this guy above knows some @#$% lol kudos!

Doesn't know squat. A die shrink doesn't up performance. I guess the ipad2 is a bargain because its so much powerful now at 32nm even though they didn't make any tweaks. Typical galaxy tunnel vision.


...of the people, by the people, for the people...

Still not convinced??? read below

Nope. Actually your own article quotes give it all away.
''but more like increases it by 50% (at same frequency, and only for very multi-threaded tasks),'' as i've stated earlier how many apps are multi-threaded and take advantage of all 4 cores right now?

So exactly, the memory bandwidth does not increase by itself, but the smaller process node gives more room for Samsung to do so. But have they changed it? I don't see no info on their site regarding this for Exynos 4212, only for 5250. So for now, this, and clever bandwidth architecture looks like wishful thinking. And i cant find any info on this except on the site u listed. I can say same for Qualcomm's 28nm.

Samsung on their site claimed that Exynos 4212 gives 25 percent CPU improvement over 4210 in its dual core configuration and that sounds about right due to higher clock and dual channel memory.
Regarding GPU, its roughly another way of saying overclocking but in a proper manner.

But benchmarks really (if i haven't seem them enough already). u comparing HTC One X on 4.0.3 to Samsung 4.0.4, and on 2 of them HTC One X At&T beats its international variant. I'll hold on to this argument till i see both versions Of Galaxy running side by side (for now i think you reading way too much into it)

Originally Posted by Xdenwarrior

Samsung on their site claimed that Exynos 4212 gives 25 percent CPU improvement over 4210 in its dual core configuration and that sounds about right due to higher clock and dual channel memory.
Regarding GPU, its roughly another way of saying overclocking but in a proper manner.

u comparing HTC One X on 4.0.3 to Samsung 4.0.4, and on 2 of them HTC One X At&T beats its international variant. I'll hold on to this argument till i see both versions Of Galaxy running side by side (for now i think you reading way too much into it)

Where does the room for higher clock/overclocking comes from?

The higher clock(performance) was made possible only via a shrink in die size or via some changes in hardware. In this case its the die shrink.


Regarding 4.0.3 vs 4.0.4, there is no change regarding performance. afaik the latest version is only meant to fix some bug related to battery life.

you can conclude this with a simple understanding,


krait a15 =< exynos a9 @ 32nm

a15 > a9 (any design)

Originally Posted by bala_gamer

Where does the room for higher clock/overclocking comes from?

The higher clock(performance) was made possible only via a shrink in die size or via some changes in hardware. In this case its the die shrink.


Regarding 4.0.3 vs 4.0.4, there is no change regarding performance. afaik the latest version is only meant to fix some bug related to battery life.

you can conclude this with a simple understanding,


krait a15 =< exynos a9 @ 32nm

a15 > a9 (any design)

That still doesn't make sense lower mn reduces heat/power to increase higher clock speeds within an equal or lesser thermal spec which may increase performance the most your getting out of is really lower heat and increased clock rates.

No matter what quad core cpu will in benchmarks in android will always give you better benchmarks than a dual core. Benchmarks are synthetic as different voltages run through each soc. I highly doubt exynos and krait use the same voltage scaling. Especially since krait has dynamic voltage scaling.

A15 has a 40% increase in performance per clock on a9. So even on two cores alone have a performance increase on a9 Quad core arc regardless on mn. Krait is based on a 28 mn which will use even less heat and less power than the the exynos at 32 mn.

If your looking in development the Exynos >Krait because of its open source. If your not hacking rooting your phone. The Krait>Exynos for a standard usage. Assuming the drivers and software are equally optimized on each soc.

Originally Posted by bala_gamer

@ Xedenwarrior

get your points and facts right before posting for gods sake!!!

die shrink not only reduces power consumption but also increases performance!

so a quad core with 45 nm will never be equal to a quad core with 32 nm. Also dual core krait is a reference design of a15 architecture and no oem specific optmizations are made. a15 design is much better than a9 and yes a15 is the next big thing for mobiles, but krait cores are still dual cores dont forget that, only future quad cores using a15 will beat the 4412 soc(for the most of the parts)

Still not convinced??? read below


source

I thought you were somewhat of a informed user but this post just reeks of fanboyism and nonsense.

First of all a pure die shrink does NOT improve performance in any way, it only allows for increased performance due to higher clocking due to lower power consumption and you know it well. One thing does not mean the other. Apple's A5 on 32nm are not faster than their 45nm counterparts and bringing into conclusion that a process shrink provides higher performance is a idiotic statement. You're comparing apples and oranges here and that's what you are being counter-argumented for.

source

Again here we are starting to quote and source articles written by some smartass on some website with absolutely no knowledge of the thing he's writing about.

He doesn't know even at what frequencies the Mali are running at, stating ALU FLOP performance being copied from other websites to compare two different architectures; and the most hilarious contradiction:

Samsung announced that they have switched to high-k materials and metal gates (HKMG) and further claimed it can provide superior performance with less power than conventional poly-Si/SiON used at 45nm.

vs a little lower:

For Power Consumption Nvidia came out with a solid solution for this Tegra 3 SoC. Tegra 3 has an extra core (also known as 4 plus 1) which is the fifth core, this core always runs at the top and do the simple tasks, quad core jumps out when only for bigger tasks. Samsung doesn’t have these kinds of features with their Exynos 4412 but Samsung claims that they have some features for low power such as multiple power domains, dynamic voltage and frequency scaling, a technique called body-biasing, and thermal management control, but in this category my vote goes to Tegra 3.

The incompetence of this writer is astounding; not only do both SoCs have DVFS, multiple power domains, thermal management, only thing going for the Exynos is body biasing which is an analogue IC design trick; but 2 paragraphs before he had just stated that the Exynos is built on a lower power process, which is actually going to be even more effective and lower power than Tegra's +1 core, whose only difference to the other cores is that it's made on a LP optimized process versus performance optimized for the other 4 cores.

The memory bandwidth does not increases by itself!!, the die shrink provides more head room to increase the bandwidth.
source

The 4412 just has the same bandwidth as the 4210 in the Galaxy S2, it's still running dual channel 32bit (@XdenWarrior, yes, they didn't add a channel) on LPDDR2 @ 800MHz resulting in the same 6.4GB/s. We still know nothing of the internal structure of the new Exynos and this article is just speculating on leaked facts that it's supposed to have an increased internal bus, whose benefits we don't know yet until a proper architectural review.

Have you seen these benchmarks results below??? dont say its only a software optimization, nvidia's tegra3 is well optimized in that department.

Those benchmarks ARE just software optimizations, SunSpider is a Javascript engine benchmark, and every single phone is running a different version or kind of browser; and as such also the BrowserMark is a pointless benchmark since Samsung's browser has been always superior since Gingerbread, this is all an apples vs oranges all over again and for the point of a technical discussion it's completely irrelevant. From a consumer point of view for who those benchmarks are directed at, it's great, but don't try to use it as an argument here. The Vellarmo scores are actually the most believable even if people claim it's a Qualcomm biased benchmark, which in my opinion it is not. Here purely you can see the architectural gains of Krait and the superior memory controller from Samsung over Nvidia's.

Originally Posted by bala_gamer

Where does the room for higher clock/overclocking comes from?

The higher clock(performance) was made possible only via a shrink in die size or via some changes in hardware. In this case its the die shrink.

Sorry but that's a logical fallacy, just because one thing is made possible by another does not mean that it is an effect of the other. Tegra 3 isn't weaker than the Exynos Quad because it's running at 40nm vs 32nm but because of stupid architectural decisions like still running on single channel and having an outdated GPU. You can simply make a better SoC on a larger process by making the chip larger, look what apples is doing with their A5X vs the A5.

Originally Posted by Killbynature

No matter what quad core cpu will in benchmarks in android will always give you better benchmarks than a dual core. Benchmarks are synthetic as different voltages run through each soc. I highly doubt exynos and krait use the same voltage scaling. Especially since krait has dynamic voltage scaling.

Every popular SoC in the last 2 years has had DVFS so by now it's just a checkbox for marketing and PR, what's the point of bringing it up?

Originally Posted by AndreiLux

I thought you were somewhat of a informed user but this post just reeks of fanboyism and nonsense.

First of all a pure die shrink does NOT improve performance in any way, it only allows for increased performance due to higher clocking due to lower power consumption and you know it well. One thing does not mean the other. Apple's A5 on 32nm are not faster than their 45nm counterparts and bringing into conclusion that a process shrink provides higher performance is a idiotic statement. You're comparing apples and oranges here and that's what you are being counter-argumented for.

Again here we are starting to quote and source articles written by some smartass on some website with absolutely no knowledge of the thing he's writing about.

He doesn't know even at what frequencies the Mali are running at, stating ALU FLOP performance being copied from other websites to compare two different architectures; and the most hilarious contradiction:The incompetence of this writer is astounding; not only do both SoCs have DVFS, multiple power domains, thermal management, only thing going for the Exynos is body biasing which is an analogue IC design trick; but 2 paragraphs before he had just stated that the Exynos is built on a lower power process, which is actually going to be even more effective and lower power than Tegra's +1 core, whose only difference to the other cores is that it's made on a LP optimized process versus performance optimized for the other 4 cores.
The 4412 just has the same bandwidth as the 4210 in the Galaxy S2, it's still running dual channel 32bit (@XdenWarrior, yes, they didn't add a channel) on LPDDR2 @ 800MHz resulting in the same 6.4GB/s. We still know nothing of the internal structure of the new Exynos and this article is just speculating on leaked facts that it's supposed to have an increased internal bus, whose benefits we don't know yet until a proper architectural review.


Those benchmarks ARE just software optimizations, SunSpider is a Javascript engine benchmark, and every single phone is running a different version or kind of browser; and as such also the BrowserMark is a pointless benchmark since Samsung's browser has been always superior since Gingerbread, this is all an apples vs oranges all over again and for the point of a technical discussion it's completely irrelevant. From a consumer point of view for who those benchmarks are directed at, it's great, but don't try to use it as an argument here. The Vellarmo scores are actually the most believable even if people claim it's a Qualcomm biased benchmark, which in my opinion it is not. Here purely you can see the architectural gains of Krait and the superior memory controller from Samsung over Nvidia's.


Sorry but that's a logical fallacy, just because one thing is made possible by another does not mean that it is an effect of the other. Tegra 3 isn't weaker than the Exynos Quad because it's running at 40nm vs 32nm but because of stupid architectural decisions like still running on single channel and having an outdated GPU. You can simply make a better SoC on a larger process by making the chip larger, look what apples is doing with their A5X vs the A5.

Every popular SoC in the last 2 years has had DVFS so by now it's just a checkbox for marketing and PR, what's the point of bringing it up?

Well,it's good that you debunk ****ty claims,but could you also share your opinion about the 4412 with us?
I don't know about you man,but this bad boy seems like the most powerful SoC out there,until phones with quad-core Krait,Sammy's series 5xxx and TI's series 5xxx A15 stuff comes out.
And yeah,I'd take the A9 4412 over the dual-core S4 any day,even though I know it's not actually "better".Leaving the Tegra 3 out,which in my honest opinion is bull****,the "battle" here is between the last SoC of this generation and the first one of the next.

Originally Posted by AndreiLux

I thought you were somewhat of a informed user but this post just reeks of fanboyism and nonsense.

Every popular SoC in the last 2 years has had DVFS so by now it's just a checkbox for marketing and PR, what's the point of bringing it up?

Sorry i meant to add on the very different chipsets the volatage use in each every chipset is different including the exact same devices will always measure benchmarks differently. Like a lot of benchmarks are biased towards intel in the pc market. A lot of benchmarks are biased on Qualcomm/Tegra/exynos/chips.
Regardless its hard to test performance truly when ever device responds differently to kernel or SoC. What it really is based on user experience. How many agree upon it. Like I said before the Exynos is greater than Krait because of the open source architecture.

Originally Posted by Killbynature

Sorry i meant to add on the very different chipsets the volatage use in each every chipset is different including the exact same devices will always measure benchmarks differently. Like a lot of benchmarks are biased towards intel in the pc market. A lot of benchmarks are biased on Qualcomm/Tegra/exynos/chips.
Regardless its hard to test performance truly when ever device responds differently to kernel or SoC. What it really is based on user experience. How many agree upon it. Like I said before the Exynos is greater than Krait because of the open source architecture.

Different voltages on the same devices will not affect performance at all, how do you come up with that? The only thing eventually different will be power consumption. They're still running the same frequency.

Also I don't see how Krait is supposed to be inferior to Exynos due to that argument? Exynos is not an open source architecture. To the average user and even developers don't do anything remotely needing the open engineering documentation of the SoCs, which is actually even not publicly available on the Exynos. For Android's sake because they're running the Linux kernel and are obliged to release the kernel sources, that's valid for any SoC, no matter how open it is. Nothing's stopping you from changing things in Nvidia's kernel for a closed architecture, for example. Unless you're a manufacturer and you actually want to produce a device, you don't give a **** how open the architecture is as long as updated kernel sources are being posted regularly for your device.

Originally Posted by AndreiLux

Different voltages on the same devices will not affect performance at all, how do you come up with that? The only thing eventually different will be power consumption. They're still running the same frequency.

Also I don't see how Krait is supposed to be inferior to Exynos due to that argument? Exynos is not an open source architecture. To the average user and even developers don't do anything remotely needing the open engineering documentation of the SoCs, which is actually even not publicly available on the Exynos. For Android's sake because they're running the Linux kernel and are obliged to release the kernel sources, that's valid for any SoC, no matter how open it is. Nothing's stopping you from changing things in Nvidia's kernel for a closed architecture, for example. Unless you're a manufacturer and you actually want to produce a device, you don't give a **** how open the architecture is as long as updated kernel sources are being posted regularly for your device.

Hmm i thought Exynos was open like Ti omap thanks for fixing that for me.