[Ailuros]

Not the right topic anyway, but I'm really wondering for quite some time now where the 9th FLOP per ALU comes from in Series5XT GPU IP.

[wco81]

Well Samsung announced Galaxy Tab 2 at 7 and 10.1-inch screen sizes.

They have ICS and are described as 1Ghz dual-core. They don't try to compete with high-resolution screen on the iPad but they are lower in prices, especially the 7-inch model at $250.

[Ailuros]

These results for the 32nm Exynos should be final:

http://www.glbenchmark.com/phonedeta...chmark=glpro21

[Nebuchadnezzar]

Originally Posted by Ailuros

These results for the 32nm Exynos should be final:

http://www.glbenchmark.com/phonedeta...chmark=glpro21

Galaxy Note 2 model number, smashing.

[Nebuchadnezzar]

So the kernel sources for the S3 were released and the final clock on the Mali is the same as I posted several months ago, 440MHz.

While I'm pretty sure there's no driver magic going on here, the only reasonable explanation would be that the memory bandwidth is vastly improved. Bandwidth tests put the S3 at roughly 30% higher speeds in real-world metrics over the S2. I don't see any other explanation for a 95% performance increase for only 65% clock increase on the GPU. Reports have been posted that it has an "internal 128bit bus" over 64bit in the 4210, but I do not understand how exactly does this help memory bandwidth, as the memory itself remained (apparently) unchanged.

[french toast]

Originally Posted by Nebuchadnezzar

So the kernel sources for the S3 were released and the final clock on the Mali is the same as I posted several months ago, 440MHz.

While I'm pretty sure there's no driver magic going on here, the only reasonable explanation would be that the memory bandwidth is vastly improved. Bandwidth tests put the S3 at roughly 30% higher speeds in real-world metrics over the S2. I don't see any other explanation for a 95% performance increase for only 65% clock increase on the GPU. Reports have been posted that it has an "internal 128bit bus" over 64bit in the 4210, but I do not understand how exactly does this help memory bandwidth, as the memory itself remained (apparently) unchanged.

I get 797mb/s on that test using a 70mb test size-45% i would have thought that they would have stuck in some lpddr2 1066?? anyway im not running out of bandwidth anytime soon i know that much, this thing chews through anything, the only gripe is that Samsung didn't stick in 2gb ram, ive only got 780mb to play with for some reason, i expect thats ICS+touchwhizz taking that, then i reguarly find around 400mb used constantly doing nothing on the home screen, even when i close every app from task manager and clear ram i barley get it under 400mb at best, ive installed a ram booster app to clear ram when i get to only 300 free.

Ive noticed i some times get kicked out of apps like Opera when im heavy loading, i wonder if that ICS/Touchwhizz cutting apps back? you would have though they would have started with some annoying background processess first??

Anyway once ive put that auto ram booster in i don't run out of ram for my main apps, certainly even with everything running some 25 apps ive experienced only a slight stutter when ram get filled and ICS starts closing or tombstoning, Bandwidth it seems is more than fine.

Let me know if you want me to run any benchmarks, i have quite a few installed already. cheers.

[Ailuros]

Originally Posted by Nebuchadnezzar

So the kernel sources for the S3 were released and the final clock on the Mali is the same as I posted several months ago, 440MHz.

While I'm pretty sure there's no driver magic going on here, the only reasonable explanation would be that the memory bandwidth is vastly improved. Bandwidth tests put the S3 at roughly 30% higher speeds in real-world metrics over the S2. I don't see any other explanation for a 95% performance increase for only 65% clock increase on the GPU. Reports have been posted that it has an "internal 128bit bus" over 64bit in the 4210, but I do not understand how exactly does this help memory bandwidth, as the memory itself remained (apparently) unchanged.

What am I missing, where's the 95% performance difference? http://www.glbenchmark.com/compare.j...%20Galaxy%20S2 ....or are you saying that the S2 results are at 440MHz in the latter?

IF above Egypt offscreen results should be at default GPU frequencies (266 and 440MHz respectively) the results look quite reasonable.

[Nebuchadnezzar]

Originally Posted by Ailuros

What am I missing, where's the 95% performance difference? http://www.glbenchmark.com/compare.j...%20Galaxy%20S2 ....or are you saying that the S2 results are at 440MHz in the latter?

IF above Egypt offscreen results should be at default GPU frequencies (266 and 440MHz respectively) the results look quite reasonable.

65fps?! Wait a moment. ......... Okey I just ran it again in 2.1.4 and now it gives me 61fps. A few weeks ago I was getting constantly 53fps. This is silly. Either they changed something between 2.1.3 > 2.1.4 or drivers did indeed improve performance and I didn't notice it in the meantime. Bollocks. French toast can you run an Egypt test on the latest GLBenchmark? I'm still waiting on my blue S3.

[french toast]

Egypt offscreen 720p- 11064 frames 98fps 98/65*100= 51%

Pro offscreen 720p -6167 frames 123fps

Note latest run, ive edged abit higher at 99 + 125 however ive only conducted 3 tests

EDIT; As im quite new to this app it seems ive made a mistake, or at least im confused as the app says its running in 1080p offscreen, version 2.1.4?? i thought 1080p was version 2.5?

[Nebuchadnezzar]

Well well. This pretty much leaves out only viable explanation indeed driver improvements, if Kishonti didn't change much in the benchmark.

So basically scores on the S2 improved 30-40% over the last year or so..

[Ailuros]

Originally Posted by Nebuchadnezzar

Well well. This pretty much leaves out only viable explanation indeed driver improvements, if Kishonti didn't change much in the benchmark.

So basically scores on the S2 improved 30-40% over the last year or so..

Doesn't surprise me one bit; ARM isn't the only IHV with GPU IP where some driver and/or compiler tweaking brought significant performance increases over time.

[almighty]

Originally Posted by Nebuchadnezzar

Well well. This pretty much leaves out only viable explanation indeed driver improvements, if Kishonti didn't change much in the benchmark.

So basically scores on the S2 improved 30-40% over the last year or so..

Adreno released a new driver for the Adreno 220 GPU in December and I gained 2-3x the performance jump in loads of apps.

I can get 53fps in Nenamark 2 which is much higher then even the Galaxy S2.

[french toast]

Right quick update about Exynos 4412 and quad cores in general on Android.

Loaded system tuner pro (amazing peice of software) to track all four threads to see whether they acually are being used, and if so how much.
It is quite clear that that they do get used very frequently, sometimes clocking all 4 at 800mhz, some times thrashing all four at 1.4ghz (more often than you would think) with those optimisations Samsung said being evident as 3 cores shut off, or 2 or even 1, with differnent cores being able to clock at different frequencies for extra efficiency.

The minimum speed is 200mhz, and you adjust it to only go to 1ghz using power saving mode, which does very slightly impact performance even with all 4 cores available, that along with the speed you run out of ram tells me Android could easilly use another gb ram and some higher single thread performance, amazing considering my phone is now considerably faster than my netbook.

So all those nay sayers saying that 4 cores were a waste of battery (batterylife is very good) and would be a waste of resources as they would be redundant can pipe down, i have have seen my self that all 4 threads are indeed used at differing frequencies with maxx frequency used more often than you would think, Android is silky smooth as a result.

Cheers.

[Ailuros]

While there will always be naysayers for pretty much everything, it remains a fact that efficiency per core (or per thread) is way more important than a sterile amount of cores. Besides personally I wouldn't care how often N hw is really needed, but when needed how badly exactly.

[french toast]

Yes i get your point, but most of what you have said is in software, Exynos 4412 is nearly at the apex of what can be done on a mobile device, i suspect Snapdragon S4 PRO, built on 28nm HKMG would be the ultimate sumit of both performance and batterylife, but for the next 2 years apart from ram, i can't say im going to want or need any extra power in my pocket, we have got to the stage of a pc in your pocket, ridiculous.

Can i ask you, how does the Adreno 320 compare to the S3's Mali 400 mp4?

EDIT; One more thing, in head to head gaming, ie Nova 3, Tegra 3 gets trounced by Exynos 4412.

[ToTTenTranz]

Originally Posted by french toast

Can i ask you, how does the Adreno 320 compare to the S3's Mali 400 mp4?

Why don't you see for yourself?
But should you look at the synthetic benchmarks, it's very likely that the drivers will substantially improve performance on Adreno 320's side.

Originally Posted by french toast

EDIT; One more thing, in head to head gaming, ie Nova 3, Tegra 3 gets trounced by Exynos 4412.

It should. The GS3's GPU is a lot faster than Tegra 3's.
Does Nova 3 come with a benchmark?

[french toast]

Why don't you see for yourself?

I have thanks, i was asking for his personal opinion.

No benchmark, just subjective assessments from you tube HTC one X lags, as well as Transformer prime. GS3 powers through it no worries.

[Ailuros]

Originally Posted by ToTTenTranz

Why don't you see for yourself?
But should you look at the synthetic benchmarks, it's very likely that the drivers will substantially improve performance on Adreno 320's side.

Driver improvements are likelier to show up more and to a higher degree in popular synthetic benchmarks than games for good reason. That said since upcoming GLBenchmark2.5 adds quite a bit in shader complexity from what I've seen I'd say that GPUs like Adreno (USCs) might definitely benefit from it more than GPUs with non unified ALUs.

[Ailuros]

Originally Posted by french toast

Can i ask you, how does the Adreno 320 compare to the S3's Mali 400 mp4?

I'm not aware of the detailed specifics of Adreno 320, yet considering that 225 is already a GPU with 8 Vec4 USC ALUs, it remains my gut feeling that things will improve for all Adrenos as applications add in complexity in the future.

Qualcomm (as quite a few others apart from NV) need to obviously get their stuff together and get rid of what I assume to be compiler related problems.

EDIT; One more thing, in head to head gaming, ie Nova 3, Tegra 3 gets trounced by Exynos 4412.

Could you provide a link for the specific youtube video? Just make sure that any T3 based device runs the game in a comparable resolution to the 4412 based devices.

[french toast]

Originally Posted by Ailuros

I'm not aware of the detailed specifics of Adreno 320, yet considering that 225 is already a GPU with 8 Vec4 USC ALUs, it remains my gut feeling that things will improve for all Adrenos as applications add in complexity in the future.

Qualcomm (as quite a few others apart from NV) need to obviously get their stuff together and get rid of what I assume to be compiler related problems.



Could you provide a link for the specific youtube video? Just make sure that any T3 based device runs the game in a comparable resolution to the 4412 based devices.

Transformer prime 1280 x 800-Tegra 3.
http://www.youtube.com/watch?v=h9yAVGJGLHI

Galaxy Note 1280 x 800 -Exynos 4210.
http://www.youtube.com/watch?v=NYKkj...feature=relmfu


HTC ONE X ; 1280 x 720p-Tegra 3.
http://www.youtube.com/watch?v=P2tD6dCAkKo

Samsung Galaxy S3 1280 x 720p- Exynos 4412.
http://www.youtube.com/watch?v=8mupu...feature=relmfu

[Ailuros]

Originally Posted by french toast

Transformer prime 1280 x 800-Tegra 3.
http://www.youtube.com/watch?v=h9yAVGJGLHI

Galaxy Note 1280 x 800 -Exynos 4210.
http://www.youtube.com/watch?v=NYKkj...feature=relmfu


HTC ONE X ; 1280 x 720p-Tegra 3.
http://www.youtube.com/watch?v=P2tD6dCAkKo

Samsung Galaxy S3 1280 x 720p- Exynos 4412.
http://www.youtube.com/watch?v=8mupu...feature=relmfu

Judging from the comments in more than one T3 video links, it seems that Nova3 on the platform is for the moment not officially supported. No idea if it is or isn't but if it should be the case it's a rather dubious comparision.

[french toast]

Yea i didnt read the comments section, however it does look like a graphically intensive game, 2gb download is nothing to sniff at, we have already surpassed xbox 1 graphics in some games, gta3 on this gs3 is mindblowing.

Ive had a quick look at anands htc one x at&t article to see what the adreno 225 performs like in comparison, it loses by 12% on egypt, which led people to believe tegra ulp was better, however adreno does sneak a win in pro, and not to forget the other gpu tests were adreno dominates.

As adreno has more complex unified shaders i do wonder whether it is in fact the better gpu, especially in more advanced games such as nova 3.

It seems to me this exynos 4412 is a very impressive chip in every way, only really let down by lack of ram and baseband efficiency, from what i can tell its an intel gold 65nm number, cell standby is a power hog.

[Nebuchadnezzar]

Sorry for bumping, but it's the most fitting thread and there's new information out:

The Galaxy Note 2 will be coming out with a new 4412 versioned Rev 2.0, where as the one currently in the S3 is versioned Rev 1.1. The new chip will be launched at 1.6GHz default clock. What is interesting is that they have increased the base clock from 800MHz to 880MHz, most of the SoC internals feed off this clock, meaning that we're going to have 10% clock boost in the internal bus and memory speeds.

Now as a side note: One thing that I haven't understood from the press releases back in May, is that there were this "internal 128bit bus" mentioned, with some idiotic websites taking that tidbit and claiming the chip was a 128bit architecture. Whatever. Anyway, the reason for this is that the way the Samsung SoCs internally function: they are separated in a "left bus" and a "right bus". The left bus is connected to the memory controllers and is also just called the MIF/Memory Interface. The right bus is called the "internal bus" and is connected to the ARM cores and everything else. The biggest difference here between the 4412 and the previous Samsung iterations was that both these were running at the same clock. In the 4412 the internal bus is running at half the memory interface bus, this corresponds to the increase to 128bit in the internal bus.

Now I got curious due to all this talk about the A6 and this tidbit:

"K3PE7E700F-XGC2" the last two characters refer to the clock speed. The iPhone 4S was [under]clocked at 800 Mhz. "K3PE4E400B-XGC1" was the A5's part number. E4 refers to 2 Gb LPDDR2 die and because A5 features a dual-channel LPDDR2 memory with two 32-bit die. 2 GB x 2 = 512 mb of RAM. C1 was the clock speed which was 2.5ns which indicates a 400MHz clock frequency. Two channels result in the A5 clock speed of 800MHz. So the A6 has C2 which is 1.9ns which indicates a 533 MHz clock frequency. 533 x 2 is ~1066 GHz.

Both the A6 and 4412 use the same memory, only difference being what seems to be a revision serial character. I was talking a few months ago how the 4412 showed a good 30% bandwidth improvement over the 4210, and credited this to it running 1066mbps memory instead of 800mbps; but in reality that is not the case.

I went over the source code of the busfrequency driver in the S3, and found that actually there is an entry for the internal frequency to run at 266MHz (128bit), but that entry is disabled in the driver; because the memory interfaces don't exceed 400MHz. The bus speed is defined in (MIF/INT) pairs and top speed available is 400200 (400MHz memory, 200 internal). Well this is interesting we can overclock our device's memory then if there's headroom! Well that idea quickly faded as I found that the C2C (Chip-to-chip) interface to the memory isn't capable of being clocked to 533MHz because simply the C2C clock divider register simply doesn't allow a divider value needed for that frequency, only being able to run 400MHz(and lower) and 800MHz. Basically we can't use the fast memory because it seems the clock dividers don't allow it. Anyway, coincidentally the i9305 sources were released two days ago and it included all the Note 2 sources and so on, so what Samsung did was simply increase the MPLL base clock from 800 to 880MHz, actually increasing the frequency of a load of things like the camera interface and who knows what at the same time.

What this also means is that Samsung increased effective bandwidth by 30% without increasing the memory speed. This indicates much improved memory controllers, and also why it easily beats the Tegra 3 and others in memory benchmarks.

Another new addition to the REV 2.0 chip is that we'll be running 533MHz for the Mali clock by default. We were already experimenting with this on the S3 and pretty much made the GPU run up to 700MHz, of course, it gets quite warm and battery hungry, but it's neat nonetheless.

Well I hope some people found some of my rantings interesting, and if there's some questions feel free to ask.

[Ailuros]

At least I find your posts highly interesting. On a sidenote current Exynos4 devices (the SIII included) received another speedbump from 1500+ to 1700+ frames:

http://www.glbenchmark.com/result.js...chmark=glpro25

Doesn't sound like a frequency change to me, but rather another clever driver/compiler improvement possibly. I wonder where it places the so far suggestions that GLBenchmark2.5 is heavily taxing in terms of geometry. The distance is too healthy compared to the currently fastest Tegra3 SoC IMHO.

[Nebuchadnezzar]

Originally Posted by Ailuros

At least I find your posts highly interesting. On a sidenote current Exynos4 devices (the SIII included) received another speedbump from 1500+ to 1700+ frames:

http://www.glbenchmark.com/result.js...chmark=glpro25

Doesn't sound like a frequency change to me, but rather another clever driver/compiler improvement possibly. I wonder where it places the so far suggestions that GLBenchmark2.5 is heavily taxing in terms of geometry. The distance is too healthy compared to the currently fastest Tegra3 SoC IMHO.

I'm getting 1714 frames as well with relatively old drivers, on stock speeds, I don't know how those lower scores even came to be, probably power management lagging in scaling up the frequencies. Currently we have a 1 second sampling window for each frequency jump (We jump linearly between frequencies ; 160, 266, 350, 440), something that really puzzles me, I set this up to 100ms in my kernel and I'm getting much better responsiveness in a lot of things, even outside 3D environments. I didn't say the current devices will even get this speedbump even if they had a newer chip, but the upcoming Note 2 will ship with it.

PS: ARM released completely new platform drivers a few months ago, and Samsung has been experimenting with these in their Jellybean builds, but we don't know what exactly is being used as the source code isn't released yet for the i9300, as far as the i9305 is concerned, it's still using old r2pX drivers, but the r3p0 drivers are included in their sources, so eventually we might see a migration to the new driver architecture. We're talking about the platform drivers here, and not the 3D drivers, we don't have access to those since they're closed source. So I have no idea what the advantages might be, other than what appears to be superior power management and load balancing between the 4 pixel processors (They can be individually turned on/off). I'm not familiar with the logic too much as the drivers are very exotic compared to the rest of the Linux kernel, and a bit beyond my understanding.