Simba123I did solve my problems.  I set the card to -20Mhz below stock speed.  Re-installed the .Net 4.0/4.5 Framework, re-installed the core17, and most importantly (I think) fixed a cooling problem for the VRMs on my motherboard.

First of all, I can say, that that at least concerning my sys, MoBo-VRM overheating can´t be the source of the problem, becase my MoBo ist fully watercooled as my GTX Titan is too.
 
But after a lot of thinking about the problem, with all its various crashes with different error messages concerning the kernel, I had the idea, that it could be a matter of the VRAM. Like the different crashes and bluescreen error codes coming from defective (or too highly overclocked) RAM on a WindowsPC.
 
So I decided to lower the clock speed of Titans VRAM from 3000 to 2800 and -- BINGO! -- since yesterday I folded eight 8900 without any crash or even "bad state detected" error!. I even had been able to overclock the GPU to 1137 MHz clock speed using stock voltage and am testing 1150 now.
By the way, folding performace /ppd-output is not affected by the speed of VRAM.

 
This is the solution at least in my case, but I hope (and expect) that lowering the clock speed to the VRAM of your card, this will mean the same for your case... .

Has anyone seen a drop in PPD with those 8900 latlely?
 
Maybe is it just my system, I updated to W8.1 and 320.49 drivers and now I have seen a huge dump in points. Gone from 180k PPD to around 117k PPD on both my Titans. TPF is though basicly the same or around 2:42, so it seem like they changed how it is calculated.
 
*edit*
I found the soulution, my ***** mistake :). When I set up Windows 8,1 it did put in the wrong time-zone and therefor the bonus was calculated wrongly.

itsmekirill

I left my Titan running when I went on vacation about a week ago with forced max fan since I wasn't going to be around to listen to it.

Apparently it really opens up with lower temps in terms of turbo boost, when these new WU's started coming it spiked from about 35k PPD to about 170k.

http://folding.extremeoverclocking.com/user_summary.php?s=&u=520049

If the PPD is changing that much something must be wrong or
the system is in uses a lot when your around or overheating with
to low a fan speed an throttling back. PPD should be about the same at all times.

clo007

Try folding these 8900s on a GTX580 like me. That is all I have caught as of late and it only gets me 40-50k PPD. :(
I see all of y'alls PPD from the same WUs and you can tell the Fermi got nerfed.

Still a lot more then I remember months back, barely would get 33k on v7 no flags. Around 850 clock now, with advanced flags, it seems to be around 45K ppd, that's still more then my 6 core with HT would make doing SMP, least more then my w3680 at 4.3 Ghz. 
 
One of my Titans really, really hates these units. Even at stock speeds the bus width usage would just drop from ~30% (which I know it's actually working) dropping down to 0, GPU usage is still 95+%, but core temps dropped 3C and TPF seems to double, but the WU still goes on. Plus I seem to be hitting my breakers limit in my room when I try and fold now...darn summer time. 
 
Also the 1 Titan that works mostly, PPD has dropped, seeing others mention that. I completely remember getting around 180K-200K PPD, now it's around 140k PPD. Not so bad really, but it's weird that my 690 PPD is still the same on these WU's but Titan took a dive. 

watch out for 8901 WU's
they are 66% larger