PDA

View Full Version : ROM Comparison Analysis


jorge_culv
11th October 2006, 06:55 AM
Many of you guys are good at programming and can whip up cabs in your sleep--I'm not so good at C++, etc but here's my contribution, hope it helps someone.

Analysis of Wizard ROMs

Objective: Try to determine if there is performance improvements in newer ROMs. Also to see if the 12->8 MB paging size has a measureable effect in performance.

Summary of Results:
There doesn't appear to be much of a measurable difference in performance among these ROMs, with the exception of the last two ROMs tested. For the rest of the pack, the hardware responds about the same with a little noise. Summitter's 2.17 ROM appears a little above average, but this could be just a testing glitch, and its not significantly above the others. This appears to indicate that the upgrade to decrease the paging pool has no effect (to this benchmark).
I'm concerned about the consistency of results with the last two. They were performed under slightly different test conditions (at home, vs at work for the others), and there's no reason ShogunMark's at least (if not both) shouldn't be closer to the others in performance. I plan to rerun these tomorrow and see how they compare.

Method: I flashed a decent cross-section of the ROMs out there--don't feel insulted if yours isn't included. In fact, I'd be happy to add yours to the mix--just ask! For each ROM listed, I flashed then immediately recorded stats from Settings>System> About, Device Info, and Memory. Then added PC Pro Labs Pocket PC Benchmark and embedded vb runtime (required for Benchmark), removed USB cable, soft-reset and ran benchmark software once to obtain last few columns of results.

Assumptions: The biggest assumption is that this is decent benchmarking software. While it was written for WM2003, its probably not a big stretch to assess WM2005. I konw Spb has a benchmark program, but it ran for about 1.5 hours on the first ROM. Too long for me, guess I'm just impatient! Other big assumption is that higher values are better, although this appears inversely releated for the file read/write benchmarks compared to the kb/sec measurements. This may also be true for the others, butsince they're all about the same it really doesn't matter.

Additional Observations: At the very least, this was an interesting exercise to record some comparable data regarding the different ROMs. ROMs have evolved over time, and sometimes there are questions about which ROM contains what, etc. This might lead to another idea where we keep track of the ROMs in some type of registry to track consistent information about each.

Analysis Improvements: This could be improved by using more current benchmark software. Spb Benchmark is a decent candidate, but takes a while to run. Also, the analysis could be improved with many benchmark runs per ROM instead of just one run. This would average out the "noise" betweens runs and might give a better indication of slight performance trends.
Finally, the method used above will be repeated under similar conditions for the last two ROMs to see if the provided results are valid.

Terms:
BM - Benchmark
GDI - graphics display test
CPU - central processing unit
kb/s - kilobytes per second (used for read/write tests).

shogunmark
11th October 2006, 07:15 AM
wow, its like a report card for roms, lol and good work by the way...

however, is it possible to lay out the results in a better format?

shogunmark
11th October 2006, 07:34 AM
nevermind.. i noticed this new damn board wont let you upload an excel file, i did it myself.. thanks again.. and i agree it is odd that the last 2 are lower than everyone else's but still close to each other.. i would vote computer as well

markgamber
11th October 2006, 05:18 PM
Good job! This pretty much confirms what I've been thinking for a long time...there's no real difference in performance. If you're going to add or alter the list, I'd like to also see the stock T-Mobile 2.26 rom thrown in there as well. Maybe if people see that "it seems faster" doesn't mean anything and they're all pretty much the same, they might think twice about risking their expensive toy for a reason they probably can't define.

lvlolvlo
12th October 2006, 03:04 AM
zip it up and then upload it...

jorge_culv
12th October 2006, 05:15 AM
zip it up and then upload it...

Good idea! Zip includes the Excel file of results and run report for each of 8 runs performed so far with PC Pro benchmark.

Didn't get to updating this with my work computer today, but I will in the next few days.

Vladimyr
12th October 2006, 05:28 AM
hi, jorge_culv, least numbers mean quicker?

jorge_culv
12th October 2006, 05:50 AM
Good question, and I honestly don't know the answer. There's no documentation with this benchmark, even on the hosting company's website. Other benchmarks I've used had higher numbers as better. There is one clue if you look closely at the file read and write tests for each ROM. It appears the lower benchmark scores match up with the faster read and write speeds, so for those it appears lower is better, not sure if that also applies to the CPU and graphics tests.

Also, I'm real hesitent about the last 2 runs--don't read too much into those until I can do more follow up testing (hopefully in the next few days).

jorge_culv
12th October 2006, 06:00 AM
Good job! This pretty much confirms what I've been thinking for a long time...there's no real difference in performance. If you're going to add or alter the list, I'd like to also see the stock T-Mobile 2.26 rom thrown in there as well. Maybe if people see that "it seems faster" doesn't mean anything and they're all pretty much the same, they might think twice about risking their expensive toy for a reason they probably can't define.

Mark, I'll try to throw in a T-Mo ROM as well. I'm not sure if this test really proves there's no difference in performance. I might be splitting hairs, but maybe it only proves there's no difference in CPU math, graphics or file read/write--in other words, the ROM upgrades affect performance in ways not measured by this benchmark. The factors measured by this benchmark sure seem more hardware related and maybe not influenced by the ROM at all. I was really hoping for a difference between the newer "30MB" roms, as that may affect read/write speed through the smaller page file, but the verdict is still out...

markgamber
12th October 2006, 07:03 AM
True, but it might be interesting to take a look at. I don't mean to come down on people creating these roms but personally, if there's a performance difference between any of them and stock 2.26, it's so small that I've never noticed it. When you compare a minute performance difference to the boost of overclocking the cpu, you wonder why you ever bothered risking the phone in the first place. At least I wondered why and, unable to come up with an answer other than "because I can" and not really wanting to throw away $350, I stopped putting new roms on the phone and won't do it again without a damn good reason.

Andy@Contacta
12th October 2006, 03:03 PM
hi, jorge_culv, least numbers mean quicker?
Lower the number the better except for File read/write. I'm intrigued as to what gives the bottom two their boost in the benchmarks.

kamakazi17
12th October 2006, 10:31 PM
so what do you guys suggest as the best rom with both speed / batteries / stability?

pmi
16th October 2006, 06:09 PM
I have 3 ROM version by this moment:

Qtek 2.18

T-mobile 2.26

AKU 3.2 on T-mobile 2.26

And i make test of CPU load on every ROM update by Rhino Stat.

On Qtek 2.18 and AKU 3.2 on T-mobile 2.26 CPU has constantly load 8-9% in stand-by mode.

And T-mobile 2.26 ROM has 0-1% CPU load in stand-by mode.