April RK3368 True Octacore Thread

[Shomari]

Anybody can use RK's bullshit patches in any firmware, including mine, by the way. I don't support it myself, necessarily, and I don't wish to expend a single second of my time supporting RK's non-conforming, bootleg api methods.

For those that don't quite understand the reasoning or basis behind that line of thought, I sympathize with you a little. But I'm not explaining public vs. private api libraries and methods a million times every time it comes up.

Rockchip is shooting itself in the foot with a .50BMG and it's the users suffering. Simple as that.

[Shomari]

....OK. Who got the 670 single core on Geekbench today with RK3368 running at 1608mhz?

Is it someone on this thread? I know it's definitely a freaktab user. Ayo cOOLio - that you? It doesn't have your name in the build props, though.

It's not the score that's a big deal (it's actually pretty much identical to mine from about 8 months ago)...

... what I need to know is how are you getting Geekbench to complete and not crash while running tests, because mine segfaults at the 76% mark, everytime. I used to be able to set the mem alignment differently in the ramdisk and it would complete, but that changed once I altered the HMP tunables, so I suspect whoever is running Geekbench with my cpu hack is doing it with an altered - maybe even stock - boot image.

Also, look at the crypto score for my result, it's like 36 or something, compared to around 400 in the latest versions, as Geekbench didn't support the crypto in RK3368 firmware back when I was able to complete tests with it ~ version 4.0.0.

Geekbench is ridiculously artificial but that makes it good for measurements, and for testing single core performance particularly, which is much more important to quantify than multicore performance.

Other tools in combination with Geekbench is how I got my firmware to where it is today.

[Shomari]

I've tested further, and it's definitely not the boot.img.

My guess it's the firmware with the power menu fix, because it uses different cpu code that keeps the cluster topology intact, and that maintains the memory/cache coherency, allowing the test to complete.

That code can be further optimized, there's a reason I abandoned that method. . .

[Sabinus]

I got 670 Geekbench last night..
and I did it just after installed full rom, then Z4 images


And thanks for the explanations. I totally agree that RK is doing it wrong. At the same time I continue the search for the most suited fw for my need. I will try the one you pointed in the link and tell you after the results

[Sabinus]

Just tried Zidoo s X6 Pro v9.5 rom mod by Lonix.
Standard performance but works Bitsteram audio output with any material: either DD or DTS. It even manages to send TRUE HD and DTS HD MA (only ZDMC so far.. even if is not really true) to my "special" configuration:
Analog Creative Gigaworks system connected to cheapest DAC (25 dollar Audiogear 5.1 decoder).
TV is connected with sp/dif to Decoder and also with HDMI to Android box, this is why I needed bitstreaming output through hdmi, to pass multichannel audio from Android box through tv to Decoder.
You see, it was a lot of money wasted to let Gigaworks linger in a corner unused. Thus I gave them a new life without having to invest in a 350 usd multichannel receiver plus at least 100 more for a IEC 6xxx compliant device
Quality is acceptable, not great but certainly worth the 25 dollars invested

[cOOLio]

I got 670 Geekbench last night..
and I did it just after installed full rom, then Z4 images

congratz



Shomari
I was pretty sure that you, or let's say your ROM, will beat that high score pretty soon. Because as you always mention. It needs more juice, so it has more juice. But I'm already proud of what I did with less juice. And as soon as I release my new ROM (more infos on the release date can be found here) you will see the crazy shit I did. One tip. Try to mixup your T8C & OC techniques and your shut down bug will be history. The main reason for this T8C bug is that you disabled the second (litle) core.

[Sabinus]

Oh, I forgot to mention: I used the default power adapter which is a mere 2 amperes. I think it's more about luck or the quality of the assembly.. Though I didn't open the box, it didn't feel particularly hot..

[dkholm56]

Shomari,
flashed firmware you sent, then flashed stock kernel. Box boots to M8Pro flash screen and stays there. Tried it twice. I did modify build.prop as you told me then flashed the boot.img you provided. made sure the box would boot. then flashed the firmware you sent, and then stock kernel. Was that correct.
thanks

I've been trying different things to get the firmware to boot with no luck. Tried to flash the boot.img you gave and then flash different combinations of the images you gave me. Still the best I can get is to boot the M8Pro boot image successfully. Beyond that nothing works to get the box to boot the firmware. Fortunately I can go back to stock rom so I'm willing to try just about anything to get this firmware to boot.
thanks again for all the help you've given.

[Shomari]

I got 670 Geekbench last night..
and I did it just after installed full rom, then Z4 images


And thanks for the explanations. I totally agree that RK is doing it wrong. At the same time I continue the search for the most suited fw for my need. I will try the one you pointed in the link and tell you after the results

... it's gonna sound like BS now, but I had actually determined on my own that it must've been someone that was using the Z4 files I posted, because that one keeps the cluster toplogy intact, and so it doesn't have the memory/cache coherence issues that can lead to segfaults when running benchmarks.

And I'm 100% in sympathy with you that I can't provide a viable, standards compliant HD audio supporting firmware, RK just doesn't want to play ball on that end. It's not because I haven't tried.

And I will continue to try with each new release of sources.

[Shomari]

Oh, I forgot to mention: I used the default power adapter which is a mere 2 amperes. I think it's more about luck or the quality of the assembly.. Though I didn't open the box, it didn't feel particularly hot..

... actually, I was able to tune the vcore across the board on that particular build to get a pretty good, power efficient overclock going.

If you settle on a firmware and still want to keep the speed and performance of my firmware, just let me know and I'll attempt a mod of whatever firmware you settle on to add the performance enhancements and leave everything else intact

[Shomari]

congratz



Shomari
I was pretty sure that you, or let's say your ROM, will beat that high score pretty soon. Because as you always mention. It needs more juice, so it has more juice. But I'm already proud of what I did with less juice. And as soon as I release my new ROM (more infos on the release date can be found here) you will see the crazy shit I did. One tip. Try to mixup your T8C & OC techniques and your shut down bug will be history. The main reason for this T8C bug is that you disabled the second (litle) core.

... you're on a roll. And yup, we had discovered since back in November that the power menu bug was due to RK's dvfs needing to be able to 'see' the L cluster..

[Shomari]

I've been trying different things to get the firmware to boot with no luck. Tried to flash the boot.img you gave and then flash different combinations of the images you gave me. Still the best I can get is to boot the M8Pro boot image successfully. Beyond that nothing works to get the box to boot the firmware. Fortunately I can go back to stock rom so I'm willing to try just about anything to get this firmware to boot.
thanks again for all the help you've given.

I haven't forgotten about you ... I'm going to get back to finishing a port for your device very soon. It's a very tricky one, but we're gonna figure it out..

[Shomari]

What if I told you folks that there's an ancient firmware - the very first on my Z4 from 2015 - that somehow manages to have DOUBLE the r/w performance of any other RK3368 firmware I've tested, to date? I even reinstalled it just to test, and I confirmed what I remembered to be true. If I can manage to figure out how it's done, I'm baking it in to my firmware, for sure.

I might need cOOLio to help with that, since he's on a tear optimizing for performance.

[Shomari]

congratz



Shomari
I was pretty sure that you, or let's say your ROM, will beat that high score pretty soon. Because as you always mention. It needs more juice, so it has more juice. But I'm already proud of what I did with less juice. And as soon as I release my new ROM (more infos on the release date can be found here) you will see the crazy shit I did. One tip. Try to mixup your T8C & OC techniques and your shut down bug will be history. The main reason for this T8C bug is that you disabled the second (litle) core.

... they can't be 'mixed'. It's either you 'disassemble' the cluster topology (true octacore hack) or keep the clusters intact (power menu fix) ... unless you know another way, and can still preserve the performance and power efficiency of both methods.

There's one other possible way, but it's a little more involved and I never really pursued it much..

[Shomari]

dkholm56 try this system.img with your M8 Pro. Just a bare bones, quick test, to see if it boots and runs correctly.

Flash this system image along with the modded boot image I sent you, over your stock firmware.

https://mega.nz/#!LERADAqA!f4_oa4gEvuRxG0PEkWUDT_wD6oN6Ok9IWtM1bYgbn2o