GeForce Driver 352.86 WHQL {DO NOT USE DEGRADED PERFORMANCE BY 50%}

Some twenty years ago we had at least four names in the GPU market - S3, VIA, Nvidia and ATI, until S3 couldn't keep up, and got snapped up by VIA. Same goes for the CPU market - we even had different competiting architectures until 2006 when Apple moved to Intel. But everyone knows all of that. My dad had an old Power Mac, and it was ridiculously powerful back in the day. Apple did have some fair claim to having made the "most powerful personal computer on Earth".

Most of the names are now solely found in the mobile and fringe supercomputer market, where there's some decent competition - Qualcomm, Apple, Samsung, Texas Instruments, MediaTek. Unfortunately, the mobile market has also been hogging most of TSMC's fabs for the 20 nm process - which partly contributed to Nvidia delaying, and eventually announcing that they'd skip 20 nm altogether.

 

Oh there were way more than 4 players in the GPU market back then. Even as late as the early 2000s it wasn't just the big 3 we have nowadays. I remember once owning an SiS something PCI (not PCIe) graphics card back then. It was a fixed function card, probably equivalent to a GeForce 2. No happy stories about 3dfx Voodoo sadly as that was before my time. Anyway, the reason none of those other guys survived is because their products sucked (or bad business decisions in the case of 3dfx). Everyone except AMD, Intel, and Nvidia folded or got bought out hence leading to market consolidation.

Edit: Actually I think Matrox still exists in some form. Now that's a name I haven't heard in forever. They looked like they were on to something with Parhelia (I think that's what it was called) in the early 2000s before that bombed and consigned them to history basically.

http://www.techspot.com/article/650-history-of-the-gpu/

http://www.techspot.com/article/874-history-of-the-personal-computer/

And don't remind me about the old PowerPC Macs. Maybe they were fast in synthetic benchmarks or certain PPC-optimized pro apps, but they were slow as heck in real usage compared to the Pentiums and K6/Athlons I had at home. Or maybe it was because Mac OS 9/early OS X sucked back then. Who knows. I remember using the old colorful plastic Power Mac G3/G4 and original iMac in elementary school, and then in middle school we had the flat panel iMac G4 and G5. I went to school in Cupertino so all school computers were nothing but Macs. Apple switching to Intel right at the start of their Core golden age, as opposed to sticking with IBM or going to an already declining AMD, probably saved their Mac business as far as I'm concerned.

Welp now I probably just dated myself.

 

Tested newest driver 352.86 on my Aw17. 3Dmark11 Score P9566 vs P9528 from previous driver 350.12 which several have problems with. Tested with oc (1072/1490). Outstanding results with both last drivers from Nvidia. Could it be that you are using a too aggressive vbios?
http://www.3dmark.com/compare/3dm11/9835196/3dm11/9761352

 

Yeah, this driver sucks. I kept getting crashes in a number of games. Going back to 350.12.

 

Good lord what kind of temperature you are getting on these clocks?

 

Cpu 73° and gpu 67° Liquid ultra

 

That Liquid Ultra insane!!!

My GPU doesn't heat up doesn't heat up much on the clocks in my sig except in Shadow of Mordor where I hit 88c ... But ofc not 67 more like 77 & the power draw is insane.

You are using a 330w PSU?

 

I dunno guys - with GTA V, I see a clear performance drop, with the exact same settings throughout, same CPU speed and same GPU overclocks. Settings used are in the first spoiler.

Spoiler: GTA V Benchmark with 350.12

Frames Per Second (Higher is better) Min, Max, Avg
Pass 0, 35.782333, 68.642380, 56.289032
Pass 1, 31.863939, 69.579659, 50.709991
Pass 2, 10.409784, 65.972618, 53.220806
Pass 3, 36.670658, 69.782982, 58.124638
Pass 4, 20.014585, 85.921021, 52.870316

Time in milliseconds(ms). (Lower is better). Min, Max, Avg
Pass 0, 14.568259, 27.946753, 17.765450
Pass 1, 14.372017, 31.383440, 19.719980
Pass 2, 15.157804, 96.063469, 18.789644
Pass 3, 14.330141, 27.269758, 17.204409
Pass 4, 11.638596, 49.963562, 18.914206

Frames under 16ms (for 60fps):
Pass 0: 20/527 frames (3.80%)
Pass 1: 15/472 frames (3.18%)
Pass 2: 5/478 frames (1.05%)
Pass 3: 13/546 frames (2.38%)
Pass 4: 138/6043 frames (2.28%)

Frames under 33ms (for 30fps):
Pass 0: 527/527 frames (100.00%)
Pass 1: 472/472 frames (100.00%)
Pass 2: 458/478 frames (95.82%)
Pass 3: 546/546 frames (100.00%)
Pass 4: 6036/6043 frames (99.88%)

=== SYSTEM ===
Windows 8.1 64-bit (6.2, Build 9200)
DX Feature Level: 11.0
Intel(R) Core(TM) i7-4710MQ CPU @ 2.50GHz (8 CPUs), ~2.5GHz
8192MB RAM
NVIDIA GeForce GTX 860M , 2089MB, Driver Version 350.12
Graphics Card Vendor Id 0x10de with Device ID 0x1392

=== SETTINGS ===
Display: 1920x1080 (Windowed) @ 60Hz VSync ON
Tessellation: 2
LodScale: 1.000000
PedLodBias: 0.000000
VehicleLodBias: 0.000000
ShadowQuality: 2
ReflectionQuality: 1
ReflectionMSAA: 4
SSAO: 1
AnisotropicFiltering: 16
MSAA: 0
MSAAFragments: 0
MSAAQuality: 0
TextureQuality: 2
ParticleQuality: 1
WaterQuality: 2
GrassQuality: 2
ShaderQuality: 1
Shadow_SoftShadows: 3
UltraShadows_Enabled: false
Shadow_ParticleShadows: true
Shadow_Distance: 1.000000
Shadow_LongShadows: true
Shadow_SplitZStart: 0.930000
Shadow_SplitZEnd: 0.890000
Shadow_aircraftExpWeight: 0.990000
Shadow_DisableScreenSizeCheck: false
Reflection_MipBlur: true
FXAA_Enabled: true
TXAA_Enabled: false
Lighting_FogVolumes: true
Shader_SSA: true
DX_Version: 2
CityDensity: 1.000000
PedVarietyMultiplier: 1.000000
VehicleVarietyMultiplier: 1.000000
PostFX: 1
DoF: false
HdStreamingInFlight: true
MaxLodScale: 0.000000
MotionBlurStrength: 0.200000

Spoiler: GTA V Benchmark with 352.86

Frames Per Second (Higher is better) Min, Max, Avg
Pass 0, 22.939838, 62.294437, 43.533764
Pass 1, 14.467400, 58.700443, 36.710854
Pass 2, 12.079161, 62.497437, 40.147896
Pass 3, 24.989998, 63.102978, 45.724854
Pass 4, 0.085769, 166.514557, 42.779007

Time in milliseconds(ms). (Lower is better). Min, Max, Avg
Pass 0, 16.052797, 43.592285, 22.970676
Pass 1, 17.035646, 69.120918, 27.239901
Pass 2, 16.000656, 82.787209, 24.907906
Pass 3, 15.847113, 40.016010, 21.869944
Pass 4, 6.005481, 11659.209961, 23.375952

Frames under 16ms (for 60fps):
Pass 0: 0/404 frames (0.00%)
Pass 1: 0/342 frames (0.00%)
Pass 2: 0/369 frames (0.00%)
Pass 3: 1/421 frames (0.24%)
Pass 4: 30/4819 frames (0.62%)

Frames under 33ms (for 30fps):
Pass 0: 398/404 frames (98.51%)
Pass 1: 325/342 frames (95.03%)
Pass 2: 352/369 frames (95.39%)
Pass 3: 415/421 frames (98.57%)
Pass 4: 4681/4819 frames (97.14%)

On average, a 12.46 FPS decrease throughout the benchmark. And that's in a game, although I get similar 3DMark scores. The decrease is very, very noticeable, during gameplay too.

 

The folks that are having instability and lower performance need to raise hell in the GeForce Forums about it. NVIDIA needs all of the criticism you can dish out right now to keep focused. Let's do all we can to stir up hate and discontent in the forum where it will get most noticed by Mr. Green Jeans. Post it here, T|I and other forums and give them hyperlinks to the nasty things you have posted elsewhere. Bad publicity is the only thing that is going to keep these losers honest, and that takes some effort. The time for using tact and diplomacy has expired. They don't respond to that because they don't care, so it's time to play the bad publicity card again and try force them to act right out of public humiliation. If we can tank their stock prices, they will react to that.

Official NVIDIA 352.86 WHQL Game Ready Display Driver Feedback Thread (Released 5/18/15)

 

Okay, so this is interesting... Since The Witcher 3 patch, 352.86 performs better than 347.88, in game.

 

Indeed. CDPR has greatly reduced instabilities with patch 1.03 - just had a three hour run and finally left White Orchard, although there are several locations there that I haven't explored yet.

Or, it's the fact that I reduced my overvolt by 75 mV to push it back down to a +200 mV offset.

 

I've installed this version. If there's one thing that's clear it's that my GPU really doesn't like being pushed. Any more than +120MHz on the GPU clock causes the GPU driver to crash no matter the firmware or driver version. On the other hand, nothing I've been able to do causes the performance drop that others have experienced.

Edited to add: these are the notebook drivers. I haven't done anything with trying to get desktop drivers installed, yet.

 

I've been trying to OC in benchmarks with this driver and it throttles the minute you OC a bit.. On stock, it's completely fine.. when I tried a +200/200 OC on my 970M in BF4/ World of warships, it held the clocks throughout the game etc.. I've had no games crashes etc... Looks like NVIDIA wants to kill benchmarking but gaming is completely fine even with OC..WTF???

 

Nvidia is the anti-Samsung

 

A driver for the panty-waisted masses.

 

330W is a must... GPU was 66˚C degrees during the test and not 67˚

 

Are you using the notebook version or desktop version?

I ask this because I've run 2 full Heaven benchmark runs and a Fire Strike run at +135MHz/+250MHz on 347.88 and 2 full Heaven runs and 2 Fire Strike runs with the same settings on 352.86. No crashes, close enough to identical Heaven scores, but Fire Strike is interesting. It scored ~9400 on 347.88 but only ~8900 for both version 352.86 runs. My takes:

The notebook versions are dodgy. I don't know why, don't know enough about how all these bits work to identify the reason, but I couldn't set the GPU clock higher than +120MHz without crashing on the notebook drivers while all of my runs on the desktop drivers at +135MHz worked perfectly.

Version 352.86 is extra dodgy. I could not reproduce the massive degradation that has been reported but a 10% overclock gave me only a 5% better score.

 

NVIDIA developers like... "Performs better for The Witcher 3, kills everything else."

 

I switched drivers twice in five hours 352.86 -> 350.12 to play GTA V, then back to 352.86 to play TW3, because I get a forced reboot (! WTF Nvidia???) if I play it with 350.12.

 

Yeah, you have to switch drivers now. Thanks, NVIDIA.

 

so GTAV performs badly if on the latest drivers?

 

Yes. Look at this post that I made a while ago.

 

Gotta love how far back we have gone. I remember having to change drivers for specific games back when I had my ATi Radeon 9800 Pro... I didn't miss that. So annoying.

 

Thank you all for taking the time to test the drivers in various games and benches. I don´t have the energy or time for that atm so I´m staying on 347.88 which is working fine.

 

Heck, go back another generation to the first of the 3D accelerators which didn't have drivers. A game had to be coded specifically to use a given vendor's card. 3dfx Voodoo, anyone?

 

I completely wiped 347.88 from my PC using DDU, and installed a clean install of 352.86. For some reason, it works much better now. Express Installation failed before and crippled my system... Even though it's better than before, it's not as good as 347.88 in benchmarks. Now I receive about 500 points less in most benchmarks as opposed to 4000 or something. You can thank Windows 8 for that. I hate Windows 8.

Verdict: We have to switch drivers depending on what we'd like to do now. Benchmarking is much better in 344.xx through 347.88 series of drivers. Depending on the game, upgrading may be beneficial. Ever since Witcher 3 updated, 352.86 performs better with this game, and this game alone! I'll be keeping it installed only until I finish the game.

This is absolutely ridiculous for NVIDIA to be doing. Now only that, but they're now intentionally killing performance of Kepler by restricting performance in their latest drivers for the newest games to force people to upgrade. How pathetic is that?

 

I've seen something similar when installing over existing drivers. My suspicion is that it's something flakey in one or more versions of the PhysX engine but I don't think I have the chops to pinpoint it.

 

NVIDIA's new logo:

 

Well, NBC is reporting that overclocking has once again been disabled by this driver. Nvidia just ripped off our balls, then tried to stitch them back on, then ripped them off again.
Nvidia's latest drivers again disable overclocking for GTX 900M graphics cards

 

Maybe they're afraid it will hurt sales (i.e. Pascal) if people are able to squeeze 30%+ out of Maxwell, so they're cutting out the possibility.

That's the only reason I can think of at this moment. Maxwell has proven to be much more impressive than any of us had anticipated. I remember back when it was merely a rumor, people, including myself, were expecting it to be just another "re-brand" attempt. Lo and behold, mobile and desktop components have never been closer.

 

352.86 notebook driver does not block overclocking. It is quite limited compared to the desktop version (limited over-voltage range for example) but it does not outright block overclocking.

 

If I remember correctly, it does block core clock overclocking. Memory clock overclcoking is doable, but core clock, which is the more important of the two, is locked.

 

That's not the driver. It's the GPU firmware.

Edited to add: I've spent a bunch of time over the last three days with notebook driver versions 345.20, 347.88 and 352.86 for my 980M, and with stock firmware (core clock locked), Prema's stock firmware (core clock unlocked) and Prema's OC firmware (everything unlocked) trying to reproduce the performance hit in the thread title. No luck with that, but I can assure you that the driver itself does not block overclocking.

Edited again to add: For reference to where I started this delve into insanity:
http://forum.notebookreview.com/thr...-to-the-batcave.767105/page-658#post-10008473

 

The issue presents itself with an Express Install over an older driver. After I wiped all traces from my PC and clean installed the 352.86, the score was only 500~ points less than 347.88.

Happened twice with me while experimenting. Maybe it's hit and miss.

 

Old files or libraries does appear to be part of the problem but I don't think that's the whole story. I think @Ethrem is probably right that there's something in the firmware. Something twitchy in the firmware that interacts with something twitchy in the drivers that results in problems under OC conditions.

 

The root cause of it is a new hybrid GPU virus named NVIDUMUS SUCTAMUS MAXIMUS.

 

New drivers seem to work just fine on my G751JY but i don't really bother with overclocking and benchmarking. Was hoping they would turn on mobile G-Sync when combined with the new bios for my notebook model but nothing. Still hoping it will show up soon though.

 

I am able to overclock both memory and core with these drivers.

 

Not here. Only one stays.

 

Yeah, your GPU is probably from the same production run as mine so it shipped with a locked GPU clock.

 

*sigh*
350.12 and 352.86 = allow OCing, but do NOT ignore the OC block bit that nVidia demanded new vBIOSes be released with.
347.88 = allow OCing *AND* ignore "clockblock bit" in vBIOS.
Unlocked vBIOS = OC with 347.88, 350.12 and 352.86.
Clockblocked vBIOS = OC with 347.88; blocked with 350.12 and 352.86

 

Nvidia's drivers managed to mess with X-Plane as well: NVIDIA: 4 Ben: 1.

 

Nvidia just pushed out a hotfix to address the failure to remove the clock-block in 352.86 notebook driver.
https://forums.geforce.com/default/topic/834157/geforce-drivers/announcing-hotfix-driver-353-00/

 

NVIDIA's OLD Logo

 

Thank you. I am going to download and test it now with 780M. Hopefully, the crazy 400MHz throttling issue I have with 350.12 and 352.86 is gone now.

 

downloading it now!

 

I'm sitting here waiting for some test results

 

Sorry, I fell asleep, LOL.

I just downloaded and install. Will test in a few minutes.

Guess what? It has DSR... will test that as well and report findings.



New thread created:
NVIDIA GeForce Driver 353.00 Discussion Thread

Edit: NOPE! Still a crap driver for my 780M SLI setup. Massive throttling and horrible FPS with moderate overclock. Didn't bother testing stock since stock is unimportant if overclocking is throttled. Don't care if DSR works if it throttles. Testing with 980M SLI on my Clevo P570WM sometime tomorrow.


Edit: FOR THE RECORD, I just checked and it does not throttle at stock clocks. So, behavior is same at 350.12 and 352.86 for me and throttles like a son of a female canine overclocked. Worthless driver. Don't care about DSR if I cannot overclock, so I'm not even going to test to see if DSR works with stock clocks. Going back to my 345.20 DT CUDA driver mod... again.

 

Ooh heck, looks like I've missed a lot of craziness this past week. Well, I've got the new driver installed, so I'll have a little play around.

 

DSR is nice... Really puts the extra horsepower of my 980Ms to work. Other than that, it's not a good driver if you overclock and benchmarks suffer a performance hit. Kind of amusingly, the AMD optimized Bioshock Infinite gained 23FPS average for me while the nVidia optimized FFXIV actually took a hit.