Hi Bruce,
The issue isn’t locally, these are the test screenshots with nPerf:
With Hardware acceleration enabled:
Without Hardware acceleration:
and this occurs with any firmware installed on the Flint 3
Disabling hardware acceleration will result in better network speed performance, which is for the download, if this issue also exists in the standard firmware, please email to support@gl-inet.com, the technical support team will check
I wrote to support.
I thinking it was normal to see a decrease in performance if hardware acceleration was disabled, although I thought the decrease in speed was too significant.
Should be the oposite, Bruce?
I expect the hardware acceleration to improve the performance.
There is a missing question mark (?) in the above sentence. 
Enabling hardware acceleration should result in better speed performance. Disabling hard acceleration will depend on the performance of the CPU when packet processing through the CPU.
If hardware acceleration is disabled, it is possible that speed of F3 is around 1Gbps+.
May I know after you disable hardware acceleration and do not enable QOS/SQM/Data Statistics/Content Protection, what is the network speed?
In the DPI graph I would like to be able to hover over a line on the graph and have only its information shown. In this screenshot, I’m hovering over the yellow line but it’s still showing multiple traffic types. Optimally it would only show HTTP/S since sometimes the color differences are so subtle it’s hard to differentiate. It would also help those people who may have issues with colors overall.
The installed fibre has a nominal speed of 5Gbs.
The router is connected via a 2.5Gbps interface to the provider's NTE.
The entire LAN is connected with Category 8 ETH cables. Specifically, the notebook (Asus TUF A14) is connected via a USB-C adapter (Minis Forum TB4-01 docking station with 2.5Gbs ETH port) directly to the router.
With firmware 4.8.99, I had not activated any additional features, but the degradation was similar.
With firmware 4.8.x, however, AdGuard Home is active, the WireGuard server is active but no clients are connected.
As additional plug-ins, I have activated luci-app-qos and luci-app-statistics.
I think we have to click above first to (reverse) filter out the Apps we want to see. There is only this App information in the Data Usage.
Just do a comparison test to check:
Hardware acceleration is disabled and all plug-ins are not activated. What is the network speed?
Enable hardware acceleration and all plug-ins are not activate. What is the network speed?
Hardware acceleration is disabled and all plug-ins are not activated. What is the network speed?
Enable hardware acceleration and all plug-ins are not activate. What is the network speed?
Via WiFi, the performance gap is significantly larger:
with HW Acceleration disabled:
https://pic.nperf.com/r/3614566522421619-0oeM7NOi.png
with HW Acceleration enabled:
Right, but you still need to match the colors. It would be MUCH easier just to hover over the graph line you are interested in. More of a “nice to have one day” than anything immediate.
@bruce Will there also be a beta for the Flint 3e (GL-BE6500)? Thanks!
Sorry, there should be no plan about DPI test firmware for Flint 3e.
It sounds meaningful, let me collect this requirement first
You mentioned before that you also tested on v4.8.x (stable) firmware, may I know what the speed results were?
If not, please test the stable firmware to check again.
Hello,
We have not reproduced this issue locally.
An OpenSpeedTest Server is installed on the NAS, which is connected to MT6000 LAN.
MT6000 enables SQM and sets to 100/20.
The wireless client accesses the OpenSpeedTest webpage (NAS), and the speed are normal:
Please share us about these:
@bruce After your post I went and did a fresh install of .99 and ran tests on the LAN. I was getting as expected speeds running iperf3 to a small plex server via WiFi. 800ishMbit/ps iPhone 16 pro 5g. So short story shorter. Thank you for verifying, not sure how or why I had those results on the first .99 but I am running it now so weeee
Maybe SQM or QoS have set the wrong numbers?
Anyway, glad to know works fine now. If it reproduced, please keep the network environment and let us check it remotely.
Personally I think vlan management through the ui would be really nice on such product.
Currently with all the new features, the only thing what prevents me to use it are vlan management 
I already use a flint 2 as isp router which does the vlaning through the normal OpenWrt firmware, but with this new DPI features I actually have plans to make this new Brume to do the heavy lifting of both pushing vlans to encapisulate big parts of far clients to a single hop to the brume, and have the DPI working on all vlans
I guess what can be added if vlans are considered in the dpi ui, is grouping of vlans that would be very nice, kinda like dropdowns.
The extra ports are also usefull, currently my full network goes over lan1 port which is a challenge with 50+ devices and bridge priorities, but I could half the ports to the living room, and half to upstairs, then the congestion control should be better.
Thanks so much. Hopefully we can make it next year, 2026.
Could you please share some more details on this? We are also thinking of how the VLAN works together with other features.
