Another rant about peformance (SPB Benchmark included) - Tilt, TyTN II, MDA Vario III General

Hi !
I was a little suspicios about the Kaiser... So here are MY SPB Benchmark results( i didn'r run the included apps tests or the battery/storage tests)
Spb Benchmark index 311.74 (iPAQ 3650 scored 1000)
CPU index 1562.67 (iPAQ 3650 scored 1000)
File system index 123.95 (iPAQ 3650 scored 1000)
Graphics index 3474.68 (iPAQ 3650 scored 1000)
ActiveSync index insufficient number of tests were done. (iPAQ 3650 scored 1000)
Platform index insufficient number of tests were done. (iPAQ 3650 scored 1000)
The CPU index is kinda low... as expected
btw: i think the graphics index is somehow fake ( i can see tearing while the test is done, and i also see tearing in normal usage, which leads me to believe the Kaiser is cheating)
PS: I still love my Kaiser, it's just that it's not everything i thought it will be (hopefully a ROM update will show us it's true power)

gxmark benchmark
I just ran the gxmark app... this is closer to reality: Prophet and Kaiser graphics performance. This is closer to what i'm feeling using the device

Can someone run a SPB Benchmark ? But don't give me links from the internet (i saw those and they were pretty), just run one yourself if you can.

Don't forget that the Kaiser is dual core
So it's going to be a little hard doing an apples-to-apples comparison based on the current version of SPB, which isn't really designed to test that way. I also tried SK Tools benchmark tests - the Kaiser is up at the top of the Pocket PC Phones - mainly because of file system access speed rather than pure processing power.

It's not dual core, there's a separate processor for the radio, that's all

Related

Making Wizard run faster

Is there anyway to, as they say 'overclock' the wizard. It's a great PPC/phone, but runs quite slow. Thanks. [/b]
try the search button. use querry "omap"
The utility in this thread is what you need: http://forum.xda-developers.com/viewtopic.php?t=55782
It works on the Wizard, Charmer and Prophet. After installing you should find that you can run your device at 260Mhz rather than the factory set 195. 8)
Was able to download the battery-plug-in software. didn't damage anything on my wizard, but am not realy sure if it increased the speed or not. stil seems mighty slow.
You'll find this by searching too, scattered about. Programs like SPB Diary, SPB Pocket Plus make the Wizard sluggish even when you've overclocked. I learned the hard way to be very choosy about programs that run automatically. Today agenda, SPB weather, PHM Traylaunch and Batterystatus are the only things on my today screen. Smartskey insures that practically nothing is running in the background. I have the latest rom, and I used Xelencin's version (now my own version) so that I have much more space on the device. The effect of all this is that my device is very snappy, except when it decides it needs to talk with the network at length and a good 15 seconds can go by while it does something with the network. Then it just doesn't do anything for 15 seconds, which seems like an eternity when you are about to show someone something, need to check an address or phone number, etc. But 90% of my complaints went away when I eliminated automatic programs, many of which I'd spent money on. It isn't that these programs don't offer functionality, they do, but most of that functionality is also offered by programs that have a smaller footprint.
You'll find this by searching too, scattered about. Programs like SPB Diary, SPB Pocket Plus make the Wizard sluggish even when you've overclocked. I learned the hard way to be very choosy about programs that run automatically. Today agenda, SPB weather, PHM Traylaunch and Batterystatus are the only things on my today screen. Smartskey insures that practically nothing is running in the background. I have the latest rom, and I used Xelencin's version (now my own version) so that I have much more space on the device. The effect of all this is that my device is very snappy, except when it decides it needs to talk with the network at length and a good 15 seconds can go by while it does something with the network. Then it just doesn't do anything for 15 seconds, which seems like an eternity when you are about to show someone something, need to check an address or phone number, etc. But 90% of my complaints went away when I eliminated automatic programs, many of which I'd spent money on. It isn't that these programs don't offer functionality, they do, but most of that functionality is also offered by programs that have a smaller footprint.
I like SPB Diary and Pocket Plus though. Between the larger cache trick and OCing it with battery status, the speed is fine for me. It was a mite too slow without the OC.
After using it for a little less than half a day, I do feel that there is some improvement in terms of speed. However, it still pales in comparison with the xda mini which I used to use. Is this the maximum speed that the wizard can process?
omapclock
Do you have omap running? I run mine at 247MHz but some people are running as high as 260 and report stability.
Re: omapclock
prestonmcafee said:
Do you have omap running? I run mine at 247MHz but some people are running as high as 260 and report stability.
Click to expand...
Click to collapse
Yes - 260Mhz is just fine. 8)
I use omapclock and clock to 264 MHz with it, rock solid at that speed, but higher speeds are no go. The battery plugin shows 286 MHz when omapclock says 264 though
Yeah, have mine running at 260mhz, but don't see much improvement still. Am i doing something wrong? I copied the cab file into the main memory (but not directly into the windows folder), then executed it.
Yeah, have mine running at 260mhz, but don't see much improvement still. Am i doing something wrong? I copied the cab file into the main memory (but not directly into the windows folder), then executed it.
How do you know it is at 260?
If you use batterystatus, you can toggle back and forth from 260 to 195. That way you know if it is running and how fast. It doesn't need to be in Windows; mine is in My Documents/Programs. But it needs to be turned on; it is off by default even if installed, which is why I asked.

Benchmarks

I decided to start a new thread to experiment instead of posting inside the 1.3 thread.
Even with all this cache boosting, it seems the reported free memory is still above 60MB.
That's with no today plugins running.
Here are the settings I am testing:
Code:
[HKLM\System\StorageManager\FATFS]
"BufferSize"=dword:0x000400 (decimal 1024) (old:256)
"CacheSize"=dword:0x004000 (decimal 16384) (old:4096)
"DataCacheSize"=dword:0x004000 (decimal 16384) (old:4096)
"DLL"="fatfsd.dll"
"EnableCache"=dword:00000001
"EnableCache4Way"=dword:00000001 (old: didn't exist)
"EnableCacheWarm"=dword:00000001
"FatCacheSize"= 0x004000 (decimal 16384) (old:1024)
"Flags"=dword:00001006 (writethrough bla bla)
"MaxCachedFileSize"=dword:00020000 (don't cache >128Kb files)
"Paging"=dword:00000001
"PathCacheEntries"=dword:00000400 (1024) (old:256)
"UpdateAcces"=dword:00000000 (do not update access times, should have a positive impact flash wearing)
[HKLM\System\StorageManager\IMGFS]
"BufferSize"=dword:0x000400 (decimal 1024) (old:not sure)
"CacheSize"=dword:0x004000 (decimal 16384) (old:not sure)
"DataCacheSize"=dword:0x004000 (decimal 16384) (old:not sure)
"PathCacheEntries"=dword:00000400 (1024) (old:256)
"UpdateAcces"=dword:00000000 (do not update access times, should have a positive impact flash wearing, not sur if it has any effect on IMGFS)
[HKLM\System\StorageManager\Profiles\FlashDrv\FATFS]
"BufferSize"=dword:0x000400 (decimal 1024) (old:256)
"CacheSize"=dword:00000400
"EnableCache"=dword:00000001
"PathCacheEntries"=dword:00000400 (1024) (old:not sure)
[HKLM\System\StorageManager\Profiles\FlashDrv\IMGFS]
"BufferSize"=dword:0x000400 (decimal 1024) (old:256)
"CacheSize"=dword:00000400 (at the cost of 256 Kb RAM)
"EnableCache"=dword:00000001
"PathCacheEntries"=dword:00000400 (1024) (old:128)
Therewas one single test which simply failed, I don't know why. I believe this failed test is the reason why the overall score is unavailable. But the machine is stable so far, I have no complaints.
The results are the following:
Code:
reference machines (from spb benchmark website database):
(1) Compaq iPAQ 3600 Series (2000, 206Mhz)
(2) Asus MyPal A620 (2003, 400MHz)
(3) Asus MyPal A716 (2003, 400MHz)
(4) Fujitsu-Siemens Pocket LOOX 600 (2003, 400MHz)
(5) Fujitsu-Siemens Pocket LOOX 610 (2003, 400MHz)
(6) Toshiba e750 (2003, 400MHz)
(7) Toshiba e755 (2002, 400Mhz)
(8) Dell Axim x51v (WM5, 400MHz, video hardware accel)
AND...
(9) Blue Angel WM5 Helmi 1.3 beta
(1) (2) (3) (4) (5) (6) (7) (8) (9) Comments
Spb Benchmark index 1000 1573 1622 939 1610 1225 1073 - unavailable probably because of the database error
CPU index 1000 1858 1796 1202 1813 1838 1234 1646 upper half, not that far from the others
File system index 1000 1092 1205 670 1175 1128 1270 - unavailable probably because of the database error
Graphics index 1000 4034 3954 1367 3943 688 651 1207 before tweaking it was around 850
ActiveSync index 1000 1458 2192 480 1775 1751 1479 3269 WINNER! And i am using USB 1.1!
Platform index 1000 1277 1510 796 1341 1085 850 - unavailable probably because of the database error
(1) (2) (3) (4) (5) (6) (7) (8) (9) Comments
Write 1 MB file (KB/sec) 794 1285 1284 589 1282 1221 1200 630 2160 Double the 2003 write results!
Read 1 MB file (MB/sec) 18.2 28.7 27.8 16.5 27.4 26.8 20.7 3.78 4.14 Better than the x51v
Copy 1 MB file (KB/sec) 790 1281 1279 581 1279 1252 1180 600 3526 Hmmm, sweet buffers...
Write 10 KB x 100 files (KB/sec) 560 654 928 422 936 859 768 245 314 different from above results, ROM technology might explain the difference...
Read 10 KB x 100 files (MB/sec) 6.35 8.64 10.4 6.34 10.7 9.62 7.6 1.88 1.43 here we lost against x51v
Copy 10 KB x 100 files (KB/sec) 476 500 820 376 830 757 626 247 290 below 2003, but still decent.
Directory list of 2000 files (thousands of files/sec) 123 23.6 22.2 14.1 20.4 20.5 153 1.33 8.19 trouncing x51v, but humble compared to 2003
Internal database read (records/sec) 421 1549 1539 1059 1518 1391 503 1950 error THIS is the mysterious ERROR...
Graphics test: DDB BitBlt (frames/sec) 26.9 316 309 185 305 68.4 42.3 277 109 HW graphics might give x51v the edge here...
Graphics test: DIB BitBlt (frames/sec) 13.5 27.2 27.3 14.7 27.3 22.9 29.9 22.2 31.1 but not here :)
Graphics test: GAPI BitBlt (frames/sec) 216 752 725 176 722 73.2 72.4 58.9 134 neither here :D
Pocket Word document open (KB/sec) 31 44.2 105 61.9 41.6 37.7 28.3 12.7 7.3 long loading times... UPX'ed apps might have an impact here?
Pocket Internet Explorer HTML load (KB/sec) 13.1 7.88 9.27 3.74 9.49 7.28 6.67 7.05 3.61 ditto
Pocket Internet Explorer JPEG load (KB/sec) 52.8 154 239 149 245 233 105 135 53.8
File Explorer large folder list (files/sec) 515 641 598 382 625 569 291 483 58.6
Compress 1 MB file using ZIP (KB/sec) 106 263 241 152 249 243 89.1 241 195 not beating x51v every time, but still \
Decompress 1024x768 JPEG file (KB/sec) 319 613 609 426 609 607 567 657 533 a good performer...
Arkaball frames per second (frames/sec) 108 250 242 102 245 61 55.7 52.6 96.3 seems x51v video drivers are teh suxx0rz
CPU test: Whetstones MFLOPS (Mop/sec) 0.046 0.076 0.076 0.061 0.077 0.075 0.076 0.117 0.074
CPU test: Whetstones MOPS (Mop/sec) 34.1 55.3 55.4 54.3 55.5 55.2 55.4 84.9 54.8
CPU test: Whetstones MWIPS (Mop/sec) 2.98 5.01 5.02 4.04 5.03 4.99 4.94 7.53 4.82
Memory test: copy 1 MB using memcpy (MB/sec) 70.4 103 98.4 61.2 99 106 90.6 116 102
ActiveSync: upload 1 MB file (KB/sec) 115 135 203 46.9 158 157 135 - 332
ActiveSync: download 1 MB file (KB/sec) 94 250 377 67.8 367 345 274 - 409
There is only one WM5 device besides teh BA, which I collected from this review:
http://www.mobiletechreview.com/Dell-Axim-X51v.htm
Helmi, can you make a hotfix cab called 'Brazilian_Joe_UNTESTED_performance_tweaks.cab' to apply these tweaks and add it to Helmi 1.3 beta front page, with a big red fat warning?
EDIT: The 'internal database benchmark error' happens even with the default 'Helmi 1.3 beta' registry values, so it's not a by product of my settings. Just to reinstate, even with this error the machine still runs without problems.
Nice work, Brazilian Joe! But the real question is: do these tweaks affect the real performance of the BA? I mean do you notice any improvements during normal work (with all today plug-ins, etc.)
That's actually all that matters, because benchmark results don't necessarily reflect the real-life performance.
If you say the difference is noticable (I mean "really" noticable), the we all should give it a try. But if it's only a benchmark improvement, I think we should not mess with the settings...
Regards,
Martin
MHoefler said:
Nice work, Brazilian Joe! But the real question is: do these tweaks affect the real performance of the BA? I mean do you notice any improvements during normal work (with all today plug-ins, etc.)
That's actually all that matters, because benchmark results don't necessarily reflect the real-life performance.
If you say the difference is noticable (I mean "really" noticable), the we all should give it a try. But if it's only a benchmark improvement, I think we should not mess with the settings...
Regards,
Martin
Click to expand...
Click to collapse
I have the whole SBSH doing a party on my today screen under normal circumstances (PocketBreeze etc). My first benchmark (which I lost) still had all those on, andthe numbers were inequivocal.
It was not a 'whoa, look! 1KB/s write improvement!'...
when I said improved responsiveness, that's what it meant. switching tabs in today plugins, program loading, most tasks improved, the experience was more fluid.
Then I procceeded to have more barebone results to avoid having the programs distort the numbers.
That said, I completely agree with your consideration. My next steps are going to be:
1) reinstall everything
2) do another 'vanilla settings' test with everything running
3) put the tweaked settings back on
4) benchmark one more time...
So that we can verify the results.
Just working on my Settings also ..
seems I have forgotten somethin in the settings since I cant't achive the same results than earlier .. (reminds me on helmi not remembering this WiFi driver thing )
EDIT:
Test Time Speed % of iPAQ 3650* speed
Write 1 MB file 99.2 ms 10323 KB/sec 1300%
Read 1 MB file 58.1 ms 17.2 MB/sec 94%
Copy 1 MB file 39.4 ms 25990 KB/sec 3292%
Write 10 KB x 100 files 1718 ms 596 KB/sec 106%
Read 10 KB x 100 files 269 ms 3.72 MB/sec 59%
Copy 10 KB x 100 files 1362 ms 752 KB/sec 158%
Directory list of 2000 files 238 ms 8.41 thousands of files/sec 7%
Some more testing required .. will set up the same Bench as BJ next time
EDIT2:
Okay .. here is what I am doing .. atm I am only switching values in the FATFS:
BufferSize:16 or 8 try around
CacheSize:0 (System reserves Memory) or 16384
DataCacheSize: 4096
will do more testing .. but smaller buffers seems to be much faster ...
Helmi 1.3 vanilla settings, no apps installed.
Free memory: 67.39 MB
CPU index 1386.85
Graphics index 1154.58
ActiveSync index 2502.05
Test - Time - Speed - % of iPAQ 3650 speed (reference machine)
Write 1 MB file 2471 ms 414 KB/sec 52%
Read 1 MB file 440 ms 2.27 MB/sec 12%
Copy 1 MB file 2014 ms 508 KB/sec 64%
Write 10 KB x 100 files 7968 ms 129 KB/sec 23%
Read 10 KB x 100 files 960 ms 1.04 MB/sec 16%
Copy 10 KB x 100 files 7567 ms 135 KB/sec 28%
Directory list of 2000 files 2692 ms 0.743 thousands of files/sec 1%
Internal database read error
Graphics test: DDB BitBlt 9.71 ms 103 frames/sec 383%
Graphics test: DIB BitBlt 32.7 ms 30.6 frames/sec 226%
Graphics test: GAPI BitBlt 7.81 ms 128 frames/sec 59%
Pocket Word document open 43656 ms 5.97 KB/sec 19%
Pocket Internet Explorer HTML load 7958 ms 3.11 KB/sec 24%
Pocket Internet Explorer JPEG load 5221 ms 48.5 KB/sec 92%
File Explorer large folder list 35332 ms 56.6 files/sec 11%
Compress 1 MB file using ZIP 8807 ms 115 KB/sec 108%
Decompress 1024x768 JPEG file 496 ms 566 KB/sec 177%
Arkaball frames per second 10.6 ms 94.8 frames/sec 88%
CPU test: Whetstones MFLOPS 5369 ms 0.069 Mop/sec 150%
CPU test: Whetstones MOPS 1211 ms 52 Mop/sec 153%
CPU test: Whetstones MWIPS 10821 ms 4.62 Mop/sec 155%
Memory test: copy 1 MB using memcpy 10.2 ms 98.5 MB/sec 140%
ActiveSync: upload 1 MB file 4145 ms 247 KB/sec 215%
ActiveSync: download 1 MB file 3013 ms 340 KB/sec 362%
Storage card test results
Storage card "RAMdisk"
Speed index = 369.2
Test Time Speed % of iPAQ 3650* speed
Writing 1 MB file 109 ms 9377 KB/sec
Reading 1 MB file 77.7 ms 12.9 MB/sec
Copying 1 MB file to storage card 581 ms 1763 KB/sec
Copying 1 MB file from storage card 1524 ms 672 KB/sec
Writing 100 of 10 KB files 2275 ms 450 KB/sec
Reading 100 of 10 KB files 352 ms 2.84 MB/sec
Copying 100 of 10 KB files to storage card 2330 ms 440 KB/sec
Copying 100 of 10 KB files from storage card 7744 ms 132 KB/sec
Directory listing: 2000 files 247 ms 8.11 thousands of files/sec
Storage card "Storage Card"
Speed index = 344.21
Test Time Speed % of iPAQ 3650* speed
Writing 1 MB file 479 ms 2139 KB/sec
Reading 1 MB file 71.5 ms 14 MB/sec
Copying 1 MB file to storage card 865 ms 1184 KB/sec
Copying 1 MB file from storage card 1831 ms 559 KB/sec
Writing 100 of 10 KB files 2489 ms 411 KB/sec
Reading 100 of 10 KB files 449 ms 2.23 MB/sec
Copying 100 of 10 KB files to storage card 3131 ms 327 KB/sec
Copying 100 of 10 KB files from storage card 8187 ms 125 KB/sec
Directory listing: 2000 files 247 ms 8.11 thousands of files/sec
MHoefler said:
Nice work, Brazilian Joe! But the real question is: do these tweaks affect the real performance of the BA? I mean do you notice any improvements during normal work (with all today plug-ins, etc.)
That's actually all that matters, because benchmark results don't necessarily reflect the real-life performance.
If you say the difference is noticable (I mean "really" noticable), the we all should give it a try. But if it's only a benchmark improvement, I think we should not mess with the settings...
Regards,
Martin
Click to expand...
Click to collapse
It affects the real / overall / daily use performance. Try it out
Yeah, there are lies, damn lies, and statistics. and benchmarks are statistics, right?
Benchmarks do not directly translate in real world performance, but if they are done properly, there is some meaning to them, and you can expect that the performance displayed at the benchmarks will be reflected in your daily usage.
And I started those benchmarks by going the other way round:
I found some settings, tweaked them, noticed a difference, and only then I started benchmarking to back up my subjective feeling with hard numbers.
let's see how the benchmarking saga unfolds...
Blame me stupid ... installed SPB Bench in RamDisk at first .. damn ...
Rebenching now, but can confirm that with the Settings fomr BJ an mine mixed to the way you use your device in daily usage this runs veeerrry smooth ... )
MasterMerlin said:
Just working on my Settings also ..
Okay .. here is what I am doing .. atm I am only switching values in the FATFS:
BufferSize:16 or 8 try around
CacheSize:0 (System reserves Memory) or 16384
DataCacheSize: 4096
will do more testing .. but smaller buffers seems to be much faster ...
Click to expand...
Click to collapse
I have tried setting only
BufferSize:16
but after I lowered it from 256 to 16 if always froze my machine. had to hard reset to get it back to a useable state...
Code:
[HKLM\System\StorageManager\FATFS]
"BufferSize"=(decimal 16) (old:256)
"CacheSize"=(decimal 0) (old:4096)
"DataCacheSize"=(decimal 0) (old:4096)
"DLL"="fatfsd.dll"
"EnableCache"=dword:00000001
"EnableCache4Way"=dword:00000001 (old: didn't exist)
"EnableCacheWarm"=dword:00000001
"FatCacheSize"= (decimal 0) (old:1024)
results in
Write 1 MB file 2635 ms 389 KB/sec 49%
Read 1 MB file 301 ms 3.32 MB/sec 18%
Copy 1 MB file 1799 ms 569 KB/sec 72%
Write 10 KB x 100 files 7197 ms 142 KB/sec 25%
Read 10 KB x 100 files 641 ms 1.56 MB/sec 25%
Copy 10 KB x 100 files 5892 ms 174 KB/sec 37%
Directory list of 2000 files 1844 ms 1.08 thousands of files/sec 1%
seems that the smaller buffer helps while reading
Brazilian Joe said:
I have tried setting only
BufferSize:16
but after I lowered it from 256 to 16 if always froze my machine. had to hard reset to get it back to a useable state...
Click to expand...
Click to collapse
Thats weird .. since you reinstalled it fresh .. mine is up since yesterday, running Pocketbreeze and some other Today Plugins while using the Benchmark ..
Have you tweaked some other infos before in another section?
MasterMerlin said:
Brazilian Joe said:
I have tried setting only
BufferSize:16
but after I lowered it from 256 to 16 if always froze my machine. had to hard reset to get it back to a useable state...
Click to expand...
Click to collapse
Thats weird .. since you reinstalled it fresh .. mine is up since yesterday, running Pocketbreeze and some other Today Plugins while using the Benchmark ..
Have you tweaked some other infos before in another section?
Click to expand...
Click to collapse
no, I only changed that and nothing else. I think the system was starved because of the small buffer size. I had PocketBreeze, iLauncher and PocketWeather on my today screen.
EDIT. trying the Chache = 0 settings and see how they perform, without messing with the buffer
EDIT2:
MasterMerlin, looks like there is some fluctuation in the file results. can you re-run the benchmark more than one time? I am not getting a result like your 'Read 1 MB file ~300 ms'. maybe the result in that specific run was better than average?
EDIT3: definitely, buffers=16 is not friendly at least to PocketBreeze. Other apps might gasp on it either. At the moment I installed PB, my machine froze. There goes another hard reset...
PocketBreeze and PocketWeather are on my Today Screen too. Will rerun the Bench now .. went to bed yesterday
EDIT:
You are right Brazilian Joe. The Benchmark seems to fluctuate much.
Retried with:
Write 1 MB file 2566
Read 1 MB file 433 ms
Copy 1 MB file 2042 ms
Write 10 KB x 100 files 6405 ms
Read 10 KB x 100 files 913 ms
Copy 10 KB x 100 files 6292 ms
Directory list of 2000 files 2593 ms
Will test around your Settings with other Buffers and some more reg-changes and rerun another Test...since this great ROM should stay stable and speedy ...
MHoefler said:
Nice work, Brazilian Joe! But the real question is: do these tweaks affect the real performance of the BA? I mean do you notice any improvements during normal work (with all today plug-ins, etc.)
That's actually all that matters, because benchmark results don't necessarily reflect the real-life performance.
If you say the difference is noticable (I mean "really" noticable), the we all should give it a try. But if it's only a benchmark improvement, I think we should not mess with the settings...
Regards,
Martin
Click to expand...
Click to collapse
Its should make a difference... I've try it before I post the rom, but I dont test it much, but it seems our Brazilian Joe hv test it more then me.. so give it a try... I'll post a cab for it as requested.
ps: I dont think it safe to try this tweak on v1.2 or lower, but try it anyway...
If we use Brazilian Joe tweaks and later want to later go back to original configuration - how do we do that?
Manually?
or
Uninstall the cab?
hasanj4 said:
If we use Brazilian Joe tweaks and later want to later go back to original configuration - how do we do that?
Manually?
or
Uninstall the cab?
Click to expand...
Click to collapse
hmm, I think if u use it in v1.3 its safe, but this is the revert back .CAB
btw I use Platformxxx.reg method (so its ready for further tweak update from Brazilian Joe and wont erase the reg entry if u uninstall it)
Just install this cab overwrite or uninstall the previous version, both way is ok. this .CAB will revert back registry value to its original value.
read my post i te main thread, these cache tweaks shouldnt be used on thier own, the speed is shortlived.
Midget_1990 said:
read my post i te main thread, these cache tweaks shouldnt be used on thier own, the speed is shortlived.
Click to expand...
Click to collapse
Is that like the upgraded Axim/4700 filesys.exe performance issues? I didn't knew our machines were affected by it too.
Anyway, I'll keep the tweaks and see if the machine slows down over time.
Brazilian Joe said:
MasterMerlin said:
Brazilian Joe said:
I have tried setting only
BufferSize:16
but after I lowered it from 256 to 16 if always froze my machine. had to hard reset to get it back to a useable state...
Click to expand...
Click to collapse
Thats weird .. since you reinstalled it fresh .. mine is up since yesterday, running Pocketbreeze and some other Today Plugins while using the Benchmark ..
Have you tweaked some other infos before in another section?
Click to expand...
Click to collapse
no, I only changed that and nothing else. I think the system was starved because of the small buffer size. I had PocketBreeze, iLauncher and PocketWeather on my today screen.
EDIT. trying the Chache = 0 settings and see how they perform, without messing with the buffer
EDIT2:
MasterMerlin, looks like there is some fluctuation in the file results. can you re-run the benchmark more than one time? I am not getting a result like your 'Read 1 MB file ~300 ms'. maybe the result in that specific run was better than average?
EDIT3: definitely, buffers=16 is not friendly at least to PocketBreeze. Other apps might gasp on it either. At the moment I installed PB, my machine froze. There goes another hard reset...
Click to expand...
Click to collapse
Hey Joe,
The BA is a strange device, it seems, that it behaves itself in different ways So... It can be, that MasterMerlin is right. I tried the Cache=0 and Buffersize=16 on my BA, and It didn't freeze at all, and It became more speedier than with your settings. I use SPB pplus + diary + weather + time + backup on today (tabbed) + wireless + messaging + device lock on todayscreen.
The device freezes, if you try to raise Your posted setting's values.
I hope i could help a bit.. But I really don't understand why your BA freezes if you set cache to 0 and buffer to 16...
ThExSenatoR said:
Hey Joe,
The BA is a strange device, it seems, that it behaves itself in different ways So... It can be, that MasterMerlin is right. I tried the Cache=0 and Buffersize=16 on my BA, and It didn't freeze at all, and It became more speedier than with your settings. I use SPB pplus + diary + weather + time + backup on today (tabbed) + wireless + messaging + device lock on todayscreen.
The device freezes, if you try to raise Your posted setting's values.
I hope i could help a bit.. But I really don't understand why your BA freezes if you set cache to 0 and buffer to 16...
Click to expand...
Click to collapse
As I told before, SBSH PocketBreeze dislikes this setting. Anyway, I will try setting the caches to 0, but keeping the buffer size and see if it improves. Then I will proceed on halving the buffer until it does not work anymore.
EDIT:
Just to clarify, I use PocketBreeze with iLauncher, ContactBreeze and PocketWeather as tabs inside it.
Anyway, the buffers = 16 setting looked like it was working for me, but when I installed PocketBeeze it froze...
/EDIT
On the subject of benchmark fluctuation, I think the cause is the same thing which also causes video to skip, some system process might be taking up too much processing.
Since our hardware is not built for WM5, we might be suffering the same problems that the Axim/4700 users have. Do we have a filesys.exe throttler? If we don't, is there a way to do it? It would be so much easier to find the guilty process if we had a XP-style task manager, which displayed the amount of cpu time each process is taking... Anyone knows of any?

Benchmarks

Who's up to creating a table of benchmarks for ROMs like this? I would, but I don't even have time to cook much anymore... :-/
following problems:
- the benchmarks have to be amde on one device with the same settings, etc....
- every overclocking gives a new benchmark.
- sometimes 2 benchmarks with one and the same os is different after using a program....
i used sktools benchmark and often got different benchmarks which were not comparisable.
and the best benchmark in my opinion is:
"WOW that feels fast when i´m using it...."
if this is correct or not - but it makes me happy while using the device
You would run with no overclocking. Settings would make a difference only because the cook chose to set them in the ROM like that for a reason. You could have a stability section, a compatibility section (with programs (i.e. NetCF 3.5 issues)) and the benchmark section. It'd be kinda useless, but it would make me and hopefully other cooks aware of what they should work on.
I would install the ROM. Let the phone rest for 5 minutes without launching programs to let all the start up stuff finish, then check for activesync and close it. Then run the tests.

[APP][EXPERIMENTAL] MobileBooster v.a1 [UPD.:21-03-2010]

(Current version: alpha 1) DISCONTINUED!
MobileBooster is an experimental CPU-booster tool (experimental as if it may have no effect at all in worst case).
In theory this application can improve the speed and responsiveness of your device in high-load scenarios. It has no settings and is completely invisible.
Features:
- simple and clean installation and uninstalation
- boosts your multitasking capabilities by managing process priorities
- slightly boosts active application speed at the expense of background tasks
- supports all HTC devices
Requirements:
- 275kb of free space
- 148kb of RAM
- Microsoft .NET Compact Framework 3.5
Installation:
- download and install the CAB to your device
- soft reset and you're all set!
Uninstalation:
- uninstall using Remove Programs, no reset is required
- be aware that you DO need to soft reset to eliminate all boosting effects (for benchmarks/comparing speed)
The usual disclaimer:
This application is provided as-is. I take no responsibility for any damages that may occur by misuse or bugs in this software.
Project goal:
This is experimental concept and it may have no effect at all save for consuming CPU cycles. If you download and test this, please post back your impressions and opinions. You may experience increased fluidity of your device during normal use and better responsiveness under heavier load. You may experience nothing at all. You may experience decreased speed. Whatever this does for you, uninstall will wipe the application clean so this should be a risk-free test.
Note that your device is less responsive immediatelly after soft reset - this is normal WM behavior. Give it a few minutes before testing.
.
let's give it a go on my TP2 with AthineOS 23549
first impression: Opera 10 seems to render the xda forum pages much quicker. Now lets's play BoingGLES
Yeah... "SEEMS" is the keyword here, Opera indeed seems to be working faster for me, especially with running background tasks and Sense is a bit more responsive, but even though the theory behind this is valid and time proven on single-core PCs, it may all be just a placebo effect.
That's why I posted it for others to see and judge
It could just be a roll of the dice, but gscroll is actually working perfectly (on ereader) with this running. I probably shouldn't have said anything, cuz it'll decide to take a lunch-break soon.
Can't tell much else, but it doesn't seem to break anything.
each restart of the HTC HD2 it gives error cant start this application,
so i unistalled
On my TP works fine, actually seems the whole system actually speeded-up. I had an error once but never happened again. It takes 624k of RAM, so it's quite something but oh well... keep up doin this good job!
Well it's not exactly optimized for memory or CPU usage, neither does it properly handle runtime errors... this is just experimental proof of concept
I have been testing it extensively and from my experience:
1) it definitely does not slow things down.. if it seems so, it's probably because your device is still booting
2) it has little to no effect when you use only 1 application
3) it seems to help a little with concurrency of 2 or more processes
that can be seen during heavy background network traffic or heavy concurrent tasks
- Opera loading pages and rendering at the same time
- you can notice increased responsiveness while playing music in background and doing stuff
- rotating screen in various scenarios also seems a bit faster
- sound cracking under heavy load should be reduced
- basically the more things you run at once, the better improvement you should see; well until you hit the ceiling of your CPU
... I'm not talking huge improvements here but every little bit helps
4) keep on mind that part of the effect is persistent until you soft reset, even if you uninstall or kill the booster you need to soft reset if you want to bench/compare
Thanks nik3r for the app I tried it installed and uninstalled easy with no errors.
Unfortunately, it greatly slowed down my TP performance. Opera rendering was sluggish, opened apps were windows media player, contacts, WLM, tasks, calendar, activesync. As for opening and using other programs , i did not notice a dip or increase in speed.
i'll try it o my htc touch pro 2 and my HTC mogul...
Could you please post the name of your ROM and pagepool size if you know it? I think that pagepool may have huge impact on the ability to multitask.
I'm not 100% sure it has to do with mobileboost, but I've been having some issues with writing to the sd card. (This is actually kind of funny, lol). Last night, I created a shortcut to mobileboost.exe and put it on my sd card, where I have a second copy of my start menu that I access with the XTask file manager (I put stuff I'm testing out there instead of the real start menu). This morning, I scanned my sd card with Scandisk, and it found an error in the folder where I put the shortcut. Ok, no big deal. Today, I was running some SK Tools benchmarks with mobile boost, and I copied them to my sd card (and created a new folder) while MSB was running. After soft resetting and running some tests w/o it running, I looked in the sd card and the folder with the benchmarks wasn't there. I scanned the disk, and the files were recovered as .chk files. I guess it could be happening because I'm using caches applied to the sd card as well as writeback. But, I've never seen this happen before.
Code:
[HKEY_LOCAL_MACHINE\System\StorageManager\Profiles\SDMemory\FATFS]
"EnableWriteBack"=dword:00000001
"BitmapCacheSize"=dword:00000100
"FatCacheSize"=dword:00000400
"DataCacheSize"=dword:00000c00
Definitely unrelated, MobileBooster installs to the device no matter what you chose and it autostarts (from \windows\startup folder).
Make sure you safely disconnect from PC when you're in flash drive mode and don't connect/disconnect the device in flash mode during any transfers
OK, here are some benchmarks with SK Tools. The first two compare benchmarking w/ and w/o mobileboost running, each after a soft reset. I ran three tests, took the average of each three and show the standard deviations.
W/O the boost:
Test AVG STDEV Units
Integer 329.2143667 2.620626281 Moves/25 usec
Floating point 7.595333333 0.022590558 MWIPS
RAM access 537 31.48015248 Speed index
Draw bitmaps 638.6666667 7.637626158 Speed index
Main storage (write) 2366.656667 29.27902036 KB/sec
Main storage (read) 6871.67 770.7838631 KB/sec
Storage Card (write) 3385.506667 131.6158236 KB/sec
Storage Card (read) 4341.836667 1005.563271 KB/sec
File List 2606.666667 139.2850794 Items/s
FL: Storage Card 2424.666667 26.57693235 Items/s
SKTools loading 2185 86.79285685 ms
Click to expand...
Click to collapse
With the boost running:
Test AVG STDEV
Integer 329.7313333 2.010925713
Floating point 7.614 0.031240999
RAM access 442 1
Draw bitmaps 653 4
Main storage (write) 2420.056667 27.17658244
Main storage (read) 7459.18 177.8904416
Storage Card (write) 2712.48 215.2829438
Storage Card (read) 5341.15 928.2778203
File List 2512.333333 68.23733094
FL: Storage Card 2340.666667 67.82575715
SKTools loading 2187.666667 27.53785274
Click to expand...
Click to collapse
These two are really pretty similar. In my experience, there's a lot of variance in how well the RAM access and bitmap tests run; they aren't consistent at all, so seeing the faster ram access without the boost running probably means nothing.
Next, I ran a set of .mp3's over WMP and did the tests (3 for each, w and w/o the boost, with a soft reset in between the tests). This was to test multi-tasking performance. I ran the same two songs for each test (takes about 5 min per test).
W/o boost:
Test AVG STDEV
Integer 289.113 23.80731014
Floating point 7.300333333 0.015143756
RAM access 531.6666667 10.96965511
Draw bitmaps 729.3333333 4.509249753
Main storage (write) 1840.866667 46.8405992
Main storage (read) 7033.93 267.0015249
Storage Card (write) 2730.49 189.3420405
Storage Card (read) 4724.97 997.641275
File List 2279.333333 34.58805189
FL: Storage Card 2196 40.03748244
SKTools loading 2320 57.61076288
Click to expand...
Click to collapse
W/ the boost:
Test AVG STDEV
Integer 241.0114 13.07119076
Floating point 7.227333333 0.007023769
RAM access 543.6666667 13.0511813
Draw bitmaps 727.6666667 3.055050463
Main storage (write) 1898.493333 59.4344196
Main storage (read) 6885.55 481.199009
Storage Card (write) 2708.99 39.82972383
Storage Card (read) 5276.103333 174.1402014
File List 2342.333333 44.23045708
FL: Storage Card 2057.333333 43.01550108
SKTools loading 2286 11.35781669
Click to expand...
Click to collapse
The integer test is a lot faster w/o the boost running, and the floating point is slightly faster (not a ton). Both take a hit compared to not running WMP, which is a major resource hog. The RAM and bitmap tests are both faster, and this just goes to show that they're pretty random tests and I wouldn't ever get too excited about results with these two tests (unless they're consistently awful, I suppose). The only benchmark (besides the first two tests, which basically measure processor speed or effectiveness) that I care about is the last one, which is the SK Tools boot-time (you have to close the app and re-start to get different measurements). It's pretty good in all cases, and doesn't change that much.
Long story short, I'm not seeing much benefit in these tests, but if it helps to get gscroll to work (not sure 100% yet), then I might use it.
nik3r said:
Definitely unrelated, MobileBooster installs to the device no matter what you chose and it autostarts (from \windows\startup folder).
Make sure you safely disconnect from PC when you're in flash drive mode and don't connect/disconnect the device in flash mode during any transfers
Click to expand...
Click to collapse
Yeah, I dunno if it's related or not, but I was doing all the file transfers with total commander on the device. I didn't use the cab to install mobilespeedbooster, I dumped the contents with sk tools and then started it manually with a shortcut (that I made w/ TC). So, it wasn't running from startup.
Edit: now that I think about it, I've also got these keys applied to the sd card:
Code:
[HKEY_LOCAL_MACHINE\System\StorageManager\Profiles\SDMemory\FATFS\Filters\CacheFilt]
"Dll"="CacheFilt.dll"
"Order"=dword:2
"FileCacheWriteBackPriority256"=dword:FE ;shadrac's settings
"FileCacheWriteBackTimeout"=dword:A
I have a feeling that mobileboost was messing with the writeback thread, and that's why I wasn't seeing files writing properly to my card.
Thanks for the tests, even though synthetic benchmarks on WM are inherently imprecise and this app is more about perceived speed than actual raw speed. Your SD problem at least gave me more ideas, I should really treat system services separately with higher priority and try to decrease the thread scheduling overhead, which may cause the differences in benchmarks.
where can I find the setting app?
cause I cannot find it on System settings and start program....
or just install it without setting and anything else...?
There's no settings, the core is fully automated - just install and soft reset and it's online.
Improved performance on my phone
I'll like to start off by thanking you for this great app I usually just read other people who post and if it looks good I download it and I never post back but I really liked this app so I felt like it deserved some feedback. I've noticed my TP run smoother and faster esp when I'm listening to music . Also I don't know if its from your program but my phone would just go to a white screen for 5 sec when I try to go back to my home screen, but thats about the only issue I've seen so far. Keep up the good work
Mine is Exquisite ROM 4.06 ... and i think the pagepool size is 0... as the thread says...
Torakiki83 said:
Mine is Exquisite ROM 4.06 ... and i think the pagepool size is 0... as the thread says...
Click to expand...
Click to collapse
and i forgot, sense 2.5

WP BENCH - looking for Mango and Samsung phone combo guy

Hi,
becouse of the rules I can't post this thread to the development section
Anyway, I'm looking for some1 with Mango and Samsung phone combo.
I'm working on new version of WP BENCH and I need to know if there's any difference in DATA test between memory and storage part of the test.
If you would be so nice and want to help, please go to the Marketplace, download WP BENCH ($free), run the DATA test and post here your result of both its parts.
Thanks!
Robert
CPU s-8426ms p-7953ms
Data time 6966ms memory-13.07 mb/s. storage-13.07 mb/s.
Gpu 23 frame average 692 frames drawn.
Well thats really strange, the storage speed is same as memory speed and it happens on all devices, even HTCs with MicroSDs.
Looks like
a) Mango optimization is more than amazing
b) They're using some "first save to memory then write to storage in the separate thread" hack
Strange
Almost Same
CPU s-8437ms p-7788ms
Data time 6966ms memory-13.07 mb/s. storage-13.07 mb/s.
Gpu 23 frame average 693 frames drawn.
Samsung Omnia 7 with Mango beta (orange Uk)
Omnia 7 with Mango beta
CPU time-16400ms sequential-8504ms parallel-7896ms
data time-6982ms memory-13.07mb/s storage-13.07mb/s
gpu 701 frames, avg:23fps

Categories

Resources