Database Info

CompCache vs. Linux vs. Stock

I've been plugging away for the last several weeks trying to decide if CompCache is making my phone faster or slower. I've always been happy with the pure Linux swap, but everyone seems to be on the CompCache bandwagon.
I have installed compcache many times, with all different sorts of settings. No matter what, it always seems to make my phone slower. I had to prove it. However, subjective statements of opinion are not going to carry any validity in this forum. So this is what I did:
Round 1 CompCache vs. Stock
First, I installed a fresh copy of CM 3.6.7 on both of my G1's. On the black one, I installed CompCache 0.6 via userinit. It also had a Backing Swap (not linux as the video states!) of 32mb. Swappiness was set at 20. NO OTHER SETTINGS WERE TWEAKED.
On the other phone was a pure installation of CM 3.9.7. For the stress test, I alternated between a pre-loaded webpage(Valdosta.edu, which is graphics heavy but also has a very high speed connection) and SHERPA.
I Chose Sherpa because it tends to cause the browser to reload by pushing it out of memory or cache.
I filmed the whole thing and you can view it here.
I know the filming kind of sucks, but bear with me. In short, THE PHONE WITHOUT COMPCACHE WAS NOT ONLY FASTER, BUT WENT LONGER WITHOUT NEEDING A RELOAD.
Round 2 Stock G1 vs Linux swap enabled
For the next test, I used the same two phones and ROMs. On the Black phone, ONLY THE LINUX SWAP WAS ENABLED. On the white phone, I had no additional caching (stock CM).
You can watch this video here. IN short, THE LINUX SWAP ENABLED PHONE WAS FASTER THAN STOCK CM 3.9.7
This video is 7 minutes (AND yes I was smoking a cigarette) but I actually show proof that one phone is stock and one has only Linux swap by showing the ADB output of userinit(something I forgot to do with the first video).
Watch these videos and you cannot help but wonder if Compcache is slowing our phones down. Having two phones, I have tried many many times, and Compcache always makes the phone slower.
So to sum it up, a purely stock phone was faster and required less reloading than one with CompCache. However, a phone with Linux swap was faster than stock....so Linux>Stock>Compcache
In fact, for two days I carried around both phones. Whatever I did to one, I did to the other. The compcache enabled phone was ALWAYS sluggish compared to the Linux swap phone. I mean it would be fast at first,but always returned to being sluggish.
I submit that these videos don't PROVE anything. They just bring the question to light so that smarter people can actually figure out why Compcache is bogging our systems. And for those that say their unit flies with Compcache, I say you should try it with only the Linux swap.

I tend to agree...all things being equal, compcache w/ backing swap performs worse in the long run than does just a linux swap partition.
my anecdotal experiences, of course.
& that's not to say that some compcache / sysctl config out there won't make it outperform the run of the mill swap partition
nice to see some real experimentation on this issue

Thanks. I wish I had a nicer setup where I could film better.
I really believe that compcache is sucking the life out of our G1s...at least in it's current configuration.
I know there is going to be unbelievers....but that's ok too. To each his own.

i still dont like the implementation of a linux swap partition on the sdcard though. It brings about issues not only with bus bandwidth but also the eventual degradation of the sdcard.
I think a good compromise that frees up the sdcard would be to make a change to the nand partitioning and set up a 24mb partition along with all others to be used for swap. This would mean lessened life on the nand rom instead, but we flash so many roms per week, we're killing the phone already anyway.

pinetreehater said:
Thanks. I wish I had a nicer setup where I could film better.
I really believe that compcache is sucking the life out of our G1s...at least in it's current configuration.
I know there is going to be unbelievers....but that's ok too. To each his own.
Click to expand...
Click to collapse
For memory intensive applications, Compcache is definitely slower than stock, as it requires additional cpu time to decompress the data and compress the data that need to be swap. Plus additional fetch and store time to main memory. This incurs additional bus transactions. The SD as backingswap increase the bus arbitration delay ( you can also say the bus architecture is the bottleneck for G1 with CC in some situations). Overall, slows down the phone. However, for small applications, you may find it faster. so its really depends the how you use the phone.

Can you repeat your experiment with CM-3.9.9.1? There is a new helper patch I added to the kernel which improves the performance of compcache. This is the main reason why I enabled it by default.

jubeh said:
i still dont like the implementation of a linux swap partition on the sdcard though. It brings about issues not only with bus bandwidth but also the eventual degradation of the sdcard.
I think a good compromise that frees up the sdcard would be to make a change to the nand partitioning and set up a 24mb partition along with all others to be used for swap. This would mean lessened life on the nand rom instead, but we flash so many roms per week, we're killing the phone already anyway.
Click to expand...
Click to collapse
seems easier/cheaper to burn a $15-$20 sdcard than the non-replaceable(?) nand flash in the $400 phone.

cyanogen said:
Can you repeat your experiment with CM-3.9.9.1?
Click to expand...
Click to collapse
I seccond that request. 3.9.9.1 is the first build I can really feel compcache making my phone faster linux-swap (same swappiness, i.e. 35 in my case).
Thank you for taking the time to do such a thorough test!

How do you get compcache to work? Do you have to partition your SD card 3 times? 1 for Fat32. 2nd one for apps2sd. and the 3rd for compcache? I'm confused...can anyone help me?

51dusty said:
seems easier/cheaper to burn a $15-$20 sdcard than the non-replaceable(?) nand flash in the $400 phone.
Click to expand...
Click to collapse
Huh?
The ram on any electronics is designed for a lot of read/write operations...
Goog1e Phone said:
How do you get compcache to work? Do you have to partition your SD card 3 times? 1 for Fat32. 2nd one for apps2sd. and the 3rd for compcache? I'm confused...can anyone help me?
Click to expand...
Click to collapse
Compcache is automatic with CM-3.9.9.1. 2 partitions: FAT32 and EXT2/3/4

I think some of this is dependent on the ROM and how it is written. Also, you probably didn't have Compcache configured optimally. I made very minor changes to the settings and it made a significant difference in performance. Read through the xROM thread if you don't believe me. Moving the Swapiness from 30 to 28 made an improvement. Who would have thought that? If you were at 20, I'm sure you have problems because at 24 it was really slow for me.
Edit: On the ROMs, I did similar testing with xROM and I found that Linux-Swap was much slower all around on that ROM. I didn't try to tune linux swap like I did compcache because it was so bad to start with. I suppose I could do the same thing I did for linux swap to see if I can get an improvement from it. It's possible that Cyan just works better with Linux Swap than xROM does.

I think you should definately retry it like cyanogen said, if not on 3.9.9.1 than on the next stable build, 4.0.

try swapness set to 60 or more when use compcache.
I do think this will help.

I will retest this evening with 3.9.9.1....compCache vs. Swap
stay tuned!

pinetreehater said:
I will retest this evening with 3.9.9.1....compCache vs. Swap
stay tuned!
Click to expand...
Click to collapse
Why is a thread on technical issues like lunix swap and compcache suddenly more compelling than a ufc tapout match!?

Excellent experimentation, Pinetreehater! Thanks for doin a pseudo-scientific battery test so we can really try to tweak the settings. I saw that you were using 3.6.7 build, not sure if that was a typo or not??
Anyway you can repeat the test using the new 3.9.9.1 rom, as I have seen significant improvements in memory management for that build in comparison to 3.9.7 and 3.9.8.
Thanks again for your work!
EDIT: Cant wait for the retest results....thanks for taking the time to do this, as the whole community will benefit from your results

I'm starting to hate pine trees too...lots of sap that f&%*s up car paint. Awaiting your future results as I have found similar results as you, but alas, I only have one Dream to test with. Keep up the death matches!

good test well as far as burning out the microsd i doubt i have a phone long enough to burn it out i switch phones like no other, but this is rather interesting...

pinetreehater, great job. As I said a while ago in one of your previous threads, compcache was really sucking for me no matter what the swappiness, and having it disabled entirely seemed to improve the performance of my phone. I'll back you on this, you're not crazy. Anyway, yeah, very interested in your results tonight. Maybe I'll like it this time around.

Lollipop_Lawlipop said:
Huh?
The ram on any electronics is designed for a lot of read/write operations...
Click to expand...
Click to collapse
nand is not the same as ram. it's like a sdcard and gets worn out over time just the same.

[Q] Couple noob questions

I'm not necessarily a noob to android. I've flashed a few phones, dabbled with rooting and custom roms, but nothing too serious.
My first question about these Tegra 2 processors is fairly simple. I want to know if overclocking them makes them more media capable? I have a Droid Charge and had a Galaxy S before it, and both phones were perfectly capable of playing a 720p mkv with nary a stutter. It boggles my mind that I have to convert every video file before I drop it onto the Xoom. It's not a huge hassle but it's rather interesting a single core hummingbird can handle it with ease and a dual core clocked at the same speed has so much trouble.
Second question is this. Is there any development going on for the Xoom Family Edition? I haven't noticed much other than news posted regarding this version on this forum, so I'm not sure. It's obviously not the hit they were hoping it would be, there's a whole 3-4 cases made to fit it, and virtually no accessories for it like there is for the Xoom (docks and whatnot). If overclocking helps without nuking battery life, I'd be interested. I wouldn't know where to start developing, I'm definitely not a skilled programmer, but I would be happy to help any way I could (minus a potential brick, I can't afford a new tab)!

Bump. Any opinions?

Regarding the overclocking the tegra 2 , yes it will be more powerful therefore increasing media playback performance. Although battery usually drops faster you can install set CPU to set profiles for when the screen is off etc...
As for the Xoom I myself am not very knowledgeable about that device so I cannot effectivily answer your question.
Hope this helps!
Sent from my PG86100 using XDA App

Remember that the Xoom screen is so much bigger, so takes a lot more juice to playback at same resolution, but overclocked cpu does make a difference for sure. Also, if on ICS, go into developers settings menu and turn off animation or set to .5. Makes a big difference.
The Xoom FE (like the Xoom2 and the Xyboard) don't have unlockable bootloaders and noone yet has found the exploit to gain root access. Hence no development...so far. If you find the way, you will be popular among the owners of those devices.

If you overclock the Xoom it will of course consume the battery faster, as with anything that requires voltage tweaking. The main thing to remember is research the governors that will adjust what clock you're running at and when you're running it. Interactive usually gives the best performance speedwise while Ondemand gives you a mix of performance and battery life.
Your media playback issue could possibly be due to several factors. Processor, RAM, SD read rate, and the media player are the main culprits that come to mind.
If it is the processor then overclocking will most likely help. There is a minor off chance that the Tegra2 does not contain certain instruction sets included on the other devices processor that allow it to decode the Matroskiva, mkv, format as readily.
If it is the RAM, your best bet is to get a task killer and use it to kill everything before you try to play the video. You can also go into settings and go to individual apps to force kill them which tends to work better than most task killers.
If it is the SD card rate, research fairly deeply into the subject because I personally have heard many mixed reviews in regards to the Xoom and higher "class" or access rate SD cards. Eventually I plan on getting a collection of them to run some testing myself for a unified chart, but until then your best bet would be to ask experienced Xoom users, and browse these forums.
If it is the app then try looking around to see what other players are out there. Some people use different decoding codecs than others, and some tend to work better on mobile devices with limited instruction sets.
Mind you, if i remember correctly, the other two devices also display at a lower resolution, which would take less power, and the app used to play the file might not support the larger resolutions as well.
And if you have not already toyed around with your Xoom and hacking it, just as a warning, like all other devices it can be easy to brick. Make sure you have read at least 2 different tutorials on how to do it beforehand as some are much more clearly stated than others.
Hope this helps you some

I hadn't thought about the memory card being the issue. I'm not sure how fast the internal memory card is, but my external is only a class 2, so that could have an effect on load times and everything. I know the RAM isn't really a problem, I don't do much on my tablet, and I've tried killing apps and fresh boots and nothing seems to work. It only seems to use about 450mb out of 1024mb for apps, and out of that only 124 was in use last time I checked.
I have tried many other media players though, including Dice Player which seems to be unanimously the best, and nothing I've tried is able to play an mkv on the Xoom. I typically try downloaded tv shows before movies, which include 500mb Big Bang Theory mkv files and 1gb Top Gear mkv files at 720p. Neither play at all in the stock player and play very badly in Dice or other players.
I can't wait until ICS on the Xoom FE, I'm betting it fixes a lot of my issues. Such as the browser constantly force closing and my wifi slowing down the longer the tablet is on.

Can I/O issue be fixed by Software? Evidence based.

Undoubtedly this is the hottest topic on this forum (so it seems). Based on what I have seen, there has been no convincing proof on either way. The benchmarks score comparison are great, but my belief is it may not reflect day to day experience as most of time you never push the system to its max.
So my hypothesis is Transformer Infintiy's I/O issue related user experience can be fixed by software even if the underlying issue had been hardware related.
Since some people believes the opposite, and this is infinity forum. I wanted to make myself right for the sake of infinity owners. But given no previous good convincing test, I have decided to test one thing.
First video
This is one example video that Anandtech showed back in July that Infinity is indeed experiencing I/O issue. This video simply proved that while I/O is in use, the whole system slowed down. Based on the date of this post, which is 7/26/2012, I am assuming the firmware used was V9.4.5.22.
Second video
The credit of this video goes to BarryH. The reason why I included this is Galaxy note 10.1, which I owned for 3 weeks should be used as gold standard in this comparison as it is perhaps the only worth mentioning competition of TF700T and it is really great tablet of its own. You can clearly see that the downloading file in background really had no effect at all on the Galaxy Note 10.1.
Third video
The above is my very first Youtube video ever. But that's not a point. This is running stock ROM, rooted. No build prop tweaks. The only special thing I have installed here is Browser2Ram. So for this basically, I used AirDroid (WIFI file transfer application) and transferred 1G+ file over my network. While doing so, I basically opened browser, and went to the same site as Galaxy note 10.1 did in above. The little delay after opening browser is not hang in browser, but simply I forgot to set to landscape mode.
Conclusion: Based on these I believe I can conclude followings:
Compare first to third. Infinity had already significantly improved its I/O performance. How did it do? Clearly not by changing hardware, but rather software. So this proves my hypothesis of software can indeed fix I/O issue that user actually see.
*Please note that since I have browser2ram installed, this may be different for pure stock user experience. And Zeus users may even have better experience than me.
Special thanks goes to BarryH_GEG. Without him posting first two youtube clips I wouldn't have actually realized that Infinity's I/O issue had already been nearly completely fixed on my unit.
.
For clarification, basically all I am doing here is that there is indeed a way that what we call IO issue that can be alleviated by software modification. So if anyone state its all hardware and can never be fixed/improved, that is incorrect statement. However, this does not prove stock Infinity in the future will receive such update. So you can have a hope, but hope has no guarantee.
Different people use device differently so I say try the device and update to the latest firmware and see if you still face the IO issue that is significant enough for you. I can say that it is definitely not as bad as first video show. Whether you can tolerate the IO lag or not is simply your personal preference. In the end, all devices have lag to certain degree when stressed. Even PS3, Xbox 360, or high end PC games gets frame drops when you stress them. The question is whether you care, and whether you push to the limit.
If I ever get chance, I will try to turn off Browser2Ram and try do the same test again and see what the true stock experience is, though I am certain it is not as bad as Anandtech video.
Update 9/24/2012
Since some people are pointing to the issue resolution is purely due to Browser2Ram, I did same test with this time while AirDroid transferring 1.1 GB file over the WIFI, I launched Final Fantasy 3. No lag even while launching. Everything is smooth. However, if I launch Horn instead of FF3, I did notice significantly longer time to load. So this is simply proving what I have said. We can fix issue to certain degree but whether the certain degree is enough for a specific user is really dependent. I am ok not being able to play Horn while I am transferring 1GB data, but some may not.

Of course Browser2RAM will help on this, you're caching to just RAM instead of the stupidly slow NAND memory Asus decided to put in their flagship device. Try the same "test" on complete stock and you'll see that it's a vastly different experience.
Zeus ROM works around the I/O issue by reducing or even eliminating SQLite fsyncs. Risky business for your data, good for performance. I like my data without corruption, so I'll pass, but others may not see it that way. They simply want the performance this tablet should have had in the first place.
Asus did reduce the overall sluggishness slightly with the .26 update that changed from NOOP scheduler to CFQ in the kernel, but the tablet still stutters. Browsing is far from smooth, even with alternative browsers like Dolphin HD. Performance is decent right after a reboot, but once memory fills up and Android starts its memory management and closing down processes (doing a lot of fsyncs in the process) it grinds to a halt. This wouldn't be a problem if flash I/O performance was higher.
The CM-based ROMs starting to pop up for the TF700 seems to help on the performance as well, and that's still with the 2.6.39.4-based kernel. CM's always been smoother than pretty much any stock device in my experience though, so no surprise there. They can never completely get rid of it though, since it's a hardware issue. Asus were stupid and chose cheap, slow NAND that gets totally crippled if you're doing random writes. There's no magic software to just fix it.
I'm sorry, but I don't really see anything new here, and your "evidence" isn't really that, simply because you're not running stock. You're using Browser2RAM which greatly increases browsing performance by using RAM, which is several magnitudes faster than NAND flash. It's not even comparable. The TF700 (and TF201 and TF300 and TF101) I/O issue can never be completely fixed in software, only (slightly) improved, often at the cost of data safety (disable SQLite fsyncs or available RAM (Browser2RAM).
It's all well-known by now that Asus ****ed up (again!). If it weren't for the oh-so-sweet high resolution display and keyboard dock I'd get a different tablet.

Einride said:
I'm sorry, but I don't really see anything new here, and your "evidence" isn't really that, simply because you're not running stock. You're using Browser2RAM which greatly increases browsing performance by using RAM, which is several magnitudes faster than NAND flash. It's not even comparable. The TF700 (and TF201 and TF300 and TF101) I/O issue can never be completely fixed in software, only (slightly) improved, often at the cost of data safety (disable SQLite fsyncs or available RAM (Browser2RAM).
It's all well-known by now that Asus ****ed up (again!). If it weren't for the oh-so-sweet high resolution display and keyboard dock I'd get a different tablet.
Click to expand...
Click to collapse
Hi Einride,
Sorry if I directed you wrong way. But my point here is not to prove stock has already fixed issue or not even to say IO issue can be completely fixed. The latter is simply unknown. But I am just proving here that there are ways software can make difference in user front, which some people questions.
To be honest, how do you even know RAM is not bottle neck? What about GPU, which is far inferior to the new Ipad in fact its even worth than iPad 2 by far margin? We see a number, and see its less than others so concluded its the conundrum, which could be true by good chance but not a proof.
Here I am basically proved whatever the method is, there is a way to improve what people call IO issue can be alleviated by software approach.
Because Anandtech showed IO issue originally on infinity stating background 2MBps download resulted in marked degradation in browser performance. I basically had 1GB file transfer over WIFI using Airdroid.
So I basically proved here that my infinity, which clearly has not touched anything on hardwarewise, but have improved performance since Anandtech review.
As for Zeus fsync, I don't use Zeus so cannot speak for it. But if they turn off fsync and still keep the system stable without data loss, why would you care?
Having said all this, I don't know if ASUS will ever put effort in fixing issue because if they would why won't they simply install browser2ram as part of their firmware? But they are doing something as I noticed they took out pixit from background with latest firmware, which kept running in background for no reason..

Einride is on target. This is a hardware issue, so any "fix" is going to be a kludge and come with a bunch of compromise. I can't believe ASUS specced the same crazy slow flash on the 700 that they did the 201/300.
FWIW, The best/cleverest hack I have seen is the TF201 dev who has been playing with remounting the internal SD card to point at the removable microSD card. If this can be made to work smoothly, it means that if you have an external card that is specced much higher than the stock internal flash you could eliminate the issue completely.
But of course that is really hacky. It's one of those you could brick if you don't install it right kind of deals so I don't ever see it being a mainstream options for these tabs unless ASUS productized it which would be a big expense in support for them, so, again, it'll never happen.

I've got to push back a bit. I think this is firmware related, but not of the device of individual components. I think this can be fixed with "software"

tf201 said:
I've got to push back a bit. I think this is firmware related, but not of the device of individual components. I think this can be fixed with "software"
Click to expand...
Click to collapse
You can continue thinking that if it makes you feel better. I like and enjoy my Infinity but as a digital design engineer, to me this seems to clearly be a design issue @ the hardware level and will always limit the overall system to some degree. There are hard limits in any hardware software system. <Shrug>
My best advice to an end user is what I would give for any device issue: to decide whether it is a deal breaker for you *as it stands right now* and not sit around and wait for or bank on some kludgy cure that may be worse than the disease.
It's not too bad for me, I don't do tons of random I/O or web browsing. In fact, I'd say it would have taken me a long time to notice this on my own without the benchmarking and threads here... If it was unacceptable I would cut my losses and sell the tab and get a different device. Life is short, guys.

zenaxe said:
Einride is on target. This is a hardware issue, so any "fix" is going to be a kludge and come with a bunch of compromise. I can't believe ASUS specced the same crazy slow flash on the 700 that they did the 201/300.
FWIW, The best/cleverest hack I have seen is the TF201 dev who has been playing with remounting the internal SD card to point at the removable microSD card. If this can be made to work smoothly, it means that if you have an external card that is specced much higher than the stock internal flash you could eliminate the issue completely.
But of course that is really hacky. It's one of those you could brick if you don't install it right kind of deals so I don't ever see it being a mainstream options for these tabs unless ASUS productized it which would be a big expense in support for them, so, again, it'll never happen.
Click to expand...
Click to collapse
Well, again (and I now added to opening post) my point is not denying IO hardware issue existence. I am just simply providing proof here that it can be improved by software (to certain degree). Whether it is to the extent of users complete satisfaction would be different issue. However, to be honest, if you see my video and it was original retail packaged Infinity fully on stock and from the day 1 performance, don't you think less people would raise a voice? It's because software workaround can change what you see as user, and that's all I am proving here. So yes. this should give a hope to someone who thought, it's hardware related so never can be fixed. However, this does not confirm or deny ASUS to take a step and fix this because I am sure it will take some effort on their part, which clearly they have not spent so far if this had been a problem existed since TF201.
To put into extreme, they can modify OS such that all front end basic programs such as stock browser, movie player or whatever to actually run on RAM if that is what makes the difference. But would they do it? Absolute not, because they won't spend money on such big project for device that had already sold well and gained essentially best Android tablet metacritic reviews (I did not take actual poll but just following Infinity news daily, it seems like pretty much most site gives the highest score for android tablet).
So yup. I don't disagree with you guys that IO issue there. But my point here was to help people clarify that there are indeed ways to make better by software. Whether happens or not is out of my control and would simply be guess for anyone.

I'm not just a noob either. Here's why I think this is software related:
1) The performance is so bad that it precludes just the hardware. Maybe the hardware sucks but there is alot of performance lost somewhere cheap NAND from 2 years ago outperforms SQLite operations by >10x.
2) Performance seems to degrade with time. This is indictive of a wear leveling and writing algorithm which may or may not be able to be adjusted with firmware.
3) SQLite fsync performance appears to be tied to T3 frequency, that suggest there is something with the T3 drivers that could be tweaked vs NAND hardware limitations.
4)...
I'll also mention the OP is right. ASUS could do things with caching data before writing, and writing in chunks the NAND is best with limiting Virtual Ram etc.

HoushaSen said:
Hi Einride,
Sorry if I directed you wrong way. But my point here is not to prove stock has already fixed issue or not even to say IO issue can be completely fixed. The latter is simply unknown. But I am just proving here that there are ways software can make difference in user front, which some people questions.
To be honest, how do you even know RAM is not bottle neck? What about GPU, which is far inferior to the new Ipad in fact its even worth than iPad 2 by far margin? We see a number, and see its less than others so concluded its the conundrum, which could be true by good chance but not a proof.
Here I am basically proved whatever the method is, there is a way to improve what people call IO issue can be alleviated by software approach.
Because Anandtech showed IO issue originally on infinity stating background 2MBps download resulted in marked degradation in browser performance. I basically had 1GB file transfer over WIFI using Airdroid.
So I basically proved here that my infinity, which clearly has not touched anything on hardwarewise, but have improved performance since Anandtech review.
As for Zeus fsync, I don't use Zeus so cannot speak for it. But if they turn off fsync and still keep the system stable without data loss, why would you care?
Having said all this, I don't know if ASUS will ever put effort in fixing issue because if they would why won't they simply install browser2ram as part of their firmware? But they are doing something as I noticed they took out pixit from background with latest firmware, which kept running in background for no reason..
Click to expand...
Click to collapse
No misleading here, don't worry
You are just describing software workarounds, though. None of which can permanently fix it entirely since it's a hardware limitation.
A proper "fix" would greatly increase I/O performance with no downsides. Browser2ram helps browsing, nothing else. Disabling SQLite fsyncs increases risk of data corruption or data loss at the cost of better overall performance.
Sent from my ASUS Transformer Pad TF700T using Tapatalk 2

tf201 said:
I'm not just a noob either. Here's why I think this is software related:
1) The performance is so bad that it precludes just the hardware. Maybe the hardware sucks but there is alot of performance lost somewhere cheap NAND from 2 years ago outperforms SQLite operations by >10x.
2) Performance seems to degrade with time. This is indictive of a wear leveling and writing algorithm which may or may not be able to be adjusted with firmware.
3) SQLite fsync performance appears to be tied to T3 frequency, that suggest there is something with the T3 drivers that could be tweaked vs NAND hardware limitations.
4)...
I'll also mention the OP is right. ASUS could do things with caching data before writing, and writing in chunks the NAND is best with limiting Virtual Ram etc.
Click to expand...
Click to collapse
I agree that they can probably make improvements with these kind of tweaks. The extent of the fix will always be limited by the low spec hardware in the end, though. So, I wouldn't encourage people to expect ginormous strides. For the most part, I expect more of the same with some evolutionary improvements but I doubt they will ever make a quantum leap. I would be angry if they made it worse but I'm an optimist at heart so I at least expect some slow improvement over time.
IMHO, as it stands it is usable and I'm hoping they can take it to "decent" (say the level pepole are seeing in Zeus). But to folks who are banging their heads against this constantly and unsatisfied as a result, I would still say there will be no true fix and you should bail on this device. It's a personal choice.
If you are willing to forgo your warrantly, I guess you could demo one of the custom ROMs. That probably shows you the extent of a software fix. But beware some of those fixes (cached writes) do put data integrity at risk. There is always a trade off.

Einride said:
Disabling SQLite fsyncs increases risk of data corruption or data loss at the cost of better overall performance.
Sent from my ASUS Transformer Pad TF700T using Tapatalk 2
Click to expand...
Click to collapse
If I understand the fsync thing correctly it can only cause issues if the device suddenly powers off, correct? While that may be a risk, it's a *very* small risk and well worth the performance improvement, in my opinion... I've used ROMs that disable fsync for quite some time and have never had a single issue...
I tend to go by real-world results instead of "theory"...
Same thing with "browser2ram" - it can only cause an issue if the device suddenly powers off and even then - so you lose some web cache data - so what?? Who cares if you lose your browser cache - it's just a browser cache!
Besides, if your device is powering off suddenly, you have much bigger problems than worrying about your cached web data!
I truly agree with the OP. People get so caught up on benchmarks and "what could happen" (even though in practice, it really doesn't)... Truth is, we all just want a better end-user experience - if they can "work around" hardware limitations via software then it makes sense to do that.
Obviously, the hardware isn't going to change, so complaining about that will never help whereas implementing software tweaks to work-around these hardware limitations *does* actually help...
Just my two cents!
By the way, I've been *very* happy with the performance of my TF700 since installing the Zeus ROM - another perfect of example of someone using software to get around the hardware limitations - and it works very well! Another example of "real world" results - that's all of the evidence that I need!
Sent from my ASUS Transformer Pad TF700T using Tapatalk 2

Unfortunately random reboots are not simply theory in case of the Infinity, especially if you're doing some memory-aggravating stuff on yours. I'm getting one every few days.

d14b0ll0s said:
Unfortunately random reboots are not simply theory in case of the Infinity, especially if you're doing some memory-aggravating stuff on yours. I'm getting one every few days.
Click to expand...
Click to collapse
While I agree that random reboots are not simply theory, what does that have to do with what we are talking about here?
However, since you brought it up, I've personally never had a single random reboot on my TF700, which may be yet another example that most of this stuff can be fixed by software (since that does not occur on Zeus ROM)... So that just goes further to illustrate my point. I'm assuming that you are not running Zeus?
Sent from my ASUS Transformer Pad TF700T using Tapatalk 2

jtrosky said:
another example that most of this stuff can be fixed by software (since that does not occur on Zeus ROM.
Click to expand...
Click to collapse
This is from the Quadrant thread. The rest of the conversation describing what SQLite is and the role it plays in overall IO performance is in that thread. All Zeus' ROM does is tweak the SQLite settings and if you read more in the other thread you'll understand why that plays a relatively minor role in overall IO performance. Bottom line is whether or not someone has memory and/or IO issues is more determined by what they do with their device than the s/w running on it. Which is why some people running stock are perfectly content while others are pulling their hair out.
P.S. - Sorry HoushaSen, the lack of information on what SQLite is and the obsession with Quadrant brought me back in to the discussion.
BarryH_GEG said:
To show you what Zeus' impact is, here's a comparison to a Note and TF300 on JB. Red is perecent slower than the Note, green faster. After tweaking SQLite, the remainder of the TF700's IO scores remain significantly below that of the Note (or SGS3 or One X) and some are worse than stock. If you use the TF300 on JB as a proxy for how the TF700 would perform after the update the column on the far right shows the difference between the two.
Click to expand...
Click to collapse

Wait until you see Zeus on Jelly Been. It's going to be so smooth and snappy and you will never think about I/O issue again!!!!

BarryH_GEG said:
This is from the Quadrant thread. The rest of the conversation describing what SQLite is and the role it plays in overall IO performance is in that thread. All Zeus' ROM does is tweak the SQLite settings and if you read more in the other thread you'll understand why that plays a relatively minor role in overall IO performance. Bottom line is whether or not someone has memory and/or IO issues is more determined by what they do with their device than the s/w running on it. Which is why some people running stock are perfectly content while others are pulling their hair out.
P.S. - Sorry HoushaSen, the lack of information on what SQLite is and the obsession with Quadrant brought me back in to the discussion.
Click to expand...
Click to collapse
No problem. In the end this is still gray zone that (at least in my opinion) nobody knows what the end result is, and I think you stated well. It's really depends on user experience.
For those missed my last couple line update on opening thread. I actually had 1.1GB file download going on my infinity, and launched FF3 game, which was a breeze. No lag. So IO issue had been fixed partly from original already (assuming IO was really poor from get go, but I cannot confirm this because I do not have original firmware, and did not play FF3 when I had one). But I definitely noticed significantly far less ANR, which was one major reason I originally returned my infinity and hesitant to come back from Galaxy Note 10.1. However, if I launch Horn instead of FF3, it takes forever and even got ANR, which I hadn't seen for a while on my Infinity. Whether this is related IO or memory cap of 1GB or some other internal limitation is unknown, but since all I had was AirDroid transferring file and Horn is only other thing running, I am assuming 1GB is sufficient; hence, most likely related to IO issue. But having said this, how many would really complain about this? Not sure. Because even on my PC (which is not that high end) but I can basically get same issue. If I encode video and try to run high graphic PC game, the machine stalls, and even gives freeze. Would everyone complain about this? Some would and say that's why you get better PC. The other accepts it is what is, and simply don't encode, and play high end game at same time.
I am pretty satisfied with Infinity as all the concern I had before coming back to Infinity from Note seemed to be solved (at least for me) and got back to Full HD screen; however, there are clearly still people out there concerned of IO performance thus the topic continues to arise. Once everyone gets satisfactory IO result, I believe we will see significantly less discussion about this (if ever happens).
The fact we know are:
1. IO hardware on Infinity is last generation and not as fast as main stream current generation expensive tablets.
2. Software can change what user see (whole point my this particular thread)
Fact nobody knows
1. Degree of how much software can change user experience. Whether enough to completely hide relatively poor performance of underlying IO hardware. Or opposite extreme is basically just soften up a little without true effect on most users.
2. Whether ASUS will even try fixing it.
Benchmark number is great to assess, but I really don't think that's what users are really interested unless someone who just want to say "hey my benchmark score is high!" If this is the case, nobody probably would ever get Apple because they are usually not after benchmark of individual component but rather they use decent hardware, minimize bottleneck by deciding all the hardware on their own, and write optimized OS. But individual pieces are not cutting edge for its price.
And in all honest, I am a bit lost at this point after writing this thread, what is it exactly that we are calling IO issue? Browser2Ram improved my browser speed but even Final Fantasy 3 runs fine from launch while I download a big file at its maximum speed. So I don't think it's browser2ram that did trick here, but rather ASUS already fixed IO issue for downloading file completely hogs the system. If the issue is slow stock browser, that may not be IO related. It may simply be ASUS did not optimize the stock browser. Maybe my system runs so well because I have turned all the bloatware off and many stock user aren't?

I don't see Browser2ram in the store. Got a link?
Sent from my Rooted TF700T

i wish i saw this thread before i bought my Infinity + screen protector + case.
hope that JB will help with some of this, but i didn't realize it was just cheap-ass NAND.
ugh.

xPSYCHOTRONx said:
I don't see Browser2ram in the store. Got a link?
Click to expand...
Click to collapse
http://mark-tech.blogspot.com/2012/06/browser2ram-fixes-transformer-prime.html

Searching for a ROM with low storage footprint

Hi guys,
my girlfriend has the first generation Moto G with only 8gb of storage. I've put CM12 on it a long time ago, and everything worked very well. But lately shes been running into problems concerning the phones storage. She regularly can't download apps because there is no space left on the phones internal memory. She has barely any music, videos and even apps on it, so these things can't be reduced any further.
The only solution i came up with so far is to use a ROM that still offers Dalvik instead of ART. But that's probably not enough for a long term solution.
Does anybody know ROMs that uses the smallest possible amount of storage on the device? I appreciate all other solutions as well.
Thanks a lot!

Best solution is to get a better phone with 32 or 16 GB.. but still considering your request
you can try SLIM rom even though they stopped making 4.4.4 roms and are currently back making 5.1.1 roms and their roms are <200mb (as their name suggests "SLIM").
but still i dont think flashing a low data consuming rom is going to help out.. i guess its time to get a new device

LG G7 Ram Management is still a mess for me after the 10f update.

So by now I've mentioned a few times already how awful my experience with the ram management has been.
If I have chrome open I can go look for something on other app and when I come back chrome will be wiped out of memory.
This is for me a normal everyday use of my phone, I'm not trying to look to some absurd way of doing speed tests or whatsoever.
This videos is based on something that happened today.
Scenario: I had music playing on GPM, I was browsing the web, and somebody answered to me on a Dutch forum asking why I thought the G7 haptics were so awesome (and yes it is), so I wanted to go to the camera app so I could just confirm what parts of the software would have a feedback attached to it after playing a bit I accidentally pressed the shutter button, at this moment the music stream should lower the volume for the shutter sound, well the music simply stopped and GPM was force killed, and so was chrome .
It isn't lack of ram, 4GB is more than enough for a flagship phone to work, besides the phone reports having over 400MB of free memory.
What have I done so far:
-Because after I got the phone I got an OTA update, I actually did a full reinstall via LG bridge and a software reset
-10f update was also done via LG bridge
-I have kept the installed apps at a minimum, and no app shows excessive ram or battery usage.
https://youtu.be/o5gs1n13jt0

Could it just be that you have a faulty? Because I never get this kind of app killing

xtrocata said:
Could it just be that you have a faulty? Because I never get this kind of app killing
Click to expand...
Click to collapse
I assure you, his device isn't defective.

I find the app killing to be MUCH stronger than I was used to after two years on the OnePlus 3. (A wonderful, smooth device, even two years later! Yes, I still miss it.) I was thinking it was the drop from 6 to 4GB of RAM, but it just isn't taking them out of RAM, it's entirely background killing apps from the recents list too. I used to have apps in my recents list I hadn't opened for WEEKS on the OnePlus. So I'm also inclined to believe it's aggressive RAM management on LG's part.
It's not unliveable for me, but definitely noticed, and a bit shocking that a flagship two years newer isn't as good in pretty much any respect as something 'budget' from two years back.

LG bloat and everything is terrible on this device. I wish when they release the android one G7 that they make it available to install on the flagship G7.

Henaway said:
I find the app killing to be MUCH stronger than I was used to after two years on the OnePlus 3. (A wonderful, smooth device, even two years later! Yes, I still miss it.) I was thinking it was the drop from 6 to 4GB of RAM, but it just isn't taking them out of RAM, it's entirely background killing apps from the recents list too. I used to have apps in my recents list I hadn't opened for WEEKS on the OnePlus. So I'm also inclined to believe it's aggressive RAM management on LG's part.
It's not unliveable for me, but definitely noticed, and a bit shocking that a flagship two years newer isn't as good in pretty much any respect as something 'budget' from two years back.
Click to expand...
Click to collapse
And what version of Android were you running before on the OP3? Just curious...

I also can assure you that this is Not normal! I get nothing like this on my G7. I can multitask between many many apps all day long and none close, music plays no matter what etc etc

No problems here either, I migrated from my old phone yesterday and had 110 apps installing and configuring them. No unwanted force closing as far as I could see, listened to YT Music for a long time and no glitch, all notifications works just fine. I'm on 10f.

Anarchist77 said:
And what version of Android were you running before on the OP3? Just curious...
Click to expand...
Click to collapse
It was (is) on Android 8.0. We got it on the OP3 last year. They're skipping 8.1 and going straight to 9.0 on them though .. and that'll be the last official update for the 3/3T series. But they're unlocked and really easy to flash, so the ROM community will look after future Android versions for quite some time I'd imagine. The specs of those phones will keep them going strong for a long time. Basically until something fails catastrophically.
I'm hoping LG will be a BIT timely with updates to the G7 ThinQ. They really don't have a great reputation for it though, so I'm not holding my breath.

Maybe the next update (10g) will fix it.

I'm having problems with the poor ram management too. I used to have an xperia XZ Premium with same 4gb of ram where i never encountered these ram problems.
I've yet debloated the G7 which improved but didn't improve to a satisfying point. Has anyone tried some of the ram booster apps and do any of them work as expected?
Greets

I had this issue with the HTC U11 on stock. It was just a complete pain. So, I installed the latest unofficial Lineage ROM on it and it's been working pretty well since. I don't currently have this problem on my g7 yet.