| HDD GURU FORUMS http://forum.hddguru.com/ |
|
| Samsung 870 Evo -- is it a dud model? http://forum.hddguru.com/viewtopic.php?f=10&t=42876 |
Page 2 of 2 |
| Author: | csava [ September 8th, 2023, 19:43 ] |
| Post subject: | Re: Samsung 870 Evo -- is it a dud model? |
fzabkar wrote: csava wrote: However, other brands with a history of poor practices, such as Samsung and Western Digital, might only release firmware updates. These updates could adjust the offset of the temperature sensor (actual temperature 80 degrees Celsius, displayed as 65 degrees Celsius, offset -15 degrees Celsius)... Are you referring to the "composite" temperature? IIUC, this is a value that is derived by computing a weighted average of all the temperature sensors on the drive. As such, it would be very subjective. For example, a manufacturer could lower the reported composite temperature by applying a lower weighting to the flash core without actually falsifying the core temperature reading. No, I'm referring to the SMART reported by the controller, where the C2 option represents temperature. The lower 4 bits represent degrees Celsius, and typically, this temperature refers to the current temperature of the disk controller. A fully functional controller will have a thermocouple installed near the NAND flash on the PCB to measure the current temperature of the NAND flash. Controllers that typically support external temperature sensors will calibrate an option defined by the manufacturer. However, in practice, regardless of which value it is, controller manufacturers can freely change it to a fixed value, modify the offset, modify the multiplier, or determine an upper limit. This also applies to the 09 option reported by SMART. |
|
| Author: | fzabkar [ September 8th, 2023, 20:40 ] |
| Post subject: | Re: Samsung 870 Evo -- is it a dud model? |
I was under the impression that current NAND flash has an on-die temperature sensor. |
|
| Author: | csava [ September 9th, 2023, 8:17 ] |
| Post subject: | Re: Samsung 870 Evo -- is it a dud model? |
fzabkar wrote: I was under the impression that current NAND flash has an on-die temperature sensor. I believe that even the latest NAND flash chips do not have internal temperature sensors. My basis for this is that NAND flash cannot change its state based on temperature readings; this part of the processing is handled by the controller. The controller can adjust the current frequency and power control based on the temperature of the NAND flash. It can also set different values in the chip's RR register based on different temperature ranges. However, I have not seen any method for obtaining the current temperature of NAND flash in the specifications of any manufacturer. But I believe that SSDs, UFS, and eMMCs with BGA packaging may include internal temperature sensors. |
|
| Author: | fzabkar [ September 9th, 2023, 14:07 ] |
| Post subject: | Re: Samsung 870 Evo -- is it a dud model? |
Micron's L06B Fortis Flash datasheet describes SET FEATURES (EFh) or SET FEATURES by LUN (D5h) and GET FEATURES (EEh) or GET FEATURES by LUN (D4h) commands. One of the features is "E7h Temperature sensor". Each bit is roughly 1.2degC. |
|
| Author: | fzabkar [ September 9th, 2023, 16:19 ] |
| Post subject: | Re: Samsung 870 Evo -- is it a dud model? |
Here is a Spectek datasheet with a temperature sensor: http://www.kindlehk.com/updatefile/file/Spectek%20Nand%20Flash%20FBNL05B128G1KDBABJ4.pdf Another one ... https://caes.com/sites/default/files/documents/Datasheet-UT81NDQ512G8_0.pdf |
|
| Author: | csava [ September 9th, 2023, 20:33 ] |
| Post subject: | Re: Samsung 870 Evo -- is it a dud model? |
As you mentioned, Micron's flash memory indeed integrates a temperature sensor within the chip. I had overlooked this previously. I referred to the ONFI 5.1 specifications, which did not mention it as a mandatory requirement. I also consulted Toshiba's BICS3 manual and searched for all temperature-related content but did not find any mention of this feature. However, without exception, they all mentioned the impact of temperature on input-output impedance and provided recommended ZQ calibration reference values at different temperatures. Therefore, I believe that the integration of a temperature sensor in NAND flash is currently a specific implementation of Micron's chips. |
|
| Author: | fzabkar [ September 9th, 2023, 20:40 ] |
| Post subject: | Re: Samsung 870 Evo -- is it a dud model? |
I would heat and cool the NAND chips and monitor the temperature sensor(s). The NVMe spec allows for 8 of them (Get Log Page – SMART / Health Information Log, bytes 200 - 215). |
|
| Author: | csava [ September 9th, 2023, 20:45 ] |
| Post subject: | Re: Samsung 870 Evo -- is it a dud model? |
I am curious whether controller manufacturers are currently utilizing temperature sensor data from Micron NAND flash for any specific purposes. I will keep an eye out for this information. |
|
| Author: | fzabkar [ September 9th, 2023, 22:35 ] |
| Post subject: | Re: Samsung 870 Evo -- is it a dud model? |
This Sandisk patent talks about on-die NAND temperature sensors and proposes a method by which a particular NAND IC can be independently thermally throttled. https://patentimages.storage.googleapis.com/3a/2d/db/cabea4fd3a6b29/US9811267.pdf |
|
| Author: | fzabkar [ September 10th, 2023, 0:17 ] |
| Post subject: | Re: Samsung 870 Evo -- is it a dud model? |
See this thread (WD/Sandisk SN770 SSD): https://superuser.com/questions/1754770/what-are-the-3-temperatures-composite-and-sensors-reported-by-a-wd-black-sn770 WD's Dashboard reports 3 temperatures: https://i.stack.imgur.com/65Wcj.png The drive does not have an external temp sensor: https://www.techpowerup.com/ssd-specs/western-digital-black-sn770-1-tb.d648 It has a single NAND. https://technuovo.com/wp-content/uploads/2022/08/wd_black-sn770-NVMe-drive-1.jpg The implication is that there are two on-die sensors, one for the controller and one for the NAND. If you have an SSD like this, you could allow it sit at a stable idle temperature, then heat or cool the NAND and see which, if any, of the two temp sensors changes. |
|
| Author: | csava [ September 14th, 2023, 9:29 ] |
| Post subject: | Re: Samsung 870 Evo -- is it a dud model? |
https://www.youtube.com/watch?v=7k0yQAvradA I have conducted preliminary testing on Samsung NAND flash known for premature degradation and developed a new read retry mode for them. It is called preliminary testing because the test conditions were not rigorous, and there were not many samples available. However, based on my previous knowledge of Samsung NAND flash, these chips are indeed versions with relatively lower quality in their history. I speculate that the chips used in these flashes fall within the range of fabrication processes between below 19nm and above 14nm. In the era of 3D NAND flash, manufacturers no longer disclose the fabrication processes of individual chips but instead focus on the number of stacking layers. Therefore, my assessment is based solely on my understanding of 2D NAND |
|
| Page 2 of 2 | All times are UTC - 5 hours [ DST ] |
| Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |
|
