GeForce 466.47 WHQL driver download & discussion

i semi agree, but theres a point where the user can't go further because of things like the nvidia symbols being unavailable publicly.


Hyperthreaded CPU's
Hyperthreaded/Multicore cpu's with C states enabled
Certain Motherboards with digital/adaptive EPU managers (some can be swapped back to analog/static with an available utility to resolve this)
Faulty Monitors / Cables
Faulty storage
Faulty IO Controller (SB/NB/PCIEonCPU) (not necessarily all slots/lanes are affected - seen it with a P6X58D)
Unusual error correction on the bus that could be any of the devices that make up the chain.
Outdated Realtek audio and lan drivers
Bad sata driver
Windows itself
A certain Kingston A series SSD (yep, im going there )
PS4 controllers (these are notorious for DPC in the hid driver)


It would be safe to assume that only a couple of those listed have any possibility of being interactive with the system when a depth format changes, i've bolded those.

atleast one of the affected users has a 6700K, so ReBAR can be ruled out as a factor.

There is the chance of some sort of bad interaction or errors on the pcie bus, but i don't have any personal history with 6700k's experiencing that.
Motherboard configuration of the PCIE bus, features enabled, among other things, are a possibility.

Hey, did we all forget nvidia enabled gpu passthrough on recent drivers? has anyone tested disabling Intel VTd/AMD IOMMU?

did you happen to make a trace etl?

 

I can dig that ofcourse! there are some exceptions to what end users are capable of them self drivers and code wise.

 

Was he on recent driver when you tested?
Remember certain drivers like 460.89/457.51 don't trigger the dpc latency spike.

 

If I remember correctly neil has a 2k card and has been vocal about having similar spikes in certain drivers after 457.51

 

There was a user over at Geforce forums a few weeks ago describing what you guys are discussing.
Basically high latency in any driver past 457 and possibly a mem leak in 465, 466, which doesn't occur with 462.
Maybe that's why Nvidia Studio, Quadro & Tesla are still 462.
I get a very slight pop on my phones whenever I'm selecting options in the menu of GTA 5, which started with 466.
So could very well be the driver.

 

Hi Kman, it happens on both the 980 GTX and the RTX 2070... One is a I7 2600k the other an I7 4970K with two completely different motherboards and memory set ups. Happens both with the onboard and soundcard

 

pascal and maxwell from my testing are not affected for whatever reason on the systems they are running on.

did someone try turning off vt-d yet?

 

I have vt-d at disabled (since forever). As soon as I switch to 8 Bit colors the spikes are there in latencymon (just idling at the desktop Powersetting to "High"). I haven't seen this spikes at 10 Bit (HAGS off for me).

But please, can someone else test this again? As one single systems findings must not mean anything.

 

I just played Doom Eternal for 10 minutes with Latencymon running on a second monitor and didn't have any nvlddmkm.sys spikes (usually when the spikes happen I get a half second hitch in games) but as soon as I opened this screenshot in Paint3D the latency shot up to 1100.



I know that this isn't the intended usage for Latencymon but it's definitely helped identify a link between these weird half second hitches I've been having intermittently while gaming and the nvlddmkm.sys spikes happening at the same time. The spikes don't always occur while playing games though and happen seemingly randomly even with no 3rd or 1st party apps running and Windows left showing the desktop.

 

doesn't look like you're affected, running the old driver?

 

It's 466.55, the latest hotfix driver.

 

- redacted -