Hi everyone after reading what I could and tried everything I could. I am looking for help to restore my DL2100. I found a tut for pr4100 “How to Boot from a USB Stick”
(How to Boot from a USB Stick), but I was wondering anybody can help me restore my NAS dl2100.
the main issue after 3 years with a good run my NAS stopped working and the Blue LED is flashing slowly. 4 seconds and 40 seconds will not work.
Thanks for the tip. But could I try anything Else or try the USB Boot before I loose hope on this NAS. If so how do I recover my data. Should I make a Linux partition on my PC I don’t know really.
Thanks I recovered my data.
Now if you believe it’s the CPU. Can I just change the motherboard or CPU. It’s possible, did anybody have done it before.How hard is it! But again how do I make sure it’s the CPU.
I thought replacing the CPU on NAS is like replacing it on a PC?
Other possibility, is to purchase another motherboard that can fit DL2100 and can install a free NAS. Anybody
FWIW, neither the DL2100 nor the DL4100 are susceptible to the Intel Atom issue, due to the fact that the affected pins (lpc_clk0, lpc_clk1) are not connected to anything in these designs.
The Sentinel DX4200 from which the DL4100 and DL2100 were derived, is vulnerable since those pins are connected and used in the design.
My understanding is that there is no way to determine whether a given chip is vulnerable, without removing the heatsink and physically examining the marking on the pkg substrate, since Intel didn’t bother to update any internal registers to indicate whether the chip is susceptible or not. I suppose that you could deduce it by examining the mfg date of the NAS itself. Anything prior to 2017, definitely is vulnerable… if the pins are being used.
I gleaned this fact when I removed the CPU from a DL2100 circuit board and noted that balls AG51 (lpc_clk0) and AM49 (lpc_clk1) had no traces connected to them.
Later when installing that CPU onto a DL4100 circuit board, I observed that it too, had no traces connected to those pins.
While it’s certainly possible to create a through hole via straight down from a solder ball pad into another layer, it was easy to see on this board, that was not done. That is, there was no drill hole in the middle of the ball pads for these 2 pins. It’s generally bad practice to do that, since solder wicking into the via, makes it hard to control the solder to ball contact. It is sometimes done on super dense layouts, but not here. In this design, every place a via was needed to another layer, there is a very short trace from the ball pad to the via.
Additionally, tdr measurement on the ball pads in question, showed no connect, as did simple diode mode test.
There is no connection to the lpc_clk0 or lpc_clk1 pins on this board design.
Hi User000001,
thanks for your very helpful reply.
So if the LPC Bus / Clock is not used I have to assume that the clock is still needed internally inside the CPU to work, right?
Do you were succesful with replacing the whole CPU?
From a system architecture standpoint, the LPC clock is not used for anything except the Low Pin Count bus. As such, we could conclude that since there are no connections to the LPC clock or the other pins of the bus, that it’s not needed.
There is a remote possibility that it could still be needed though. If the bios tries to query to see what might be present on the LPC bus, the clock could be a problem. It all depends on how Intel designed this I/O. If they have a driver circuit that sends the clock out continuously and the bios decides to look for one thing or another on the LPC bus, if circuit is designed to loop back the clock directly from the pin to internal circuitry, that needs to see it, then that could cause a hang condition. That is, assuming that the LPC clock pin driver has degraded sufficiently.
The fix of course, barring reprogramming the bios to not go anywhere near the LPC, would be to install a pull-up resistor on the circuit board. Unfortunately, this is well nigh to impossible on the DL4100 or DL2100 since there are no traces from the LPC clock pin. There is certainly a copper pad that the pin is soldered to, but no actual traces. You’d have to remove the CPU and somehow add a rework wire to a pull-up resistor. If you’re going go to the trouble of removing the CPU, you should just replace it with a later stepping of the CPU where this issue is fixed. Alternatively, if you’re really good and can locate the precise coordinates of the pin from the other side of the circuit board, you might be able to drill a hole straight down until you hit solder and then somehow attach a wire to a pull-up resistor. I might try that for a laugh, if either of my DL4100s goes down.
It was much easier to add the pull-up resistor to my 2 DX4200s since there are traces from the LPC clock pin on the DX4200s.
Keep in mind that I’m only speculating as to the possibility of the CPU needing the LPC clock looped back from the pin itself. I don’t actually know if it does or does not. Intel could probably tell us, but they don’t seem to want to talk about it.
Hi,
thanks for your reply. I sacrificed my dl2100 to get a better understanding. Think I found the right ball. But tbh, it’s a real pain with the datasheet (https://www.mouser.com/datasheet/2/612/atom-c2000-microserver-datasheet-334978.pdf). A simple ball map where each ball just have a the number on it… . Anyway, had the same idea with drilling a hole. Not sure if there is an massive ground plane in one layer but I will check that. I now try to adopt the position of the hole to the DL4100. Sadly the CPU is angled on the dl2100…
As you already know, the trick here is to find a reference point that is present on both sides of your DL4100 board. Then you can measure the x/y offset on the component side of the board and use that to establish the drill point on the back side of the board.
What is the board number of your DL4100? It should be a number like 4065-705177-001 or similar. If yours is the same as one of the DL4100s that I have, I might be able to make some x/y measurements for you.
I looked at one of my DL4100 boards (4060-705194-000) and I can establish the offset needed, but looking at this under x-ray, this looks to be a 10 layer board and aside from the expected power and gnd planes, there are also traces passing directly under the place where you need to drill. This is for LPC_CLK_0. There is also an LPC_CLK_1. I don’t know if that one is needed to be pulled up too. I doubt it considering it wasn’t needed for the similar Synology NAS fix.
So, for this board, there isn’t any way that I can see to connect a pull-up resistor to LPC_CLK_0 short of desoldering the CPU itself and somehow connecting a rework wire to the pad on the board and then re-installing the CPU.
Hi,
sorry for delayed response. Oh wow, great you have the possibility to even x-ray this. With 10 layers, changes are pretty high to either cut or touch any trace within the board. That’s sad news as we have another two boxes trashed… But yep, with this information I will give up at this stage. Also not sure about LPC_CLK_1, on other affected NAS, the header / tp often is referenced to _0 but maybe they are connected together. Never checked this.
For what it’s worth, examining my old DL2100 board, under x-ray, it would appear that it is possible to drill the board and not hit anything, as the DL2100 board is only a 6 layer board and I couldn’t see any traces in the way. If anyone wants to give this a try, I can workout the measurements for the place to drill from the back side of the board.
I can’t stress enough, the delicate nature of attempting this. Were I to attempt it, I’d probably put a dvm in diode mode (negative) to the spindle of the drill press and positive on the board ground. Drill very slowly until you see some continuity that doesn’t look like ground or power. Then stop and poke a wire down through the hole to make contact with the CPU pin.