 8051 service processor would only affect management functionality (it isn't part of the main switch-fabric data path at all), except for a dumb decision on Netgear's part to configure the switch registers to send a copy of *all* HTTP (tcp port 80) traffic, originating from any port, to this tiny management CPU.
This has the effect of badly crippling the layer-2 Flow-Control feature, causing any and all HTTP traffic flowing through the switch to be bottlenecked to about 10Mb/s whenever Flow Control is switched on. The reason is that flow-control rate limiting kicks in whenver any port receiving the traffic gets overloaded, the weak management CPU effectively connects to a internal 9th, on-chip port that seems to run at only 10Mb/s, AND all web traffic from anywhere to anywhere (even when bearing an 802.1q VLAN tag!) gets uncontrollably copied to the phantom port-9.
This wouldn't be so bad if the web interface could be moved to a less important port than tcp/80, set to listen to requests only from one specific switch port, or disabled entirely (until next power-cycle, say, or even semi-permanently until a factory reset), but none of these options are provided.
Netgear does seem to ship these switches with Flow Control turned OFF by default, masking the performance problem, and in many cases Flow Control is undesirable anyway, but it can be useful in cases where a node on your network can't keep up with full Gigabit rates, but can do better than 100M. Many low-power single board computers fall into this category and can benefit from FC.
The management controller can also be a big security hole, since it doesn't respect VLAN boundaries at all, and copies of its transmissions get relayed to ALL ports, in-the-clear and untagged, regardless of VLAN membership status. The switch effectively reverts to just a dumb hub whenever the hidden Port 9 is involved.
HARDWARE MOD:
If you don't mind voiding your warranty with a small bit of soldering, it's thankfully possible to disable this ill-behaved web interface and completely disconnect the BCM53128's 8051 management controller from its switch fabric, eliminating its security exposure and letting Flow Control work as intended, with no more weird slowdowns affecting port 80 . Here are the steps:
1. First Get everything configured as you like in the web interface, setting up and testing all VLANs, etc. Reboot the switch and verify it comes up in the desired configuration.
2. Open the cover and find tiny surface-mount resistor R75, between the main chip and the ports, near the crystal oscillator. Desolder this resistor. See my first photo, where it's already been removed, leaving bare pads.
R75, which I measured at 4.4k in-circuit, pulls BCM53128 pin 43, "EN_8051_TxRx" high, to 3.3V. It CAN just be left floating if you don't mind losing the web (and Windows-config-tool) interface permanently after initial setup.
3. If you want to be able to toggle the web interface off and on, solder a very fine-gauge wire to the removed R75 resistor's pad nearest the main chip, and another such wire to a 3.3V power pin anywhere on the board-- you could use the other side of R75, but it's easier to grab this voltage from a less closely-spaced area. I chose to use the power pin (pin 8) of U5, the 8pin serial EEPROM at upper-right.
4. drill a hole on the back panel somewhere to mount a small toggle switch. Solder one of its terminals to the R75 pin (BCM531128 pin43, EN_8051_TxRx signal) through a 3.3k-ohm resistor. Solder the other toggle switch terminal to any convenient 3.3V pin, through a 1k-ohm resistor (optional - these two resistors in series approximate the original 4.4k-ohm pullup, but anything in the ballmark should work)
Note that EN_8051_TxRx is only latched during reset, so after flipping the switch you have to power-cycle the switch for it to take effect. At least this Netgear is fast to reboot, but having to do so makes its port traffic & error counters effectively useless, unless you leave management enabled all the time.
If you want to add a RESET button also, solder a fine wire to the right-hand (near the coil) terminal of resistor pad R7, which should be empty to begin with. This goes to the Shutdown terminal of the 3.3V switchmode regulator, and grounding it (through a 1k-ohm resistor) even briefly will cause a clean reset via 3-pin power-supervisor chip U3. That IC actively drives the BCM53128 RESET pin both high and low, so you can't safely pull RESET down directly.
Rather than mounting a physical toggle switch, since my GS108e is in a hard-to-reach spot I decided to bring out these control signals (EN_8051_TxRx, +3.3V, RESET-via-regulator-shutdown) to a 4-pin header, which plugs into GPIO outputs on a nearby router, allowing management-enable and reset functions to be controlled remotely. Anyone going this route should ensure both systems share a common logic ground, and take care to never drive either signal to more than 3.3V Documentation on the BCM5128 is hard to come by, but I very much doubt it's 5V-tolerant.
Frequently bought together
- +
Total price:
To see our price, add these items to your cart.
Choose items to buy together.
Customers also search
Page 1 of 1Page 1 of 1
From the manufacturer
|
|
|
|
|
|
|---|---|---|---|---|---|
| GS105NA | GS105E | GS108 | GS108E | GS116NA | |
| # Gigabit Ethernet Ports | 5 x 1G | 5 x 1G | 8 x 1G | 8 x 1G | 16 x 1G |
| # PoE Ports | ---- | ---- | ---- | ---- | ---- |
| # SFP Fiber Ports | ---- | ---- | ---- | ---- | ---- |
| Management Type | Unmanaged | Plus | Unmanaged | Plus | Unmanaged |
| Management Layer | ---- | L2 | ---- | L2 | ---- |
| Advanced Network Features | ---- | Basic VLAN & QoS IGMP Storm Control Auto DoS | ---- | Basic VLAN & QoS IGMP Storm Control Auto DoS | ---- |
| Fanless Housing | ✓ | ✓ | ✓ | ✓ | ✓ |
| Mounting Options | Desktop or Wall | Desktop or Wall | Desktop or Wall | Desktop or Wall | Desktop or Wall |
Compare with similar items
 | TP-Link TL-SG105 | 5 Port Gigabit Unmanaged Ethernet Network Switch, Ethernet Splitter | Plug & Play | Fanless Metal Design | Shielded Ports | Traffic Optimization | Limited Lifetime Protection | NETGEAR 5-Port Gigabit Ethernet Unmanaged PoE Switch (GS305P v2) - with 4 x PoE+ @ 63W, Desktop or Wall Mount | NETGEAR 5-Port Gigabit Ethernet Unmanaged Switch (GS205) - Desktop, Ethernet Splitter, Plug-and-Play, Silent Operation | TP-Link Litewave 5 Port Gigabit Ethernet Switch | Desktop Ethernet Splitter | Plastic Case | Unshielded Network Switch | Plug & Play | Fanless Quiet | Unmanaged (LS1005G) | TP-Link TL-SG108 | 8 Port Gigabit Unmanaged Ethernet Network Switch, Ethernet Splitter | Plug & Play | Fanless Metal Design | Shielded Ports | Traffic Optimization | Limited Lifetime Protection | |
|---|---|---|---|---|---|---|
| Customer Rating | (11897) | (111072) | (1463) | (6004) | (25914) | (35207) |
| Price | $27.99 | $15.99 | $39.99 | $23.99 | $12.99 | $19.79 |
| Shipping | FREE Shipping. Details | FREE Shipping on orders over $25.00 shipped by Amazon or get Fast, Free Shipping with Amazon Prime | FREE Shipping. Details | FREE Shipping on orders over $25.00 shipped by Amazon or get Fast, Free Shipping with Amazon Prime | FREE Shipping on orders over $25.00 shipped by Amazon or get Fast, Free Shipping with Amazon Prime | FREE Shipping on orders over $25.00 shipped by Amazon or get Fast, Free Shipping with Amazon Prime |
| Sold By | Amazon.com | Amazon.com | Amazon.com | Abe's Electronics Center | Amazon.com | Amazon.com |
| Connectivity Technology | wireless | Ethernet | — | — | Ethernet | ethernet |
| Data Transfer Rate | 1000 Mb per second | 1000 Mb per second | — | 1000 Mb per second | 1000 Mb per second | 1000 Mb per second |
| Item Dimensions | 4.02 x 1.06 x 3.9 inches | 3.9 x 3.9 x 1 inches | 6.2 x 4 x 1.1 inches | 9.25 x 5.51 x 2.4 inches | 3.54 x 2.83 x 0.91 inches | 6.22 x 3.94 x 0.98 inches |
| Item Weight | 1.08 lbs | 7.76 ounces | 2.27 lbs | 0.68 lbs | 2.80 ounces | 0.86 lbs |
| Number of Ports | 5 | 5 | 5 | 5 | 5 | 8 |
| Size | Unmanaged | 5 Port | Unmanaged | 05 Port | — | 8 Port |
| Wattage | 3.5 watts | 2.43 watts | — | 3.56 watts | 3.7 watts | 4.5 watts |
Videos
Page 1 of 1Start OverPage 1 of 1
Videos for this product
4:29
Click to play video
How to Set Up a NETGEAR Nighthawk Switch
NETGEAR
Videos for this product
3:32
Click to play video
Unboxing the NETGEAR Nighthawk S8000 Gaming Switch
NETGEAR
Videos for this product
2:18
Click to play video
NETGEAR ProSAFE 10-Gigabit Ethernet Switches | XS708Ev2 & XS716E
NETGEAR
Videos for related products
1:00
Click to play video
NETGEAR Smart Managed Plus PoE 5 & 8 Port Ethernet Switches
Netgear Inc
Videos for related products
1:49
Click to play video
UNBOXING & QUICK OVERVIEW: Netgear 5-Port Gigabit Ethernet Switch
Hammer and Home
Videos for related products
1:40
Click to play video
NETGEAR 8-Port Gigabit Ethernet Unmanaged Switch Unpacking
Trui
Videos for related products
2:52
Click to play video
TP-Link 8 Port Gigabit Plug & Play Switch Ethernet Splitter
Caesar
Videos for related products
0:59
Click to play video
NETGEAR 5-Port Gigabit Ethernet Unmanaged PoE Switch
Merchant Video
Customer reviews
Nice BCM53128 switch with flawed management interface (can disable via hardware hack)
Reviewed in the United States on August 8, 2018
PROS: - true 802.1q VLAN tagging support, at a very low price - excellent wire-speed performance, tested @ 950+Mb/s in iperf on 6 ports at once, so long as Flow Control is disabled (see below). The main chip is a Broadcom BCM53128, which seems slightly faster and more efficient than the Realtek RTL8380 used on competing switches from TP-Link and Zyxel. - very low power use, approx. 1W - 3W dependent on ports in use (~ 0.25W DC / 0.45W AC base plus 0.3W per connected 1000M port, or 0.2W per 100M port). Traffic levels seem not to affect this much, though cable length might, since it tries to use lower transmit power on short cables. The switch-matrix core runs on 1.2V, with 3.3V for I/O, both provided by efficient switchmode buck regulators (unlike some switches that use a hot-running, linear step-down from 3.3V to 1.2 or 1.8V, at < 50% efficiency). - silent operation, with no fan or coil whine. Despite the lack of a fan, mine never even gets noticeably warm. There is a reasonably sized heatsink on the main chip, and a thermal pad under the PCB, allowing the case bottom to serve as an additional heatsink. - good build quality. Just one electrolytic capacitor, which is a Nichicon (reputable Japanese brand) 220uF on the power input, over-rated at 25V when operating at 12V. Main switch chip is a Broadcom BCM53128. Case is solid metal, compromised slightly by a stick-on platic piece around the ports for labeling, which another reviewer complained about. - happily runs on the variable 12V battery bus (10.5-14.4V) from an offgrid DC power system - starts up quickly, from cold-powerdown to forwarding traffic in about 8 seconds, far faster than any "true" managed switch (Catalyst, Juniper, Procurve etc.)CONS: - management interface, whether via web or Windows app is limited and cumbersome, especially when configuring VLANs, though is most cases this is only a one-time annoyance, set-and-forget - typical of switches in this class, there's no CLI management, nor SNMP, so tracking port activity with MRTG, etc. is not possible. Omitting these probably does reduce both cost and power use, though. - The management controller, integrated within the BCM53128 is a weak CPU derived from the 1980s-vintage, 8-bit Intel 8051, which is easily overloaded. This explains the lack of HTTPS SSL support, occasional dropped HTTP requests, etc. It's actually impressive they managed to squeeze an IP stack and web interface onto such a small CPU at all. - This limited 8051 service processor would only affect management functionality (it isn't part of the main switch-fabric data path at all), except for a dumb decision on Netgear's part to configure the switch registers to send a copy of *all* HTTP (tcp port 80) traffic, originating from any port, to this tiny management CPU.This has the effect of badly crippling the layer-2 Flow-Control feature, causing any and all HTTP traffic flowing through the switch to be bottlenecked to about 10Mb/s whenever Flow Control is switched on. The reason is that flow-control rate limiting kicks in whenver any port receiving the traffic gets overloaded, the weak management CPU effectively connects to a internal 9th, on-chip port that seems to run at only 10Mb/s, AND all web traffic from anywhere to anywhere (even when bearing an 802.1q VLAN tag!) gets uncontrollably copied to the phantom port-9.This wouldn't be so bad if the web interface could be moved to a less important port than tcp/80, set to listen to requests only from one specific switch port, or disabled entirely (until next power-cycle, say, or even semi-permanently until a factory reset), but none of these options are provided.Netgear does seem to ship these switches with Flow Control turned OFF by default, masking the performance problem, and in many cases Flow Control is undesirable anyway, but it can be useful in cases where a node on your network can't keep up with full Gigabit rates, but can do better than 100M. Many low-power single board computers fall into this category and can benefit from FC.The management controller can also be a big security hole, since it doesn't respect VLAN boundaries at all, and copies of its transmissions get relayed to ALL ports, in-the-clear and untagged, regardless of VLAN membership status. The switch effectively reverts to just a dumb hub whenever the hidden Port 9 is involved.HARDWARE MOD:If you don't mind voiding your warranty with a small bit of soldering, it's thankfully possible to disable this ill-behaved web interface and completely disconnect the BCM53128's 8051 management controller from its switch fabric, eliminating its security exposure and letting Flow Control work as intended, with no more weird slowdowns affecting port 80 . Here are the steps:1. First Get everything configured as you like in the web interface, setting up and testing all VLANs, etc. Reboot the switch and verify it comes up in the desired configuration.2. Open the cover and find tiny surface-mount resistor R75, between the main chip and the ports, near the crystal oscillator. Desolder this resistor. See my first photo, where it's already been removed, leaving bare pads.R75, which I measured at 4.4k in-circuit, pulls BCM53128 pin 43, "EN_8051_TxRx" high, to 3.3V. It CAN just be left floating if you don't mind losing the web (and Windows-config-tool) interface permanently after initial setup.3. If you want to be able to toggle the web interface off and on, solder a very fine-gauge wire to the removed R75 resistor's pad nearest the main chip, and another such wire to a 3.3V power pin anywhere on the board-- you could use the other side of R75, but it's easier to grab this voltage from a less closely-spaced area. I chose to use the power pin (pin 8) of U5, the 8pin serial EEPROM at upper-right.4. drill a hole on the back panel somewhere to mount a small toggle switch. Solder one of its terminals to the R75 pin (BCM531128 pin43, EN_8051_TxRx signal) through a 3.3k-ohm resistor. Solder the other toggle switch terminal to any convenient 3.3V pin, through a 1k-ohm resistor (optional - these two resistors in series approximate the original 4.4k-ohm pullup, but anything in the ballmark should work)Note that EN_8051_TxRx is only latched during reset, so after flipping the switch you have to power-cycle the switch for it to take effect. At least this Netgear is fast to reboot, but having to do so makes its port traffic & error counters effectively useless, unless you leave management enabled all the time.If you want to add a RESET button also, solder a fine wire to the right-hand (near the coil) terminal of resistor pad R7, which should be empty to begin with. This goes to the Shutdown terminal of the 3.3V switchmode regulator, and grounding it (through a 1k-ohm resistor) even briefly will cause a clean reset via 3-pin power-supervisor chip U3. That IC actively drives the BCM53128 RESET pin both high and low, so you can't safely pull RESET down directly.Rather than mounting a physical toggle switch, since my GS108e is in a hard-to-reach spot I decided to bring out these control signals (EN_8051_TxRx, +3.3V, RESET-via-regulator-shutdown) to a 4-pin header, which plugs into GPIO outputs on a nearby router, allowing management-enable and reset functions to be controlled remotely. Anyone going this route should ensure both systems share a common logic ground, and take care to never drive either signal to more than 3.3V Documentation on the BCM5128 is hard to come by, but I very much doubt it's 5V-tolerant.
Reviewed in the United States on August 8, 2018
Images in this review
Top reviews from the United States
There was a problem filtering reviews right now. Please try again later.
Reviewed in the United States 🇺🇸 on August 8, 2018
- true 802.1q VLAN tagging support, at a very low price
- excellent wire-speed performance, tested @ 950+Mb/s in iperf on 6 ports at once, so long as Flow Control is disabled (see below). The main chip is a Broadcom BCM53128, which seems slightly faster and more efficient than the Realtek RTL8380 used on competing switches from TP-Link and Zyxel.
- very low power use, approx. 1W - 3W dependent on ports in use (~ 0.25W DC / 0.45W AC base plus 0.3W per connected 1000M port, or 0.2W per 100M port). Traffic levels seem not to affect this much, though cable length might, since it tries to use lower transmit power on short cables. The switch-matrix core runs on 1.2V, with 3.3V for I/O, both provided by efficient switchmode buck regulators (unlike some switches that use a hot-running, linear step-down from 3.3V to 1.2 or 1.8V, at < 50% efficiency).
- silent operation, with no fan or coil whine. Despite the lack of a fan, mine never even gets noticeably warm. There is a reasonably sized heatsink on the main chip, and a thermal pad under the PCB, allowing the case bottom to serve as an additional heatsink.
- good build quality. Just one electrolytic capacitor, which is a Nichicon (reputable Japanese brand) 220uF on the power input, over-rated at 25V when operating at 12V. Main switch chip is a Broadcom BCM53128. Case is solid metal, compromised slightly by a stick-on platic piece around the ports for labeling, which another reviewer complained about.
- happily runs on the variable 12V battery bus (10.5-14.4V) from an offgrid DC power system
- starts up quickly, from cold-powerdown to forwarding traffic in about 8 seconds, far faster than any "true" managed switch (Catalyst, Juniper, Procurve etc.)
CONS:
- management interface, whether via web or Windows app is limited and cumbersome, especially when configuring VLANs, though is most cases this is only a one-time annoyance, set-and-forget
- typical of switches in this class, there's no CLI management, nor SNMP, so tracking port activity with MRTG, etc. is not possible. Omitting these probably does reduce both cost and power use, though.
- The management controller, integrated within the BCM53128 is a weak CPU derived from the 1980s-vintage, 8-bit Intel 8051, which is easily overloaded. This explains the lack of HTTPS SSL support, occasional dropped HTTP requests, etc. It's actually impressive they managed to squeeze an IP stack and web interface onto such a small CPU at all.
- This limited 8051 service processor would only affect management functionality (it isn't part of the main switch-fabric data path at all), except for a dumb decision on Netgear's part to configure the switch registers to send a copy of *all* HTTP (tcp port 80) traffic, originating from any port, to this tiny management CPU.
This has the effect of badly crippling the layer-2 Flow-Control feature, causing any and all HTTP traffic flowing through the switch to be bottlenecked to about 10Mb/s whenever Flow Control is switched on. The reason is that flow-control rate limiting kicks in whenver any port receiving the traffic gets overloaded, the weak management CPU effectively connects to a internal 9th, on-chip port that seems to run at only 10Mb/s, AND all web traffic from anywhere to anywhere (even when bearing an 802.1q VLAN tag!) gets uncontrollably copied to the phantom port-9.
This wouldn't be so bad if the web interface could be moved to a less important port than tcp/80, set to listen to requests only from one specific switch port, or disabled entirely (until next power-cycle, say, or even semi-permanently until a factory reset), but none of these options are provided.
Netgear does seem to ship these switches with Flow Control turned OFF by default, masking the performance problem, and in many cases Flow Control is undesirable anyway, but it can be useful in cases where a node on your network can't keep up with full Gigabit rates, but can do better than 100M. Many low-power single board computers fall into this category and can benefit from FC.
The management controller can also be a big security hole, since it doesn't respect VLAN boundaries at all, and copies of its transmissions get relayed to ALL ports, in-the-clear and untagged, regardless of VLAN membership status. The switch effectively reverts to just a dumb hub whenever the hidden Port 9 is involved.
HARDWARE MOD:
If you don't mind voiding your warranty with a small bit of soldering, it's thankfully possible to disable this ill-behaved web interface and completely disconnect the BCM53128's 8051 management controller from its switch fabric, eliminating its security exposure and letting Flow Control work as intended, with no more weird slowdowns affecting port 80 . Here are the steps:
1. First Get everything configured as you like in the web interface, setting up and testing all VLANs, etc. Reboot the switch and verify it comes up in the desired configuration.
2. Open the cover and find tiny surface-mount resistor R75, between the main chip and the ports, near the crystal oscillator. Desolder this resistor. See my first photo, where it's already been removed, leaving bare pads.
R75, which I measured at 4.4k in-circuit, pulls BCM53128 pin 43, "EN_8051_TxRx" high, to 3.3V. It CAN just be left floating if you don't mind losing the web (and Windows-config-tool) interface permanently after initial setup.
3. If you want to be able to toggle the web interface off and on, solder a very fine-gauge wire to the removed R75 resistor's pad nearest the main chip, and another such wire to a 3.3V power pin anywhere on the board-- you could use the other side of R75, but it's easier to grab this voltage from a less closely-spaced area. I chose to use the power pin (pin 8) of U5, the 8pin serial EEPROM at upper-right.
4. drill a hole on the back panel somewhere to mount a small toggle switch. Solder one of its terminals to the R75 pin (BCM531128 pin43, EN_8051_TxRx signal) through a 3.3k-ohm resistor. Solder the other toggle switch terminal to any convenient 3.3V pin, through a 1k-ohm resistor (optional - these two resistors in series approximate the original 4.4k-ohm pullup, but anything in the ballmark should work)
Note that EN_8051_TxRx is only latched during reset, so after flipping the switch you have to power-cycle the switch for it to take effect. At least this Netgear is fast to reboot, but having to do so makes its port traffic & error counters effectively useless, unless you leave management enabled all the time.
If you want to add a RESET button also, solder a fine wire to the right-hand (near the coil) terminal of resistor pad R7, which should be empty to begin with. This goes to the Shutdown terminal of the 3.3V switchmode regulator, and grounding it (through a 1k-ohm resistor) even briefly will cause a clean reset via 3-pin power-supervisor chip U3. That IC actively drives the BCM53128 RESET pin both high and low, so you can't safely pull RESET down directly.
Rather than mounting a physical toggle switch, since my GS108e is in a hard-to-reach spot I decided to bring out these control signals (EN_8051_TxRx, +3.3V, RESET-via-regulator-shutdown) to a 4-pin header, which plugs into GPIO outputs on a nearby router, allowing management-enable and reset functions to be controlled remotely. Anyone going this route should ensure both systems share a common logic ground, and take care to never drive either signal to more than 3.3V Documentation on the BCM5128 is hard to come by, but I very much doubt it's 5V-tolerant.
Style: 8 Port with Enhanced FeaturesVerified Purchase
Nice BCM53128 switch with flawed management interface (can disable via hardware hack)
By packetrat on August 7, 2018
PROS:By packetrat on August 7, 2018
- true 802.1q VLAN tagging support, at a very low price
- excellent wire-speed performance, tested @ 950+Mb/s in iperf on 6 ports at once, so long as Flow Control is disabled (see below). The main chip is a Broadcom BCM53128, which seems slightly faster and more efficient than the Realtek RTL8380 used on competing switches from TP-Link and Zyxel.
- very low power use, approx. 1W - 3W dependent on ports in use (~ 0.25W DC / 0.45W AC base plus 0.3W per connected 1000M port, or 0.2W per 100M port). Traffic levels seem not to affect this much, though cable length might, since it tries to use lower transmit power on short cables. The switch-matrix core runs on 1.2V, with 3.3V for I/O, both provided by efficient switchmode buck regulators (unlike some switches that use a hot-running, linear step-down from 3.3V to 1.2 or 1.8V, at < 50% efficiency).
- silent operation, with no fan or coil whine. Despite the lack of a fan, mine never even gets noticeably warm. There is a reasonably sized heatsink on the main chip, and a thermal pad under the PCB, allowing the case bottom to serve as an additional heatsink.
- good build quality. Just one electrolytic capacitor, which is a Nichicon (reputable Japanese brand) 220uF on the power input, over-rated at 25V when operating at 12V. Main switch chip is a Broadcom BCM53128. Case is solid metal, compromised slightly by a stick-on platic piece around the ports for labeling, which another reviewer complained about.
- happily runs on the variable 12V battery bus (10.5-14.4V) from an offgrid DC power system
- starts up quickly, from cold-powerdown to forwarding traffic in about 8 seconds, far faster than any "true" managed switch (Catalyst, Juniper, Procurve etc.)
CONS:
- management interface, whether via web or Windows app is limited and cumbersome, especially when configuring VLANs, though is most cases this is only a one-time annoyance, set-and-forget
- typical of switches in this class, there's no CLI management, nor SNMP, so tracking port activity with MRTG, etc. is not possible. Omitting these probably does reduce both cost and power use, though.
- The management controller, integrated within the BCM53128 is a weak CPU derived from the 1980s-vintage, 8-bit Intel 8051, which is easily overloaded. This explains the lack of HTTPS SSL support, occasional dropped HTTP requests, etc. It's actually impressive they managed to squeeze an IP stack and web interface onto such a small CPU at all.
- This limited 8051 service processor would only affect management functionality (it isn't part of the main switch-fabric data path at all), except for a dumb decision on Netgear's part to configure the switch registers to send a copy of *all* HTTP (tcp port 80) traffic, originating from any port, to this tiny management CPU.
This has the effect of badly crippling the layer-2 Flow-Control feature, causing any and all HTTP traffic flowing through the switch to be bottlenecked to about 10Mb/s whenever Flow Control is switched on. The reason is that flow-control rate limiting kicks in whenver any port receiving the traffic gets overloaded, the weak management CPU effectively connects to a internal 9th, on-chip port that seems to run at only 10Mb/s, AND all web traffic from anywhere to anywhere (even when bearing an 802.1q VLAN tag!) gets uncontrollably copied to the phantom port-9.
This wouldn't be so bad if the web interface could be moved to a less important port than tcp/80, set to listen to requests only from one specific switch port, or disabled entirely (until next power-cycle, say, or even semi-permanently until a factory reset), but none of these options are provided.
Netgear does seem to ship these switches with Flow Control turned OFF by default, masking the performance problem, and in many cases Flow Control is undesirable anyway, but it can be useful in cases where a node on your network can't keep up with full Gigabit rates, but can do better than 100M. Many low-power single board computers fall into this category and can benefit from FC.
The management controller can also be a big security hole, since it doesn't respect VLAN boundaries at all, and copies of its transmissions get relayed to ALL ports, in-the-clear and untagged, regardless of VLAN membership status. The switch effectively reverts to just a dumb hub whenever the hidden Port 9 is involved.
HARDWARE MOD:
If you don't mind voiding your warranty with a small bit of soldering, it's thankfully possible to disable this ill-behaved web interface and completely disconnect the BCM53128's 8051 management controller from its switch fabric, eliminating its security exposure and letting Flow Control work as intended, with no more weird slowdowns affecting port 80 . Here are the steps:
1. First Get everything configured as you like in the web interface, setting up and testing all VLANs, etc. Reboot the switch and verify it comes up in the desired configuration.
2. Open the cover and find tiny surface-mount resistor R75, between the main chip and the ports, near the crystal oscillator. Desolder this resistor. See my first photo, where it's already been removed, leaving bare pads.
R75, which I measured at 4.4k in-circuit, pulls BCM53128 pin 43, "EN_8051_TxRx" high, to 3.3V. It CAN just be left floating if you don't mind losing the web (and Windows-config-tool) interface permanently after initial setup.
3. If you want to be able to toggle the web interface off and on, solder a very fine-gauge wire to the removed R75 resistor's pad nearest the main chip, and another such wire to a 3.3V power pin anywhere on the board-- you could use the other side of R75, but it's easier to grab this voltage from a less closely-spaced area. I chose to use the power pin (pin 8) of U5, the 8pin serial EEPROM at upper-right.
4. drill a hole on the back panel somewhere to mount a small toggle switch. Solder one of its terminals to the R75 pin (BCM531128 pin43, EN_8051_TxRx signal) through a 3.3k-ohm resistor. Solder the other toggle switch terminal to any convenient 3.3V pin, through a 1k-ohm resistor (optional - these two resistors in series approximate the original 4.4k-ohm pullup, but anything in the ballmark should work)
Note that EN_8051_TxRx is only latched during reset, so after flipping the switch you have to power-cycle the switch for it to take effect. At least this Netgear is fast to reboot, but having to do so makes its port traffic & error counters effectively useless, unless you leave management enabled all the time.
If you want to add a RESET button also, solder a fine wire to the right-hand (near the coil) terminal of resistor pad R7, which should be empty to begin with. This goes to the Shutdown terminal of the 3.3V switchmode regulator, and grounding it (through a 1k-ohm resistor) even briefly will cause a clean reset via 3-pin power-supervisor chip U3. That IC actively drives the BCM53128 RESET pin both high and low, so you can't safely pull RESET down directly.
Rather than mounting a physical toggle switch, since my GS108e is in a hard-to-reach spot I decided to bring out these control signals (EN_8051_TxRx, +3.3V, RESET-via-regulator-shutdown) to a 4-pin header, which plugs into GPIO outputs on a nearby router, allowing management-enable and reset functions to be controlled remotely. Anyone going this route should ensure both systems share a common logic ground, and take care to never drive either signal to more than 3.3V Documentation on the BCM5128 is hard to come by, but I very much doubt it's 5V-tolerant.
Images in this review
Reviewed in the United States 🇺🇸 on November 29, 2022
Style: 5 Port with Enhanced FeaturesVerified Purchase
Reviewed in the United States 🇺🇸 on November 24, 2022
Form, fit, and function are successful.
Aesthetics are a disaster because of power cord plugin placement. Sometimes engineers and bean counters should not get there way.
Nice stamped metal housing.
Style: 8 portVerified Purchase
Gets the job done unattractively
By Cat4life on November 23, 2022
The power cord is on the front. It's designed I suppose primarily for wall mounting. Ugly and awkward if you don't do that. I just need it to sit below the bedroom TV. Goofy looking with the power cord front and center.By Cat4life on November 23, 2022
Form, fit, and function are successful.
Aesthetics are a disaster because of power cord plugin placement. Sometimes engineers and bean counters should not get there way.
Nice stamped metal housing.
Images in this review
Top reviews from other countries
Adrian Walsh
Just missing POE
Reviewed in the United Kingdom 🇬🇧 on October 7, 2021Style: 8 portVerified Purchase
paulington
Just works.
Reviewed in the United Kingdom 🇬🇧 on December 15, 2021Style: 8 portVerified Purchase
Amazon Customer
Simple but effective
Reviewed in the United Kingdom 🇬🇧 on December 29, 2020Style: 8 portVerified Purchase
Amazon Customer
Does what it was supposed to do, but the management UI needs work
Reviewed in Canada 🇨🇦 on December 30, 2021Style: 8 Port with Enhanced FeaturesVerified Purchase
nidaniel
Malwear center
Reviewed in Canada 🇨🇦 on June 7, 2020Style: 5 Port with Enhanced FeaturesVerified Purchase
