Lenovo Thinkcentre/ThinkStation Tiny (Project TinyMiniMicro) Reference Thread
- Thread starter Parallax
- Start date
-
- Tags
- lenovo tiny tinyminimicro
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
your riser looks great. how's it working so far? what speeds are you getting off those nvme's? what's heat like for those drives up there? sent you a pmProduction batch of the dual-M.2 risers has arrived!
View attachment 36695
View attachment 36696
View attachment 36697
The details: this will only work in an M920q or an M920x, and provides two extra M.2 connectors (each with 4x PCIe lanes) plus an open-backed x4 slot for a NIC and a 12-volt fan header. With an M920x (or an M920q modded to be an M920x), you'll get a total of four M.2 slots, or three with a regular M920q. It does require some soldering, two 0402 resistors need to be installed on the motherboard to enable PCIe bifurcation.
I'm selling 'em for $50 if you want to print the mounting bracket yourself or $55 with the brackets included. Send me a PM
I have a m720q and a p330 and with a bios update, running headless they reach C9, C8 with a connected PIKVM.
Being limited to C3 sounds like a problem with one of the extra devices. Like PCIe cards not supporting ASPM, do you have any?
In my case, both Mellanox ConnectX-3 and Google Coral M2 TPU caused this behavior.
I switched a CX3 to an Intel X710-DA2, and now my P330 system reaches C7.
Bios update to something custom? Or just the latest stock bios?
I do have a 10Gb SolarFlare NIC, but it says it supports ASPM. I'll have to try taking it out. It was so much lower power than the other 10Gb NICs that I'm not sure I'd do better swapping it out.
What does your M720q idle at?
Edit: I took the NIC out and idle power dropped from 10W to 3.6W!! :O I think 4W of that should be the NIC itself, but still, with the X710s I should be able to hit 6.9W, maybe lower if I shut off some ports.
Last edited:
The latest official bios.
In my case adding X710-DA2 added around 3W in total to power consumption at idle. P330 tiny running headless idles at 6W for me: post-402456, post-416869
Last edited:
I'm getting about 2 GByte/sec read speeds from the NVMEs in the riser. Haven't benchmarked writes but I imagine they're similar. The temperatures are all around 40 celsius at idle, climbing to around 70 during a zfs scrub. No complaints so faryour riser looks great. how's it working so far? what speeds are you getting off those nvme's? what's heat like for those drives up there? sent you a pm
HiHey all, just wanted to send this along.
I've got 4 of the M90q Gen3 65w i5-12500's with x16 risers and have added the Dell branded Mellanox Connectx-4 LX cards to mine.
Initially they were just floating around the back but decided to have my hand at modifying a 3D model someone made for a different card to see if I could make one compatible with these.
The result is a fully functional baffle that holds the card firmly without any clearance issues and also uses the original screws from both the card, and the chassis.
Here you go.
Lenovo M90q Gen3 ConnectX-4 LX (Dell CX4121C) Dual 25G Baffle Remixed by Chewza
I have a m90q gen3 12500 and the same Mellanox Connectx-4 LX (Dell 0MRT0D) inserted in x16 raiser. Have you measured the power consumption in this configuration?
I have without SFP card - In POWERTOP main heatpack C9-C10 - consumption - 4-6W
When inserted Mellanox Connectx-4 LX main heatpack C2 - consumption 24-26W.
Please see what C-state you have in powertop
I suspect a similar problem as here
Using x16 pcie slot disables low package c states ASPM. Alderlake and Raptorlake
Hello, I have a question regarding using INTEL X710-DA2(pcie gen3, x8) in pcie slot 1 which connects directly to cpu. Is it a design choice to prevent cpu to go to lower package C states when using x16 slot connected directly to cpu? I have a Asus B760 motherboard with an i3-13100, where if...
community.intel.com
The same card in m90q gen1 10500 - C6-C7 and power consumption is 13-15W.
Can anyone suggest anything who has a m90q gen 3?
Last edited:
your riser looks great. how's it working so far? what speeds are you getting off those nvme's? what's heat like for those drives up there? sent you a pm
+1 for the riser. I got similar - 2.6 GBps read, writes were about 1.3 GBps, limited by the drive itself. Keep in mind that the slot connected to the PCH will be limited by the fact that the PCH only has 4x lanes to CPU. So if you have the M.2 on the bottom (and the second one, if you do the mod), all 3 of those drives will share that bandwidth. The M.2 connected directly to the CPU will always get the full speed though, subject to some processing throughput of the kernel.I'm getting about 2 GByte/sec read speeds from the NVMEs in the riser. Haven't benchmarked writes but I imagine they're similar. The temperatures are all around 40 celsius at idle, climbing to around 70 during a zfs scrub. No complaints so far
I need some help with bios recovery.
I bought a m920q Mainbord and the bios is totally ****ed.
The pre owner tried to manipulate the bios and did a bad dump then he modified it and programmed it back. Now it's impossible to boot, no led no fan nothing.
The board has two bios chips how can I bring it back in a working condition?
I read about it but I can't find a solution and I don't want to buy pre programmed bios chips.
Much appreciated for helping me out.
I bought a m920q Mainbord and the bios is totally ****ed.
The pre owner tried to manipulate the bios and did a bad dump then he modified it and programmed it back. Now it's impossible to boot, no led no fan nothing.
The board has two bios chips how can I bring it back in a working condition?
I read about it but I can't find a solution and I don't want to buy pre programmed bios chips.
Much appreciated for helping me out.
I'm getting about 2 GByte/sec read speeds from the NVMEs in the riser. Haven't benchmarked writes but I imagine they're similar. The temperatures are all around 40 celsius at idle, climbing to around 70 during a zfs scrub. No complaints so far
Haven't tried it. But you could try putting the jumper on the mainboard to 1-2 position (CLR_CMOS) which is apperantly also MAINT mode. Then the system should be able to update the bios from either CDROM or bootable USB. Lenovo offers a ISO or USB BIOS package on their support page.
See reference information here: English Community-Lenovo Community
There is only one BIOS flash chip, the other is the EC/Intel ME flash chip.
You need to find good known dumps from both chips, most likely you can request them at the bad-caps and the win-raid forums and someone will have it.
Then you need to desolder each chip and program then, programming in place almost always leads to a corrupted flash since the other components interfere with the data lines, plus you aren't just putting 3.3V into parts of the motherboard.
OK this sounds good, I have a working m920q Mainbord here.
Soldering is the smallest problem
Have all tools here and will mod it to m920x if it's working.
How can I add the right serial number and Mac address for the ethernet port?
Any guide for this?
I'm having trouble getting past GRUB on my Tiny m90q without a keyboard plugged in. This might be a GRUB/TrueNAS problem, but it might be related to a BIOS setting which is why I'm asking here. Anyone else have trouble running TrueNAS headless?
If I boot with a keyboard plugged in, I get to GRUB and the normal 10s countdown will occur. If I unplug the keyboard during this, the countdown disappears and it stays on the grub screen. If I boot without a keyboard installed at all, it wont display a countdown at all. When I plug a keyboard in and hit 'enter' it will boot just fine.
In the BIOS security secure boot is off, not sure if USER vs Deployed mode makes a difference. There are a lot of security options in the BIOS. Any help would be greatly appreciated!
If I boot with a keyboard plugged in, I get to GRUB and the normal 10s countdown will occur. If I unplug the keyboard during this, the countdown disappears and it stays on the grub screen. If I boot without a keyboard installed at all, it wont display a countdown at all. When I plug a keyboard in and hit 'enter' it will boot just fine.
In the BIOS security secure boot is off, not sure if USER vs Deployed mode makes a difference. There are a lot of security options in the BIOS. Any help would be greatly appreciated!
For anyone who wants to use a 3.5" HDD with their M920q, I put together a quick write-up of an easy mod I did to get 12V to the drive (the stock connector only supplies 5V). There's also some info in there about how much current the 12V and 3.3V supplies can drive, as well as possible soldering points for both, which could be used for the HDD or something like an extra fan.
github.com
GitHub - kaysond/m920q-3.5in-drive-mod: Simple mod to use a 3.5" drive in a Lenovo M920q
Simple mod to use a 3.5" drive in a Lenovo M920q. Contribute to kaysond/m920q-3.5in-drive-mod development by creating an account on GitHub.
github.com
The chip is soldered to the motherboard, the clip will power anything that is connected to the 3.3V rail.
The data, cs, rd, etc pins are also directly connected to the chipset, so you will be "forcing" voltage into all those components when using a clip.
I mean, I guess... but since you're connecting the clip directly to the chip and all of those components are on the other side of the connection between the chip and the clip, I don't think you have to worry about interference from those components to the point where unsoldering it from the board is the better answer.
Don't think I've run into an issue caused by this in the past, but who knows... in any case it's always been easier/safer to just reflash with the clip then getting involved with soldering chips.
Personal opinion of course.
Anyways, this is probably an off-topic philosophical tangent I guess.
That's not how it works*. Everything is shorted together, and with power to the chipset, the chipset could be trying to drive any of the pins to a different voltage than the programmer. Both will have some maximum rated current that they can source / sink out of a pin, but unless they have active current protection mechanisms, which I doubt they do, they will pull / push as much current as they can, even in excess of their ratings. This could easily cause you to burn something out, and could cause flashing issues because the voltages will not be fully driven to supply or ground.
I'm guessing that in reality, the chipset isn't actually active (it might need a supply other than 3.3V), and even if the chipset has power, without other signals from SMBus, etc, it won't actually turn on the outputs. It's still definitely a risk though.
* with very high frequencies , the signals travel in "waves", so there's some truth to the physical location of the components mattering. In practice, however, the routes are short enough, and the clock frequency low enough that in this case it doesn't actually make a difference.