NVIDIA ForceWare 368.39 drivers causing problems

I’ve been having hard times figuring out random “The stub received bad data” errors as well as random failing of Task Scheduler service.

Event Viewer was also getting filled with “Faulting application name: svchost.exe_wuauserv” in connection to NVIDIA’s “nvwgf2umx.dll”.

With more digging I’ve uncovered that ForceWare 368.39 drivers are indeed bugged (probably entire 368.xx series), but didn’t know what exactly.

With more Googling and later testing myself, I can confirm it’s caused by “MFAA” setting. If you leave it off (default state when drivers are installed clean) everything will be fine. But if you change it, it will corrupt the settings and cause all these problems.

NVIDIA is already aware if it and they are working on it, till they release a fix, best way to get this stupid crap resolved is to disable MFAA. If that doesn’t help, reinstall drivers, use reset settings to default during installation and never turn it on (for now at least). You can adjust everything else, just leave MFAA alone.

I’ll keep you posted about the updated driver which includes an actual fix for this mess.

Fix DisplayPort not working (No Signal)

I keep on randomly experiencing this, either after changes I do to system or just when restarting the system and it all of a sudden decides to fuck itself up for no real reason and based on Google, it seems to be a very common thing with NVIDIA graphic cards and high end ASUS monitors like my ASUS VG248QE (144Hz gaming monitor). Though I’ve seen reports with other brands as well.

It’s a really annoying one because you often don’t think it’s “just this” and you try to exclude countless other things, wasting time when solution for the issue takes just 1 minute of sitting still. Yeah…

When I was experiencing this issue, I’ve disassembled half of my system, tested with backup graphic card (which only happens to have HDMI output, not DisplayPort so I was basically just testing if the rest of the system is working), reinstalled my ASUS Strix GTX 980 and the damn thing wouldn’t output image through DisplayPort to my ASUS VG248QE. When it was connected through HDMI, image was there. Connected through DisplayPort, “No Signal”. I was going mad WTF is going on. Was testing different DisplayPort ports on graphic card, turning monitor on and off using Power button, changing input on monitor, nothing. But nothing changed because one variable in the problem wasn’t really getting changed or “reset”…

Then I figured out yet another dumbest solution to seemingly complex issue…

Unplug monitor from power for 1 minute and then plug it in again. Voila, DisplayPort magically starts working again! How retarded is that eh?

It seems like there is some electronic in monitor that goes haywire here and there and only way to get it unstuck is to cut power to the monitor till all capacitors in it fully discharge. This apparently resets the “problematic” state and it then manages to sync with a graphic card correctly and starts outputting the image again through DisplayPort.

Hope this helps. Just try to remember this as the first “go to” solution when it happens again so you won’t waste time testing everything else… Try this first since it only takes a minute.

Jokes on NVIDIA’s broken GTX 970 :D

Hitler’s reaction to GTX 970 being broken…

Interview with NVIDIA engineer on GTX 970 issues…

And some more from their friends at AMD…

GeForce GTX 970 has dark secrets…

As users have uncovered suspicious things regarding GeForce GTX 970, it has now turned out that NVIDIA was in fact lying to everyone and they are now admitting it around the corners. GTX 970 doesn’t actually have 64 ROP units. It only has 56. Despite them advertising it as 64 ROP cards. Bus width? Nope, 224+32 instead of full-blown 256bit. 3,5 + 0,5GB of VRAM instead of 4GB.

Now I’m going to place tinfoil hat on my head and draw some sort of conclusion on it. You buy a GTX 970 today, all is well, games work fine. What will happen when users start using 4K more in more demanding games? If users already noticed something isn’t right now, who really thought it would go unnoticed 1 year later? To me it seems like NVIDIA engineers said, so we have a GTX 980 core here, we can use crippled ones and make GTX 970. And we’ll somehow patch the memory thing and no one will notice it anyway. And so they’ve done that…

Now, why is this important? Today, games don’t use that much memory to really go past 3,5GB very often, but things will change and this might happen very soon with GTA5 and other demanding games on the horizon. But when they will use it, that 0,5GB segment will do shitty stuff to the performance. In computers, systems are as fast as the weakest link in the system. In this case this 0,5GB of VRAM is the weak link. If the rest of GPU and VRAM has to wait for data to be fetched from this crappy part of slow VRAM, the fact is, the rest will stall. And while this doesn’t necessarily mean unplayable, it will result in a framerate drop and other issues like the reported stuttering.

Call me biased because I’m using AMD Radeon and because I’ve decided to wait for R9-380X to see what’ll be better, but I’m kinda laughing here, knowing how people are defending a broken piece of hardware just because they bought it and no one ever wants to admit he made a mistake. GTX 970 is a hacked crap that works well when conditions are good, but when they won’t be quite soon, it won’t be so good anymore.

At least GTX 980 is a proper card that isn’t hacked, but it’s a lot more expensive and leaves a bad aftertaste knowing they were hacking things and covering them up from consumers until someone caught them and now they are making excuses and explaining how it doesn’t affect anything. Do whatever conclusion you want, but I don’t like it when companies do that…

GeForce GTX 970 suffering from dementia

Well, not quite, but apparently GeForce GTX 970 is suffering from a design flaw that causes all weird shit to happen when VRAM usage goes beyond 3,3GB. It starts to lose memory…

Not much is known so far and people are still trying to figure out things, but from the looks of it, NVIDIA chopped the GTX 980 chip in order to make it fit for the lower model segment aka GTX 970 and they sort of quickly patched it in order to seemingly make it work fine with 4GB of on-board memory. In most conditions it’s fine, but apparently things go haywire when you try to allocate more than 3,3GB of VRAM.

There are some rather lame excuses in terms of “segmented memory where that extra 0,7GB is added as an extra” and “well, you’re taxing the graphic card, that’s why it’s working so badly”, but the reality is, you just tax it with a bit more textures to spill over the 3,3GB barrier and things go south despite cards supposedly having 4GB of VRAM. No such thing on a full-fledged GTX 980 cards and these aren’t all that faster than GTX 970 to automatically explain why this isn’t happening on GTX 980.

While it all looked like NVIDIA made a really decent card, it’s starting to turn out it was a fast and sloppy job. Good thing I haven’t pulled the trigger on GTX 970 and I was like hours away from transferring the money. Luckily R9-380X news came out and I’ve postponed the purchase. Which now seems even better decision seeing how GTX 970 is less than ideal thing and that waiting for R9-380X looks like a reasonable thing to do. No one knows if AMD will have problems of their own, but so far it’s a better option either way…

You can follow this link in order to see how things will unfold:

http://www.techpowerup.com/forums/threads/geforce-gtx-970-design-flaw-caps-video-memory-usage-to-3-3-gb-report.209205/