FORUMS

PSA: You Can Optimize Your Note 4’s Recents Menu & RAM

The Note 4 never had the fastest Recents Menu, and despite its 3GB of RAM, … more

XDA Picks: Best Apps of the Week (July 25 – Aug 1)

Apps are at the front and center of any smartphone experience, and with over a … more

Voices Of XDA: Orbiting The Earth With Android

Editor’s note: This week’s feature has been written by forum … more

CloudPlayer: DIY HiFi Music Streaming Solution

In our Helpful Guide to Music Streaming Services, we mentioned several different services … more

[TOOL] G2x Nullifier AROMA [UPDATED!] 2 August, fully customizable: Made with AROMA!

1,092 posts
Thanks Meter: 541
 
By Волк, Senior Member on 17th March 2012, 11:24 PM
Post Reply Subscribe to Thread Email Thread
23rd April 2012, 01:41 AM |#301  
mansa_noob's Avatar
Senior Member
Flag Markham
Thanks Meter: 153
 
Donate to Me
More
Quote:
Originally Posted by Волк

Please read above post, again and more carefully. Once you're done, read it again, then stop and think about how quickly one would accumulate 1k writes, even merely on stock. And, yes it would take a damn long time to accumulate 1m writes.

Yes, I agree. But if the flash in the g2x is the 100k type then over time cache would suffer.

Lets do some math shall we: first page of cache written all the time, say 50 times a day would result in failure of that page in 2000 days. If that is moderate use the heavy user could easily achieve that in 1/10th the time. 200 days is < 1 year. The page will fail, but a smart os would just reallocate and hence give longer time.

As I said, I used to work with these things... In heavy use you will see failures!

BTW I read what you wrote. It was very nice.

Sent from my LG-P999 using XDA
 
 
23rd April 2012, 02:04 AM |#302  
Senior Member
Thanks Meter: 48
 
More
Quote:
Originally Posted by mansa_noob

Yes, I agree. But if the flash in the g2x is the 100k type then over time cache would suffer.

Lets do some math shall we: first page of cache written all the time, say 50 times a day would result in failure of that page in 2000 days. If that is moderate use the heavy user could easily achieve that in 1/10th the time. 200 days is < 1 year. The page will fail, but a smart os would just reallocate and hence give longer time.

As I said, I used to work with these things... In heavy use you will see failures!

BTW I read what you wrote. It was very nice.

Sent from my LG-P999 using XDA

you would only see 100k on SLC, not MLC, as most of these NAND should be.
http://www.anandtech.com/show/5067/u...ing-tlc-nand/2
at the 2x node, MLC only lasts for about 2k-3k cycles, and that's on average. the worst crap that might only last for 1k or less cycles go to the cheap flash drives, the middle crap might go to the sd cards. The best ones (the ones that really last 2k-3k or marginally more) go to SSDs. I should know, I work in a company that makes these.
The flash won't keep writing to the same page every time, even though it would think it was writing to the same page of cache like your example , as the controller inside should have write leveling logic to it, and will spread the writes evenly around. But you are still correct, under heavy use the flash will definitely see failures, and will happen sooner than one might expect.
Last edited by menting; 23rd April 2012 at 02:10 AM.
The Following User Says Thank You to menting For This Useful Post: [ View ]
23rd April 2012, 03:04 AM |#303  
mansa_noob's Avatar
Senior Member
Flag Markham
Thanks Meter: 153
 
Donate to Me
More
Quote:
Originally Posted by menting

you would only see 100k on SLC, not MLC, as most of these NAND should be.
http://www.anandtech.com/show/5067/u...ing-tlc-nand/2
at the 2x node, MLC only lasts for about 2k-3k cycles, and that's on average. the worst crap that might only last for 1k or less cycles go to the cheap flash drives, the middle crap might go to the sd cards. The best ones (the ones that really last 2k-3k or marginally more) go to SSDs. I should know, I work in a company that makes these.
The flash won't keep writing to the same page every time, even though it would think it was writing to the same page of cache like your example , as the controller inside should have write leveling logic to it, and will spread the writes evenly around. But you are still correct, under heavy use the flash will definitely see failures, and will happen sooner than one might expect.

Awesome. I was hoping an expert would step in.

+1. And you are absolutely right that logic is required to ensure same bits are not stressed over and over again!

IMHO this should settle the debate

Sent from my LG-P999 using XDA
23rd April 2012, 03:43 AM |#304  
Волк's Avatar
OP Senior Member
San Francisco
Thanks Meter: 541
 
Donate to Me
More
Ok, enough thread jacking. I could care less where you work or how old you are. Stating how long you've been around or where you work is just a sign of arrogance. Nand failure from too many writes is a myth with no proof, I see a bunch of if's, if this if that, like I said if it were feesable we'd have seen it spread widely on other devices by now, with or without the black-listing of bad sectors.
23rd April 2012, 04:10 AM |#305  
Senior Member
Thanks Meter: 48
 
More
Quote:
Originally Posted by Волк

Ok, enough thread jacking. I could care less where you work or how old you are. Stating how long you've been around or where you work is just a sign of arrogance. Nand failure from too many writes is a myth with no proof, I see a bunch of if's, if this if that, like I said if it were feesable we'd have seen it spread widely on other devices by now, with or without the black-listing of bad sectors.


WOW...you think it's a myth? Hardly.

I have a dead first generation (but not released) dead SSD from my company in my possession due to firmware not write leveling sector 0 and it wore out. The team leader for that project told me that all the first generation hard drives died about the same time as mine. He even checked it through our internal tools. That proof enough?

Want more proof? If it is a myth, why do SSD hard drives have a write endurance spec? Why do SSDs bother to overprovision? Why are NAND manufacturers worried about write endurance on new process nodes? I mean, if it's a myth, it shouldn't be something to worry about right? Apparently you know more about NAND than those manufacturers.
Here's even more proof. http://www.xtremesystems.org/forums/...e-25nm-Vs-34nm

I'd get you the write endurance distribution graphs from work if I could, but those numbers are proprietary. I can guarantee you though, it DOES wear out from writes.

To me, arrogance is one who doesn't work in the industry saying the person who does work in the industry is wrong without any proof.
Last edited by menting; 23rd April 2012 at 04:14 AM.
The Following User Says Thank You to menting For This Useful Post: [ View ]
23rd April 2012, 04:12 AM |#306  
Волк's Avatar
OP Senior Member
San Francisco
Thanks Meter: 541
 
Donate to Me
More
Didn't I *just* say something about staying on topic?
23rd April 2012, 04:17 AM |#307  
Senior Member
Thanks Meter: 48
 
More
yes you did, and if you stopped at that I wouldn't have replied. But insulting someone's knowledge by saying a fact is a myth went a bit too far for me.
23rd April 2012, 04:19 AM |#308  
Alakratt's Avatar
Senior Member
Thanks Meter: 10
 
More
On my next flash I'll try this tool, haven't tried it since I get a little freaked out because of the word 'null', I am programmer so...yeah. I've looked most of the pages on this thread, I admit not all of them, I have a question; I know this wipes what needs to be wiped for ROM flashing, but does it wipe only that or the whole SD? Also, does it also wipe the internal SD?
23rd April 2012, 04:19 AM |#309  
owain94's Avatar
Recognized Developer / Contributor
Flag The Hague
Thanks Meter: 6,061
 
More
What is exactly the point of "nullifing" the cache partition on a device where the cache partition isn't used??

Sent from my LG-P990 using xda premium
The Following User Says Thank You to owain94 For This Useful Post: [ View ]
23rd April 2012, 04:36 AM |#310  
da-pharoah's Avatar
Recognized Contributor
Seacoast near Joppa
Thanks Meter: 1,605
 
Donate to Me
More
Quote:
Originally Posted by owain94

What is exactly the point of "nullifing" the cache partition on a device where the cache partition isn't used??

Sent from my LG-P990 using xda premium

? ok and how is it not used? and if it "isn't used" why would we need to format it upon doing any update on something you say 'isnt used?"
23rd April 2012, 04:54 AM |#311  
owain94's Avatar
Recognized Developer / Contributor
Flag The Hague
Thanks Meter: 6,061
 
More
Quote:
Originally Posted by da-pharoah

? ok and how is it not used? and if it "isn't used" why would we need to format it upon doing any update on something you say 'isnt used?"

Well the cache partition is normally used for the dalvik cache
But since our cache partition is so small we store the cache on the data partition (/data/dalvik-cache)
By defining dalvik.vm.dexopt-data-only=1 in the prop files or even the inits

And cache doesn't need to get wiped...
You can even boot your phone up with a corrupted (so unmounted) cache partition (yes is happened me once caused by that buggy touch recovery)

Also davik cache doesn't needs to be wiped on a update since android checks the dalvik and rebuild it when it is not matching the the system apps or frameworks hashes

So yeah cache partition is useless on this device

I have resized my partitions using NVflash too use the cache partition for the dalvik cache that is why i know it is normally useless

I don't know what you store on it but the 2 folders (lost+found and recovery) are not really needed for the system

Flashing new ROMs only need a full wipe if you change between major releases like between gingerbread and ics

I never wiped data / dalvik / cache since i am running ics and still no failures or any problems at all

Asked arcee once about it he also never wiped dalvik or cache in his device

You can search for the post google did make about how dalvik works in android if you do not believe me

Sent from my LG-P990 using xda premium
Last edited by owain94; 23rd April 2012 at 05:06 AM.
The Following User Says Thank You to owain94 For This Useful Post: [ View ]

Read More
Post Reply Subscribe to Thread
Previous Thread Next Thread
Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes