ext2 does help! score 2221 + NAND INTACT *EDITED 21AUG 2018HKT

Search This thread

ykkfive

Inactive Recognized Developer
Jul 25, 2010
3,175
2,990
ykkfive.blogspot.com
I know that many people worry about frequent reads/writes will damage the NAND, so I began a new test days before without using the NAND

this time, I switched from rfs to ext2, with some mods to the kernel to load the ext2 partitions, no OC

note that this is only serve as a real case to show u ext2's performance and to prove that rfs is one of the problem areas. u guys can implement it your ways with further improvements, perhaps with ext3, ext4, yaffs, etc


EDIT: 21AUG 2018HKT
- some ppl said that it is easy to cheat the score obtained in quadrant by moving around the quadrant's data files. yes, i agree that. but it wil be much easier to edit it in photoshxp for even 30k score. and although some ppl said the score itself is meaningless, which i agree with it also, but it provides a mean to let u know how ur phone is performed. it is not an indicator for whether ur videos or sounds can be played smoothly since it depends on a lot of things as well
- also note that if u formatted the /sdcard in different file format, some of the recovery mode functions wont work since it will only mount rfs/vfat, so u have to modify the stock recovery or rom manager's recovery manually, or by mounting the partitions via shell



ext2.png
 
Last edited:

mbr01

Member
Sep 7, 2008
45
5
Nice to see the score split up in the different parts. It shows that the inital I/O of GS wasn't really that bad. Also i guess we can expect about 750 points of CPU score with Froyo (looking at the Nexus), so maybe a 3000 score is within reach :).
 

ykkfive

Inactive Recognized Developer
Jul 25, 2010
3,175
2,990
ykkfive.blogspot.com
Nice to see the score split up in the different parts. It shows that the inital I/O of GS wasn't really that bad. Also i guess we can expect about 750 points of CPU score with Froyo (looking at the Nexus), so maybe a 3000 score is within reach :).

right, i saw one with 27xx or something already with old tricks. so 3000 should not be a problem with 2.2
 

g00ndu

Retired Recognized Developer
Apr 22, 2008
2,782
273
right, i saw one with 27xx or something already with old tricks. so 3000 should not be a problem with 2.2
ykk_five, first, many thanks for the tip. The score of 2711 was done during your first initial suggestion, where the fix was temporary. In additional, I happen to have the overclock kernel (by raspdeep) installed. Hence, I just ran it to see how much mileage we can get from the SGS.

Having said that, RyanZA and Chainfire had implemented excellent lagfixes. With your latest test, I'm really not sure how much more improvement we can get. Many of us are now very happy with what they have, and I am not sure if there are anymore juice to be squeezed out from this.

Let's hear it from the more experienced users, and perhaps this latest discovery might be as exciting as your first. :)

Greetings from Singapore. :)
 

ykkfive

Inactive Recognized Developer
Jul 25, 2010
3,175
2,990
ykkfive.blogspot.com
ykk_five, first, many thanks for the tip. The score of 2711 was done during your first initial suggestion, where the fix was temporary. In additional, I happen to have the overclock kernel (by raspdeep) installed. Hence, I just ran it to see how much mileage we can get from the SGS.

Having said that, RyanZA and Chainfire had implemented excellent lagfixes. With your latest test, I'm really not sure how much more improvement we can get. Many of us are now very happy with what they have, and I am not sure if there are anymore juice to be squeezed out from this.

Let's hear it from the more experienced users, and perhaps this latest discovery might be as exciting as your first. :)

Greetings from Singapore. :)

thx a lot, i really hope that u guys will enjoy it!
 

Frostfree

Senior Member
Apr 8, 2008
274
30
I have a consistent score of 990+ without any lag fix on a froyo ROM. Previously I used a lag fixed JM5 which score 2000+, however with everyday use the froyo ROM seems snappier. Good to see all efforts to speed things up and eliminate the lag, but I'm not sure higher benchmarks score translate to a better user experience.

What I'm hoping for now is a fix than enables smooth scrolling of contacts & call logs seen on the desire.

Good work though!
 

ykkfive

Inactive Recognized Developer
Jul 25, 2010
3,175
2,990
ykkfive.blogspot.com
I have a consistent score of 990+ without any lag fix on a froyo ROM. Previously I used a lag fixed JM5 which score 2000+, however with everyday use the froyo ROM seems snappier. Good to see all efforts to speed things up and eliminate the lag, but I'm not sure higher benchmarks score translate to a better user experience.

What I'm hoping for now is a fix than enables smooth scrolling of contacts & call logs seen on the desire.

Good work though!

u can compare the IO stat, although it's a roughly comparison
 

toca79

Senior Member
Oct 12, 2009
114
4
ykk_five, first, many thanks for the tip. The score of 2711 was done during your first initial suggestion, where the fix was temporary. In additional, I happen to have the overclock kernel (by raspdeep) installed. Hence, I just ran it to see how much mileage we can get from the SGS.

Having said that, RyanZA and Chainfire had implemented excellent lagfixes. With your latest test, I'm really not sure how much more improvement we can get. Many of us are now very happy with what they have, and I am not sure if there are anymore juice to be squeezed out from this.

Let's hear it from the more experienced users, and perhaps this latest discovery might be as exciting as your first. :)

Greetings from Singapore. :)

This kind of solution should be a lot better as it doesn't use any space (or so i undertand).
 

bub181

Senior Member
Jul 23, 2007
254
8
So what exactly did you do to get these new scores? Did you format the /data partition as EXT2 to get these scores?
 

fantacies

Member
Aug 2, 2010
7
0
I understand someone may not like to spend time to answer noob's questions. But what is the point to show just the Quadrant score and then tell others you have done something in this Development board? Should it be presented in a more technical way?
 

bub181

Senior Member
Jul 23, 2007
254
8
Yes please give some more details on what exactly you did.

Sent from my GT-I9000 using Tapatalk
 

Tyxerakias

Senior Member
Jun 19, 2010
522
300
London, UK
I wonder if Froyo will bump this score even higher.

I just performed a test on the HTC Desire, with newly installed Froyo (2.2). On the scale above, it scores between the Nexus and the Droid, but closer to the Droid. What intrigued me is that the reason the score was lower than the Nexus, was the CPU score. The other colours were about the same as the Nexus One. The CPU part was still bigger than anything else (up to the 1/4 of the green part on the droid below it) but about 4/5 that of the Nexus. Sorry I couldn't post a screenshot.
Which is really weird, since the two devices share the same CPU. (are the reference scores accurate?)

EDIT: I tried running it again, and I got a higher CPU score this time, so now it almost matches the Nexus (up to 1/2 of the green part on the Droid) Overall score 1301. So scratch the above comment, I guess the reference scores are the highest you can get.

Now in the picture of the Galaxy above, the biggest part is the IO score, which is logical because of the modification, but the CPU score is very small. So assuming that Froyo will bump up the cpu part like on the Nexus and the Desire, we'll be looking one hell of a record score, right?

Has anyone with a JP firmware tried this lagfix?
 
Last edited:

OrionBG

Senior Member
Sep 21, 2004
1,345
242
Plovdiv
Samsung Galaxy S21 Ultra
The better CPU scores on FROYO are due to the JIT Compiller! The JP firmwares Have JIT but it is not active still! Or atleast not to the point it is on the N1! I hope that the final firmware will have full speed.
 

distortedloop

Senior Member
Feb 14, 2010
3,200
336
Los Angeles, CA
@ykk_five - I'm not really sure what NEW point you're making, but I love that you're still working on this!

Didn't we pretty much already know that rfs was the appaent problem with the i/o? You proved that early on as far as I can tell. Paul O and mimocan assumed it was slow hardware with their fixes, but you showed that a different fs fixed it on the same memory space. Some of us have been clamoring for a reformat of the internal rom space as the solution since you first posted your findings.

What I find more interesting is that the i/o on the stock SGS in Quadrant isn't that bad - it appears comparable to most of the other phones, except the Motorola Droid X! I wonder why it's so fast on that one.

My own earlier READ-speed testing done with reading a 200mb data file from the command line with the dd command showed that read speeds on RFS internal /data partition vs ext4 on sdcard partition were pretty close. Seems like RFS can serve up the reads pretty darned fast (Class 6 speeds or better). Well over 13Mb/s from my post over on Modaco:

Code:
BEST SPEED ON EXTERNAL SDCARD EXT 4 = 15.79 MB/S
BEST SPEED ON EXTERNAL SDCARD FAT32 = 15.6 MB/S
BEST SPEED ON INTERNAL SDCARD FAT32 = 13.1 MB/S
BEST SPEED ON INTERNAL ROM AREA = 13.6 MB/S

Since we're now doing the loop mount ext2 on top of RFS fix, I thought I should test the read speeds there. I ran four tests, they ranged from 1.1MB/s to 8.1MB/s:


Code:
BEST SPEED ON LOOP MOUNTED EXT2 /DATA/DATA = 8.1MB/S

That's right, it's 50% SLOWER to read files off the ext 2 /data/data partition (with RyanZ's 2-3 method) than just raw RFS on the best read, 1/13th of the speed in some cases!

Seems hard to imagine how this fix speeds things up other than cacheing writes, which shouldn't impact the opening of files, unless the lag in the first place was some sort of delay reading the files while other things were written to the disk.

With that in mind, apparently the Quadrant speedup is in the WRITE part of the i/o tests, wouldn't you agree? That would be because of the cached writes ext2 does vs RFS?

If this is the case, then high Quadrant scores aside, none of these fixes should really do much to improve the normal look and feel of using the phone, as you alluded to with your comment about playing back videos/music. RFS or ext will serve up the data at more than we need it.

It makes me wonder why so many of us assert that the phone is indeed much snappier on opening apps and stuff. I'm sure it's not a placebo for me, the phone definitely feels faster with any of the lag fixes we use.

What we really need is to see a breakdown in Quadrant's i/o benchmark that distinguishes between the read and writes.

I really don't like that with all these fixes, no one really knows why they speed the phone up.

I'm tempted to go back to just stock JM5 and see what happens.
 
Last edited: