The upgrade that made everything worse is a classic — and it is almost never the NAS's fault. Work through these checks in order. The first two explain most cases, and every one of them is a five-minute fix once found.
Yesterday: one cable, boring, a steady 110 MB/s. Today: two cables, a bond, big expectations — and 40 MB/s with pauses. Extra ports create extra ways for the two ends of a link to disagree with each other. Diagnose in order: physical speed, mode agreement, address layout, frame size, then the host.
Open the Network Interface list in DSM and read the negotiated speed of every port, not just its connected status. A damaged cable or a tired plug drops a gigabit port to 100 Mbps while everything still looks plugged in — and a 100 Mbps leg announces itself with copies capped near 11 MB/s.
Swap that leg's cable for a known-good one and watch the port renegotiate. This check takes two minutes and ends a remarkable share of all "bonding is slow" stories before they start.
A bond is a promise made by two devices. If DSM is set to IEEE 802.3ad but the switch has no LAG group — or the switch runs a static group while DSM expects LACP — the links technically connect and then fight: traffic lands unevenly, sessions drop, and pauses appear inside big copies.
The mirror-image mistake is just as common: switch aggregation switched on while DSM uses Adaptive Load Balancing, which DSM's help explicitly warns against. Lay the mode table from this series next to your switch settings and make the pair match.
If instead of bonding, both ports were simply given DHCP addresses inside the same network, you built a guessing game. Replies leave by whichever interface answers first, some sessions ride the slow or wrong path, and shares connect beautifully on Monday and sulk on Tuesday.
The cure is a decision, not a tweak: either bond the ports into one interface with one address, or run them as a deliberate split — static IPs in different subnets with named jobs, like the video's LAN and WAN layout. Two loose addresses in one range is the one arrangement to never keep.
Jumbo frames are an all-or-nothing setting. If the NAS interface is set to an MTU of 9000 while the switch or the PC sits at the standard 1500, large frames get dropped or chopped into retransmissions. The symptom is distinctive: small files feel fine and big copies crawl.
The rule is simple — every device on the path shares one MTU, or everything stays at 1500. A half-jumbo network is slower than an honest standard one, and it fails weirdly enough to waste weekends.
Virtual adapters are only as fast as the host's real network port, and that port may be shared with the host itself and every other virtual machine on it. Three virtual LAN cards on the Synology can all be queueing behind one physical gigabit cable on the VMware host.
Also confirm the VMware-side discipline from the video still holds: each adapter on a manual, unique MAC address. An adapter that picked up a fresh automatic MAC during an edit makes interfaces swap identities — a fault that looks exactly like random slowness and vanishing shares, and is not a DSM problem at all.
Compare your build against the video once more — the faults on this page are almost all differences between the two.
▶️ Watch Full VideoThat is a 100 Mbps link introducing itself. Somewhere — a cable, a wall socket, a switch port — one leg negotiated Fast Ethernet instead of gigabit. DSM's Network Interface list shows the real negotiated speed of each port.
Yes. Connected only means the cables have link. If the mode on the NAS and the aggregation setting on the switch disagree, traffic still crawls or drops. Agreement, not connection, is the requirement.
Often that is the fastest cure. Remove the bond, test one port alone and confirm full speed, then rebuild with the correct mode. A clean single-port baseline tells you whether you are chasing a network problem or a NAS problem.