Rome 2: The way it's meant to be played (tm) - for NVIDIA owners: performance boost and finally SLI support (updated 22.04.2014)

[alQamar]

Hello TWC members,

many of you should know me being technically interested and releasing reports about Rome 2 performance.

I made a remark on my thread, where we discuss about a new Rome 2 engine and I said that DirectX 12 could completely resolve the performance issues for many games, which use the GPUs very extensively while also creating a high amount of DirectX overhead (additional amount of data to translate Rome 2 source code from CPU to GPU using DX9-DX11). Rome 2 suffers very much of this problem as it works with a lot polygons and textures and additionally to the CPU overhead the DirectX is causing, it causes a CPU bottleneck on one core.

What does this "CPU bottleneck" / overhead problem mean?:
Rome 2 can utilize up to 8 cores (or more) but the utilization cannot be even and is not balanced, so if one core is busy he won't swap to other cores which has less to do. The first CPU core (Core 0) seems to be responsible for AI and graphics. This led many to assume Rome 2 is not multicore capable, and just runs on one core, which is technically wrong.

However the first of all available cores is loaded the most.

Basically any game / DirectX application will use the first core for the DirectX instructions. 3Dmark does the same here. So when this core is loaded up to 100% he will cap the information sent to the graphics card (GPU). This is what we call a bottleneck, so the while the GPU could still produce higher framerates as it is not completely busy internally, the CPU is limiting the input. This is at least something we can assume to be the case for Rome 2 and many DirectX games.

We will never know the exact functions and core utilisation distribution without a response from CA in this case. So far they did answer on this question, that these is internal information they cannot pass.

What is DirectX 12:
DirectX 12 API has been annouced some weeks ago by Microsoft to reduce this CPU overhead massively. So far MS said no new hardware will be needed to support it. Maybe they limit it to be compatible with Windows 8.x or Windows 9, so excluding Windows 7 but that's hypothetic.


What does Mantle:
Some of you may have noticed that AMD meanwhile tried to address those problems prior the release of DirectX 12 introducing a new low level API for developers called Mantle.

Mantle offers any developer a way of communication with supported AMD GPUs natively without using the DirectX software API, thus reducing (by excluding) the DirectX overhead. The developers can address the instructions to the AMD GPU faster and directly via the GPUs driver, and so reducing the work the CPU has to do for transforming the instructions to a language the driver interprets.

So while Mantle uses a different low level API, so not DirectX API, the game needs to support it technically. Mantle needs to be supported by the game.

What does NVIDIA here:
The advantage in NVIDIA's approach is their planned optimisation runs with the use of the existent DirectX API, so in opposite to the Mantle support, CA do not need to make any efforts here, changing their sourcecode of the game to support it.

NVIDIA planned the release a new driver in a few days (7th April) with the version 337.50 as a public beta, which will also use an driver based optimization to address the CPU overhead of DirectX 9 - 11 has.
In their release notes NVIDIA states insane performance increases up to 64% excusively for Rome 2 and finally a support of SLI configurations!


I still think this is not the end of the ladder and future release can make more differences across different hardware setups and then again we should see an additional boost when DirectX 12 comes and Rome supports it.

I really hope CA will turn and fix their messed up LoD issues, if this driver really can deliver what it promises in the first stage.

have a look on the pictures (english)
http://www.computerbase.de/2014-04/n...ownload-april/

You can expect me to update my thread about Rome 2 performance then.

I have one question to anyone of you who have the intention to install the new drivers:

Please run the ingame benchmark with the current settings (current NVIDIA driver) AND AGAIN with the same settings (new 337.50 driver) and post both screenshots (Windows 8.x: Windows key + PrintScr Key) in the benchmark thread, please. Please state your graphics card (GPU) and processor model (CPU). If you want to Upgrade from Windows 8.1 to 8.1 update 1 on the 10th April another run pre Update 1 / after Update 1 would be feasible.

Thank you very much. I really look forward to many feedback and comparison screens.

edit, 22.04.2014: lastest driver 337.60
http://nvidia.custhelp.com/app/answers/detail/a_id/3491

[RedFox]

This actually sounds pretty exciting, alQamar If we can finally get some decent performance out of Rome II, it would be a total game changer for a lot of people (aside from gameplay issues, but that's off-topic).

As for Direct3D12, I think it might be a little bit too optimistic to say that CA can/or will implement it even if it comes out late next year. Judging how D3D API's have completely changed between graphics generations and from what Microsoft released in their Build conference, the API will be different from the current D3D11 API. The developers will have direct access to graphics card resources and in order to reap the most performance gain, the developers would have to redesign their resource system accordingly.

I'm glad the nVidia driver updates improved performance of existing D3D11 games though - it makes me even more eager to get a new GTX card.

[Ishan]

In their release notes NVIDIA states insane performance increases up to 64% excusively for Rome 2 and finally a support of SLI configurations!

Yeah AMD said the same thing (50%) for mantle but that was for top end R9 290 cards only & here it's only for GTX 700 series. Would like to see how much boost a GTX 770 gets on 1080p. I'm not sure what's the benefit here really, since GTX 700 series was performing well with the game from the very start:
(Oct 8, 2013)

So 64% on this would mean 100+ fps, what's the benefit of that? Unless of course someone is playing at 1440p or above. Or using a GTX 760 on 1080p.

[baptistus]

So 64% on this would mean 100+ fps, what's the benefit of that? Unless of course someone is playing at 1440p or above. Or using a GTX 760 on 1080p.

You will have a "boost" on scenes with high cpu demand (close look on battle, very large battles ect...). So instead of to have 20 fps on a close look on a battle, you could have 25-30 fps.
But It depends of the cpu you have, the gpu you have, the graphic setting of the game, the scene of the test ect...

[Ishan]

You will have a "boost" on scenes with high cpu demand (close look on battle, very large battles ect...). So IMHO you will not have 100fps with a good cpu, but instead of to have 20 fps on a close look on a battle, you coukd have 25-30 fps.
But It depends of the cpu you have, the gpu you have, the graphic setting of the game, the scene of the test ect...

If the CPU gets bottleneck GPU wouldn't be used to its full potential in the first place, so one has to assume that CPU is holding back the GPU in such scenarios which isn't the case for everybody if they have high end CPUs. Anyways 100+ was an exaggeration on my part. The exact figures as per the advertised performance boost on GTX 770 should be around 85 fps and in graphical intense situation you can divide that by 2 to have 42+ min fps. My point is that 700 series is the top end series which can run the game quite well already. People who're having performance issues mainly have average rigs, so they shouldn't expect anything from this new driver if they have older cards like GTX 600 series etc.

[The Wandering Storyteller]

Yeah AMD said the same thing (50%) for mantle but that was for top end R9 290 cards only & here it's only for GTX 700 series. Would like to see how much boost a GTX 770 gets on 1080p. I'm not sure what's the benefit here really, since GTX 700 series was performing well with the game from the very start:
(Oct 8, 2013)

So 64% on this would mean 100+ fps, what's the benefit of that? Unless of course someone is playing at 1440p or above. Or using a GTX 760 on 1080p.

I have an I5 3570 processor with 16gb and GTX 760.

I hope some improvement comes - I'm playing with 30 different mods and DEI is CPU intensive on my game, though am finding solutions to reduce the lag/statterings.

Waiting to see what this update brings.

[SamueleD]

I hadn't heard about this massive performance increase for Rome II yet, but I had heard news that nVidia was planning a new driver to further reduce CPU overhead in DX11, and not for single games but for all DX11 applications, which would amount to performance improvements in CPU bottlenecked DX11 games, like TW, but I did not expect such a massive improvement, this is great news for nVidia users!

[Ygraine]

Can I get some clarification please - is this increase only for 700-series cards or will other cards (like my 670 Windforce) also get some kind of performance increase? Thanks!

[SamueleD]

Can I get some clarification please - is this increase only for 700-series cards or will other cards (like my 670 Windforce) also get some kind of performance increase? Thanks!

The 600 and the 700 series are based on the same architecture, so I would assume you should get similar benefits on TW. Though probably all cards in recent generations should get some improvement, since the main improvements are on the software side and are not game specific: nVidia drivers are now more CPU optimised (it's a simplified version, but that's basically what's happening).

[Ishan]

Can I get some clarification please - is this increase only for 700-series cards or will other cards (like my 670 Windforce) also get some kind of performance increase? Thanks!

Yeah 700 only:

GeForce GTX 700 Series (Single GPU):

Up to 64% in Total War: Rome II
Up to 25% in The Elder Scrolls V: Skyrim
Up to 23% in Sleeping Dogs
Up to 21% in Star Swarm
Up to 15% in Batman: Arkham Origins
Up to 10% in Metro: Last Light
Up to 8% in Hitman Absolution
Up to 7% in Sniper Elite V2
Up to 6% in Tomb Raider
Up to 6% in F1 2013
GeForce GTX 700 Series (SLI):

Up to 71% in Total War: Rome II
Up to 53% in Sniper Elite V2
Up to 45% in Aliens vs. Predator
Up to 31% in Sleeping Dogs
Up to 20% in CoD: Black Ops 2
Up to 10% in Hitman Absolution
Up to 9% in F1 2013
Up to 7% in Far Cry 3
Up to 6% in Metro: Last Light
Up to 6% in Batman: Arkham Origins

http://videocardz.com/50176/nvidias-...ased-april-7th

Have to wait for more info once they're released. Benefits for lower models will be obviously much lower than 700.

[Sir_Pee_Alot]

http://www.geforce.com/drivers/results/74732
-----------------------------------------------------

Spoiler Alert, click show to read: 

New in GeForce 337.50 Beta drivers

• Performance - Introduces key DirectX optimizations which result in reduced game-loading times and significant performance increases across a wide variety of games.Results will vary depending on your GPU and system configuration. Here are some examples of measured gains versus the previous 335.23 WHQL driver:

GeForce GTX 700 Series (Single GPU):

• 
  
  • Up to 64% in Total War: Rome II
  • Up to 25% in The Elder Scrolls V: Skyrim
  • Up to 23% in Sleeping Dogs
  • Up to 21% in Star Swarm
  • Up to 15% in Batman: Arkham Origins
  • Up to 10% in Metro: Last Light
  • Up to 8% in Hitman Absolution
  • Up to 7% in Sniper Elite V2
  • Up to 6% in Tomb Raider
  • Up to 6% in F1 2013
  
  

GeForce GTX 700 Series (SLI):

• 
  
  • Up to 71% in Total War: Rome II
  • Up to 53% in Sniper Elite V2
  • Up to 45% in Aliens vs. Predator
  • Up to 31% in Sleeping Dogs
  • Up to 20% in CoD: Black Ops 2
  • Up to 10% in Hitman Absolution
  • Up to 9% in F1 2013
  • Up to 7% in Far Cry 3
  • Up to 6% in Metro: Last Light
  • Up to 6% in Batman: Arkham Origins
  
  

• SLI Technology
  
  • Total War: Rome II – added profile
  • War Thunder – added profile
  • Watch Dogs – updated profile
  • Diablo III – updated profile
  
  

• Gaming Technology
  
  • Supports GeForce ShadowPlay™ technology
  • Supports GeForce ShadowPlay™ Twitch Streaming
  
  

• SHIELD
  
  • Supports NVIDIA GameStream™ technology
  
  

• 3D Vision
  
  • Supports new “3D Compatibility Mode” for 3D Vision that enables us to improve the 3D experience for many key DirectX 10 and 11 games.
  
  

• 
  
  • 3D Vision Profiles
    
    • Path of Exile – rated “Good”
    • KickBeat – rating now “Excellent”
    
    
  
  

• 
  
  • 3D Compatiblity Mode Profiles
    
    • Assassin’s Creed Liberation – previously “Not Recommended”, now rated as “Excellent”
    • Sniper Elite: Nazi Zombie Army – previously “Good”, now rated as “Excellent”
    • Sniper Elite: Nazi Zombie Army 2 – previously “Good”, now rated as “Excellent
    • Strike Suit Zero – previously “Not Recommended”, now rated as “Good”
    • Watchdogs – rated as “Good”
    
    
  
  

[alQamar]

http://www.geforce.com/drivers/results/74732
-----------------------------------------------------

Spoiler Alert, click show to read: 

New in GeForce 337.50 Beta drivers

• Performance - Introduces key DirectX optimizations which result in reduced game-loading times and significant performance increases across a wide variety of games.Results will vary depending on your GPU and system configuration. Here are some examples of measured gains versus the previous 335.23 WHQL driver:

GeForce GTX 700 Series (Single GPU):

• 
  
  • Up to 64% in Total War: Rome II
  • Up to 25% in The Elder Scrolls V: Skyrim
  • Up to 23% in Sleeping Dogs
  • Up to 21% in Star Swarm
  • Up to 15% in Batman: Arkham Origins
  • Up to 10% in Metro: Last Light
  • Up to 8% in Hitman Absolution
  • Up to 7% in Sniper Elite V2
  • Up to 6% in Tomb Raider
  • Up to 6% in F1 2013
  
  

GeForce GTX 700 Series (SLI):

• 
  
  • Up to 71% in Total War: Rome II
  • Up to 53% in Sniper Elite V2
  • Up to 45% in Aliens vs. Predator
  • Up to 31% in Sleeping Dogs
  • Up to 20% in CoD: Black Ops 2
  • Up to 10% in Hitman Absolution
  • Up to 9% in F1 2013
  • Up to 7% in Far Cry 3
  • Up to 6% in Metro: Last Light
  • Up to 6% in Batman: Arkham Origins
  
  

• SLI Technology
  
  • Total War: Rome II – added profile
  • War Thunder – added profile
  • Watch Dogs – updated profile
  • Diablo III – updated profile
  
  

• Gaming Technology
  
  • Supports GeForce ShadowPlay™ technology
  • Supports GeForce ShadowPlay™ Twitch Streaming
  
  

• SHIELD
  
  • Supports NVIDIA GameStream™ technology
  
  

• 3D Vision
  
  • Supports new “3D Compatibility Mode” for 3D Vision that enables us to improve the 3D experience for many key DirectX 10 and 11 games.
  
  

• 
  
  • 3D Vision Profiles
    
    • Path of Exile – rated “Good”
    • KickBeat – rating now “Excellent”
    
    
  
  

• 
  
  • 3D Compatiblity Mode Profiles
    
    • Assassin’s Creed Liberation – previously “Not Recommended”, now rated as “Excellent”
    • Sniper Elite: Nazi Zombie Army – previously “Good”, now rated as “Excellent”
    • Sniper Elite: Nazi Zombie Army 2 – previously “Good”, now rated as “Excellent
    • Strike Suit Zero – previously “Not Recommended”, now rated as “Good”
    • Watchdogs – rated as “Good”
    
    
  
  

You made me to get out of my warm bed mate, unfortunately this link is dead. How sad I though I could just start the needed benchmarks.

Can I get some clarification please - is this increase only for 700-series cards or will other cards (like my 670 Windforce) also get some kind of performance increase? Thanks!

Yeah 700 only:
Have to wait for more info once they're released. Benefits for lower models will be obviously much lower than 700.

Hi Ishan thanks for supporting this discussion. I think I have very much to agree with SamueleD's technical information here.

Most of the 700s series cards are pimped ones from the 600's series, so their GPU chips are technically identical. Sorry if I raised too much expectations but the word "up to" is a holy grail in the marketing language.

Spoiler for
Comparing GTX600 / GTX700 cards GPU architecture: 

Red marks "old" and "new" 600 / 700 cards. Green framed cards are based on revised architechture (GK110) so a developed chip of the GK104. The major changes between the 600 and 700 card in the red frames are more internal available shader cores as well higher clockrates.

We will certainly see how much this figure of 64% is worth across other cards. Even if the performance boost is less significant, let's say 40% and we consider the advantages take place on low FPS rather on high FPS I hope this game gets far better playable to me and others not having 7-8 low fps in certain conditions. The max fps have been never a serious problem for the playability of this game.

I also would like to remind: NVIDIA has a good reason to state only the 700's cards in their patch logs. They can gain revenues by selling these 700's cards if anyone is naive enough to believe those increases would only happen to these new cards and not the previous ones, which should not be the case for given reasons.

If this case comes true, I bet my non existent hat again, they then read your GPU card model and only "unlock" the performance if you have certain 700's cards even if those are technically based on the same GK104 chipset, as their previous 600's pendants. If we can successfully prove those speculations, I think this may lead a real problem of their reputation very quickly.
I would not hesitate to buy a 770 card for a limited time for testing purposes only or again ask PrussianPrince to assist me here to gather reliable data as needed.

As I do trust NVIDIA, even they seem at least to try to led people into temptation letting them think the increases are only possible with their current family, I will not recommend anyone to replace a 670 or similar card before we haven't tested the stuff thoroughly enough.

So let's see and wait what we get on the 7th. I am back to bed

[ToonTotalWar]

nice find and informative +reps

[-iceblade^]

Wow... Wonder driver indeed LOL. I hope the 64% isn't just in 1 specific circumstance, as it usually is. Even if it's generally 30% or so, it'll still be something since there really hasn't been much on nVidia's end for improving performance here.

[AngryTitusPullo]

Moved from General Discussions.

[AnthoniusII]

So let me see if I understand it well..
Those with W8.0 and with gc of 700 series -atleast- will benefit...
Other gc (lets say my GTX-660) wont have any performance change...right?
So there wil be no need for people like me to download the new gc drivers.

[alQamar]

So let me see if I understand it well..
Those with W8.0 and with gc of 700 series -atleast- will benefit...
Other gc (lets say my GTX-660) wont have any performance change...right?
So there wil be no need for people like me to download the new gc drivers.

As explained in post #11 I am quite sure this driver will not only address the 700's series. We will see it tomorrow for sure. It would be very cool if you can do the steps as described in my OP, so basically running forest benchmark with both drivers and same settings.

[Mr.Freud]

I really hope it will work for 600 generation too. I have 2 gtx 680s.
But how exactly will it work? I mean - will GPU render some of the stuff that has been rendered with the help of CPU?

[alQamar]

I really hope it will work for 600 generation too. I have 2 gtx 680s.
But how exactly will it work? I mean - will GPU render some of the stuff that has been rendered with the help of CPU?

No not that, just the amount of overhead causing a higher CPU load should be reduced by this driver. Can you please run both comparisons I asked for in the OP? This would be really great as you not only have a 600 series card but also SLI which is additionally added.

[TylerTR95]

I have a 660 and use DEI mod, so any optmization would be wonderful, even if its not the 70% like on the 700 cards, since the battles are so much bigger with DEI.