Memory speed issue on the Droid?

[Elysian]

I loaded UD1.0 on my phone earlier, and immediately went to Quadrant to see how it does... My score went down a touch, but the interesting thing is the reason why... The CPU score is up, quite a bit actually, a good 400 points, 3D is also up. The kicker is, the memory score has been split in half... I've tried 4 different kernels, tried with compcache off and on, tried my old kernel(which I used in my previous Quadrant attempts), and nothing has changed it, which leads me to believe it's an issue with the ROM... Heres some screenshots, the most recent Quadrant is the first image, the old high score on the second. My OG Droid is clocked at 1.1ghz for both of these, I believe the 1521 was done on UD7, the kernel was the ChevyNo1 LV 1.1ghz kernel... Running the Sholes kernel for the most recent screenshot.

Ideas?

 

[Dave12308]

Could be some sort of difference between FRG22-based ROMs and FRG83-based ROMs. Things like this show why synthetic benchmarks are somewhat less than useful.

BTW, how do you get Quadrant to break down the scores with the color-coding? Mine just shows an overall score.

Edit: Nevermind, just realized that I need the paid version to break the scores down

 

[Elysian]

Could be some sort of difference between FRG22-based ROMs and FRG83-based ROMs. Things like this show why synthetic benchmarks are somewhat less than useful.

BTW, how do you get Quadrant to break down the scores with the color-coding? Mine just shows an overall score.

Edit: Nevermind, just realized that I need the paid version to break the scores down

I think it's a good enough measure to tell if there is anything in the system not acting properly, like the memory controller in this instance... Comparing sythetic benchmarks to other phones doesn't really tell the whole story, for sure.

 

[Dave12308]

I think it's a good enough measure to tell if there is anything in the system not acting properly, like the memory controller in this instance... Comparing sythetic benchmarks to other phones doesn't really tell the whole story, for sure.

The thing is, it's only useful for that purpose with an established baseline. For example, an FRG22D based ROM runs at X speed with Y kernel. If I flash Z kernel, I can tell whether it's faster or slower.

However, in this case we're comparing what is likely FRG22D based with UD1.0.0 which is FRG83. So it's a bit hard to extract meaningful data.

I use it to check for stability. If a given OC causes an FC issue, freezes the phone, or initiates a sudden reboot; I figure it's overclocked too highly.

 

[Elysian]

I think it's a good enough measure to tell if there is anything in the system not acting properly, like the memory controller in this instance... Comparing sythetic benchmarks to other phones doesn't really tell the whole story, for sure.

The thing is, it's only useful for that purpose with an established baseline. For example, an FRG22D based ROM runs at X speed with Y kernel. If I flash Z kernel, I can tell whether it's faster or slower.

However, in this case we're comparing what is likely FRG22D based with UD1.0.0 which is FRG83. So it's a bit hard to extract meaningful data.

I use it to check for stability. If a given OC causes an FC issue, freezes the phone, or initiates a sudden reboot; I figure it's overclocked too highly.

The issue with this train of thought is that memory bandwidth should not change, unless there is an issue. It especially shouldn't be halved, this is hardware we're talking about after all. I did not experience this sort of issue with UD8, the benchmarks were consistent with my UD7 benches, so an issue has definitely arisen. I just happened to screenshot my UD7 results and not my UD8, since there was no issue.

 

[haledroid]

I've been running ud roms since 2.1, this is by far the best one out of all. Battery life has improved, everything runs smooth. Im using the 1ghz kernel by slayer, compcache enabled and cpu set at 900mhz. Quadrant can suck my left nut. This is a great rom!

Sent from my Ultimate Droid 1.0.0 Revolution Theme Luigi90210 using Tapatalk

 

[haledroid]

To add you should disable compcache and run another benchmark. I doubt quadrant is setup to understand how compcache handles system memory. Also read up on compcache. You'll find your answer.

Sent from my Ultimate Droid 1.0.0 Revolution Theme Luigi90210 using Tapatalk

 

[Elysian]

To add you should disable compcache and run another benchmark. I doubt quadrant is setup to understand how compcache handles system memory. Also read up on compcache. You'll find your answer.

Sent from my Ultimate Droid 1.0.0 Revolution Theme Luigi90210 using Tapatalk

Like I said in the OP, I tried with and without compcache enabled, as well as a kernel that doesn't even support compcache. No difference.

I also don't know why you're assuming I'm knocking the rom? It's good, sure, but if I've found a legitimate issue, it could very well be even better.

 

[Dave12308]

The issue with this train of thought is that memory bandwidth should not change, unless there is an issue. It especially shouldn't be halved, this is hardware we're talking about after all. I did not experience this sort of issue with UD8, the benchmarks were consistent with my UD7 benches, so an issue has definitely arisen. I just happened to screenshot my UD7 results and not my UD8, since there was no issue.

I guess what I am saying is that we're talking about using SOFTWARE to measure the performance of the HARDWARE. Perhaps there is something different in FRG83 that causes the SOFTWARE to incorrectly report the performance of the HARDWARE.

In other words, i'm not saying something changed between FRG22D and FRG83 to make the hardware physically underperform; rather that something in FRG83 is causing the Quadrant app to incorrectly measure that particular portion of the benchmark score. I'm basically thinking that Quadrant is somewhat incompatible with this particular build.

If UD1.0.0 was truly halving memory performance, i'm thinking it would be noticable in circumstances other than just while running Quadrant. I haven't noticed any other apps underperforming at all.

 

[Elysian]

The issue with this train of thought is that memory bandwidth should not change, unless there is an issue. It especially shouldn't be halved, this is hardware we're talking about after all. I did not experience this sort of issue with UD8, the benchmarks were consistent with my UD7 benches, so an issue has definitely arisen. I just happened to screenshot my UD7 results and not my UD8, since there was no issue.

I guess what I am saying is that we're talking about using SOFTWARE to measure the performance of the HARDWARE. Perhaps there is something different in FRG83 that causes the SOFTWARE to incorrectly report the performance of the HARDWARE.

In other words, i'm not saying something changed between FRG22D and FRG83 to make the hardware physically underperform; rather that something in FRG83 is causing the Quadrant app to incorrectly measure that particular portion of the benchmark score. I'm basically thinking that Quadrant is somewhat incompatible with this particular build.

If UD1.0.0 was truly halving memory performance, i'm thinking it would be noticable in circumstances other than just while running Quadrant. I haven't noticed any other apps underperforming at all.

xScope 6 has been running pretty terribly for me since moving to 1.0, noticed that last night while I was out.

 

[Elysian]

Loaded UD8 back on here, and using the same kernel as the UD1 Quadrant screenshot, the memory benchmark score has doubled... There is obviously an issue...

 

[luigi90210]

im also getting the same memory issue

 

[Elysian]

UD 2.0 fixes the memory bandwidth issue, glad I can move on past UD8

 

[Lars1]

Good news!