QNAP QXG-2G4T-I225 ?During their trundling around the webz I don't suppose anyone has seen an actual 4 port Intel 2.5Gb PCIe card, have they? I've seen Realtek 4 x 2.5Gb ones but not Intel.
That'll do, bloody well done!QNAP QXG-2G4T-I225 ?
What about this one @dums? https://www.aliexpress.com/item/1005004678867769.htmlQNAP QXG-2G4T-I225 ?
Features
• Supports PCI Express revision 2.0
• Support 2.5G and 1G Lite mode
Would probably do the trick, but I wouldn't call it cheap. Also I've never heard of IOCREST brand (but this should not be a problem for bunch of guys that buy TOPTON routersWhat about this one @dums? https://www.aliexpress.com/item/1005004678867769.html
What does 1G Lite mean?
1G lite is a mode implemented by Realtek so that Gigabit NICs can connect at 500Mb/s using just 2 pairs in cat5 or newer.What about this one @dums? https://www.aliexpress.com/item/1005004678867769.html
What does 1G Lite mean?
7.12. Giga Lite (500M)
The RTL8111H/RTL8111HS supports Giga Lite (500M) mode that allows two link partners that both
support 1000Base-T and Giga Lite mode to transmit at 500Mbps data rate if only two pairs (AB pairs)
can be detected in the CAT.5 UTP cable. This feature is a Realtek proprietary feature and it conforms to
the 802.3az-2010(EEE) specification.
So how does fit in with our “io-crest” NIC that is meant to be using Intel i225 version 3 chips?1G lite is a mode implemented by Realtek so that Gigabit NICs can connect at 500Mb/s using just 2 pairs in cat5 or newer.
From RTL8111 datasheetCode:7.12. Giga Lite (500M) The RTL8111H/RTL8111HS supports Giga Lite (500M) mode that allows two link partners that both support 1000Base-T and Giga Lite mode to transmit at 500Mbps data rate if only two pairs (AB pairs) can be detected in the CAT.5 UTP cable. This feature is a Realtek proprietary feature and it conforms to the 802.3az-2010(EEE) specification.
I wonder if any switch ( even unmanaged ) support this ...
Probably "copy+paste" from the specs of the other model that uses RTL8125B chips.So how does fit in with our “io-crest” NIC that is meant to be using Intel i225 version 3 chips?
Wait, want to make sure I understood you right.I've wanted to install M.2 -> 2xSATA converter in my V3 box. It actually worked just fine but blocked NVME port, so I've tried to use similar ribbon cable from aliexpress:
And while BIOS was still able to detect SATA disks connected to SATA adapter, box would not boot anymore from attached SATA diskSo my experience with ribbon M.2 extender was negative.
Also I discovered that it would be difficult to cram extra cable/adapter into already limited space inside the box (especially if you're adding extra fan). I think adding riser studs on 3D printing custom bottom lid would be better options.
This one looks interesting. Anybody knows if similar I226V based NICs are out?What about this one @dums? https://www.aliexpress.com/item/1005004678867769.html
What does 1G Lite mean?
In my case I'm positive it was ribbon cable:Wait, want to make sure I understood you right.
With two M.2 NVMEs, second one installed with ribbon cable extension, your third SATA SDD shows up in BIOS but you can't boot off of it?
Could you confirm all three drives were visible in BIOS?
From why I recall, on certain machine second NVME and SATA ports of often multiplexed, meaning they share the same high-speed lane. So this could be BIOS related issue not the ribbon cable.
Got it, that makes lot more sense now.In my case I'm positive it was ribbon cable:
Case 1:
I've removed NVME drive and installed this card in slot "under" NVME:
Connected SATA ssd to one of SATA ports on adapter card.
Was able to install and boot Ubuntu from ssd (NVME still removed since SATA adapter is now blocking second M.2 port).
Case 2:
NVME drive still removed, but SATA adapter card is connected via ribbon cable. BIOS can see ssd drive connected to adapter, but Ubuntu crashes during the boot.
Case 2 is identical to Case 1 with exception of ribbon cable. Case 2 crashed during the boot.
Going back to Case 1 (no reinstall), Ubuntu boots just fine.
I've had/tested just one cable, so it might be just bad luck with poor QC from our China friends.
I have installed sunon MF40101V21000UA99 fans in 2 of the units that will run at places without AC.Anybody here done experiments to quantify temperature delta with and without the fan?
root@thu-gw1:~ # sysctl -a | grep temperat
hw.acpi.thermal.tz0.temperature: 27.9C
dev.cpu.3.temperature: 49.0C
dev.cpu.2.temperature: 49.0C
dev.cpu.1.temperature: 48.0C
dev.cpu.0.temperature: 48.0C
root@thu-gw1:~ # smartctl -A /dev/nvme0 | grep Celsius
Temperature: 67 Celsius
root@a-vpn1:~ # sysctl -a | grep temperat
hw.acpi.thermal.tz0.temperature: 27.9C
dev.cpu.3.temperature: 43.0C
dev.cpu.2.temperature: 43.0C
dev.cpu.1.temperature: 41.0C
dev.cpu.0.temperature: 41.0C
root@a-vpn1:~ # smartctl -A /dev/nvme0 | grep Celsius
Temperature: 50 Celsius
What kind of samsung drive are you using?If you cram some samsung NVME in there, which seem to be running A LOT hotter than other vendors, you might *have* to use a fan, especially because samsung starts thermal throttling relatively early and aggressively...
in those networking appliances? none. *because* they run _much_ hotter and have much higher power consumption. I'm using WD SN530s in those topton units.What kind of samsung drive are you using?
Thanks, this is super useful info.I have installed sunon MF40101V21000UA99 fans in 2 of the units that will run at places without AC.
I had to switch to FreeBSD for now, due to botched ACPI on the BIOS side that endlessly fires ~9000 interrupts per second on OpenBSD (GPE L6F - a problem relatively common with other cheap-ass mainbords/vendors like ASrock...). FreeBSD seems to have mitigated this already; otherwise it would have been very simple to just override the AML.
Another side effect of running FreeBSD (apart from not having 70-80% interrupt load on one core): thanks to powerd the CPU is running on much lower average clock (I already reduced the graphics unit to 200Mhz at the BIOS; woud have completely disabled it if possible...) and thus lower temperatures compared to OpenBSD.
Unit without fan:
with fan:Code:root@thu-gw1:~ # sysctl -a | grep temperat hw.acpi.thermal.tz0.temperature: 27.9C dev.cpu.3.temperature: 49.0C dev.cpu.2.temperature: 49.0C dev.cpu.1.temperature: 48.0C dev.cpu.0.temperature: 48.0C root@thu-gw1:~ # smartctl -A /dev/nvme0 | grep Celsius Temperature: 67 Celsius
Both are running at ~23°C room temperature.Code:root@a-vpn1:~ # sysctl -a | grep temperat hw.acpi.thermal.tz0.temperature: 27.9C dev.cpu.3.temperature: 43.0C dev.cpu.2.temperature: 43.0C dev.cpu.1.temperature: 41.0C dev.cpu.0.temperature: 41.0C root@a-vpn1:~ # smartctl -A /dev/nvme0 | grep Celsius Temperature: 50 Celsius
So its a ~7° difference for the CPU and ~15-20° for the NVMe.
I actually only cared about the NVMEs - the CPU would be perfectly fine at 70° or 80° as there is still plenty of headroom to the actual TJmax of the SoC (105°C), so there is absolutely no need to add a fan just for the CPU.
If you cram some samsung NVME in there, which seem to be running A LOT hotter than other vendors, you might *have* to use a fan, especially because samsung starts thermal throttling relatively early and aggressively...
I am using a 970 Evo Plus and a Seagate BarraCuda 510.What kind of samsung drive are you using?
In traditional useage patterns then I agree that the drives do not get overly hot but in a system where for example you perform a parity check and read the drive from beginning to end in one go, then even connected at Gen3 x1 can cause the drives to bake (although probably not throttle).About your comment on using fan with NVME, these tiny N5105 box only run PCI in x1 Gen3 lane configuration. Do you really think without a fan, drive will go into thermal throttle mode given it's running only at 25% of stated transfer capability?
Ouch on >75C.In traditional useage patterns then I agree that the drives do not get overly hot but in a system where for example you perform a parity check and read the drive from beginning to end in one go, then even connected at Gen3 x1 can cause the drives to bake (although probably not throttle).
My 970 will hit >75deg during the 30min parity check and the 510 will tend to settle in at 65deg.
Both drives have heatsinks attached but in a case with only passive cooling, heatsinks behave as the name suggests. They only act as a sink for the thermal load. The peak temp is within 1-2deg with/without, but without a heatsink you hit the peak in a few min. With a heatsink it takes 10-15min.
Just my 2c.![]()