Kernel 53.22 problem with SamFlex [SOLVED]

[samo79]

@DAX

Wow what a test, bravo !

I really like to do some more test for myself, now that i put up my A1 again i can be able to do also some more test-comparison between the two, of course don't forget that we have the same Sam440 model so eventually we can help each other and share our test to discover any new eventual issue.

I really hope that we will find support from other official betatesters and programmers

[Spectre660]

No freezes for two days thus far . 800mhz Sam Flex .Update 3 kernel .Only Moire and Forrest enable in Screenblanker

[DAX]

@Samo79
Let us know if you find strange behaviours of any kind!

I really hope that we will find support from other official betatesters and programmers

Now that would be a nice next step

@Spectre660
As I said, the crash problem is due to extra stress (and thus "heat" I presume) generated by the new kernel, but 1) it only happens with taxing tasks and 2) (the most important part) not all 800Mhz Sams are born equal and some have high resistance while others do not (apparently mine and Xenic's one at the moment).
A CPU fan solves the problem and the system becomes very stable, however it's no longer silent and we have no explanation as of why Kernel 53.12 allows any kind of abuse (and even some additional overclock) without problems and without any need for a fan, while the new one needs a CPU fan on certain "less fortunate" 800Mhz Sams.

Also we have no explanation as of why Virtual memory doesn't work with 53.22 while it does with 53.12, and last (but not least) why the Sam440_SetUp file doesn't yeld the same good results it manages with the older Kickstart (values comparison in a previous post).

[ChrisH]

@DAX
It sounds like the new kernel is more optimised, so it can push the CPU harder, and so generates more heat (enough to break your borderline stable overclocked CPU).

Non-working virtual memory is a separate issue, and will be a real pain when I want to recompile PortablE itself (512MB is not *quite* enough).

[Spectre660]

@DAX

Would something like the distributed.net client trigger the problem ?

[K-L]

Dax : As ChrisH, I think the new kernel is much more optimized for the Sam440 than the previously one.

Don't forget that Update 3 has brought many improvements for the Sam440 (DMA acceleration amongst others) and I presume they are now used fully with all DSP instructions being used to get the maximum of the 440ep.

I'm using a Sam440ep MiniITX 667 that I've overcloked to 733Mhz (FSB=147). I've been compelled to add a fan since I'm using it for reading videos (thanks to DvPlayer) and playing games (PrBoom, etc...), and at 733Mhz, the Sam always crashed at one moment or another.

Knowing that the 733 Mhz and 800 Mhz models are overclocked 667 Mhz versions, no doubt that if some heavy load is pushed on the 440, it will crash. There are now very silent fans and I could only advise Sam440's owners to put one in the machine if they are using overclocked versions.

[DAX]

@ Spectre660
I don't know, I installed the fan now and the system is rock solid to make other crash tests I would have to remove it (shivers).

@K-L, ChrisH
To think that my first install on top of OS4.1.2 went perfectly well and I had no problems aside from the sudden crash , if the cpu-fan was advisable it certainly would have been nice to hear so in the "read me" (made two clean installs ).

Alas, I'm sure that the different behaviour of the many 800Mhz Sams out there must be a surprise for Acube which couldn't anticipate it, some grief stays as this system was sold this way "from factory" as a fan-less system, and was guaranteed to work fine like that (ie: I did not overclock it myself). Not anymore it would seem.

At this point it would be advisable to get back the higher performance for the video bus etc. (same as what I get with Kickstart 4.1.1 + Sam_440SetUp) as even now, when I flip the old Kickstart back, I get less sound hiccups in E-UAE games. I use Turrican 2 intro sequence as a first test to measure performance (it's very sensible to minor changes) and I remember the happiness I felt when I heard the soundtrack sounding 1:1 with my A2000 (instead of being interuppted by random micro stutters) , now with Kick4.1.3+ Sam440_SetUp current results, the micro-stutters are back.

[xenic]

@DAX
Well I've been operating with Kernal 53.12 long enough to be convinced that the freezes are related to the kernal. However, my erratic mouse problem is also an erratic keybaord problem. When I was writing a post in another forum my keyboard input got very slow and erratic. Once that happened, the mouse pointer was also slow and erratic as with previous occurances of this problem. After I restarted USB I checked the USB log and debug buffer which both had repeated lines of this:
Debug buffer = [OHCI] Missed interrupt (status=0x00000046)!
USB Log = OHCI Controller Task Unit 0 | OHCI | Missed interrupt for controller 0!

Since these USB interrupt failures are happening at about the same frequency as the freezes with the new kernal, I wonder if interrupt failures are at the root of the freezes with the new kernal. I wonder if overclocking could affect interrupt synchronization. Could different CPU and memory access clocking affect the interrupts?? It's beginning to look like my only solution is to revert to Update2 and wait for X1000 to be publicly available.

[cha05e90]

this system was sold this way "from factory" as a fan-less system

Yep - but to be truthful, I never really trusted this, 'cos there are more things than a "hot" cpu. The first two years my SAM440ep (mini-itx, 667MHz) resided in a really small, fanless housing. But burning DVDs with my slimline DVD writer and having a 2,5 inch harddisk spinning above the CPU (and getting rather "warm" while operating) made be overthink the situation. Now my 667-MHz-system is in a bigger housing with an active (but rather silent) fan - moving away all that warm/hot air from board, pico PSU, hd and DVD writer. I can sleep better now.

BTW: My SAM440ep (mini) 667 MHz seems to be rockstable up to now. Everythings works as expected, mouse&keyboard are o.k (both A-Eon AmigaOne) and I just made a full backup of my partitions (some dozens of GB) to an external USB hd with full USB2.0 speed.

[ChrisH]

Since these USB interrupt failures are happening at about the same frequency as the freezes with the new kernal, I wonder if interrupt failures are at the root of the freezes with the new kernal.

I get stuff like:

DoControlXfer | Destall failed: io_Error = -32
...
HID Server | Destalled EndPoint 130 (Ifc 0x584EABD0)
...
HID Server | Interrupt req. returned -35
...