Benchmarking scores conflict each other - General Questions and Answers

I bought an Android Tablet running an ARM11 (v6) processor. Not sure what the GPU is. However, I ran the NeoCore benchmark for the GPU and got an average of about 13.2 FPS. I compared this against Droid Incredible (Qualcomm Snapdragon) and Droid X (TI OMAP 3630). Both of which as you know that are Cortex A8 variance. All are running at 1.0Ghz.
I also ran Softweg's Benchmark on those 3 devices.
For my tablet, I got scores of about 98 for the GPU and 936 for the CPU.
For the Droid Incredible I got NeoCore score of 26 FPS and for the X, 42 FPS. However, they only scored 27 & 30 respectively on Softweg's GPU test.
I would believe the NeoCore score as I am sure the GPU is poor on the tablet. Why would Softweg's Benchmark app show higher scores on my GPU versus those more advanced Android phones?
My CPU score is also higher than the phones which is not possible. Thank you.

interesting , but u know what , maybe its that neocore bench renders the same amount of data all the time , and the other bench resizes it to the size of the screen , so if ur tab has resolution lower than 480*800 thats the reason why

Actually the resolution is 1024x600. I could understand why the NeoCore score is low which is what I would expect. However, I do not know why my CPU & GPU scores under Benchmark would be higher than a better Cortex processor.

Related

Samsung Captivate -- Wrong Processor?

I hope I am wrong or got the wrong phone or maybe the phone just dynamically clocks itself down to 800 mhz instead of the advertised 1000mhz. When I use Quandrant Standard Edition my Captivate is listed at 800 mhz with a max of 1000mhz. But the cpu is ARMv7 Processor rev 2? Is that right? BogoMIPS is 797.9, and hardware is SGH-I897? is is not supposed to be SGS? Please let me know what you all are getting and what Quadrant Score...my friends Droid X destroys my quadrant score with ~1200, I get only 867. The benchmark really drags during the I/O sections. Help please and share your specs, cause I am feeling a little disappointed.
Dynamically clocking down.
nevermind...I just saw the other thread with this issue.
THREAD ENDED
If this processor is similar in it's capabilites as the qualcom Snapdragon than the processor has the ability to slow down if the higher clock speed is not needed. My N1 can switch it's frequency between 245Mhz and 1Ghz (or 1.1Ghz when overclocked with an upgraded kernel).
Or the software you are using is not reading the correct speed.
ronpinoy253 said:
I hope I am wrong or got the wrong phone or maybe the phone just dynamically clocks itself down to 800 mhz instead of the advertised 1000mhz. When I use Quandrant Standard Edition my Captivate is listed at 800 mhz with a max of 1000mhz. But the cpu is ARMv7 Processor rev 2? Is that right? BogoMIPS is 797.9, and hardware is SGH-I897? is is not supposed to be SGS? Please let me know what you all are getting and what Quadrant Score...my friends Droid X destroys my quadrant score with ~1200, I get only 867. The benchmark really drags during the I/O sections. Help please and share your specs, cause I am feeling a little disappointed.
Click to expand...
Click to collapse
No the specifications are right .. i too ran the benchmark and was puzzled yesterday. You can find that thread in this forum.
Check this video. it explains why captivates score is less than Droid X .. In fact as per the video even though the score is less captivate is in fact a faster phone than Droid X and i am inclined to agree watching video
http://www.youtube.com/watch?v=zgoQHxy-0mM
----------------
itzz(AN)dRoiD

Tegra 2 faster than Exynos, A5, and PYrami Dual-core

I am sick of everyone thinking the upcoming dual-core devices will blow away tegra 2.
Tegra 2 vs Dual Core A5 (Ipad 2)
A lot of talk about Andntech OpenGL benchmark trumping Tegra 2, but what about Stockfish and Benchit Pi where A5 got slaughtered (PC Magazine)? With half the RAM and lower clock I don't see this thing smoking Tegra 2 in all benchmarks, or real life CPU situations.
Tegra 2 vs Exynos (Some Galaxy S2)
Lower benchmarks in Smartbench Gaming. Plus there is early benchmarks of Quadrant scores of 2100 tablets running the Exynos 4210. There is a reason why Samsung Galaxy S2 is including Tegra 2 in some regions.
Androidevolution.."One negative surprise on the S2 so far has been the level of GPU performance. So far, most of the early benchmark shows that Exynos 4210 isn’t up to par when it comes to the GPU performance. This is strange given that Samsung was leading the market when they introduced the previous generation SoC ...... Smartbench 2011 GPU numbers are once again, very disappointing"
Tegra 2 vs Dual Core-Snapdragon (HTC Pyramid)
This thing got smoked in Smartbench with gaming and productivity.
" Their tests confirm that the Pyramid indeed houses a dual-core chip, but the popular Smarbench 2011 shows a CPU and GPU that simply don’t hold up to the Tegra 2 chip found in the LG Optimus 2X and Motorola Atrix 4G"
Something to remember, Tegra may be the fastest, but........ Just like a computer in your house the slowest component is the driver or the memory storage is usually the one that slows it down or the motherboard. So having the fastest cpu makes your bench look good but in practical use it is not better performance.
If the sole purpose is to game then, yeah ok that may be a winner, but get real, using the phone for gaming is such a waste of technology
Mind you synthetic benchmarks aren't the best way to show performance, quadrant and linpack are both easy to inflate scores just by changing a single value. As for ram, it's hard to compare ram seeing that Ios and Android handles memory differently.
vbetts said:
Mind you synthetic benchmarks aren't the best way to show performance, quadrant and linpack are both easy to inflate scores just by changing a single value. As for ram, it's hard to compare ram seeing that Ios and Android handles memory differently.
Click to expand...
Click to collapse
+1. Synthetic benchmarks are not a good indicator of real world performance.

[INFO/Q] HTC Sensetion only 1900 points with

smartbench 2011 Productivity test
http://smartphonebenchmarks.com/ind...11:Productivity&filter_cpu=all&filter_gpu=all
gpu score i might understand why its low cos the high res but why the Productivity is so low ?
i guess HTC didnt put faster NAND ROM
Evo3D did 2000
someone maybe know what the problem or cause ?
Proz00 said:
smartbench 2011 Productivity test
http://smartphonebenchmarks.com/ind...11:Productivity&filter_cpu=all&filter_gpu=all
gpu score i might understand why its low cos the high res but why the Productivity is so low ?
i guess HTC didnt put faster NAND ROM
Evo3D did 2000
someone maybe know what the problem or cause ?
Click to expand...
Click to collapse
The reason is...
The CPU is cortex 8.
Tegra 2 and the new Samsung processors are Cortex 9.
Coretex 9 is a PRETTY big improvement over cortex.
Once again HTC is going for garbage hardware
What is in the sensation is 2 Desire HD CPUS oC to 1.2 Ghz + better GPU.
What is in the SGS2 is 2 MUCH better Hummingbird CPUs OC to 1.2 + MUCH better GPU
the cpu is neither a cortex a8 nor a cortex a9. it will provide plenty of performance and will be competitive with other dual cores.
the adreno 220 gpu that comes with the sensation is faster than the mali gpu that comes with the sgs2 when looking at preliminary tests done by anandtech.
whether it will be the fastest or slowest dual core soc will have to wait until its released, and benchmarks often only tell part of the story. but certainly it will provide far more performance than any of the single core soc's we have right now and will provide much satisfaction from its owners.
kaiserkannon said:
the cpu is neither a cortex a8 nor a cortex a9. it will provide plenty of performance and will be competitive with other dual cores.
the adreno 220 gpu that comes with the sensation is faster than the mali gpu that comes with the sgs2 when looking at preliminary tests done by anandtech.
whether it will be the fastest or slowest dual core soc will have to wait until its released, and benchmarks often only tell part of the story. but certainly it will provide far more performance than any of the single core soc's we have right now and will provide much satisfaction from its owners.
Click to expand...
Click to collapse
Huh? I'm confused.
Is the cpu not based on arms cortex a8? Just a slightly modified version. It is identical to the Single core Snapdragon in the Desire HD.
The benchmarks so far don't make it seem too be as competitive as the Tegra 2 OR orion.
Samsung has said that the Mali 400 is MUCH faster then the current hummingbird GPU. Current benchmarks say that it is infact SLOWER...
I doubt samsung would release the Orion with a GPU SLOWER then its previous gen... that just makes no sense. If that is the case then Tegra might be king. If the Mali 400 IS much better tho, samsung will have the best SoC.
The CPU in the Sensation is ROUGHLY... 2.4 ghz. Compare that to the Desire HD stable OC of 1.8 ghz.
What is left to be seen is how much the CPU can be OC'd.
I would think that it would be less then 1.8 ghz each core. But thats yet tooo bee seen.
Regardless of what you think... the HTC sensation CPU will be slower then the competitions.
EDIT: Forgot to mention that the Sensation CPU should have the same battery life as the current single core Snapdragon... however it is pushing more pixels sooo..
Samsung should have mated its Orion to Hummingbird gpu. Hummingbird was great
Sent from my MB860 using XDA App
Maedhros said:
The benchmarks so far don't make it seem too be as competitive as the Tegra 2 OR orion.
Click to expand...
Click to collapse
Dunno where you got your information from, but it's very competitive with the Tegra 2. (8660 is the CDMA version of the Sensation's 8260). From these benchmarks, we also know that an overclock of at least 1.5GHz will be perfectly viable--the chip was designed for that anyhow.
Debating A8 vs A9 is a trivial matter, because it's a tiny fraction of the entire picture.
Wondering if cm7 can help the score
First, that Anandtech benchmark is not a good measuring stick. Anandtech benched the MDP that had the 8660 running at 1.5 GHz and 800x480 so the results are higher than what Sensation can achieve because Sensations runs at a lower clock and higher resolution.
Second, Qualcomm 8260/8660 is A8 Cortex. Tegra 2, OMAP4 and Exynos are A9 Cortex based. Claims that Qualcomm doesn't use the ARM architecture is a lie.
Never trust smartbench. Period.
GLbenchmark is more trustworthy.
Sent via psychic transmittion.
t-mizzle said:
First, that Anandtech benchmark is not a good measuring stick. Anandtech benched the MDP that had the 8660 running at 1.5 GHz and 800x480 so the results are higher than what Sensation can achieve because Sensations runs at a lower clock and higher resolution.
Second, Qualcomm 8260/8660 is A8 Cortex. Tegra 2, OMAP4 and Exynos are A9 Cortex based. Claims that Qualcomm doesn't use the ARM architecture is a lie.
Click to expand...
Click to collapse
The scorpion core in snapdragon socs use the arm v7 instruction set that both the a8 and a9 use, but it is not an a8 or an a9, it is qualcomms own design.
And personally I like comparing the different chips in these phones at the same resolution to see which chip has better performance on a level playing field. But yeah the sensation will have a bit worse performance thanks to higher resolution. Like the atrix vs optimus 2x. But to me the higher resolution is completely worth the hit in performance.
TeroZ said:
Never trust smartbench. Period.
Click to expand...
Click to collapse
Would you care to elaborate on this please?
GLbenchmark is more trustworthy.
Sent via psychic transmittion.
Click to expand...
Click to collapse
GLBench is a decent 3D benchmark app, but it is just that - it tests only the GPU. Smartbench was designed to test both CPU (inc. dual-core ones) and GPU, hence reporting two numbers. IMO, you are not comparing apples to apples unless you were only referring to the GPU portion of the test.
kaiserkannon said:
The scorpion core in snapdragon socs use the arm v7 instruction set that both the a8 and a9 use, but it is not an a8 or an a9, it is qualcomms own design.
And personally I like comparing the different chips in these phones at the same resolution to see which chip has better performance on a level playing field. But yeah the sensation will have a bit worse performance thanks to higher resolution. Like the atrix vs optimus 2x. But to me the higher resolution is completely worth the hit in performance.
Click to expand...
Click to collapse
Stop spreading FUD. MSM 8260/8660 is not capable of out of order execution. Cortex A9 supports this feature, A8 does not.
MSM 8260/8660 Pipeline Depth is 13 stages, therefor it's clearly a A8 Cortex.
A9 was a successor to the A8 and it's a significant improvement over it.
t-mizzle said:
Stop spreading FUD. MSM 8260/8660 is not capable of out of order execution. Cortex A9 supports this feature, A8 does not.
MSM 8260/8660 Pipeline Depth is 13 stages, therefor it's clearly a A8 Cortex.
A9 was a successor to the A8 and it's a significant improvement over it.
Click to expand...
Click to collapse
qualcomm disagrees with you though. they state that it is not based on the a8 and has partial out of order execution. it also has a 128 bit wide neon data path for neon instructions in comparison to the 64 bit wide path in a8 and a9 designs. while there are some similarities to the a8 as you pointed out, the scorpion is not qualcomm's implementation of an a8. and it has some advantages over both a8 and a9. and some disadvantes to an a9. overall the a9 will probably be a bit faster clock for clock, but the scorpion cores in the snapdragon dual cores are clocked faster.
this is very much the same as amd and intel. they both use the same instruction set (x86), but their processors are not the same. qualcomm simply licenses the instruction set (armv7) and builds its own processor. while other companies like nvidia, TI, and samsung buy the cortex a8 or a9 design from ARM and build a copy of it.
Acei said:
Would you care to elaborate on this please?
GLBench is a decent 3D benchmark app, but it is just that - it tests only the GPU. Smartbench was designed to test both CPU (inc. dual-core ones) and GPU, hence reporting two numbers. IMO, you are not comparing apples to apples unless you were only referring to the GPU portion of the test.
Click to expand...
Click to collapse
You are right. But smartbench rank scorpion+adreno205 lower than DX with [email protected] is definitely nonsense.
For gpu, go glbenchmark or nenamark or an3dbench whatever but smartbench.
For cpu, crunching pi or linpack is more reliable.
Smartbench does not reflect any real world performance.
Sent via psychic transmittion.
Thracks said:
Dunno where you got your information from, but it's very competitive with the Tegra 2. (8660 is the CDMA version of the Sensation's 8260). From these benchmarks, we also know that an overclock of at least 1.5GHz will be perfectly viable--the chip was designed for that anyhow.
Debating A8 vs A9 is a trivial matter, because it's a tiny fraction of the entire picture.
Click to expand...
Click to collapse
Based on glbenchmark score the anand tests might be suspect. It was score 6% higher than tegra 2 not double like anand's test. Or qcomm might be monkeying with things.If that is the case I am going to have a big problem with qcomm products.
Maybe smartbench is right and the nand quality is poor?
The sense experience on it wasn't done. It would have to score higher than the mytouch and previous devices its dual core. Most likely a crappy engineering build on it.
Sent from my HTC Glacier using XDA Premium App
TeroZ said:
You are right. But smartbench rank scorpion+adreno205 lower than DX with [email protected] is definitely nonsense.
Click to expand...
Click to collapse
There are other benchmark apps that rank your combo in the same order as Smartbench in graphical tests. Plus, please do look at the productivity tests for Smartbench 2011 more carefully. Typical Scorpion based phone score slightly higher results on Scorpions than DX. Even games like Dungeon Defender (a graphically heavy game) ranks both as "mid-range", while ranking Galaxy S series as "high-end".
For gpu, go glbenchmark or nenamark or an3dbench whatever but smartbench.
For cpu, crunching pi or linpack is more reliable.
Smartbench does not reflect any real world performance.
Click to expand...
Click to collapse
Calculating Pi is a very very simple, narrow, and one-dimensioned test. Linpack is heavy on floating point calculations. If that is what you want to know, then I have no issues with that. But do your day-to-day tasks on your phones translate to pure floating point calculations on your phones? They don't. That's why I've included several tests and will be including more as new versions are updated in the future. Plus, I believe none of them uses more than 1 core.
I'm open to suggestions and criticisms - but please do provide more details.
Latest benchmarks made by a retail GSII which has an ORION Exynos talks by themselves
http://forum.xda-developers.com/showpost.php?p=13096662&postcount=383
Exynos at "only" 1.2Ghz is even better than adreno 220 SCORPION 1.5Ghz chip as it score 41 fps whereas the latter is scoring 38 fps in GLBenchmark EGYPT standard test
http://images.anandtech.com/graphs/graph4243/36161.png
http://nsa25.casimages.com/img/2011/04/21/110421112944690206.png
So the HTC Sensation which is underclocked to 1.2Ghz and have a bigger resolution will look like shayt, SGSII With Exynos will rule for a long long time...
touness69 said:
Latest benchmarks made by a retail GSII which has an ORION Exynos talks by themselves
http://forum.xda-developers.com/showpost.php?p=13096662&postcount=383
Exynos at "only" 1.2Ghz is even better than adreno 220 SCORPION 1.5Ghz chip as it score 41 fps whereas the latter is scoring 38 fps in GLBenchmark EGYPT standard test
http://images.anandtech.com/graphs/graph4243/36161.png
http://nsa25.casimages.com/img/2011/04/21/110421112944690206.png
So the HTC Sensation which is underclocked to 1.2Ghz and have a bigger resolution will look like shayt, SGSII With Exynos will rule for a long long time...
Click to expand...
Click to collapse
Thanks for this.
Looks like this is another HTC phone with a disappointing CPU & GPU

antutu benchmark gpu

i just want to ask if its normal to only be getting 1000 ~ 1100 gpu score on antutu benchmark???
mali 400, adreno 220 and the ulp on my one x all get atleast 1300 ~ 1600 score
what bothers me is that the adreno 220 should have a lower clock speed than the ulp on the tegra 2 but still gets higher marks...
i get around 1115~ but still it does make sense that the tegra 3 in a one x is more powerful
and clock speed is not everything at all. Efficency and architecture matter a lot in determining the power of a gpu
Eg an i7 clocked at 4ghz will be a huge ammount more powerful than a phenom II clocked at 4ghz

First independent ARM A15 Benchmark - Exynos 5250

Some exciting news, the first real-world benchmark has appeared for an ARM A15 chip, in this case the Samsung Exynos 5250, which has been launched in the latest Chromebook.
Chip Info - dual-core A15 @ 1.7 GHz & Mali T604 GPU.
http://www.samsung.com/global/busin...t/application/detail?productId=7668&iaId=2341
The benchmark is Sunspider, which is not multi-threaded, i.e. does utilise multiple cores, so you can evaluate the actual performance (javascript) of a single-core., now we can see the performance improvement ARM has baked into their latest hardware
Courtesy of Gigacom, Sunspider on the ARM version of Google Chrome that comes installed on the Chromebook = 660ms (Lower is better). Compared to the current King of the Hill ARM A9 device the Galaxy Note 2 (Exynos 4412), which is clocked at 1.6 GHz, it achieves 972 ms accorded to GSM Arena, other sites have similar figures.
http://www.gsmarena.com/samsung_galaxy_note_ii-review-824p5.php
LOWER IS BETTER
Exynos 5250 - A15 @ 1.7 Ghz = 660 ms
Exynos 4412 - A9 @ 1.6 Ghz = 972 ms
The 5250 is clocked 6% higher than the 4412, so if we adjust the results for CPU frequency parity
Exynos 5250 = 660 ms
Exynos 4412 @ 1.7 Ghz = 914 ms
This is not an exhaustive performance test!, but we can see that in this one popular benchmark that ARM A15 is ~30% faster than the A9 architecture when adjusted for clock speed.
To sweeten the deal further A15 SoC will run at a higher clock than A9s, Tegra 4 (T40) is stated to run @ 1.8 GHz with a bump to 2 GHz after a couple of quarters, just like Tegra 3. Samsung has the even mightier 5450, a quad-core variant of the chip in this test, rumored to run @ 2 GHz, combined with much more powerful GPU, and Android's software optimisations 2013 is going to be one hell of year for tech fans:victory:
Source:
http://gigaom.com/mobile/video-hands-on-with-googles-new-249-chromebook/
Nice find. I am also looking for Mali-T604 results. GLbenchmark results will be interesting. 72GFLOPs does sound very good.
EDIT: I think he says 620ms in video. Also, I am sure it will get better as the Chrome OS code is optimized for ARM. This is just first release. Exynos 4 has been optimized to limit. They can't push it any further now, at least not by a big margin.
hot_spare said:
Nice find. I am also looking for Mali-T604 results. GLbenchmark results will be interesting. 72GFLOPs does sound very good.
Click to expand...
Click to collapse
You may have to wait a while, ChromeOS can't run Android apps like GLbenchmark, only webapps. The reason Sunspider is a good test in this case, is that they both use the ARM version of Chrome, which uses the same underlying technology (Webkit & V8 Javascript engine)
Edit, there some unverified benchmarks from ES 2.0 Taiji, but there are v-sync limited to 60 fps, so we don't know how powerful the T-604, from that bench.
http://www.phonearena.com/news/Sams...i-T604-graphics-pops-up-in-benchmarks_id34681
True. I think have to wait for SGS4 for those benchmarks. More interested in browsermark, peacekeeper, google octane numbers. google itself mentioned that sunspider is outdated.
http://sunspider-mod.googlecode.com/svn/data/hosted/sunspider.html
hot_spare said:
EDIT: I think he says 620ms in video. Also, I am sure it will get better as the Chrome OS code is optimized for ARM. This is just first release. Exynos 4 has been optimized to limit. They can't push it any further now, at least not by a big margin.
Click to expand...
Click to collapse
In the video he mentions 620 ms, but in the comments he states 660 ms for Sunspider when asked the question, I chose the 660 ms to be conservative.
Antutu benchmark!
I kept looking, and found something interesting now.
"Supposedly" first Antutu benchmark for Exynos 5250. Now the values show it's running at 1.5GHz. For a dual-core SoC, 14185 score sounds very good.
The most interesting part is the 3D graphics numbers. This is 3x compared to 4412 SoC.
Source: http://www.antutu.com/view.shtml?id=2718
With more optimization, this can be really powerful.
Looks like this chip will also end up in the Nexus 10
Turbotab said:
Looks like this chip will also end up in the Nexus 10
Click to expand...
Click to collapse
That's going to be a monster tablet.
Peacekeeper browser benchmark for Exynos 5250 gets more than 1200:
https://plus.google.com/u/0/+JoeWilcox/posts/8LrBK9CKJG4
Better than any other mobile SoC so far.
This chip rapes every other chip out there, even the s4 pro and apple a6. look here- http://www.androidauthority.com/exynos-5-dual-benchmarks-125134/
prajju123 said:
This chip rapes every other chip out there, even the s4 pro and apple a6. look here- http://www.androidauthority.com/exynos-5-dual-benchmarks-125134/
Click to expand...
Click to collapse
Dude please don't use the word rape, an ugly word. But we must wait for the a GL Benchmark results of the Mali T-604 against the Apple A6 & A6X, I hope it beats them, but it won't be easy Apple used a lot of die space to create them.
Hoping for a Exynos 5450 (5 Quad) by March or April of 2013
Is it the same chip they use in the new Chromebook?
lz2323 said:
Is it the same chip they use in the new Chromebook?
Click to expand...
Click to collapse
Exactly the same, dual-core Exynos 5250 - Mali T-604.

Categories

Resources