Welcome to Jaehyek's Blog!
Here is my front-end learning path-
Open-source NLP
Open-source NLP toolkits like Stanford CoreNLP or OpenNLP
please refer to https://github.com/datanada/Awesome-Korean-NLP
-
eMMC UFS Issues 05
Component Feature
- 8GB eMCP eMMC
- Controller :
- Hy
Failure Symptom
- Using Packed Command while SPO (sudden power off ) caused data disappeared
Failure procedure
- Normal Case
- At first picture, issue packed command
- At second picture, write data seperately
- Crash Case
- At first picture, power-off at command 7
- At second picture, command 6 got no data
Cause
- Normal Case
- First, Packed Data is written to SRAM, not to NAND
- If same address data comes in, merging will be done with existing same address data at SRAM
- and then, written to NAND
- Crash Case
- First, Packed Data is written to SRAM, not to NAND
- The addresses of 5th and 7th data is same, and then, the two data is merged at SRAM
- and then while written to NAND, SPO(sudden power off ) occured
- the 6th data was not written to NAND
-
eMMC UFS Issues 04
Component Feature
- 8GB eMCP eMMC
- Controller :
- Hy
Failure Symptom
- UECC
- EOL ( end of Life )
- system can not boot up
Failure procedure
- Writing a fixed pattern ( all zero pattern ) Continuously
- Almost page blocks have UECC
Cause
- The Randomize Algorithm caused to make failure
- Random seed of each page is assigned to each different value, but fixed
- Writing all “O” pattern data causes writing to same page and same location.
- It make the cell to be faded out
-
eMMC UFS Issues 03
Component Feature
- 16GB eMCP eMMC
- Controller : Venice
- SS
Failure Symptom
- Writing via Command Queuing, and then time-out, No Reset.
Failure procedure
- Writing a mass of data such Video, picture(jpeg)
Cause
- This product use a common buffer for Meta Data and User data Garbage Collection.
- If a common buffer Meta Data was already assigned, additional buffer assignment caused to fail.
-
eMMC UFS Issues 02
Component Feature
- 32GB UFS
- Controller : Siena
- SS
Failure Symptom
- Sudden-Power-Off Test
- Booting fail
Failure procedure
- Check the Boot LU Size ( BOOT1:32MB, BOOT2:8MB )
- Meta Open if the Boot LU Size is over 8MB
- While Meta Open, read the boot data for Read-Reclaim processing for Garbage Collection
Cause
- Before fDeviceInit,
- Race-condition occurs for Buffer allocation between Garbage Collection and Boot Data Read
-
eMMC UFS Issues 01
Component Feature
- 8GB eMCP, TLC
- Controller : VPX
- SS
Failure Symptom
- Can not perform the emmc command “CMD23”
Failure procedure
- Known data write
- ( random write 4K x 3500 times Sleep Resume) x 50K times
- UECC check
- Sequential write
Cause
- was caused by EOL ( End of Life)
- was caused by frequently Random Write
- was caused by frequently MetaData Update
SS Quality Criteria
Test Pattern of DWL @ Samsung(Write Amount Ratio)
Classification Chunk Size Write Amount Ratio Random Write 4K/8K 5% Seq Write 16K/32K~4M 35% Download 4M 60% WAI(Wear Acceleration Index) – Index which indicate Meta lifespan by host write amount
item WAI with Samsung TBW Pattern WAI with 4K Random Write Pattern SLC WAI 0.28 22.96
Recent Posts
- Install Tensorflow and OpenCV on Raspberry PI 3
- Install Tensorflow on Windows
- Tensorflow 0.x => 1.0 Migration Guide
- eMMC UFS Issues 17
- Deep Learning summary from http://hunkim.github.io/ml/ (4)
- Deep Learning summary from http://hunkim.github.io/ml/ (3)
- Deep Learning summary from http://hunkim.github.io/ml/ (2)
- Deep Learning summary from http://hunkim.github.io/ml/ (1)
- Summary of CNN base cs231n.github.io
- BOOK-TCP/IP 쉽게 더 쉽게-03
Categories
Tags