Overclock.net banner
1 - 20 of 177 Posts

·
TotallydubbedHD on YT
Joined
·
11,564 Posts
Discussion Starter · #1 ·
Hey guys!

So first of all - here's the main thread:
http://www.overclock.net/t/1479451/various-geforce-337-50-wonder-driver-beta-release

The reason I'm posting this in a separate thread, is so that I can manage replies more easily and furthermore be able to have this information nicely presented in an OP, rather than being lost in a (very fast) moving paced thread.

Recent Nvidia drivers supposedly have HUGE FPS increases. I was extremely sceptical, so I thought to put my non-bias tests to good use.

In order to eliminate too much deviation, I decided to do all my tests in an empty server. Siege of Shanghai was the chosen map (due to it being a Vanilla map, more people will be able to test for themselves).

DISCLAIMER:
I would like to state that I did the tests in a controlled environment, as it allowed me to repeat the tests over and over again. Instead of going into SP or even a full MP game, I could design a test that would stress different aspects (lighting, sudden explosions, scenery etc) - however in SP for example, I'm not only limited to what I can do, but ALSO I cannot do the same tests (no matter if I try and play the same way) - furthermore SP tests would be much longer to do and much harder to carry out.

These benches I did are NOT to see what my 680s can handle nor is it what mantle can handle.
This was purely to test, how the new driver compared to the old driver. The new driver performed so well, that it had "mantle-esk" (I even say that in my video) performance gains. I think that's why there's been a huge rave about it and furthermore how Nvidia compared it to Mantle.

A driver: You would normally expect 1 -5% gains. NOT 14-30% gains. thus why it's so shocking to see

Initially when I posted this thread, I conducted only 1 set of tests. Due to the popularity of this thread and the views it got (awesome sauce response from everyone) - I decided to fortify the results.
Thus conducted 4 tests extra PER scenario. Thus giving me 5 tests per scenario. This gives a much better picture of what to expect and eliminates "deviation".

Furthermore, I started these benchmarks on Win7 ONLY - but due to popular demand - I decided to include them on Win8.1 too.
My rig and OC runs exactly the same on Win7 as it does on Win8.1.
The Only differences were:
On Win7 (home edition) - I have a Corsair Force GT SSD
On Win8.1 (pro) - I have a SATA2 HDD
I have previously found absolutely no differences in FPS with different HDD/SSD speeds. The ONLY difference between the two, are load times.

I actually double checked that as I was asked if a HDD would make a difference in the FPS vs an SSD: http://www.overclock.net/t/1480050/new-nvidia-337-50-drivers-battlefield-4-benchmarks-ht-on-vs-off-win8-1-vs-win7/70#post_22087031

Finally, on a small note:
Whilst on Win7 - I was the CN team
Whilst on Win8.1 (different server) - I was the US team
Thus the choppers were different. Weirdly enough, the US chopper took longer to "suicide" than the CN one.

Test environment:
-i7 3770k @ 4.5ghz
-SLI GTX680s 2GB VRAM (see below for GPU-Z screenshots)
-Asus Sabertooth Z77
-16GB Corsair Vengeance 9-9-9-24 2N 1600mhz RAM
-Windows 7 64 bit Home edition (On SSD) / Windows 8 64bit Professional Edition (On HDD)
-QNIX QX2710 1440p Monitor OC'ed to 96hz - fully stable


BF4 Video settings:


Thus everything is on ultra, apart from the FPS killer the MSAA which kills my 2GB VRAM limit on 1440p. I also don't see much of a difference of it on 2x or 4x on 1440p.

Scenario

Video uploaded to get a much better idea:
Written list:
  • Start
  • Spawn in chopper
  • Start benchmark
  • Exit chopper
  • Run up and down of tower
  • Shoot RPG to wall 2x
  • Reload RPG whilst jumping around and looking around map
  • Get into chopper
  • Fly low in outside view
  • Fly around buildings (and main tower) in 1st person view
  • Crash into building 2-3x and die of explosion
  • Spawn in car
  • Put 3x C4 on car, blow up from close range, blowing yourself up
  • Spawn in IFV
  • Exit IFV, put C4
  • Walk to safe distance, blow C4 and IFV
  • Shoot debris of IFV or tank/car - full auto with AWS
  • Get into tank
  • Shoot with outside view
  • Shoot with inside view
  • View scenery with tank whilst zoom
  • End

Windows 7 Results:
NEW = 337.50 Drivers
OLD = 335.23 Drivers
Avg = (Average)
Min = Minimum
Max = Maximum

All number attached were in FPS.



Percentage Changes:
Avg (Average) FPS:
New vs OLD HT ON: +14.45%
New vs OLD HT OFF: +7.85%



Min FPS:
New vs OLD HT ON: +24.60%
New vs OLD HT OFF: +8.68%



Max FPS:
New vs OLD HT ON: +6.22%
New vs OLD HT OFF: +1.79%



Monitoring CPU Threads and GPU Load on Windows 7
Screenshots of CPU Usage + RAM + GPU clocks & temps etc
PLEASE right click, open in new tab the below screenshots. Unless you got a microscope
tongue.gif


New 337.50 HT OFF


New 337.50 HT ON


Old 335.23 HT OFF


Old 335.23 HT ON


PS.
Fresh Boot CPU Usage + RAM - as everyone will have different RAM usage (my RAM usage is a little higher than most people: Around 500mb more than usual):


Windows 8.1 Results:
NEW = 337.50 Drivers
OLD = 335.23 Drivers
Avg = (Average)
Min = Minimum
Max = Maximum

All number attached were in FPS.



Percentage Changes:
Avg (Average) FPS:
New vs OLD HT ON: +2.42%
New vs OLD HT OFF: +1.84%



Min FPS:
New vs OLD HT ON: +13.64%
New vs OLD HT OFF: -2.93%



Max FPS:
New vs OLD HT ON: +2.09%
New vs OLD HT OFF: +2.17%



Monitoring CPU Threads and GPU Load on Windows 8.1
Screenshots of CPU Usage + RAM + GPU clocks & temps etc
PLEASE right click, open in new tab the below screenshots. Unless you got a microscope
tongue.gif


New 337.50 HT OFF


New 337.50 HT ON


Old 335.23 HT OFF


Old 335.23 HT ON


PS.
Fresh Boot CPU Usage + RAM - as everyone will have different RAM usage (my RAM usage is a little higher than most people: Around 500mb more than usual):



Windows 7 VS Windows 8.1

In here you'll see the averages of each data (min, max, avg) - put against each other.
You're going to be absolutely surprised of the results. Windows 7 comes out on top - by a HUGE margin.
The figures shown are the differences that Win8.1 has OVER win7.
Thus if it is positive - it's how win8.1 is POSITIVELY better than win7. If however it's negative, then it means win7 comes out on top.

The values at the bottom are displayed by percentages.



Percentage Changes between Win8.1 and Win7 (Win8.1 being seen if it is better or worse than Win7):
Thus if it is positive - it's how win8.1 is POSITIVELY better than win7. If however it's negative, then it means win7 comes out on top.
Avg (Average) FPS:
New Win8.1vs Win7 HT ON: +3.27%
New Win8.1vs Win7 HT OFF: +3.61%
Old Win8.1vs Win7 HT ON: +15.40%
Old Win8.1vs Win7 HT OFF: +9.73%

Min FPS:
New Win8.1vs Win7 HT ON: -28.34%
New Win8.1vs Win7 HT OFF: -24.33%
Old Win8.1vs Win7 HT ON: -21.43%
Old Win8.1vs Win7 HT OFF: -15.29%

Max FPS:
New Win8.1vs Win7 HT ON: -3.08%
New Win8.1vs Win7 HT OFF: +2.28%
Old Win8.1vs Win7 HT ON: +0.84%
Old Win8.1vs Win7 HT OFF: +1.90%

I have to say these results completely shocked me. I was expecting results such as the Average FPS ones or even MAX FPS ones - but the MIN FPS ones completely threw me off my chair.
Despite Win8.1 being better on "average" the minimums (which really effect fluidity) were ridiculously worse on Win8.1 over Win7.

It should be said though, both OS's did gain FPS% with the new driver - especially Win7. However I was expecting Win8.1 to gain even more % due to it utilising a newer version of DirectX.

Unbelievable results!

Personal Input:
I personally think these new drivers are amazing. They are aimed at people with high-end rigs. (not trying to undermine anyone with any other rig, just saying the truth of what these drivers are aimed at).
Even other sites I saw - all seem to point in the same direction:
If you have a powerful GPU + CPU - you'll see higher optimisation for it.
SLI rigs, seem to be targeted too (happy days for me
biggrin.gif
)

I think this Nvidia driver won't effect everyone in the same way.
Some people with more powerful PCs will see more gains, rather than people with lower-end rigs that might not see as bigger percentage gains.

Should AMD be concerned? Should Nvidia do better?
Personally I think this is better than what Mantle currently offers.
I'm not saying that as an Nvidia user, as I honestly couldn't give a single crap about fanboy "sides". I bought my GTX660ti back in the day, due to better driver support. I then bought the GTX680, because it was affordable. Then SLI'ed it as I wanted my rig to run 1440p comfortably at max graphics.

If I had known back 2 years ago, that something called Mantle would have been coming out, I would have gone for AMD. Especially seeing as my favourite franchise, battlefield, teamed up with AMD.

So why say you prefer 337.50 over mantle?
Simple: Mantle is a great low-level API utilisation of GPUs, however it can only be "enabled" in certain games. It doesn't effect every single thing.
337.50 is nothing new, just a new driver. However it is a driver for DirectX. Thus can be utilised and used throughout every game that uses it (almost everything on windows is based on it -> as windows is behind DirectX).
If you ask me, nothing has changed in terms of low-level codec etc - it's just better optimisation of the GPUs with the OS. Thus it isn't game bound. It's OS bound.

I should also state that these results aren't to be compared with a mantle system (I think it's almost impossible to do so) - however the performance GAINS that were achieved by Nvidia's latest drivers are incredible - they provide "mantle-esk" performance gains. Huge gains in either mins, max or avgs. Exactly as Mantle did, when it first came out.

What you should note out of all the results, is the huge increases in MIN FPS. This allows smoother gameplay and much better experiences for gamers.

All in all my reaction to these drivers is quite easily shown by this picture:

Keen-Clapping-After-Loss.gif

If you follow BF4 Dev - you'll know that BF4 was built on top of win8, and not win7 - thus naturally you get better FPS results on win8 - although as some people wildly claimed 40% differences- I found it to be less than 5% difference:
http://www.overclock.net/t/1433904/comparison-of-windows-7-vs-windows-8-i7-3770k-with-ht-enabled-vs-disabled-on-battlefield-4

I hope this information was useful!
-Dubbed
 

·
Registered
Joined
·
1,444 Posts
Quote:
Originally Posted by Totally Dubbed View Post

Hey guys!

So first of all - here's the main thread:
http://www.overclock.net/t/1479451/various-geforce-337-50-wonder-driver-beta-release

The reason I'm posting this in a separate thread, is so that I can manage replies more easily and furthermore be able to have this information nicely presented in an OP, rather than being lost in a (very fast) moving paced thread.

Recent Nvidia drivers supposedly have HUGE FPS increases. "Mantle killer" type performances. I was extremely sceptical, so I thought to put my non-bias tests to good use.

In order to eliminate too much deviation, I decided to do all my tests in an empty server. Siege of Shanghai was the chosen map (due to it being a Vanilla map, more people will be able to test for themselves).

Before I get into anything, here's a DISCLAIMER.
My results were controlled and all done in the same manor, with the SAME amount of applications running in the background. However, these are only single tests. I did redo some benches, and they ended up close (although not perfectly) similar. Thus you SHOULD expect variances in results. I would give or take 5% of my end results, in order to cater for random deviation. results, incl deviation will be included and are still impressive.

Test environment:
-i7 3770k @ 4.5ghz
-SLI GTX680s 2GB VRAM (see below for GPU-Z screenshots)
-Asus Sabertooth Z77
-16GB Corsair Vengeance 9-9-9-24 2N 1600mhz RAM
-Windows 7 64 bit Home edition (I didn't bother doing tests on win8.1, however I might do that at a later stage)
-QNIX QX2710 1440p Monitor OC'ed to 96hz - fully stable


BF4 Video settings:


Thus everything is on ultra, apart from the FPS killer the MSAA which kills my 2GB VRAM limit on 1440p. I also don't see much of a difference of it on 2x or 4x on 1440p.

Scenario
  • Start
  • Spawn in chopper
  • Start benchmark
  • Exit chopper
  • Run up and down of tower
  • Shoot RPG to wall 2x
  • Reload RPG whilst jumping around and looking around map
  • Get into chopper
  • Fly low in outside view
  • Fly around buildings (and main tower) in 1st person view
  • Crash into building 2-3x and die of explosion
  • Spawn in car
  • Put 3x C4 on car, blow up from close range, blowing yourself up
  • Spawn in IFV
  • Exit IFV, put C4
  • Walk to safe distance, blow C4 and IFV
  • Shoot debris of IFV or tank/car - full auto with AWS
  • Get into tank
  • Shoot with outside view
  • Shoot with inside view
  • View scenery with tank whilst zoom
  • End

Results:
Avg (Average) FPS:
New 337.5 HT OFF: 115.08
New 337.5 HT ON: 118.48
Old 335.23 HT OFF: 111.486
Old 335.23 HT ON: 97.989

Min FPS:
New 337.5 HT OFF: 48
New 337.5 HT ON: 62
Old 335.23 HT OFF: 44
Old 335.23 HT ON: 47

Max FPS:
New 337.5 HT OFF: 190
New 337.5 HT ON: 202
Old 335.23 HT OFF: 185
Old 335.23 HT ON: 183

Percentage Changes:
Avg (Average) FPS:
New vs OLD HT ON: +20.91%
New vs OLD HT OFF: +17.44%

Deviation included:
New vs OLD HT ON: 15.91-25.91% increase
New vs OLD HT OFF: 12.44-22.44% increase

Min FPS:
New vs OLD HT ON: +31.91%
New vs OLD HT OFF: +2.13%

Deviation included:
New vs OLD HT ON: 26.91-36.91% increase
New vs OLD HT OFF: -2.87% decrease to 7.13% increase

Min FPS:
New vs OLD HT ON: +10.38%
New vs OLD HT OFF: +3.83%

Deviation included:
New vs OLD HT ON: 5.38-15.38% increase
New vs OLD HT OFF: -1.17% decrease to 8.83% increase

Personal Input:
I personally think these new drivers are amazing. They are aimed at people with high-end rigs. (not trying to undermine anyone with any other rig, just saying the truth of what these drivers are aimed at).
Even other sites I saw - all seem to point in the same direction:
If you have a powerful GPU + CPU - you'll see higher optimisation for it.
SLI rigs, seem to be targeted too (happy days for me
biggrin.gif
)

I think this Nvidia driver won't effect everyone in the same way.
Some people with more powerful PCs will see more gains, rather than people with lower-end rigs that might not see as bigger percentage gains.

This is ALMOST the complete opposite of Mantle.
Mantle helped lower-end rigs to perform far more superior. This new driver, seems to have boosted powerful rigs, even further. As if Nvidia was aiming to beat the max benches, not the min benches. Ie. 290X gets 120FPS, we get 200 with a 780TI - both being the top-end gaming cards from both companies.

Should AMD be concerned? Should Nvidia do better?
Personally I think this is better than Mantle. I'm not saying that as an Nvidia user, as I honestly couldn't give a single crap about fanboy "sides". I bought my GTX660ti back in the day, due to better driver support. I then bought the GTX680, because it was affordable. The SLI'ed it as I wanted my rig to run 1440p comfortably at max graphics.
If I had known back 2 years ago, that something called Mantle would have been coming out, I would have gone for AMD. Especially seeing as my favourite franchise, battlefield teamed up with AMD.
So why say 335.7 > mantle
Simple: Mantle is a great low-level API utilisation of GPUs, however it can only be "enabled" in certain games. It doesn't effect every single thing.
335.7 is nothing new, just a new driver. However it is a driver for DirectX. Thus can be utilised and used throughout every game that uses it (almost everything on windows is based on it -> as windows is behind DirectX).
If you ask me, nothing has changed in terms of low-level codec etc - it's jsut better optimisation of the GPUs with the OS. Thus it isn't game bound. It's OS bound.
So it will be interesting to see if I get the same results over on Wn8.1 - where the DX is newer, thus maybe less or even better optimised with the Nvidia GPUs.

If you follow BF4 Dev - you'll know that BF4 was built on top of win8, and not win7 - thus naturally you get better FPS results on win8 - although as some people wildly claimed 40% differences- i found it to be less than 5% difference:

I hope this information was useful!
-Dubbed
FTFY
 

·
Premium Member
Joined
·
10,772 Posts
Quote:
This is ALMOST the complete opposite of Mantle.
Mantle helped lower-end rigs to perform far more superior. This new driver, seems to have boosted powerful rigs, even further.
Think you've got this a bit backwards, at Mantle and this new driver both help in the same cases.

They add more FPS to the higher end before your CPU becomes a limit - that's the gains you're seeing here

They also make weaker CPU's able to achieve the same FPS (the performance increases you mention here with lower end rigs)

If you clock your cpu to 2.5ghz and run the same test, you'll see the same or greater gains

Good job testing though! I know what it's like sitting for hours for some numbers
thumb.gif
 

·
Premium Member
Joined
·
8,166 Posts
Quote:
Originally Posted by ABD EL HAMEED View Post

FTFY
Can you not quote the entirety of his post please?
tongue.gif


Very nice testing mate, it's appreciated. I don't suppose there's any chance of you testing some other games?
 

·
Premium Member
Joined
·
10,772 Posts
Oh also:
Quote:
Results:
Avg (Average) FPS:
New 337.5 HT OFF: 115.08
New 337.5 HT ON: 118.48
Old 335.23 HT OFF: 111.486
Old 335.23 HT ON: 97.989
Is this backwards on here, and on the chart too?

The Old HT off and On seem to be reversed
biggrin.gif
 

·
Twitch Streamer
Joined
·
2,531 Posts
Isnt it 337.50 the driver? Not 335.7?
 

·
TotallydubbedHD on YT
Joined
·
11,564 Posts
Discussion Starter · #8 ·
Quote:
Originally Posted by xSociety View Post

Brb, going to go turn HT back on.
thumb.gif


+rep! Thanks for all the hard work!
Should have always been on!
Blasphemyyyyyy haha
tongue.gif


Quote:
Originally Posted by Cyro999 View Post

Think you've got this a bit backwards, at Mantle and this new driver both help in the same cases.

They add more FPS to the higher end before your CPU becomes a limit - that's the gains you're seeing here

They also make weaker CPU's able to achieve the same FPS (the performance increases you mention here with lower end rigs)

If you clock your cpu to 2.5ghz and run the same test, you'll see the same or greater gains

Good job testing though! I know what it's like sitting for hours for some numbers
thumb.gif
Agreed, they help in both cases. but from what I recollected, lower AMD cards and/or lower-end CPUs benefited more than Higher-end rigs for Mantle gains.
I can edit it out if needs be - just what would you suggest there or your thoughts on the two?
smile.gif


Quote:
Originally Posted by twerk View Post

Can you not quote the entirety of his post please?
tongue.gif


Very nice testing mate, it's appreciated. I don't suppose there's any chance of you testing some other games?
I would prefer testing BF4, as I know the game quite well - thus I can cater for certain scenarios. If I go for another game, there might be other issues that might deviate the results.
This was quite time consuming (believe it or not) - but my mian aim now would be to get it running on win8.1 again and see results there. I think that would be interesting and much more relevant as then we're taking a newer and improved DX build over the other.

PS. Thanks for your concern, you have my permission, as the OP to edit that gentleman's quote - makes the first page easier to read
smile.gif
 

·
Premium Member
Joined
·
1,900 Posts
Nice results, here is a quick bench I ran with Thief on a single GTX 780 Ti and 3770K (stock) at 1080P

Thief
335.23 = 65FPS
337.50 = 68.4FPS

Surely with a higher OC on the CPU or SLI, the performance gain should be even greater.
 

·
TotallydubbedHD on YT
Joined
·
11,564 Posts
Discussion Starter · #11 ·
Quote:
Originally Posted by Cyro999 View Post

Oh also:
Is this backwards on here, and on the chart too?

The Old HT off and On seem to be reversed
biggrin.gif
Believe it or not, they're actually correct.

Here's another one with HT ON with old drivers:
2014-04-07 19:29:51 - bf4
Frames: 12532 - Time: 126002ms - Avg: 99.459 - Min: 46 - Max: 178

To make it easy for you:
Result used HT ON + old driver (used in OP):
2014-04-07 19:43:56 - bf4
Frames: 13076 - Time: 133443ms - Avg: 97.989 - Min: 47 - Max: 183

HT OFF + old driver:
2014-04-07 19:22:34 - bf4
Frames: 15578 - Time: 139730ms - Avg: 111.486 - Min: 44 - Max: 185

I got all the excel frametimes, if you're interested
tongue.gif

Quote:
Originally Posted by Aesthethc View Post

Isnt it 337.50 the driver? Not 335.7?
Yeah, typo brother - I'll correct that now - thanks for pointing out - that's what happens when you type out all this thing in one go haha
Quote:
Originally Posted by EVGA-JacobF View Post

Nice results, here is a quick bench I ran with Thief on a single GTX 780 Ti and 3770K (stock) at 1080P

Thief
335.23 = 65FPS
337.50 = 68.4FPS

Surely with a higher OC on the CPU or SLI, the performance gain should be even greater.
Nice thanks for sharing
smile.gif
!
 

·
Vermin Supreme 2020
Joined
·
25,788 Posts
decent results for a first run. I expect more from nvidia on the WHQl.
 

·
Father, Husband and Gamer
Joined
·
1,135 Posts
great job! + rep
 

·
TotallydubbedHD on YT
Joined
·
11,564 Posts
Discussion Starter · #15 ·
Quote:
Originally Posted by nasmith2000 View Post

great job! + rep
thanks!
Quote:
Originally Posted by xFALL3Nx View Post

WOO! Wonder what this will do for my 760, TIME TO ULTRA MODE!
If you can test - the more data the merrier
smile.gif


On that note guys - if you guys conduct the same tests, feel free to share I'll link it in the OP
 

·
Premium Member
Joined
·
10,772 Posts
I only tested sc2 - results were too small to blind test, think 5 maybe 7 percent if there was a difference (it's very hard to say exactly without spending hours and rolling around drivers)

so.. still will be dipping to 20fps in team games, with 5% gpu load
biggrin.gif
 

·
TotallydubbedHD on YT
Joined
·
11,564 Posts
Discussion Starter · #17 ·
Quote:
Originally Posted by Cyro999 View Post

I only tested sc2 - results were too small to blind test, think 5 maybe 7 percent if there was a difference (it's very hard to say exactly without spending hours and rolling around drivers)

so.. still will be dipping to 20fps in team games, with 5% gpu load
biggrin.gif
yup for the very few results I had I spent a good majority of 2hrs doing it.
Another 1-2hrs writing it and composing all the data.

If I went to the length of my win7 vs win8.1 comparison - we're talking over 1 week of solid benching lol

I'll definitely am going tog et more results, but maybe tomorrow or something
smile.gif

incentive:
This article got a shout-out by EVGA's product manager:
https://twitter.com/EVGA_JacobF/status/453297338904100864
 

·
PC Evangelist
Joined
·
47,612 Posts
Quote:
Originally Posted by Cyro999 View Post

I only tested sc2 - results were too small to blind test, think 5 maybe 7 percent if there was a difference (it's very hard to say exactly without spending hours and rolling around drivers)

so.. still will be dipping to 20fps in team games, with 5% gpu load
biggrin.gif
I think there is a problem there. SC2 should see 0% improvement. After all its a DX9 game.
 

·
TotallydubbedHD on YT
Joined
·
11,564 Posts
Discussion Starter · #19 ·
Quote:
Originally Posted by ZealotKi11er View Post

I think there is a problem there. SC2 should see 0% improvement. After all its a DX9 game.
Oh Cyro getting caught out on placebo?
jerry.gif
 

·
Registered
Joined
·
1,174 Posts
So obviously theres no real way to benchmark WoW but I do know that in crowded zone , my minimum can easily drop to 40fps. Now with the new drivers, It doesnt drop below 50fps (vsync off) and I did a quick LFR (25man raid) and my minimum droped to 35ish a few time but quickly went back up to 50-100fps (depending on the situation), usually it could stay around 35ish for the whole fight (in 25man). In 10man its buttery smooth , no fps drops what so ever.
 
1 - 20 of 177 Posts
Top