Two RK2818 Tabs & Two Wildly Different Benchmark Scores

Huggy · **Joined:** Tue Apr 27, 2010 6:06 pm **Posts:** 939

Well, wildly might be slight exaggeration, but the discrepancy we encountered running Linpack for Android benchmark on two different RK2818 tablets was enough to cause a stir. To get you all up to speed, we’ve had the opportunity to play with a couple of ‘generic’ RK2818 7-inch Android 2.1 tablets and just see how they work/function etc. For the most part, the experience has been smooth and uneventful… except for one thing. Over time, we began to notice one of the tablets felt a touch slower than the other but we couldn’t really place it. It was hard to find differences in playback of 720p HD video. Browser scrolling speed is subjective and unreliable. So we decided to do a benchmark. The 3D benchmark variants of Quadrant and Neocore didn’t work at all, so we used Linpack. The results are telling… it seems one of the generics has significantly better Android implementation than the other, despite the exact same hardware. What could the difference be in the ROMS? Well, speculation is meaningless at this point so just take a look at the scores yourself, after the break. One implication to consider, though, is this: if you’re in the market for an RK2818 tablet, how do you know you aren’t getting the worse one, especially with all the generic no-names floating about? The answer is you don’t. Yet.

A 3 second difference to complete the benchmarks with roughly .4~.5 less MFLOPS, the lower pictured RK2818 unit scored significantly lower, confirming our suspicions of a slower tablet. It ‘felt’ slower, but we couldn’t quantify it by eye like you can between, say a MIDnite and an aPAD. But the synthetic benchmark helped us to quantify that difference.

Note: All processes were killed using Advanced Task Killer prior to running the benchmarks. Both units are powered by RK2818 Rockchip SOC, with 256MB DDR2 RAM. Both run Android 2.1.

maderin1 · **Joined:** Sun Sep 30, 2007 2:50 pm **Posts:** 20625

so are the definitely the same hardware, i mean raz did mention that some 2818 were being made with 128mb instead??
would like to see you have a play with the londge padone as this, on the face of it, to be a more refined implementation of Android, it reminds me of the rocktab/acho and looks like it could have been done by the same people

thanks for the test - out of interest, compared to the rocktab whats the general feeling, is the better developed rocktab a smoother runner despte its supposedly lower hardware setup??

Huggy · **Joined:** Tue Apr 27, 2010 6:06 pm **Posts:** 939

mad - yes both had confirmed 256MB memory, as per "about device" within Android. So I suppose take that for what you will. We had suspected one had less RAM too but it turns out one didn't.

And yes, compared to the RockTab's, these do "next-gen" rockchip a disservice. I have no doubt that at some point the RK2818 will outdo RK2808 devices, but these two didn't hold a candle to the RockTab in terms of general usability. It's not to say they were unstable... they were just kind of slow and far less refined. They also sported standard 2.1 UI's rather than any kind of customisation a la RockTab.

However, we've got to remember that the RK2808 chip isn't that different from the RK2818. The RK2818 only adds better hardware decoding for 720p videos (which we couldn't really "notice") and support for Android 2.1 / DDR2 ram instead of SD ram (as well as typically 256MB versus 128MB). I mean, the differences between them just aren't massive enough to overcome immature Android implementation versus refined. At least, that's my take.

I would also be very interested in playing with the Londge PadOne... unfortunately don't own one.

maderin1 · **Joined:** Sun Sep 30, 2007 2:50 pm **Posts:** 20625

cheers Huggy
this is a imp3 review of the padone and the linpack results for it are
http://bbs.imp3.net/viewthread.php?tid= ... ght=padone

Product Type: Rui Zhipadone1
System Version: 2.1
Linpack (MFLops): 4.276
nbench: Memory: 0.821 Integer: floating-point 0.193 1.825
Jbenchmark: Write: 0.52MB / s Read: 0.82MB / s
Neocore: 1.6FPS
GPU bench: not work
Benchmark: Graphics: 81 CPU: 320 Memory: 157 File System: 56

Huggy · **Joined:** Tue Apr 27, 2010 6:06 pm **Posts:** 939

Wow very interesting. Seems to have a good Linpack score, slightly better than our "good" RK2818.

I didn't know that so many different ROMS were going around for this, if we are to account ROM difference as the culprit.

Could inferior PCB/memory chips also be a contributor?

Further, obviously the Londge tablet has some better Android because we couldn't get Neocore to run here. I might try find some of the other benchmarks they used.

maderin1 · **Joined:** Sun Sep 30, 2007 2:50 pm **Posts:** 20625

some stuff here
http://bbs.imp3.net/viewthread.php?tid=908557 not sure if any use

does seem to be very diff implementations, much like the different ones for the 2808, you only have to look at the rocktab compared to the apads to see that, seems that rocklchip provide a base and its up to whoever to refine that how they see fit and to better match it to their hardware, like the rocktab has done

Huggy · **Joined:** Tue Apr 27, 2010 6:06 pm **Posts:** 939

Thanks mad - that should be useful for some further testing down the line.

Raz · **Joined:** Fri Jul 25, 2003 5:53 am **Posts:** 22200

maderin1 wrote:

some stuff here
http://bbs.imp3.net/viewthread.php?tid=908557 not sure if any use

does seem to be very diff implementations, much like the different ones for the 2808, you only have to look at the rocktab compared to the apads to see that, seems that rocklchip provide a base and its up to whoever to refine that how they see fit and to better match it to their hardware, like the rocktab has done

What pretty much happens is that dev houses get the Android port from Rockchip and then they customize it, some do a lot of work, others do little to none, this customization can lead the Android port to being better or worse, hence the difference in performance. However build and quality of materials being used can also affect performance too.

MP4 Nation Forums
Back to MP4 Nation - The MP4 Nation Blog - The Raz Special - Themes