@ofxyad thinking in the planet, and freedom maybe you should wait better than buy a trashable router.
I have mine 2020 since …2020 and run as perfecly as i just can not update to 6.x because of an unresolved bug in pkgupdate process
i will just keep cool and try to resolve it.
Just for information i’v never feel more free than after buying this near “local” router.
But my 70 000 kb/s link is just enougth to do all i need,video streaming, web browsin, hosting multiple private fediverse serveces for like a normal family.
I you want it for a long time maybe you could wait a little time and save some money. Maybe just keep , and take a (mint from your own production)
I ran my internet only on two routers since 2005. The first one was Linksys WRT54GL which I bought and immediately flashed with dd-wrt. As internet got faster and faster I ended up replacing it with an Omnia. The old one had limited bandwidth and could not handle routing and had just the 802.11b WiFi.
With Omnia being quite upgradeable the limit is more likely to be the software. Hopefully the team grows and keeps supporting it.
I am hoping that the CPU in Omnia II is also substantially faster than the current Omnia CPU. Use case: Samba file server with mPCIe SATA add-on in my Omnia. When /etc/samba/smb.conf.template is configured to use smb encrypt = desired, throughput on the local LAN drops from 102MB/s to 11MB/s. (Tip: /etc/samba/smb.conf.template with socket options = TCP_NODELAY is what gets me from ~90MB/s to ~102MB/s on my Omnia) Years ago, I benchmarked samba on Omnia at 121MB/s, which pretty much saturates a GB connection, and software/kernel updates since have taken a toll – I am guessing kernel security paranoia for Spectre and Meltdown (unconfirmed). I am OK with 102MB/s, but not ok with 11MB/s. In the future, I would like to use smb encrypt = desired. For now, I use physically isolated networks to better protect the samba file server.
Do you have evidence of that or are you pontificating? And does it matter if it is hardware or software if my Omnia CPU runs at 100% CPU usage when samba runs at 11MB/s with smb encrypt = desired ? Encryption requires processing power and samba is mature, widely-used open-source software, so without further evidence it seems unlikely to me that the majority of the problem is software performance. (Modern linux kernels and using KTLS in samba might help: a quick grep of samba source code, which uses gnutls, did not find use of gnutls_record_send_file().)
Well, this is why I wrote may. I don’t know what Samba does under the hood. It’s not a 5 minute problem to find out and determine whether the hardware is too slow or if the software could use the hardware better. I also support faster CPU.
Putting my pontification hat on, maybe consider separating demanding server duties and routing/firewalling either on dedicated devices or use sufficiently powerful VMs to on beefy hardware if you want to consolidate hardware.A dual core a9, while clearly one of ARM’s better cores*, without encryption offloads/assists might be better used as pure router. Personally I do not expect the next omnia to beat x86 solutions if the goal is to run heavy VMs, but I might be wrong.
*) No sarcasm the a9 IMHO is an excellent core, but it was introduced 2007 according to wikipedia, that it might not fully deliver on a demanding 2022 workload should not be held against it.
NXP, octa-core, 10Gbps?
It looks like they’ll be using an NXP LS2088A:
Features
Core Complex
Eight 64-bit Cortex-A72 CPUs
Up to 2 GHz,
Clusters of two cores sharing 1 MB L2 cache
1 MB L3 platform cache
Two 64-bit DDR4 SDRAM memory controllers
ECC and interleaving support,
Up to 2.1 GT/s
Networking Elements
Wire Rate I/O processor, featuring:
8x 10 GbE
8x 1 GbE
L2 switching on the Ethernet interface
XAUI/XFI/KR and SGMII
MACSec on up to four 1/10 GbE
2x SATA 3.0
4x PCIe Gen 3 controllers
SR-IOV support, Root Complex
Accelerators and Memory Control
Accelerated I/O processing (AIOP) at up to 20 Gb/s packet processing
20 Gb/s SEC crypto acceleration
10 Gb/s Pattern Matching Engine
20 Gb/s Data Compression Engine
One secondary 32-bit DDR4 memory controller for the accelerated packet processing layer, up to 1.6 GT/s
4 PCIe controllers (Gen3) supporting 1x8, 4x4, 4x2, 4x1 lane configurations
Soo… we will inevitably need to upgrade to new hardware to be able to use 64bit applications.
And to be honest. The power of Turris Omnia insufficient to handle traffic in a heavily loaded environment. layer_cake.qos puts it to the limit regularly, affecting both download and upload capacity.
Please offer al least 4 GB of RAM (upgradable) and 4 Cores x86 CPU with SMT for the new model.
Yes, but how else would that work? Sure there are simulators that allow you to execute 64bit code on 32bit CPUs but these are not really intended to run applications. However almost all application to install on an omnia come from the turris repository and there the appropriate bit-ness will be compiled so 32/64 bit is mostly a non-issue (with the exception of applications that drop their 32bit support and require a 64bit OS and CPUs).
In my tests in the past I got my turris omnia to traffic shape with cake up to 550/550 Mbps with bi-directionally saturating traffic, which is pretty impressive. Sure there more stuff you run on the omnia in addition (looking at you pakon ) the fewer CPU cycles are available for SQM eating into the shapeable throughput a bit but in my case the omnis does fine with PPPoE, NAT, firewalling, pakon, smb, SQM on a 116/37 Mbps link. With >= 4 A72 cores the expected successor to the omnia should be able to traffic shape 1/1 Gbps relatively easily, but I am not sure whether these CPUs will be able to run cake on those 10/10 Gbps links that enthusiasts will be using in the near future. (Personally I do not care all that much, before switching to the omnia I was shaping my 116/37 link down to 49/30 as that was the most my old router could reliably traffic shape with SQM, and in my testing 49/30 with SQM was more usable than 116/37 without, YMMV).
The “modular RAM up to 64 GB” part in the announcement indicates that you will be able to upgrade your RAM yourself (and it is possible they sell the device “clean” that is without any RAM).
Also it is pretty much a “done deal” that they will use an ARM CPU, so no x86, sorry. IMHO that is not really a problem modern ARM cores are pretty decent (it is not ARM’s fault that most consumer routers switched from really old and (by modern standards) weak MIPS cores to rather underwhelming ARM SoCs*). A72 as seen in the raspberry pi4B is a pretty decent core (just like the original omnia’s a9) that gets a lot done with low power demands, but like the A9 it will not be the newest nor fastest core out there once the omnia hits the store shelves.
*) Most CPE manufacturers seem to bet on using acceleration engines in their SoCs/SDKs to make up for somewhat under-powered (but likely energy-efficient) CPUs…
and run as perfecly as i just can not update to 6.x because of an unresolved bug in pkgupdate process 
, 
and take a 
(mint from your own production) 
) the fewer CPU cycles are available for SQM eating into the shapeable throughput a bit but in my case the omnis does fine with PPPoE, NAT, firewalling, pakon, smb, SQM on a 116/37 Mbps link. With >= 4 A72 cores the expected successor to the omnia should be able to traffic shape 1/1 Gbps relatively easily, but I am not sure whether these CPUs will be able to run cake on those 10/10 Gbps links that enthusiasts will be using in the near future. (Personally I do not care all that much, before switching to the omnia I was shaping my 116/37 link down to 49/30 as that was the most my old router could reliably traffic shape with SQM, and in my testing 49/30 with SQM was more usable than 116/37 without, YMMV).
