New NVIDIA Geforce WHQL Driver 353.06

You kind of confuse things when you show me a link of the 970 vs 780 Ti when you're talking about the 960 beating the 780 in pCARS. Looking at it from the proper perspective, which is 960 vs 780 benchmarks, I see what you mean.

http://www.gamespot.com/articles/review-nvidia-s-gtx-960-is-a-good-not-great-1080p-/1100-6425427/

This is an even better showing of the distance between those cards.

http://techreport.com/review/27702/nvidia-geforce-gtx-960-graphics-card-reviewed/8

 

960 vs. 780, 970 vs. 780 Ti...what difference does it make? Point is Kepler is underperforming. 970 owns 780 Ti in pCARS too even though 780 Ti previously beat 970 in every test.

 

I want to see where Kepler performance has degraded in the same games it dominated in. In other words, a before and after, to prove it isn't just Maxwell optimization. Oh wait I have a 780 Ti, I'll go look for myself.

 

Alright look, as cynical as I am and as much as I dislike nVidia, I'm not cynical enough (yet anyway) to think they'd intentionally cripple Kepler performance. Depending on how cynical you are, they've either abandoned Kepler, or there's no more optimization to be done for Kepler. BUT that still doesn't excuse the 960 creeping up on the 780 in 4 recent games (COD:AW, Far Cry 4, pCARS, TW3) when in almost every other game tested it's a good 20%+ behind.

The other explanation is that those 4 games make use of some Maxwell-specific technology that Kepler doesn't benefit from, and thus suffers a large performance penalty.

 

I don't think Nvidia is intentionally crippling Kepler. I think they're putting their boot down and saying "no more optimization for Kepler" regardless of whether there is still more performance to be squeezed from it. If that's true, this is the fastest I've seen Nvidia abandon an architecture since GeForce FX.

 

Tisk tisk what a shame. After a clean install my experience with 353.06 echoes that of @Mr. Fox, I'm throttling down to base clocks in benchmarks and games. Oh well, I guess it's back to 347.88 for me.

 

Yes, and it does the same thing. I have used stock, SVL7, my own that I normally use. Nothing changes with the behavior at all. 345.20 is the only driver that does not throttle.

I just modded the vBIOS I normally use and set all P5 core and memory clocks and voltage to match P0 and set all of the core and memory clock states equal, and then set the power table minimum to 67% (stock minimum is set to 1mW and I changed the minimum to 100000mW) using the Kepler tool. This helped minimize the throttling a lot, but it still throttles. Having removed 405Mhz from the clock table I am puzzled that it still goes there. The lowest clock should be 850MHz. I was able to fix the memory and stop it from throttling to 400MHz... holds solid at 1250 (or my normal 1500MHz memory overclock) where it was dropping to 400MHz. It is absolutely a power throttle. If I can keep the GPUs below 70% of TDP it does not throttle. As soon as it exceeds 70% and stays there a couple of seconds the core drops to 405MHz.

This scary part of this is, this is EXACTLY how 980M behaves in the M18xR2. It's like deja vu.

 

Isn't 405MHz where the drop is for thermal throttle? I ran into it a number of times with the 880M......

Firing up a few games and seeing if the behavior continues would help too. If you ask me, it sounds like the system thinks the cards are at 95C.

That has been my argument. Nvidia put a lot of development into Maxwell, maybe these games are taking advantage of Maxwell improvements.

Not good. Are other Alienware owners having this issue? I don't really head over to that forum now that Dell sold out.

 

It guess that is possible. Maybe some kind of sensor in the drivers after 345.20 is giving off a false reading. The core temps are running between 65-71°C under full load.

I have no desire to use this new driver. Even if it does not throttle in games, I will still know it is bugged and it will drive me absolutely batty. I'm back on 345.20 again and I'm so mad I could smash something. NVIDIA sucks. So does AMD, so that leaves us up the creek affectionately known as feces, without a paddle.

I don't know. I'm guessing those with single 780M are OK. I checked with SLI disabled and 353.06 did not throttle. Again, points to some kind of cancer in the driver trying to cut power because it thinks my machine is using too much. That fits with the GPU-Z PerfCap reason as well. I did raise the minimum power value in the vBIOS from 320 to 540 and that did not change anything.

 

Spoiler

 

I WANT G-SYNC!!! I think gt72 with the lg display will support g-sync, because it supported that modded driver that enabled g-sync on some notebooks

 

Man... Do I really want to mess with my working 780 Ti or not...

 

Yeah, no kidding... that was a short ride. I don't think they're intentionally doing it either. I think they have made a huge mess of things and their driver engineers are just too stinking stupid to know how to fix it. At least we can give them credit for making my 780M setup run like 880M... gee, thanks. *reaches for butcher knife* Hey, NVIDIA... come over here, I wanna show you guys something.

 

Yep 347.88 works like a charm, no throttling. It's a good thing I have such a backlog of older games to play then isn't it? I don't have to worry about missing out on optimization for the latest titles I'm not playing anyways.


Sent from my iPad using Tapatalk

 

347.88 runs best on my 980Ms too.

 

If it ain't broke...

On another note, is the consensus that 347.88 is still the most usable driver for 780M SLI?

 

I don't know but it was the best for my 880Ms and my 980Ms love it too

 

god I really love the witcher 3 so much, I mean I love the whole series but this game really takes the cake. Such a beautiful and immersive game, the best game I have ever played by far.

 

After very short test ... Not a single frame difference compared to 345.20 desktop drivers ... Will test longer later tonight but atm. expecting nothing

 

Not for me. Has the same problem as all the newer drivers. That version was the first to cause throttling and every single release since then has been the same.

 

Dito, 345.20 FTW

 

OK, now this is weird. I had two 0x124 BSOD trying to install 353.06 on my M18xR2. This is normally associated with overclocking the CPU with voltage too low. I thought that was really odd because the CPU was only clocked to 4.3GHz using settings that are 101% stable 24/7, but I persisted and got it installed.

Now, I have been fighting this driver for about an hour on my Clevo P570WM and it is doing the same thing. Four times (so far) 0x124 BSOD on the Clevo also. I finally got it installed, or so I thought, and things (settings) are missing. NVCP says I do not have permission to change any settings and every time I reboot Windows scans my drives for errors. Something really screwy is going on. Attempted to install it under Windows 8 since Windows 7 is giving fits, and it is doing the exactly the same thing. Without any mods at all I have to disable Driver Signature Enforcement for Windows 8. It crashed during installation now trying to go into Safe Mode (Windows 8) to run DDU and start over, I have a black screen. Cancer drivers from hell.

I am going to try the desktop version on my Clevo and see if that fares any better.

 

Oh gods, well let us know if the desktop version works out any better for you @Mr.Fox


Sent from my iPad using Tapatalk

 

Wow. I missed a whole lot here.

I'm kind of afraid to try these drivers now. I think I'm gonna sit on 350.12 straight.

I still think that the driver issues are because the new drivers are being written against the BIOS in the Alienware models because I haven't seen any MSI or Clevo owners complaining of hardcore throttling like Mr. Fox sees... but these drivers are something else. Never heard of constant BSODs... I'll wait it out and see.

 

Not gonna lie, anybody else complaining about constant BSODs and I would've assumed user error. Yikes.

 

The driver problem is with sli and not single graphic card.

 

*points to 780M SLI in sig*

I should note that while Mr. Fox has had throttling issues for quite a while, every driver he has had throttling issues on his AW with that he did not have on his Clevo, I have not had throttling on in my Clevo (excepting 352.86/353.00 which I have not installed). So when he's getting problems in his Clevo... I want to stay far away.

 

It was 337.88, not 347.88 for the 880M. 347.88 and 344.75 are my go to drivers for the 980M. Sorry, the driver names are too close.

That BSOD thing sounds crazy. Maybe I'll try the drivers on both my desktop and my laptop since apparently my body has decided that I do not need more than 3 hours of sleep since it woke me up shortly after 3AM...

EDIT: this popped up over Roseanne. I guess it is destiny.

 

@Mr. Fox

Just run the benchmark and I got 200 points more than the 345.20 driver bro:

Stock settings in nVIDIA Control Panel / stock clocks / modified vBIOS



Val!date M3H

I wasn't able to get the OSD to display no matter how hard I tried, I installed the latest MSI AfterBurner and activated the graphcs I want but they wouldn't even show the OSD for me to post a screenshot of the live clocks but what I know for sure is that it gave me a higher benchmark

 

Seems to work fine for me...

http://www.3dmark.com/compare/3dm11/9882156/3dm11/9180838

My physics scores have always been all over the place but the graphics score got a 531 point boost.

Got a boost all the way around on my 780 Ti

http://www.3dmark.com/compare/3dm11/9882214/3dm11/8705493

 

Thank you for your f33db4ck, rep added

 

I also added my 780 Ti.

http://www.3dmark.com/compare/3dm11/9882214/3dm11/8705493

Smaller boost but still a boost across the board compared to the 340.52 drivers I was using previously.

 

Let's see what @Mr. Fox has to say about this, maybe it's a problem in his laptop lying somewhere?

 

After playing the Witcher 3, I can safely say that these drivers are excellent for my particular machine.

Spoiler

Frames, Time (ms), Min, Max, Avg
185872, 2076204, 31, 410, 89.525

Ultra with all the bells and whistles, including the hair on all characters.

The clocks boost at 1126 the entire time. The dips are when I went into a scene that wasn't demanding (like the big dip where I was playing the card game). I was playing with vsync on for most of the game. Where you see the utilization spike through the roof is when I got annoyed with the stutter it caused and killed it.

 

Installed these this morning on my 970m. Installed fine, no issues. Haven't had time to try them in Witcher 3 yet, but did run a 3DMark11 run before work. I picked up +670 P points over 352.86 at all stock clocks for GPU/CPU.

http://www.3dmark.com/3dm11/9883103

Seems like they are definitely worth trying. I'll report back with any gaming issues this evening but not expecting any problems.

EDIT:

Was curious what a +135mhz GPU OC would do, P10110, which happened to be my fastest recorded run on this machine, even besting the 344.75 fastest by about +P100
http://www.3dmark.com/3dm11/9883173

 

Welp. Seems like it's time to upgrade drivers. And hope I get no BSODs. XD

 

I think imma give these drivers a shot, even though i doubt it'll improve much performance on my 750m but it wouldnt hurt to try

 

I just ran a little bit of 3DMark Firestrike and my overclock held; no other issues with the drivers itself so far. Have not tried any CPU overclocks. @Ethrem think you can run a little 4.2GHz for a short while and see if you're stable? I don't think anyone else here who's tried it is a heavy OCer like Mr. Fox, so I'm wondering if his 24/7 OCs are the reason.

 

It's possible, but I find it unreasonably oppressive, immoral and unacceptable if NVIDIA is doing something in their drivers that fiddles with system power utilization and upsets an otherwise rock-solid system with a CPU overclock. This is overstepping their boundaries and they would deserve a massive lawsuit over their lack of discretion for taking liberties with that kind of improper influence. I will not compromise my ability to OC just to have the luxury of running a cancerous driver with a few game optimizations. The Jolly Green Giant won't own me like that... no way.

Still having issues with stability using this driver on my Clevo. I ran DDU using the "install new video card" option to block Windows from scanning for new drivers and that stopped the BSOD issues trying to install it. I still cannot get into Windows 8.1.1 Safe Mode on the Clevo now, even after removing all drivers, so I have to run DDU in regular mode for Windows 8. Not sure why the screen is going black in Safe Mode now. It does not go black with Windows 7 Safe Mode, just Windows 8.1 Safe Mode. I don't like what these drivers are doing. Even if they were working correctly knowing they have hooks this deep is objectionable.

On the Alienware M18xR2 and Clevo P570WM I have random freezing during Windows logon with this driver installed. I found the Killer Network Manager (present on both machines) was contributing to that. The hard freeze on both machine was happening when the WiFi was attempting to connect. After removing the Killer Network Manager the freezing during logon happens in about 1 out of 5 Windows startup sequences, where it was literally 1 out of 10 that would load without freezing before.

I may try a clean OS install on both machines, but I am not looking forward to spending hours installing Windows 7/8.1 dual boot from scratch on two machines. There has to be some kind of explanation.

Side note: installed 353.06 on my M18xR1 with a single 780M and it works flawlessly. Performance is identical to 345.20, no throttling, same max OC, no BSOD... using same modded vBIOS, too. Ran 1175/1500 on the M18xR1 without ever seen a drop and pulled 365W in 3DMark 11 test #1 on my UPS digital meter with a single 780M, LOL.

I may pull the Ultimate-N 6300 WiFi card from the M18xR1 and try it on both machines before reinstalling Windows. Now, wouldn't that be interesting to find out that freezing and throttling are being caused by a WiFi card or it's drivers?

Edit: another observation on my Clevo... with this driver the Windows 8.1 Clevo Hotkey program crashes when I try to launch it. On 345.20 it does not do that. That may also be Killer related, since it controls WiFi and Bluetooth. Also interesting that I have the same WiFi card and drivers on the M18xR2 and it has similar stability problems with newer NVIDIA drivers. More to follow as time permits testing.

Bottom line: even if changing to a different WiFi card or drivers fixes these problems for me, this is NVIDIA's fault. Nothing changed on my M18xR2 or P570WM, other than newer drivers from NVIDIA. This is not OK.

 

Oh, do not get me wrong. My attempt is simply troubleshooting, not to say it's right or wrong.

I can say that the killer cards I've seen have proven to be problematic in very many ways, but I have never seen them so bad. While it does not exactly fix the problem that drivers have, if the wifi cards are indeed the issue, then the 7260ac wifi cards might be a good replacement (at least for your Clevo).

My Clevo Control Center does not crash, further lending to your idea. I hope that you find the solution.

 

Where are the DRS settings? Don't see it in the control panel. GTX 970m.

 

This is on my 780GTX x2 SLI

 

I think if we have Intel + Nvidia solution (optimus) we don't have DSR...

This sucks.

 

I formatted and installed Windows 7 today just to play 3DMark

Here are my scores (all stock settings / no overclocking)

353.06:



345.20:




BUT, the 345.20 score was before doing the modded vBIOS so let me do it again......

 

My laptop has Intel HD Graphics 4600 + Dual 780GTX and I have DSR but they're not running together in Opticrap mod, you have to switch between either discrete or integrated graphics via the FN keys then you don't see the other card once you're on one

 

I have optimus + 970m. Sager NP 8268. No DSR settings .

 

Ok so here is the 345.20 test again but this time with the modded vBIOS, a few more points extra.......

again...... listen to what @Mr. Fox says, he knows his stuff inside out........stick to 345.20 if you care about every bit of performance

345.20 / no overclocking of CPU nor GPU ( http://www.3dmark.com/3dm11/9884494 ):

 

I tried it and now it's only using the Nvidia GPU apparently (not totally sure), via the FN key but it still doesn't show up.

I should post something on nvidia forums maybe, they might reply (never done it before so I don't know...).

 

That's kind of the point of Optimus. You are running off Intel graphics. Intel does not support DSR. The nVidia GPU is simply a number cruncher. That's it. It's the sacrifice you make for battery life.

Older Alienwares and the MSI GT72 were the only machines that had MUX switches... most normal laptops do not, including all the Gigabyte, Lenovo, Dell, HP, Acer, ASUS, Razer, previous MSI and Clevo laptops that weren't in SLI configurations.

If you don't have a MUX switch you don't get DSR. You need to be able to run your machine in dGPU only. The intel iGPU must be fully disabled. Unless you have a MSI GT72 or an Alienware 17 R1, Alienware 18, or any of the M17x and M18x machines, you don't have a MUX switch for disabling the iGPU.

 

And now same benchmark with 345.20 but slight overclock on the CPU to 4.1 GHz on all cores thanks to @Mr. Fox :

( http://www.3dmark.com/3dm11/9884524)