Support NeoGAF

[Buggy Loop]

https://www.nvidia.com/en-us/geforc...0-game-ready-driver/?=&linkId=100000155345874

In the last year, Game Ready Drivers have given GeForce gamers optimum performance in over 80 new releases, and have continued to optimize and improve the games you’re still playing. Most recently, our engineering team has designed new DirectX 12 enhancements that accelerate performance by double digits in a wide array of games.

If you’re using a previous-generation GeForce RTX graphics card or laptop, you’ll also want to install ASAP to benefit from new DirectX 12 optimizations that can improve performance by double digits in popular titles.

These improvements include shader compilation optimization, reduced CPU overhead, and Resizable BAR profiles for Forza Horizon 5 and F1Ⓡ 22. Many of these optimizations are most beneficial in CPU-bound scenarios, enabling your graphics card to be utilized to a greater degree. In Assassin's Creed Valhalla, Cyberpunk 2077, and Forza Horizon 5, performance improved by up to 24%:
Our DirectX 12 optimizations apply to GeForce RTX graphics cards and laptops, though improvements will vary based on your specific system setup, and the game settings used. In our testing, performance increases were found in a wide variety of DirectX 12 games, across all resolutions:

• Assassin’s Creed Valhalla: up to 24% (1080p)
• Battlefield 2042: up to 7% (1080p)
• Borderlands 3: Up to 8% (1080p)
• Call of Duty: Vanguard: up to 12% (4K)
• Control: up to 6% (4K)
• Cyberpunk 2077: up to 20% (1080p)
• F1Ⓡ 22: up to 17% (4K)
• Far Cry 6: up to 5% (1440p)
• Forza Horizon 5: up to 8% (1080P)
• Horizon Zero Dawn: Complete Edition:up to 8% (4k)
• Red Dead Redemption 2: up to 7% (1080p)
• Shadow of the Tomb Raider: up to 5% (1080p)
• Tom Clancy’s The Division 2: up to 5% (1080p)
• Watch Dogs: Legion: up to 9% (1440p)

Holy shit, atomic bombs dropping. This basically fucks up every benchmarks taken in past days for 4090, but also including any 3000 or 2000 series comparison.

Also curious if the shader optimization will fix the dx12 typical stuttering?? Would be fucking awesome.

 

[Fahdis]

How much power gain is this in terms of Gamecubes ductaped together?

 

[Fess]

Lmao and they release this now just after they’ve just released the 4090. Could make lots of people decide to not upgrade. Trying to ruin the market for scalpers?
Anyhow, will install asap

 

[adamosmaki]

Sounds just about every driver release after nvidia and Amd launch new gpus

 

[Fess]

Sounds just about every driver release after nvidia and Amd launch new gpus

What about the graphs on reviews though, can we still look at them and expect that performance boost if we upgrade?
Edit: hmm it boost the 4090 too and CPU bound games

 

[kiphalfton]

Convenient they didn't do this before the release of the RTX 4090. All those numbers comparing previous gen to current gen are now useless, as the numbers will be skewed.

 

[FireFly]

 

[sertopico]

Convenient they didn't do this before the release of the RTX 4090. All those numbers comparing previous gen to current gen are now useless, as the numbers will be skewed.

Of course. Now they can keep selling all those 3xxx series leftovers.

 

[Fess]

Convenient they didn't do this before the release of the RTX 4090. All those numbers comparing previous gen to current gen are now useless, as the numbers will be skewed.

My initial thought too.
But it sounds like it affect the 4090 too and if it’s CPU bound games the graphs for lower res could get a healthy boost and making the card relevant even if you still have a sub-4K screen.

 

[THE DUCK]

I now a few are 4k, but a whole lot of them are old games at 1080p at less than 10%. Doesn't seem this driver lives up to the headline of this post. "HUGE boost!!!!!" LOL.

 

[HerjansEagleFeeder]

Holy shit, atomic bombs dropping. This basically fucks up every benchmarks taken in past days for 4090, but also including any 3000 or 2000 series comparison.

Why? Didn't reviewers have access to these drivers when they compared the 30 series cards to the 90?

 

[MAX PAYMENT]

Where does Nvidia find this secret sauce? Did my GPU have this sauce in it the whole time? I thought secret sauce was console exclusive.

 

[GymWolf]

I heard this is only useful at 1080p and only for 3000 series.

 

[Buggy Loop]

Why? Didn't reviewers have access to these drivers when they compared the 30 series cards to the 90?

Maybe for the 4090, but not the previous cards.

They don’t get all cards on a bench every time, they have a bench setup and a benchmark suite and older cards might be weeks or months benched before the new product. Unless it’s a very small sample of cards.

Like techspot for example where they have over 20 cards and over 20 games, no they don’t bench that within a few days because an 4090 review.

If they change the bench setup like a newer CPU then yea, a poor bastard is working hard to re-bench everything.

 

[Hugare]

Up to 20% in Cyberpunk

This is nuts. Cant remember a driver update improving FPS like this before.

Damn you, Nvidia. Hate you mfckers for your prices, but you always deliver.

 

[Hugare]

I heard this is only useful at 1080p and only for 3000 series.

Not true.

Look:



People with 2060's and etc. reporting FPS gains. Also people playing at 1440p/4K.

 

[Buggy Loop]

I heard this is only useful at 1080p and only for 3000 series.

No, peoples are benching the Turing cards with higher performances. Some have gains at 4K, 1440p too. It’s really dependent on your setup and game.

A few examples from reddit

EDIT: I decided to test it myself here’s what I found

Cyberpunk AVG fps : 86 -> 95 (+10.5%)
Cyberpunk min fps : 36 -> 46 (+28%)
1440p high + dlss Q

Forza Horizon 5 AVG fps : 84 -> 92 (+9.5%)
Forza Horizon 5 min fps : 77 -> 82 (+6.5%)
1440p ultra

This is using a 5800X paired with a 3060ti on Windows 10

It’s only 2 games but pretty nice overall!

Here’s some before and after in-game benchmarks I ran at 4K on my 5900X/3080Ti (DLSS on where available). Before driver was 517.48

Cyberpunk: Avg 62.09 -> 75.92 = +22%

Horizon Zero Dawn: Avg 110 -> 139 = +26%

Deus Ex Mankind Divided: Avg 80.2 -> 92.2 = +15%

Metro Exodus Enhanced: Avg 73.88 -> 79.25 = +7%

Shadow of the Tomb Raider: Avg 87 -> 102 = +17%

Red Dead Redemption 2: Avg 79.9 -> 94.4 = +18%

Forza Horizon 5: Avg 83 -> 99 = +19%

Watch Dogs Legion: Avg 67 -> 79 = +18%

Fast Turing benchmark on RTX 2080ti two game Control, Cp 2077 and there is improvement.

• Cyberpunk 2077 517.48 : min/max - 64fps-79fps,
• Cyberpunk 2077 522.25 : min/max - 72fps-96fps

In Control on same save point fps goes from 92fps to 95.

I know it’s a small number to test, but you can see the jump and it’s not negligible, especially in CP 2077, it can’t be accidental

So yes it working on Turing after all no worries.

Some guy is working on publishing pascal cards benchmarks some time today. By end of day we should have a good picture for pascal, Turing, ampere

Nvidia does mention that ampere cards benefited the most from this optimization.

 

[GymWolf]

Thanks for the heads up guys, i'm gonna download the driver when i return from work.

 

[Mr Moose]

Just installed it, going to try the one game I have installed

Nothing changed for that game.

 

[V1LÆM]

i don't play any of those games. i plan on replaying cyberpunk but not until the expansion is out. only games i have installed are Overwatch 2, Rocket League, Fortnite, and Genshin Impact . I don't exactly struggle with them.

also i just updated my drivers last week. i don't upgrade unless there is a new game out i want to play. only reason i upgraded was because of Overwatch 2. i can't be arsed updating my drivers again. maybe if we get drivers this month for Persona 5 Royal then I'll upgrade but we'll see...

edit: it's only for 3000 cards? well fuck me i guess for having a peasant 2000 card

 

[Kilau]

I think Flight Sim mentioned this driver in the last update. That’s the one game I play that is punishing on my system but even then it’s not pushing my cpu much.

 

[crackajack]

If nVidia can offer that much, where people say their drivers are great, better than competition, how much potential lies in the intel cards if they get their shit together?
I guess this gen step is a bit different in many ways, isntead of solely focusing on their new toys nVidia seems to have big interest to position their unsold 30 series as the low midrange cards with still good value until we get the actual 4070/4060/4050.

 

[Filben]

Also curious if the shader optimization will fix the dx12 typical stuttering?? Would be fucking awesome.

That would be actually the holy grail because it'll safe PC gaming for me. So many titles I didn't get for PC because of DX12 incompetence of devs. I'm sceptical that a driver alone will fix this for every game, but I hope it fixes at least some games.

 

[Sentenza]

Someone should test this with Total War Warhammer.

 

[buenoblue]

Up to 20% in Cyberpunk

This is nuts. Cant remember a driver update improving FPS like this before.

Damn you, Nvidia. Hate you mfckers for your prices, but you always deliver.

I remember when I had sli 780 ti's back in 2014/15 there was a driver that boosted games like 20%

 

[01011001]

20% in Cyberpunk? Imma need to test that in CPU bound scenes

 

[SmokedMeat]

Why sure, I’ll take some free performance upgrades.

 

[01011001]

20% in Cyberpunk? Imma need to test that in CPU bound scenes

just launched it and it seems to indeed run smoother in the intersection that usually is CPU bound on my PC when RT is enabled

my CPU utilization also goes way higher it seems, when it was usually around 80% at max before the patch, now it's actually sometimes hitting 90+% even 100%, which tells me that my CPU is used way more effectively now

 

[Hugare]

Someone should test this with Total War Warhammer.

Saw someone on Reddit that tested it

6% improvement, I think

Biggest upgrades will be in CPU bound games

 

[Black_Stride]

Holy shit, atomic bombs dropping. This basically fucks up every benchmarks taken in past days for 4090, but also including any 3000 or 2000 series comparison.

Also curious if the shader optimization will fix the dx12 typical stuttering?? Would be fucking awesome.

This IS the driver that reviewers were using to review the 4090.
Press gets driver versions earlier than general public.
Between Press -> Retail driver there is never any real difference.

Maybe for the 4090, but not the previous cards.

They don’t get all cards on a bench every time, they have a bench setup and a benchmark suite and older cards might be weeks or months benched before the new product. Unless it’s a very small sample of cards.

Like techspot for example where they have over 20 cards and over 20 games, no they don’t bench that within a few days because an 4090 review.

If they change the bench setup like a newer CPU then yea, a poor bastard is working hard to re-bench everything.

Thats on the reviewer.
Because, literally every driver drop has improvements, thats why pick and chose your "trusted" reviewers carefully.
Some will rebench all their cards with the latest driver versions to make it fair, some wont.
If Techspot dont rebench their cards, then thats a disservice to pretty much everyone and they really shouldnt include cards on old build in their charts as it will benefit newer cards more than older cards.
So even an RTX3050 would end up looking better because its being tested with the latest driver build vs maybe a 3060Ti that would have been tested with launch drivers.

If its your job to do these reviews then do them right.
Its why the likes of techpowerup and others will sometimes only include a few cards on the review because they only tested those cards vs whatever they are review.
Even DF will rebench cards when reviewing a new card....thats the only way it makes sense.
Listing cards that might be out of spec Gamebuild/ Driver build/ OS build, isnt very professional or accurate.

 

[LiquidMetal14]

I'll install the driver tomorrow when I get the 4090 but that's great to see.

I love when optimization continues to happen and we get more out of our HW.

 

[BennyBlanco]

wow, good shit

 

[Buggy Loop]

This IS the driver that reviewers were using to review the 4090.
Press gets driver versions earlier than general public.
Between Press -> Retail driver there is never any real difference.


Thats on the reviewer.
Because, literally every driver drop has improvements, thats why pick and chose your "trusted" reviewers carefully.
Some will rebench all their cards with the latest driver versions to make it fair, some wont.
If Techspot dont rebench their cards, then thats a disservice to pretty much everyone and they really shouldnt include cards on old build in their charts as it will benefit newer cards more than older cards.
So even an RTX3050 would end up looking better because its being tested with the latest driver build vs maybe a 3060Ti that would have been tested with launch drivers.

If its your job to do these reviews then do them right.
Its why the likes of techpowerup and others will sometimes only include a few cards on the review because they only tested those cards vs whatever they are review.
Even DF will rebench cards when reviewing a new card....thats the only way it makes sense.
Listing cards that might be out of spec Gamebuild/ Driver build/ OS build, isnt very professional or accurate.

I understand, but do you recall a single reviewer mentioning the gains on ampere cards with the driver update? It’s not something you can just neglect, especially when you still have AMD thrown into the mix, it should be a clear step up for 3000 series in games where it used to be strongholds for AMD like AC Valhalla.

Techpowerup didn’t even have the 522.25 drivers for the 4090 and the other Nvidia are on an even older build. Confirming a bit what I was saying.

Drivers:	RTX 4090: 521.90 Press Driver NVIDIA: 516.94 WHQL AMD: 22.8.1 Beta Intel Arc: 101.3433 Press Driver​

 

[Black_Stride]

I understand, but do you recall a single reviewer mentioning the gains on ampere cards with the driver update? It’s not something you can just neglect, especially when you still have AMD thrown into the mix, it should be a clear step up for 3000 series in games where it used to be strongholds for AMD like AC Valhalla.

Techpowerup didn’t even have the 522.25 drivers for the 4090 and the other Nvidia are on an even older build. Confirming a bit what I was saying.

Drivers:	RTX 4090: 521.90 Press Driver NVIDIA: 516.94 WHQL AMD: 22.8.1 Beta Intel Arc: 101.3433 Press Driver​

I dont recall a reviewer mentioning driver update optimizations ever.
This isnt exactly the first time drivers have improved performance.
Testing the 4090 at 1080p will lead to nothing but CPU bottlenecks or engine limits.
At 4K the 3090s were gaining what 10% at best?

Its not exactly going to invalidate their findings if they were on the latest WHQL driver (Im assuming they were testing 4090s in September).
Id be somewhat shocked if when rebenching RTX30s they didnt use the Press Driver, unless they do clean installs for every bench, imma guess they tested with the press driver that has all the features of 522.25.

 

[kiphalfton]

Where does Nvidia find this secret sauce? Did my GPU have this sauce in it the whole time? I thought secret sauce was console exclusive.

Maybe they were limiting the cards to begin with. Not like it hasn't happened before.

 

[Kuranghi]

This is great but hHnnngggggg not sure whether to buy a 4080 after those benchmarks showed it not being able to hit 60fps at native 4K with RT in Plague Tale Requiem. Probably best to save some more money into next year and get a 4080 Ti, the 4090 is lovely but its too much to spend on a GPU... I think anyway, maybe I should just buy that. I DON'T KNOW.

 

[rofif]

HUGE BOOST
up to 5% in control !

 

[Chiggs]

This is a dagger in AMD's side.

Radeon held the edge with CPU-limited games. Well played, Nvidia!

 

[GustavoLT]

it´s just a 3090 with a different driver!

 

[poodaddy]

Went from average FPS of 43 at max settings 1440p with DLSS performance on Cyberpunk, all settings maxed including all ray tracing, to an average FPS of 51. With G Sync, that's a damn smooth frame rate. This is with my Gigabyte RTX 2080 btw, not a 2080 Super, just a regular ole 2080. Pretty damn good boost man.

 

[Fahdis]

GeForce Now will also be getting this not bad nVidia. I still hate you for extortion prices.

 

[yet_another_alt]

Getting 55+ fps on CP2077 at 1440p with all settings maxed and DLSS Performance with a 3070ti and 70+ at Ultra Performance. Pretty impressive. Well, except for the part where it glitches up the map. Then again this is like my 5th playthrough so I don't really need the map anymore.

 

[sertopico]

just launched it and it seems to indeed run smoother in the intersection that usually is CPU bound on my PC when RT is enabled

my CPU utilization also goes way higher it seems, when it was usually around 80% at max before the patch, now it's actually sometimes hitting 90+% even 100%, which tells me that my CPU is used way more effectively now

My 3080 runs also much hotter normally I get about 65 degrees C at this time of the year, now it's 70-71.

 

[Alexios]

How's it for GTX cards? Is this the point we should start avoiding updates unless a game we really want to play explicitly requires them because they're bound to start making performance worse for older models as usual or it's still alright?

 

[alucard0712_rus]

20% in Cyberpunk? Imma need to test that in CPU bound scenes

Basically new drivers are better in exactly CPU bound scenarios. I have zero difference in DLSS Quality at 1440p, but in performance mode or 1080p I definitely see a fps boost. 12600k, 3070.

 

[MidGenRefresh]

Played Cyberpunk just now. 3080 at 4k with DLSS set to Performance. Game is running much faster. ~10FPS more, easily.

 

[Hawking Radiation]

Lol how am I getting now between 50-60 fps in Cyberpunk (with a crap ton of mods) and everything dialed up to 11
This is crazy

I used to be hovering inbetween 35-40 frames before this latest geforce update.

Fuck I need to make my way down to Jig-Jig Street

EDIT: 45 -55fps and sometimes reaching 60fps running around Jig-Jig Street with everything at max

 

[BadBurger]

Gonna update and try out Cyberpunk at 4K again later. Nice.

 

[TheUsual]

Getting 55+ fps on CP2077 at 1440p with all settings maxed and DLSS Performance with a 3070ti and 70+ at Ultra Performance. Pretty impressive. Well, except for the part where it glitches up the map. Then again this is like my 5th playthrough so I don't really need the map anymore.

Glad I wasn't the only who's map is fucked up as well. Started a 2nd playthrough not too long ago and it's been about a year since I last played so I don't remember much of anything. I rolled back to my previous driver and the map is functional again.

So bizarre everything was fine except for the map.

 

[yet_another_alt]

Glad I wasn't the only who's map is fucked up as well. Started a 2nd playthrough not too long ago and it's been about a year since I last played so I don't remember much of anything. I rolled back to my previous driver and the map is functional again.

So bizarre everything was fine except for the map.

I imagine the game will be patched soon for that new DLSS stuff in the 4xxx cards. They'll probably fix the map issue for us plebs then.