Samsung Orion SoC - dual-core A9 + "5 times the 3D graphics performance"
- Thread starter Mike11
- Start date
I have a hard time with acronyms
SGS2 is Samsung Galaxy S2 using Samsung Orion (Mali400MP)
G2x is T-mobile made by LG using Tegra2
Is this correct?
If I'm right I don't understand the reference to OMAP4.
Yes, it was the G2X. I was mistaken about the chipset inside; it's Tegra 2, not OMAP4.
Ok so my memory isn't playing any tricks on me. I recall the Odroid A being somewhere in the 4400 frames league (give or take) at 1360*768 and the Exynos smart-phone in the 5500 league at 800*600. First decreased while the latter increased.
I think with GLbenchmark in general, as many of the results are "community upload", you will always have big inconsistancies, as you have no idea how many multi-tasks were going on at the time etc.Also its not out of the question that the new and old odroid-A results were with different drivers or indeed with different frequencies given that the platform is a development one.
Similarly, with android based devices, people might root them and install various Oses and under/over clock, or indeed just totally change the device name that gets reported back to try to be "clever".
One just has to look at the "unprocessed results" to see all sorts of strangely named devices, the site seems to clear out all the clearly bogus ones from time to time.
Its also not inconceivable that handset providers/graphics IP providers might do customised drivers, that never get released due to instability, just so they can get an improved benchmark.
One might be best to only reference the "team tested" results which might have a bit more integrity and are likely to have been acheived using a retail device. Or indeed the apple results which have limited and very well defined OS updates (and even then you have the jailbreak issue and possibly people faking device names).
Similarly, with android based devices, people might root them and install various Oses and under/over clock, or indeed just totally change the device name that gets reported back to try to be "clever".
One just has to look at the "unprocessed results" to see all sorts of strangely named devices, the site seems to clear out all the clearly bogus ones from time to time.
Its also not inconceivable that handset providers/graphics IP providers might do customised drivers, that never get released due to instability, just so they can get an improved benchmark.
One might be best to only reference the "team tested" results which might have a bit more integrity and are likely to have been acheived using a retail device. Or indeed the apple results which have limited and very well defined OS updates (and even then you have the jailbreak issue and possibly people faking device names).
arjan de lumens
Veteran
Sure - except that the "team tested" results don't seem to include any ARM Mali based products at all.
At this time, thats probably a fair indication of the retail experience (with the exception of the GS2), in that Mali has not made it into major brands that the team thought significant to test ? I imagine it shouldn't be too long before they'd formally test GS2.
Mind you maybe they've just totally stopped doing "team tests" and left it to the community, which would be dissapointing.
I see the first proper "GL benchmark team tested" result has been posted for the galaxy S2, with a score of 4582. One assumes this is with an over the counter unmodified phone,set up in a reference state.
It is in contrast to the results being uploaded by the "community" which were hitting over 6600. Overclocked processor and/or drivers optimised for the test spring to mind as possible reasons.
The comparison with the google Nexus S (the only official results for the S5PC110 chip) are interesting, and I've included the ipad2 as well.
http://www.glbenchmark.com/compare....D2=Samsung GT-i9100 Galaxy S2&D3=Apple iPad 2
It is in contrast to the results being uploaded by the "community" which were hitting over 6600. Overclocked processor and/or drivers optimised for the test spring to mind as possible reasons.
The comparison with the google Nexus S (the only official results for the S5PC110 chip) are interesting, and I've included the ipad2 as well.
http://www.glbenchmark.com/compare....D2=Samsung GT-i9100 Galaxy S2&D3=Apple iPad 2
Wishmaster
Newcomer
Finally, some reliable results. I was getting sick with all those results achieved with modified rom's and driver's.
I hope that now they'll benchmark HTC sensation so we could compare them both on stock ROM's.
They are not normalized for screen resolution. They are each run at the native resolution of the display.
Blah IMG really needs to get its MP drivers in better shape. Albeit not yet team tested the Motorola Droid3 (SGX540@307MHz) gets over 3k frames in 960*540.
Google Nexus and Samsung Galaxy SII run at 800*480, while the Apple iPad2 at 1024*768.
Well 4 fragment cores and 4 TMUs are nothing to sneeze over even if the frequency isn't as high as many of us expected.
The GLBenchmark Team retested Exynos and verified its very high Community results.
The mobile market is ripe for some high-end games pushing more geometry and heavier shading now, and a clearer picture of how all of these SoCs stack up would then be possible if the game dev included some benchmarking modes.
Apple's A5 would have more room to push forward, and Exynos obviously wouldn't remain capped either.
The mobile market is ripe for some high-end games pushing more geometry and heavier shading now, and a clearer picture of how all of these SoCs stack up would then be possible if the game dev included some benchmarking modes.
Apple's A5 would have more room to push forward, and Exynos obviously wouldn't remain capped either.
Similar threads
- Deleted member 13524
- Mobile Devices and SoCs