Windows 10 Anniversary update - Nvidia Drivers 369.09

[JohnLuke]

If you use Windows 10, recently were "updated" (ahem) to the Anniversary edition, use an Nvidea card and am experiences game freezing, you may want to check out these threads. There is a new driver from Nvidea (369.09) that is supposed to rectify the problem. My PC was fine after the Windows update until I played E-D using the Vive. I got freeze ups after just a few minutes.

https://www.reddit.com/r/Vive/comments/ ... niversary/

http://steamcommunity.com/app/250820/di ... 346408283/

http://forums.evga.com/Windows-10-Anniv ... 27280.aspx

I installed the new drivers using the method described above, and E-D & Vive seem to be playing well together again.

[thebs]

Just FYI, 10 Anniversary Edition has issues when you launch applications from other drives than Windows is on (e.g., D:\Program Files [(x86)] and not C:\Program Files [(x86)] when you have a C:\Windows).
- https://www.thurrott.com/windows/window ... eze-issues

I actually understand the internal NT issues with this, which goes back many releases, and it looks like Microsoft re-introduced a long (even if only partially) solved regression. I.e., the architects and engineers that figured this out are long gone, and the Indian and outsourced developers didn't know about it when it came to the Update, and re-introduced it. I know where the problem is at, and I know why it's going on. It wouldn't surprise me if they broke all of those added abstraction layers in NT with the new update as well. This is really bad, I mean ... really, really bad.

Frankly, I'm so tired of the FAT-based / Drive Letter approach in general, which Microsoft has known has been a problem since the early '90s on OS/2, and kept promising to solve it with CarioFS (Later WinFS) at some point for NT. But that's what you get from a company that bought an Operating System from a guy in his garage who ported CP/M from the 8080 to 8086/88 without a license, and it didn't even have directories, just drive letters. Heck, Microsoft swiped the added directory code form Xenix by SCO, who they owned a 20% stake in (and later resulting in cross-licensing to solve the lack of a license to use aforementioned code) for DOS 2.x.

[CMDR Abil Midena]

Must be beta drivers. I still see July 14 / 368.81 on nvidia's website (14 Aug 11 PM EST). It's curious that this is the first time I can remember when going through Win to update a driver pulls a NEWER driver than is available on the manufacturer's website.

I'm running 10 anniv with no problems, but I guess it's a VR issue.

[thebs]

Must be beta drivers. I still see July 14 / 368.81 on nvidia's website (14 Aug 11 PM EST). It's curious that this is the first time I can remember when going through Win to update a driver pulls a NEWER driver than is available on the manufacturer's website.
I'm running 10 anniv with no problems, but I guess it's a VR issue.

Could be just some VR with nVidia.

Just be aware that Windows 10 Anniversary Update has a crapload of issues in general ... enough that Microsoft has officially admitted several of them, and recommends downgrading. The volume code is really pissing me off, because this is not the first time they've had this, and it's a total regression.

Having to deal with NT's volume code v. Linux's VFS is like the dark ages v. post-industrial revolution. I have had too many bugs over the years that boils down to NT volume code assuming everything is on %BOOTDRIVE%. Seems prefetch and other NT code itself is incapable of reliably referencing anything other than C: drive.

[Liberty Prime]

Just FYI, 10 Anniversary Edition has issues when you launch applications from other drives than Windows is at (e.g., C:\Windows).
- https://www.thurrott.com/windows/window ... eze-issues

I actually understand the internal NT issues with this, which goes back many releases, and it looks like Microsoft re-introduced a long (even if only partially) solved regression. I.e., the architects and engineers that figured this out are long gone, and the Indian and outsourced developers didn't know about it when it came to the Update, and re-introduced it. I know where the problem is at, and I know why it's going on. It wouldn't surprise me if they broke all of those added abstraction layers in NT with the new update as well. This is really bad, I mean ... really, really bad.

Frankly, I'm so tired of the FAT-based / Drive Letter approach in general, which Microsoft has known has been a problem since the early '90s on OS/2, and kept promising to solve it with CarioFS (Later WinFS) at some point for NT. But that's what you get from a company that bought an Operating System from a guy in his garage who ported CP/M from the 8080 to 8086/88 without a license, and it didn't even have directories, just drive letters. Heck, Microsoft swiped the added directory code form Xenix by SCO, who they owned a 20% stake in (and later resulting in cross-licensing to solve the lack of a license to use aforementioned code) for DOS 2.x.

Thanks for the heads-up. Love your posts btw.

[thebs]

Just FYI, 10 Anniversary Edition has issues when you launch applications from other drives than Windows is at (e.g., C:\Windows).
- https://www.thurrott.com/windows/window ... eze-issues

I actually understand the internal NT issues with this, which goes back many releases, and it looks like Microsoft re-introduced a long (even if only partially) solved regression. I.e., the architects and engineers that figured this out are long gone, and the Indian and outsourced developers didn't know about it when it came to the Update, and re-introduced it. I know where the problem is at, and I know why it's going on. It wouldn't surprise me if they broke all of those added abstraction layers in NT with the new update as well. This is really bad, I mean ... really, really bad.

Frankly, I'm so tired of the FAT-based / Drive Letter approach in general, which Microsoft has known has been a problem since the early '90s on OS/2, and kept promising to solve it with CarioFS (Later WinFS) at some point for NT. But that's what you get from a company that bought an Operating System from a guy in his garage who ported CP/M from the 8080 to 8086/88 without a license, and it didn't even have directories, just drive letters. Heck, Microsoft swiped the added directory code form Xenix by SCO, who they owned a 20% stake in (and later resulting in cross-licensing to solve the lack of a license to use aforementioned code) for DOS 2.x.

Thanks for the heads-up. Love your posts btw.

I'm still trying to confirm a lot of details, but this isn't the first time it's happened.

E.g., the root cause prior was in Prefetch code, long story. I've been posting more on my Facebook/Google+, because I've been into this code before.

But in-a-nutshell, binaries/libraries not on %BOOTDRIVE% (where \WINDOWS is at) will have their prefetched references (disk references to the object code that can be loaded, so it doesn't have to search the disk -- easily a 1,000x increase used by other OSes as well -- e.g., Linux ldconfig) be incorrect, corrupted if you will, pointing to something that doesn't exist or, worse, is junk. So it blindly loads it, and executes it ... and the NT executive has a bad day.

► Show Spoiler

Then combined with how the GDI-WinForms code works rooted in every program (part of the reason portable .NET code, designed for Mono, doesn't use WinForms -- long story), with the Task Switching in EXPLORER.EXE (NT may be pre-emptive, but the GUI is not, the bane of Windows and why a single, unprivileged program can crash the whole system), along with how programs "root" themselves in a directory (usually the program's directory -- something unique to DOS/NT/Windows), it basically causes a total hang. Especially if it's causing an infinite I/O wait on the load, which prefetch keeps feeding over and over and over, so everything trying to hit that disk is unavailable.

In Microsoft's defense, I/O wait can hit any OS, and eventually hold up access and/or exhaust resources. But in NT, with GDI-WinForms and EXPLORER.EXE, you have so many great 'attack vectors' to choose from! And the bonus, especially with home users and gamers, is that Administrator privilege is usually required for software to work, or at least enabled in their accounts by default.

► Show Spoiler

I remember the NT5.x (2000/XP/2003) study that identified that 8% of all Windows crashes were due to the Spooler -- yes, the printer subsystem! If your printer didn't respond to NT constantly tapping it on the shoulder, NT would nuke itself (in-a-nutshell ... worked on some Samba code that had to emulate the "I'm fine" that was really a "go away, I really didn't check" 9 out of 10 times). It's why Microsoft deprecated the Spooler printing, TWAIN scanning, etc... in NT6.x (Vista/7/2008/8/8.1/2012), although there's still a lot of legacy (and missing replacements -- hence why most corporations use the open source stuff that's better, including the Windows versions).

I just have to shake my head at stuff like this ... especially because it was reported during the insider testing, and Microsoft chose to ignore it anyway. If I was a conspiracy theorist, I'd say it was the arbitrary "Anniversary Date." But I'm not willing to go there, largely because it detracts from the bugs themselves.

[smartroad]

I accidentally pressed the wrong button on a shutdown on my desktop and it installed the update for me Not hit any major issues but not played ED since the update. I hate Microsoft, I really wish Valve would push SteamOS more, it would be great for Linux gaming in general. Also if they could come up with an API interface for DirectX that could mean playing Windows games on Linux.

[Roger Wilco Jr]

Just be aware that Windows 10 Anniversary Update has a crapload of issues in general ... enough that Microsoft has officially admitted several of them, and recommends downgrading.

I hope this means that they've stopped the upgrades. At least I haven't had this abortion of an upgrade pushed on me yet.

[Dudley]

Errr - let us just say, *cough*, for the less PC savy of the forums, *cough*, what should we/I, do if prompted to update windows..



o7 Dawg

[thebs]

Just be aware that Windows 10 Anniversary Update has a crapload of issues in general ... enough that Microsoft has officially admitted several of them, and recommends downgrading.

I hope this means that they've stopped the upgrades. At least I haven't had this abortion of an upgrade pushed on me yet.

For non-roll-up updates, Windows 10 updates are forced. But since Anniversary Update is a roll-up update, it's not, even if some of its components still get installed. So for now, just skip any offering to upgrade to Anniversary Update.

BTW, my personal favorite yet ...
- https://www.reddit.com/r/Windows10/comm ... ed_system/

More and more "optimization" software -- like those from NAND EEPROM (SSD) and other hardware vendors -- is disabling System Restore. Why? Benchmarks. Now Microsoft has gotten into the game, purposely nuking System Restore for ... yep, you guessed it, benchmarks! I've about had it with how Microsoft tags its updates, and sets its defaults.

I.e., Microsoft literally now cares about only number targets and feature-performance reviews, than stability and recoverability. They've utterly returned to their '90s self, after actually improving things in the '00s (other than NT6.0/Vista). I don't know one security researcher that believes Microsoft cares about accurately and appropriately labeling their own security patches and updates any more. Not one. They've become a joke.