The Droid2/X use the same graphic processor as Droid 1, which is PowerVR SGX 530. According to the datasheet, this core is designed to run at 200Mhz with power of rendering 14M triangles/sec. But our Droid/Milestone runs underclocked at 110Mhz(7M tri/s) while D2/X at 200Mhz. That leads to major UI responsiveness&gaming difference between D2&D1.
I wonder if there's any possibility to overclock the GPU as well?
Thanks in advance.
Sent from my Milestone using XDA App
TeroZ said:
The Droid2/X use the same graphic processor as Droid 1, which is PowerVR SGX 530. According to the datasheet, this core is designed to run at 200Mhz with power of rendering 14M triangles/sec. But our Droid/Milestone runs underclocked at 110Mhz(7M tri/s) while D2/X at 200Mhz. That leads to major UI responsiveness&gaming difference between D2&D1.
I wonder if there's any possibility to overclock the GPU as well?
Thanks in advance.
Sent from my Milestone using XDA App
Click to expand...
Click to collapse
As far as I know this has been tried (overclocking), but with no results (constant reboots)
Imagination Technologies (PowerVR) defines the GPU internals and sells the "plans" for the part, to be included in SOCs like TI's OMAP.
But PowerVR does not, however, define the exact clocks at which the parts should run, nor other things like number of memory channels, memory speed, etc.
Texas Instruments are the ones who defined the GPU clocks. The OMAP 34xx chips (Droid 1, Milestone, XT720, Flipout, etc) are made using 65nm process, and that determines a certain power consumption using certain clocks, hence why they defined a ~100MHz clock for the GPU and ~600-800MHz for the CPU.
The OMAP 36xx (Droid X, Droid 2, Defy, etc) are made using a newer, smaller 45nm process, which allows them to run at higher speeds while spending approx. the same power, which is why Texas Instruments decided to clock the GPU at ~200MHz and the CPU at ~1-1.2GHz.
So it's not like the Milestones and Droids have their GPUs underclocked, those are just their factory clocks.
Of course, overclocking the GPU would be nice and it could be possible. If someone found out how to change the GPU's voltage and clocks, I'm sure it could come in handy in future games.
However, right now, the 1st gen Milestones/Droids are running every high-end HD game from gameloft at full speed, and I bet it'll even do Infinity Blade and other UE3 games when they're out for Android.
Every "HD" Android game has to be compatible with the 1st-gen Snapdragon's GPU, the Adreno 200, which is a lot slower than the SGX530 @ 100MHz, so we're sitting confortably above the base spec for now. And with all the Windows Mobile 7 phones coming with a 1st-gen Snapdragon (mandatory requirement), it'll be like this for a while.
So there's really not a big need for overclocking the GPU right now, except for getting higher scores in mobile benchmarks (some of them terribly unoptimized, like GLBenchmark 1.1 and 2.0).
Furthermore, I it seems the first factor to limit the 1st-gen Droids in games will be the RAM amount.
The first UE3-based game for Android is already out, and it requires 512MB of RAM.
So the game runs on Nexus One and not on a Droid/Milestone, which has far superior graphics performance.
(I'm pretty sure this has something to do with the fact that Android doesn't allow graphics buffering in the main memory, though, which could be resolved in future firmware revisions).
Then again, overclocking the GPU would be cool, and I'm pretty sure getting our SGX530 to work @ ~200MHz would significantly increase the gaming longevity of our phones for quite a while.
Thanks for your useful and important reply.
"The Manhattan Project" on Galaxy S Series just made me curious about Droid's gpu oc, because SGS also use a PowerVR gpu. But things isn't easy due to a fact that one is made by TI while another is made by samsung, the structure inside both SoCs may be completely different.
But I still hope someone capable would try something on this.
That's really cool and significantly lengthen the lifetime of our Droid and Milestone.
Thx again for your reply!
PS: I also felt strange why the UI(not games) on N1 is faster than an OCed droid, could it be the optimization problem?
Sent from my Milestone using XDA App
TeroZ said:
PS: I also felt strange why the UI(not games) on N1 is faster than an OCed droid, could it be the optimization problem?
Sent from my Milestone using XDA App
Click to expand...
Click to collapse
Definitely part of the optimization --a fast ROM with a good theme like the Droid X theme on the GOT 2.2.1 ROM has as fast a GUI as I've encountered on Android, even without overclock.
Also take in consideration that all the current 2.1 and 2.2 roms have a cap of 30fps in 2D, perhaps when the final 2.2 update arrives there will be some perfomance gain
Sent from my Milestone using Tapatalk
Let me start off by saying that I know that all benchmarks should be taken with a grain of salt, and that Quadrant is a poor and outdated benchmark. But, that does not matter. I'm sure we've all ran at least one Quadrant before. Well, there is an issue with the planets test. The D3 only gets 12fps on it, and there is artifacting throughout the test.
It has been this way for every OMAP based Moto phone. The OG Droid, Droid 2, Droid 2 Global, Droid X, and so on. They ALL run the planets test at 12fps with the same exact artifacts. No matter the GPU or clock speed. I feel like I'm the only one who has noticed this. Go and perform a Quadrant on your D3 and observe the planet test. If you have any other phones, benchmark them too. Compare a D3, D2, and Inc.
The OMAP 4430 has the PowerVR SGX 540 gpu. This was the GPU in the previous gen Hummingbird. On the planets test, it would constantly be at the framerate cap of 56 with no artifacts. The Droid Incredible, even with the pathetic Adreno 200 GPU, would run the planets test well above 30fps. It is not an intense graphics test and nearly any mobile GPU can run it above 30fps.
The only Moto phones that run the planet test without issue are those with the Tegra 2. Keep in mind that Tegra 2 based devices all use nvidia's own proprietary drivers. Sure, the Droid 3 is qHD, and it won't score as well as a Galaxy s or Optimus 3D with the same GPU pushing less pixels. The X2 is also qHD and the Droid 3 typically gets a higher framerate on other GPU benchmarks-except for the planets test. In theory, the Droid 3 should score the same, if not, slightly better, on the planets test.
All of the evidence leads me to believe that it is an issue with Moto's drivers. It isn't TI, since the Optimus 3D has the exact same CPU, but runs the test without a hitch. Is anyone able to provide any insight on this, perhaps someone who can speak with someone at Moto that would know? Will this affect games, or other benchmarks? Can it be fixed in an update? Something is obviously not right here, and I'd like to find out why.
EDIT: I'm on my phone and I was in general when I hit new thread. I don't know how this ended up in development, but I apologize. Will a mod please move this?
Games and emulators on the Droid 3 don't have the FPS problem as compared to other devices. I think it's a coding issue in Quadrant, not a problem in the drivers. It would be like running 3DMark 2000 on today's hardware and having it crash out or run slower than a P3 with a 3DFX card (which it actually does in many cases).
elkay said:
Games and emulators on the Droid 3 don't have the FPS problem as compared to other devices. I think it's a coding issue in Quadrant, not a problem in the drivers. It would be like running 3DMark 2000 on today's hardware and having it crash out or run slower than a P3 with a 3DFX card (which it actually does in many cases).
Click to expand...
Click to collapse
A coding issue IS possible, but if it was a coding issue, then how is it that only specific phones from one manufacturer with CPU's and GPU's from the same manufacturers are the only ones with issues? The only Moto phones that don't have the issue are those with the Tegra 2, which uses proprietary coding and drivers from NVIDIA. There are phones from other manufacturers with the same SoC, and very similar hardware otherwise, that don't have the issue. It seems extremely likely that it is an issue with Motorola's software and coding.
GoogleAndroid said:
A coding issue IS possible, but if it was a coding issue, then how is it that only specific phones from one manufacturer with CPU's and GPU's from the same manufacturers are the only ones with issues? The only Moto phones that don't have the issue are those with the Tegra 2, which uses proprietary coding and drivers from NVIDIA. There are phones from other manufacturers with the same SoC, and very similar hardware otherwise, that don't have the issue. It seems extremely likely that it is an issue with Motorola's software and coding.
Click to expand...
Click to collapse
I'll respectfully disagree. If the same FPS problem exhibited itself in any games on the market, then I would lean toward agreeing with you. However, I have pretty much every emulator on the market and about 75-80 games on my phone, and not one has shown any performance problems and all outperform my original Droid X by a very noticeable margin.
Or perhaps it's both? Quadrant could use something in OpenGL that isn't supported very well by Motorola's drivers, and it could be a feature that isn't widely used in other apps, so it's why you're not seeing any issues in them.
Pokelover980 said:
Or perhaps it's both? Quadrant could use something in OpenGL that isn't supported very well by Motorola's drivers, and it could be a feature that isn't widely used in other apps, so it's why you're not seeing any issues in them.
Click to expand...
Click to collapse
By all means, continue to investigate. I think the problem will end up lying in how Quadrant makes its OGL ES calls. The only one that would help right now is the developer of Quadrant, which if you can get to answer the question, would be great. Not stone hard fact, but a step toward solving this.
Quadrant is not optimised for dual core processors let alone any device that has recently come out in the last year including Tegra devices, the tests are not that complicated in the bench mark yet devices are scoring sub 30 fps values.
Use a different bench mark as Quad is old hat and in need of an update or 3.
-smc
I thought I should note that the PowerVR SGX 540 in our Droid 3's are not the same as the SGX 540 in Samsung's Hummingbird SOC. It's quite a bit faster!
The SGX 540 in the OMAP 4430 is clocked 100MHz higher and gets 4.8GFLOPS vs the Hummingbird's 3.2GFLOPS.
snowblind64 said:
I thought I should note that the PowerVR SGX 540 in our Droid 3's are not the same as the SGX 540 in Samsung's Hummingbird SOC. It's quite a bit faster!
The SGX 540 in the OMAP 4430 is clocked 100MHz higher and gets 4.8GFLOPS vs the Hummingbird's 3.2GFLOPS.
Click to expand...
Click to collapse
Yes, but the higher resolution screen counterbalances that.
GoogleAndroid said:
Yes, but the higher resolution screen counterbalances that.
Click to expand...
Click to collapse
Well, not quite. The screen resolution only increased by 35% while the GPU is 50% more powerful than the SGX 540 in the Hummingbird SOC. In theory the OMAP4 should still perform better than the Hummingbird even at the higher resolution.
I would have to say that Quadrant is at least partly responsible for the low FPS. It seems likely that Quadrant is using an odd method of method of rendering for the planets test that in combination with Motorola/OMAP drivers causes a massive performance drop. Fortunately this performance issue has not been seen in other apps/games.
My main reason for posting was just to point out that the SGX 540 in our D3's is much faster. I would hate to think our GPU is no better than the Galaxy S'
this is a problem with quadrant, all motorola devices(since the mb200 aka dext/cliq we have this issue, always got the same results in 2d test and the planet) give the same score in the tests. but the new sgx 540 for dual core have a dedicated gpu for 2d graphics, i spent 3 days searching about that. so, the problem is with the app, not the phone.
can somebody send me a benchmark from D3 of smartbench, Mandrobench, linpack, antutu and cfbench. im thinkin in getting one =p
Too many variables in synthetic tests to come to a conclusion with Quadrant. One parameter can be off and that will bias the weighted result. Speaking of weight, the dev still has yet to quantify the weighting in the app for each parameter.
D3 plays N64 and PSX games equal for most and better for some (if already 60fps, will inherently be equal) than the Tegra 2 devices I had or have now. In a way, the D3 is the device the Sony Play should have been (except the control pad is better than the keyboard, of course).
guidoido004 said:
can somebody send me a benchmark from D3 of smartbench, Mandrobench, linpack, antutu and cfbench. im thinkin in getting one =p
Click to expand...
Click to collapse
On stock I got ~70mflops on average on multi-threaded Linpack test. With some of the modifications I can do with root, Antutu gets me 5112, and on stock I got somewhere around 4700 I believe. I don't have any of the other benchmarks you listed, so I couldn't tell you for them.
guidoido004 said:
this is a problem with quadrant, all motorola devices(since the mb200 aka dext/cliq we have this issue, always got the same results in 2d test and the planet) give the same score in the tests. but the new sgx 540 for dual core have a dedicated gpu for 2d graphics, i spent 3 days searching about that. so, the problem is with the app, not the phone.
can somebody send me a benchmark from D3 of smartbench, Mandrobench, linpack, antutu and cfbench. im thinkin in getting one =p
Click to expand...
Click to collapse
Antutu- 4967
Smartbench 2010-1263, 2849
Smartbench 2011- 3612, 2586
Linpack- 42, 66
CF-Bench-9397, 2636, 5340
Quadrant- 2000-2500
I couldn't find Mandrobench.
I heard, read and saw some things about the nvidia drivers for tegra 3 being sluggish.
Is it possible that our poor results "tegra3" facing exynos soc could be consequence of the nvidia drivers, since many games are optimized for qualcomm and only few for tegra3. So is it possible to mess with the gpu drivers and get a little extra juice out of it?
Compwxr said:
I heard, read and saw some things about the nvidia drivers for tegra 3 being sluggish.
Is it possible that our poor results "tegra3" facing exynos soc could be consequence of the nvidia drivers, since many games are optimized for qualcomm and only few for tegra3. So is it possible to mess with the gpu drivers and get a little extra juice out of it?
Click to expand...
Click to collapse
If you compare (and i have it done...) the process power of the Tegra 3 vs Exynos, on, for example, ANTUTU, you will see that in PROCESS power we have more much potential, on the GPU side, our LG 4X HD has a freq cap on the GPU that make it runs max to 416Mhz, and that is to low, the HTC One X+ runs at 520Mhz and at that freq, Tegra 3 can reach the GPU power of the MALI 400 GPU, but in processing power, we have more...
Sorry for this question, but I'm very confuse about gpu of I9070,some people said "s advance have single gpu" but on website official novathor u8500 said "multi core gpu process graphics 2d and 3d" So s advance have single or dual gpu ? Thanks http://developer.sonymobile.com/knowledge-base/technologies/novethor-u8500/
S Advance has a single core Mali 400 MP GPU. And as far as I know, the Galaxy S2 also has a Mali 400 MP GPU - but its dual core (instead of just one core). If you read about Mali-400 MP in ARM's website (link), this is what you'll see:
Scalable from 1 to 4 cores the Mali-400 MP enables a wide range of different use cases, from mobile user interfaces up to smartphones, tablets and DTVs, to be addressed with a single IP. One single driver stack for all multi-core configurations simplifies application porting, system integration and maintenance. Multicore scheduling and performance scaling is fully handled within the graphics system, with no special considerations required from the application developer
Click to expand...
Click to collapse
So this shows that different phones can use the same GPU but with different number of cores.
PS: Anyone is free to correct me if I'm wrong.
Sami Kabir said:
S Advance has a single core Mali 400 MP GPU. And as far as I know, the Galaxy S2 also has a Mali 400 MP GPU - but its dual core (instead of just one core). If you read about Mali-400 MP in ARM's website (link), this is what you'll see:
So this shows that different phones can use the same GPU but with different number of cores.
PS: Anyone is free to correct me if I'm wrong.
Click to expand...
Click to collapse
Thanks for answer. It's very stranger because my galaxy S advance runs n.o.v.a 3 smooth and fast (mali400), and I have one tablet based allwinner A13 (mali 400), sometimes I got hard lags ( yes I always optimize my mb ram).