Samsung Orion SoC - dual-core A9 + "5 times the 3D graphics performance"
- Thread starter Mike11
- Start date
Some Galaxy S2 benchmarks have hit Glbenchmark.
http://www.glbenchmark.com/phonedetails.jsp?D=Samsung+GT-i9100+Galaxy+S+2&benchmark=glpro20
egypt high is 38.8 V's 21.4 for nexus S
egypt std is 41.5 Vs 26.8
pro high is 59.2 Vs 36.2
pro std is 59.3 Vs 46.5
Pro tests appear to be meaningless, as they look like they have hit a frame rate limiter.
So 80% quicker on egypt high, and 55% quicker on egypt standard, than the previous gen samsung phone nexus S ,with hummingbird S5PC110 and SGX540. Some of that of course might be due to the newer chip having increase CPU performance as its a dual core. Overall glbenchmark score is 4693 Vs 3025, which is about 55% improvement, so miles from the x5 improvement samsung suggested.
In a comparison with the ipad2 graphics, the ipad 2 is about 25% quicker on egypt high, and about 15% quicker on tegypt std, however the ipad2 has more that twice the screen pixels.
http://www.glbenchmark.com/phonedetails.jsp?D=Samsung+GT-i9100+Galaxy+S+2&benchmark=glpro20
egypt high is 38.8 V's 21.4 for nexus S
egypt std is 41.5 Vs 26.8
pro high is 59.2 Vs 36.2
pro std is 59.3 Vs 46.5
Pro tests appear to be meaningless, as they look like they have hit a frame rate limiter.
So 80% quicker on egypt high, and 55% quicker on egypt standard, than the previous gen samsung phone nexus S ,with hummingbird S5PC110 and SGX540. Some of that of course might be due to the newer chip having increase CPU performance as its a dual core. Overall glbenchmark score is 4693 Vs 3025, which is about 55% improvement, so miles from the x5 improvement samsung suggested.
In a comparison with the ipad2 graphics, the ipad 2 is about 25% quicker on egypt high, and about 15% quicker on tegypt std, however the ipad2 has more that twice the screen pixels.
Last edited by a moderator:
D
Deleted member 13524
Guest
Those benchmark results can't be only due to the 20% CPU clock increase. Given Anand's preliminary benchmarks for Exynos, there seems to be a huge difference in either driver performance, GPU clocks or both.
Exynos now stands confortably above Tegra 2 in 3D performance, and it sounds like it'll be above the 300MHz SGX540 from OMAP4.
I wonder if they managed to only overclock the vertex shader, as that seemed to be the bottleneck.
As rumours surge about Apple not releasing a new iPhone this year with A5, I guess Exynos may stand as the highest-performing smartphone SoC for 2011.
Exynos now stands confortably above Tegra 2 in 3D performance, and it sounds like it'll be above the 300MHz SGX540 from OMAP4.
I wonder if they managed to only overclock the vertex shader, as that seemed to be the bottleneck.
As rumours surge about Apple not releasing a new iPhone this year with A5, I guess Exynos may stand as the highest-performing smartphone SoC for 2011.
Last edited by a moderator:
http://www.glbenchmark.com/compare....Samsung GT-i9100 Galaxy S 2&D2=Google Nexus S
Google Nexus S (SGX540@200MHz) vs. Exynos. I don't see any hypothetical defeat against a 300MHz SGX540 in theory (until the LG gets the latest driver the Nexus S carries), nor do I see any great geometry results either for the latter considering it's definitely clocked as low as 200MHz.
If a OMAP4460 or Tegra3 won't be shipping in actual devices this year, Exynos has the chance to be the 2nd fastest smart-phone after iPhone5.
Google Nexus S (SGX540@200MHz) vs. Exynos. I don't see any hypothetical defeat against a 300MHz SGX540 in theory (until the LG gets the latest driver the Nexus S carries), nor do I see any great geometry results either for the latter considering it's definitely clocked as low as 200MHz.
If a OMAP4460 or Tegra3 won't be shipping in actual devices this year, Exynos has the chance to be the 2nd fastest smart-phone after iPhone5.
D
Deleted member 13524
Guest
Really?! Look at the Egypt High scores:
Nexus S: 21.4 FPS
Galaxy S II: 38.8 FPS
Even assuming a linear increase in performance with the GPU clock, 21.4x1.5 = 32,1 FPS.
So at best, we're looking at an OMAP4430 that is ~20% slower than Exynos 4210 in the Egypt High test.
The latest Electopia benchmark raises that discrepancy even higher.
What LG are you talking about?
Well yes, with the single vertex shader it's widely known that the Mali 400MP4 wouldn't have good geometry results.
Tegra3 might be reserved for tablets in 2011, and there are rumours claiming there won't be an iPhone5 this year, just a re-launch of the white iPhone4. Nonetheless, we'll be sure of that in a month.
Furthermore, we don't know if an A5 actually "fits" into a smartphone, or if it'll have the same CPU\GPU clocks as the tablet part.
I'd certainly expect an A5 phone to be at lower clocks, as the A4 is in iPhone 4. But I'm sure it can manage in a portable, if Sony's NGP can at the same process and with twice the SGX cores. Sure, that'll have a bigger battery, but an iPhone doesn't really need to let you game for 6+ hours either.
Apple would be pretty foolish to not release an A5-based iPhone 5 this year. A4 is at a clear disadvantage vs all of the dual cores out now and soon, and they'd be missing out on offering a huge gaming edge. Despite being a new SoC A4 is really not an awful lot different from what was in iPhone 3GS, with the iPhone 4's success riding more on its display. Now that display has to contend with SuperAMOLEDs and what have you, so it's not quite as impressive.
Apple would be pretty foolish to not release an A5-based iPhone 5 this year. A4 is at a clear disadvantage vs all of the dual cores out now and soon, and they'd be missing out on offering a huge gaming edge. Despite being a new SoC A4 is really not an awful lot different from what was in iPhone 3GS, with the iPhone 4's success riding more on its display. Now that display has to contend with SuperAMOLEDs and what have you, so it's not quite as impressive.
If medfield ships by end of year, and contains the same graphics I/P as that in cedartrail (from the other thread), even running at 400Mhz, then I guess that would swing things for medfield (in graphics at least).
The improved SGX driver lifted the Nexus S from a score ranging around 2400 to over 3000; LG's Optimus 3D would score quite well if all of the driver improvements were in place.
Still, I expect the Exynos to have significantly more graphics potential as it should considering its die size and power consumption compared to the SGX540@300Mhz.
A5 is a cinch for a new iPhone. Its GLBenchmark runs have only scratched the surface of its performance. I don't doubt that the SoC will stand alone in performance for a while.
Still, I expect the Exynos to have significantly more graphics potential as it should considering its die size and power consumption compared to the SGX540@300Mhz.
A5 is a cinch for a new iPhone. Its GLBenchmark runs have only scratched the surface of its performance. I don't doubt that the SoC will stand alone in performance for a while.
Dont give any credence to those rumours, you can bet that they will release an iphone 5(or 4S maybe?) with the A5 chip this year. With all the other dual core phone launching, they're gonna be left far behind otherwise. Last year they released the ipad on A4 in April. This year the ipad 2 with the A5 released in March so they're a month ahead compared to last year. I think all these rumours are just smoke and mirrors to confuse people(and lets apple sell a heck of a lot of white iphone 4's). IMHO we will see the next iphone release on time according to Apple's traditional yearly refresh schedule.
Im pretty sure we'll see Tegra 3 in Tablets this year, at least in Tablets (and I for one am quite interested in an Asus Transformer with Tegra 3) Also dont count Qualcomm out yet, if Adreno 220 is 2X the performance of Adreno 205, it will also be fast. But except for the NGP, no other SoC is going to touch the A5 till the next process node.
arjan de lumens
Veteran
Another day, another set of GLBenchmark results for the Samsung Galaxy S2:
http://www.glbenchmark.com/latest_results.jsp?benchmark=glpro20
It now reaches an Egypt score of 5997, putting it comfortably ahead of the ipad2.
If, at this point, anyone wants to point out that the ipad2 has more pixels, I'd like to point to the one other Exynos device in the GLbenchmark database: the HardKernel ODROID-A; granted, ipad2 gets a 30% higher framerate than the odroid, but the odroid has 33% more pixels.
http://www.glbenchmark.com/latest_results.jsp?benchmark=glpro20
It now reaches an Egypt score of 5997, putting it comfortably ahead of the ipad2.
If, at this point, anyone wants to point out that the ipad2 has more pixels, I'd like to point to the one other Exynos device in the GLbenchmark database: the HardKernel ODROID-A; granted, ipad2 gets a 30% higher framerate than the odroid, but the odroid has 33% more pixels.
D
Deleted member 13524
Guest
I wonder where these performance bumps are coming from, and how credible they are.
I haven't heard of any update for GS II in the past few days, or any performance fixes done through custom ROMs in xda, or any kind of overclocking done to the device.
But these results do match with the Electopia benchmark results, though.
Maybe that Mali 400MP4 is clocked really high..
I haven't heard of any update for GS II in the past few days, or any performance fixes done through custom ROMs in xda, or any kind of overclocking done to the device.
But these results do match with the Electopia benchmark results, though.
Maybe that Mali 400MP4 is clocked really high..
Certainly seems fast, but is it a commerical product, as it appears to be a developer platform only ? One assumes a developer platform has little concerns for battery life ?
An interesting metric would be to see how long it runs that demo for on a full battery, on a battery capacity similar to ipad2, and compare with ipad2.
Or on the other hand, just compare real end user products, rather than clogging up the benchmark lists with developer platforms, and overclocked chips and various other things that have little real life implications.
D
Deleted member 13524
Guest
http://www.hardkernel.com/renewal_2011/shop/good_buy_view.php?lang=en&g_code=G129705660781
If anyone can buy one, I'd say it's not really a developer platform exclusive. Sure, there's a small connector for an external PCB for additional connections, but it shouldn't get in the way of the overal product ergonomy.
There's a 9000mA.h battery in there, quite a bit larger than Galaxy Tab 10.1's 6860mA.h
Comparing power efficiency with Tegra 2 seems quite possible.
And for $750, it's actually a nice price.
D
Deleted member 13524
Guest
One assumes a 9000mAh battery for a 10 inch tablet shows anything but little concerns...
There's the extra potential from Exynos I was looking for. Clearly a generation beyond Tegra 2 and SGX540.
iOS hasn't been giving competitive results in GLBenchmark for a while compared to Android. I think an A5-like SoC, such as one from Renesas, powering an Android platform device should see some great GLBenchmark scores.
iOS hasn't been giving competitive results in GLBenchmark for a while compared to Android. I think an A5-like SoC, such as one from Renesas, powering an Android platform device should see some great GLBenchmark scores.
One assumes a 9000mAh battery for a 10 inch tablet shows anything but little concerns...
we have ZERO indication of battery life. We know its a developer platform To suggest that a particular size battery gives any indication of good use of power, is a very poor understanding of the entire concept.
Update:
I few moments on the odriod support boards yields this from one of the developers:-
"We could enjoy web-surfing about 5~6hour with mid-range of LCD back-light setting."
The equivalent test of the ipad2 gets over 11hrs.
So at least now we know why they had to put a 9000mAH battery into the odriod-a, which is one of the highest capacity batteries I've seen in this format.
As I previously said, its a developer platform that thus either optimised without regard to power, or its a power hungry plafform. Therefore although the benchmarks give a good indication of the maximum performance of the platform, they are of little relevance to what to expect in a consumer ready product.
Last edited by a moderator:
D
Deleted member 13524
Guest
I think your logic is equally poor.
If they didn't care for battery life, why not go with a much cheaper ~3000mAh battery, or why carry a battery at all?
What equivalent test in iPad2??!
What web-surfing are they doing? The iPad 2 will just sit idle and do nothing whenever there's Flash content to process, so what websites have those developers visited? Were they using 3G or Wifi? Was the iPad 2 using 3G or Wifi? How big is the battery in iPad 2? How strong was the signal? Is the luminosity with the "mid-range of LCD back-light setting" in both devices even comparable?
And where is this 11-hour test?
You cannot just come up with a vague (to say the least) sentence from someone, somewhere and suddenly turn it into "hard evidence" that the ODROID-A has an enormous power consumption, only to justify your pre-made assumption that the device cannot have been made with battery longevity concerns just because it can be used as a developer platform.
The benchmarks of the ODROID-A can be practically showing the same results as a consumer available Galaxy S II would do if it had to render at the higher resolution.
Just look at the synthetic tests. Some of them are a perfect match!
Last edited by a moderator:
Mr combative once again...that's unfortunate.
Anyhow I didn't "come up" with some vague sentence, I got it from the odriod developers, which I did say.
"We could enjoy web-surfing about 5~6hour with mid-range of LCD back-light setting."
http://odroid.foros-phpbb.com/t547p20-odroid-a#3107
There's the link, users name is odriod,his postings would indicate that he is a hardkernal employee.
On the same thread two posts up he says
"please remember, this is a development platform"
He didn't say it could be used as a development platform, it *IS* a development platform.
Thread further indicates web browsing test was done using wifi.
11hr figure for web browsing came from Anandtech's testing of the ipad2 on web browsing using wifi.
Anyhow I didn't "come up" with some vague sentence, I got it from the odriod developers, which I did say.
"We could enjoy web-surfing about 5~6hour with mid-range of LCD back-light setting."
http://odroid.foros-phpbb.com/t547p20-odroid-a#3107
There's the link, users name is odriod,his postings would indicate that he is a hardkernal employee.
On the same thread two posts up he says
"please remember, this is a development platform"
He didn't say it could be used as a development platform, it *IS* a development platform.
Thread further indicates web browsing test was done using wifi.
11hr figure for web browsing came from Anandtech's testing of the ipad2 on web browsing using wifi.
Similar threads
- Deleted member 13524
- Mobile Devices and SoCs