NVidia driver 337.50 beta released!

[Tristor]

Finally, actual performance gains from SLI on Total War: Rome II. I've been watching and waiting for ages. Unfortunately every Creative Assembly release has been like this. You have to wait roughly a year after it comes out before it's actually playable. Great games, great series, terrible engine.

Now I can actually play Rome II. I already bought all the DLC and have had the game installed untouched other than benchmarking after every driver and game update. It's time to do it. I'll see you guys in 6 months ;P

 

[krumme]

Finally, actual performance gains from SLI on Total War: Rome II. I've been watching and waiting for ages. Unfortunately every Creative Assembly release has been like this. You have to wait roughly a year after it comes out before it's actually playable. Great games, great series, terrible engine.

Now I can actually play Rome II. I already bought all the DLC and have had the game installed untouched other than benchmarking after every driver and game update. It's time to do it. I'll see you guys in 6 months ;P

With future lower level api i think we can expect the performance to be implemented faster because there is no need to correct the drivers to the same degree because its solely in the hands of the devs. ..That is if the devs do not use the advantage to release the games yet earlier? Lol

 

[PinchedNerve]

Pinchednerve, what are the specs of YOUR machine. I saw the one of your child. I know you have a GTX780 but what cpu etc and is your gpu OC'd?

Sorry, all the other sites I post to allow me to have a Sig. I'm guessing I've not been here long enough, or posted enough to earn a sig here.

Desktop - Asus P8Z77-V Pro | Intel Core i7 3770k | Thermalright True Spirit 120M HSF | Gigabyte GTX 780 GHz Ed | OCZ Vertex 4 256 GB | 16 Gigs Corsair Vengeance 1866 | Corsair AX1200i | Thermaltake Chaser MK-I | Denon AVR-1913 - 2 sets of - JBL ES20 - 1 Polk Audio CS10, & 1 BIC America F12 | Asus PB278Q | Win 8.1 Pro with Media Center.


My i7 turbo's to 4.2 GHz. Sleeping Dogs is the only game I currently have installed that has benchmarking ability. I don't really want to go through the hassle of installing more games or programs just to test the driver when I have multiple machines in the house to do it on.

Edit: Was just informed I had to post at least 25 posts before Sigs were available. Since I've past that I'll update my account.

 

[guskline]

Thanks PinchedNerve. You have a solid rig.

 

[PinchedNerve]

Thanks guskline!

 

[mindbomb]

It's kinda funny that they had to set up their bf4 bench to be extremely gpu limited to make sure the 780ti was faster than the 290x, but doing so made the gains from the new driver negligible. And the new driver was the whole point of conducting the bench in the first place. It's actually a 1.7% increase for the new nvidia driver, and a 6% increase from mantle going by the graphs.

I think we can gain valuable insight into their motives with that graph, as it shows that they care more about quelling mantle hype than actually improving game performance.

 

[bystander36]

It's kinda funny that they had to set up their bf4 bench to be extremely gpu limited to make sure the 780ti was faster than the 290x, but doing so made the gains from the new driver negligible. And the new driver was the whole point of conducting the bench in the first place. It's actually a 1.7% increase for the new nvidia driver, and a 6% increase from mantle going by the graphs.

I think we can gain valuable insight into their motives with that graph, as it shows that they care more about quelling mantle hype than actually improving game performance.

That goes without saying for any company who wants to make money, but to be fair, that is the FPS range most people play at. If not lower.

 

[Atreidin]

My guess is Nvidia figures the people this PR stunt was aimed at will care more about and pay more attention to the message it delivers than the actual results. Fact checking is annoying, and fact checking when dealing with rabid fanboys is just exhausting.

 

[Gloomy]

My friends are reporting decent gains in BF4 MP. It'll be a while before I can confirm them myself though, I can't just spawn 500 dollar video cards at will.

I wish I could though. :|

Maybe PCGamesHardware will run reliable MP tests again...

http://www.pcgameshardware.de/Nvidi...-sGeforce-Treiber-33750-gegen-Mantle-1116527/

Whoops, no they didn't. I have a feeling this is going to be like Mantle all over again, and nobody is going to have reliable benches, and we'll have people say the gains are garbage, when they're actually pretty good. Figures. :|

 

[Keysplayr]

Well ofcource not. - its not an eye laser operation. Did you really miss my point?

Its just pr nonsense. Writing you are focused on one thing (dx11) and in the same sentence writing you are focused on another (dx12). When you do that you actually write you are focusing on two different things. Dx11 and dx12/mantle is as we know very different. How is that sharply focused?

When amd introduced the 14 beta driver with framepacing, mantle, dx and ggpu capabilities were they then working titanium, uber, laser focused? No it was excactly oposite. They were working on different technologies. As nv is with both higher and lower level api. So when they say laser focused its just typical nv pr because they are actually doing the exact oposite, - for good reasons.

It's obviously whatever you want it to be, Krumme. So, there you go.

 

[toyota]

Well it does look like mostly hype. I am not even going to bother getting these since the next driver is supposed to also adds similar improvements so i will just wait on it.

 

[sxr7171]

It's kinda funny that they had to set up their bf4 bench to be extremely gpu limited to make sure the 780ti was faster than the 290x, but doing so made the gains from the new driver negligible. And the new driver was the whole point of conducting the bench in the first place. It's actually a 1.7% increase for the new nvidia driver, and a 6% increase from mantle going by the graphs.

I think we can gain valuable insight into their motives with that graph, as it shows that they care more about quelling mantle hype than actually improving game performance.

They could have gone more GPU intensive by using 1440p or 1600p, but then they'd lose to AMD at those resolutions.

The way I see it, it's made some games playable at higher settings for me. Some games with less dips under 60fps. And it all came for free so I'm happy. The driver definitely needs work, especially in the stability department.

The only disappointment for me is that it doesn't really help AC4 even one bit as I noticed. That game still dips like crazy and 4X TXAA is still a no-go as it must run at or above 60fps consistently to avoid that crazy screen tearing. Even with 2x TXAA there is occasional screen tearing.

This is all we have until next gen hardware I guess.

 

[guskline]

I try to keep my drivers updated.

 

[Grooveriding]

Yeah, might as well update to these as they work well and include some SLI profile updates. If you play Rome and run SLI the drivers are a must because SLI never worked in the game until now.

The whole purported significant performance increases were total hype though. There is nothing special here. PR and hype to try and answer Mantle. The single card gains in the update notes appear to be outright fabrications as well....

 

[blackened23]

I'm trying to recall at which point and time that NV specifically stated that this driver was supposed to precisely create the same increases that Mantle can in certain scenarios. Oh, wait. They didn't.

Much like everything else, most of the hype was fan based and website generated. Nvidia did their slides to demonstrate increases in key games, at no point in their slides did they state that their driver would do everything and anything Mantle does. It increases CPU efficiency and does just that, because DX11 has always had MT optimizations - those optimizations are incredibly difficult to implement from what I've read. This is partially the reason why AMD has not implemented it at all, but i'm not saying that in a negative connotation. Simply, it has taken NV years as a work in progress to get their MT drivers better and better and AMD still has not implemented MT (to my knowledge, I could be wrong, but this was true in 2012-2013). Anyway back to the driver. The driver does work with more than 1-2 games surprisingly enough. If it were Mantle, it would be one game and one game alone with an increase. With that stated, there are sizable increases in numerous games here. I'll take it. It's free, and it's driver, and it increases performance. Alrighty, sounds good to me.

No, it isn't mantle and nvidia never said that "this driver will over-take Mantle and make Mantle obsolete". These were fan based musings. Anyway, sure, you can Mantle does more in some respects. But that's true if you play one game. Looking over this driver, I see increases in multiple games. Imagine that. Common API. Multiple games. Many games. Not one game. Mind blown.

While some of the increases aren't incredible, some of them are fairly high. Some games are showing pretty sizable increases. I'll take it, like I said, sounds good to me. But if some fans of the other side want to poo poo on that fact, I don't quite get it. The 7970 had a pretty awful driver situation at launch and later released their performance driver for multiple games. That happened something like..6 months after? If I recall. It's just free performance, even if it doesn't increase every game by 20%. Clearly anyone expecting 20-30% jumps in every game had improper expectations. I do look forward to seeing what DX12 can do, hopefully we will get some beta form near the end of this year. In the meantime, AMD will have plenty of time to do their thing with Mantle. Which is fine by me.

 

[Tristor]

With future lower level api i think we can expect the performance to be implemented faster because there is no need to correct the drivers to the same degree because its solely in the hands of the devs. ..That is if the devs do not use the advantage to release the games yet earlier? Lol

We'll see, but first they'd actually have to write or adopt a new engine for the next game instead of rehashing things again with Warscape. Warscape was a decent engine 10 years ago when the original Rome: Total War came out. But these days it's a pile. Desperately needs to be scrapped in favor of something which is native 64-bit, heavily optimized for SMP/MT, and is built with multi-GPU scaling in mind. Their games look incredible, but they're heavily CPU limited and the GPU scaling is almost non-existent in many titles. I'm hoping a DX12 native engine might be on the horizon for the next entrant into the Total War series (or a highly optimized implementation in OpenGL would be even better).

 

[PinchedNerve]

I'm trying to recall at which point and time that NV specifically stated that this driver was supposed to precisely create the same increases that Mantle can in certain scenarios. Oh, wait. They didn't.

Much like everything else, most of the hype was fan based and website generated. Nvidia did their slides to demonstrate increases in key games, at no point in their slides did they state that their driver would do everything and anything Mantle does. It increases CPU efficiency and does just that, because DX11 has always had MT optimizations - those optimizations are incredibly difficult to implement from what I've read. This is partially the reason why AMD has not implemented it at all, but i'm not saying that in a negative connotation. Simply, it has taken NV years as a work in progress to get their MT drivers better and better and AMD still has not implemented MT (to my knowledge, I could be wrong, but this was true in 2012-2013). Anyway back to the driver. The driver does work with more than 1-2 games surprisingly enough. If it were Mantle, it would be one game and one game alone with an increase. With that stated, there are sizable increases in numerous games here. I'll take it. It's free, and it's driver, and it increases performance. Alrighty, sounds good to me.

No, it isn't mantle and nvidia never said that "this driver will over-take Mantle and make Mantle obsolete". These were fan based musings. Anyway, sure, you can Mantle does more in some respects. But that's true if you play one game. Looking over this driver, I see increases in multiple games. Imagine that. Common API. Multiple games. Many games. Not one game. Mind blown.

While some of the increases aren't incredible, some of them are fairly high. Some games are showing pretty sizable increases. I'll take it, like I said, sounds good to me. But if some fans of the other side want to poo poo on that fact, I don't quite get it. The 7970 had a pretty awful driver situation at launch and later released their performance driver for multiple games. That happened something like..6 months after? If I recall. It's just free performance, even if it doesn't increase every game by 20%. Clearly anyone expecting 20-30% jumps in every game had improper expectations. I do look forward to seeing what DX12 can do, hopefully we will get some beta form near the end of this year. In the meantime, AMD will have plenty of time to do their thing with Mantle. Which is fine by me.

:thumbsup:

 

[Despoiler]

I'm trying to recall at which point and time that NV specifically stated that this driver was supposed to precisely create the same increases that Mantle can in certain scenarios. Oh, wait. They didn't.

Much like everything else, most of the hype was fan based and website generated. Nvidia did their slides to demonstrate increases in key games, at no point in their slides did they state that their driver would do everything and anything Mantle does. It increases CPU efficiency and does just that, because DX11 has always had MT optimizations - those optimizations are incredibly difficult to implement from what I've read. This is partially the reason why AMD has not implemented it at all, but i'm not saying that in a negative connotation. Simply, it has taken NV years as a work in progress to get their MT drivers better and better and AMD still has not implemented MT (to my knowledge, I could be wrong, but this was true in 2012-2013). Anyway back to the driver. The driver does work with more than 1-2 games surprisingly enough. If it were Mantle, it would be one game and one game alone with an increase. With that stated, there are sizable increases in numerous games here. I'll take it. It's free, and it's driver, and it increases performance. Alrighty, sounds good to me.

No, it isn't mantle and nvidia never said that "this driver will over-take Mantle and make Mantle obsolete". These were fan based musings. Anyway, sure, you can Mantle does more in some respects. But that's true if you play one game. Looking over this driver, I see increases in multiple games. Imagine that. Common API. Multiple games. Many games. Not one game. Mind blown.

While some of the increases aren't incredible, some of them are fairly high. Some games are showing pretty sizable increases. I'll take it, like I said, sounds good to me. But if some fans of the other side want to poo poo on that fact, I don't quite get it. The 7970 had a pretty awful driver situation at launch and later released their performance driver for multiple games. That happened something like..6 months after? If I recall. Did that driver get poo poo'ed on? It's just free performance, even if it doesn't increase every game by 20%. Clearly anyone expecting 20-30% jumps in every game had improper expectations.

Classic straw man argument. NV said they could match or beat Mantle's performance. They claimed that they are improving efficiency to compete with Mantle. What they delivered only actually delivers if you ignore minimum frame rate and don't measure frame time. The tests from users and review sites that include those items shows these drivers aren't doing anything in terms of CPU bound efficiency on the level of Mantle. Like I said before, we are well beyond average and max fps being the sole measure of performance and yet here we are.

http://www.tomshardware.com/news/nvidia-driver-update-direct3d-optimization,26381.html

http://www.guru3d.com/news_story/nvidias_tba_dx11_will_be_better_than_mantle.html

 

[Gloomy]

Actually Nvidia said:

The performance data above, supplied to us by Nvidia, shows performance improvements over successive GeForce driver releases in Oxide Game's Star Swarm stress test. That test also supports Mantle, which helps put Nvidia's D3D11 optimizations in context. Tamasi conceded AMD's Mantle version "still has less slow frames" and that D3D11 "still [has] some limiting factors," but he reiterated his overarching point, which is that it's possible to "do a much better job" with D3D11.

http://techreport.com/review/26239/a-closer-look-at-directx-12/3

Some sites embellished this, because they are terrible clickbait garbage. Whatever. As you can see, they very clearly state Mantle is still more efficient, but that there's still room to optimize DX11 and they will do so.

 

[PinchedNerve]

I remember years ago (early 2000s I think) when Nvidia got caught "optimizing" its drivers for some benchmarks because its current line of cards were poo. I think it was the FX line? End result was that they were cheating and not rendering "everything" on the screen correctly.

Correct me if I'm wrong here. AMD had serious issues with frame pacing. Ignores it till it gets investigated by a few sites, then admits to it and has to rewrite drivers to fix it which took months to do and many driver revisions.

AMD launches new hardware shortly after, and some 1337 new "API" that only works on their new hardware and at its start 1 whole game..... Seems to me, this could be considered cheating and somehow isn't. Masterful slight of had work by AMD I have to say. Look here, not there!

 

[Despoiler]

Actually Nvidia said:

http://techreport.com/review/26239/a-closer-look-at-directx-12/3

Some sites embellished this, because they are terrible clickbait garbage. Whatever. As you can see, they very clearly state Mantle is still more efficient, but that there's still room to optimize DX11 and they will do so.

Sure, but isn't that what driver updates do anyways? Take what you've just provided and pretend it's a new AMD driver. What do you normally see? Performance increases in various DX games and benchmarks. Nvidia pitted this entire driver release against Mantle to try to take the wind out of Mantle's sails so to speak. Look at many of their marketing benchmark slides. It's this driver vs Mantle specifically. Look at most of the review benchmarks including their own. What do they show you? Average and max fps. What don't they show you? Minimum frame rate and frame time graphs? Why? If they did they wouldn't be able to successfully claim they can compete with Mantle. Mantle's primary purpose in being developed was to vastly improve on CPU bound situations. So it logically follows that if you want to compete with Mantle you need to show improvements in those two areas. However, we are being shown is higher max fps that in turn gives higher average FPS and ignoring the actual point of Mantle.

 

[SirPauly]

Well ofcource not. - its not an eye laser operation. Did you really miss my point?

Its just pr nonsense. Writing you are focused on one thing (dx11) and in the same sentence writing you are focused on another (dx12). When you do that you actually write you are focusing on two different things. .

Imho,

One is focused on DirectX efficiency! And clearly, nVidia is!

 

[f1sherman]

Classic straw man argument. NV said they could match or beat Mantle's performance. They claimed that they are improving efficiency to compete with Mantle.

well... with the single driver they nullified Mantle advantage in 2 out of 3 Mantle games - Star Swarm and Thief.
Not to mention whole other slew of improvements.
Folks from Oxide must be feeling real silly now - making a demo and showcasing entire new API when competition does all that through a freaking driver.
Or perhaps they didn't implement it all that well?

I don't want sound harsh when it comes to AMD, because they are the ones that got things rolling.
And when properly implemented Mantle will surely beat any NV DX specific optimization(like it does in BF4).

But it does feel like AMD is great for everyone except for themselves.
And it makes you wonder if they chose a good path; 3-4 years of R&D for a new API when competition does almost as good with mere driver release. While improving their DX path along the way.
IMHO they need new games featuring their API ASAP. + even stronger dedication to DX, else DX gap with NV will widen.

Anyway this has to be the most exciting period in recent GPU history. Competition is that good :thumbsup:

 

[Deders]

I went for a comparison in the game nvidia claimed the largest performance increase in :

Installed the latest WHQL drivers first for my first run and then used these new betas.

I got nothing performance wise. Nothing, zip, zilcho... no performance gains... Well .3 fps increase...

335.23 Driver

337.50 Driver

Nothing. Maybe the claimed performance gains come with the usual caveat that you have to run them at a specific resolution, with certain settings disabled (mine are all maxed), in a specific part of the game and while standing on one leg.

Would of expected to at least see some improvement with such huge gains claimed.

I don't think the benchmark is representative of gameplay. I'm definitely getting an improvement in large battles, but it still goes to 15fps if I zoom in on 9 testudo's all bunched up.

 

[Pandora's Box]

I don't think the benchmark is representative of gameplay. I'm definitely getting an improvement in large battles, but it still goes to 15fps if I zoom in on 9 testudo's all bunched up.

You are cpu limited big time.