NVIDIA to Launch GeForce 337.50 Beta Later Today

[btarunr]

It's now NVIDIA's turn to come up with a 'wonder driver' that introduces large across-the-board performance improvements that could affect performance standings of GeForce GTX GPUs. Bearing the canonical version number GeForce 337.50 Beta, the driver will step up DirectX rendering performance thanks to some fundamental changes in the way the the driver and API handle graphics processing loads for the metal (GPU) to chew on. The difference that makes, according to a leaked NVIDIA press-deck, is that NVIDIA GPUs will offer higher frame-rates than Mantle, while staying in Direct3D mode. Put in numbers, NVIDIA is claiming performance improvements of up to 64% in Total War: Rome II, up to 21% in Star Swarm, and up to 10% in Metro: Last Light.



View at TechPowerUp Main Site

 

[ViperXTR]

hope to see a review on this "wonder driver"

 

[johnspack]

So all I need is $750 for a gtx that can do this?

 

[arbiter]

well i already know my gtx780 is almost as fast as a 290x using mantle in theif, only about 3-4fps difference avg. this should make it more then even as nvidia tends to be little more realistic in their numbers and not claiming x but turning out to have to back pedal later.

 

[sweet]

well i already know my gtx780 is almost as fast as a 290x using mantle in theif, only about 3-4fps difference avg. this should make it more then even as nvidia tends to be little more realistic in their numbers and not claiming x but turning out to have to back pedal later.

"290x in silent mode".
Most of reviews just simply ignored this fact :|

 

[john_]

So all I need is $750 for a gtx that can do this?

Don't forget to buy also a 3930K if you want all that extra performance.

well i already know my gtx780 is almost as fast as a 290x using mantle in theif, only about 3-4fps difference avg. this should make it more then even as nvidia tends to be little more realistic in their numbers and not claiming x but turning out to have to back pedal later.

That's because you have a 4770K @ 4.5GHz. Replace it with an average i3, i5 or an AMD cpu, then rerun those tests and tell me how close you are to 290X with Mantle. With a mid range performance cpu the more expensive 780Ti with this wonder driver will need binoculars to see a 290X with Mantle. Fortunately for Nvidia NO ONE is going to test this driver with an APU or a i3.

 

[techy1]

Don't forget to buy also a 3930K if you want all that extra performance.



That's because you have a 4770K @ 4.5GHz. Replace it with an average i3, i5 or an AMD cpu, then rerun those tests and tell me how close you are to 290X with Mantle. With a mid range performance cpu the more expensive 780Ti with this wonder driver will need binoculars to see a 290X with Mantle. Fortunately for Nvidia NO ONE is going to test this driver with an APU or a i3.

you damn right - noone is gona pay 700$ for GPU and then try to combine them with sub 150$ "APU or a i3" - nor in test, nor in real life... and why would they?... if you can afford 700$ for GPU and can not afford 350$ for 4770K... then you need reduce amount you spend for GPU and increase it for CPU - simple as that... or next time someone will complai that he can not max out his 700$ gpu cuz his sub 50$ mother board is bottleneckiing him

 

[Naito]

Replace it with an average i3, i5 or an AMD cpu, then rerun those tests and tell me how close you are to 290X with Mantle. With a mid range performance cpu the more expensive 780Ti with this wonder driver will need binoculars to see a 290X with Mantle.

An i5 should be able to keep up enough to not see any real benefit from Mantle. Ephemeral is a word I'd use to describe Mantle; it is both platform limited (PC) and hardware limited (AMD) (which arguably why it is so beneficial... less overheads). Over time its benefits will be dissolved when OpenGL and DirectX (12 is looking promising) integrate similar technology, both of which are found on PCs, consoles, and mobile devices.

Fortunately for Nvidia NO ONE is going to test this driver with an APU or a i3.

I would certainly hope not. Who, spending that amount of money, would build such an asymmetrical system? There would be no point buying a 780 to throw into an i3 system, as any benefit over say a 770, would be less due to bottlenecking. Best buy an i5/770 combo, for example.

 

[john_]

you damn right - noone is gona pay 700$ for GPU and then try to combine them with sub 150$ "APU or a i3" - nor in test, nor in real life... and why would they?... if you can afford 700$ for GPU and can not afford 350$ for 4770K... then you need reduce amount you spend for GPU and increase it for CPU - simple as that... or next time someone will complai that he can not max out his 700$ gpu cuz his sub 50$ mother board is bottleneckiing him

No one? Many. You thing that anybody out there have the money to upgrade both the gpu and cpu? You think that everyone out there knows how to build a PC that it is balanced. How many out there are with an FX or a Phenom II or an i3 or an i5 and a 290X or a 780/TI? Much more than you think. MUCH MUCH MORE.

An i5 should be able to keep up enough to not see any real benefit from Mantle. Ephemeral is a word I'd use to describe Mantle; it is both platform limited (PC) and hardware limited (AMD) (which arguably why it is so beneficial... less overheads). Over time its benefits will be dissolved when OpenGL and DirectX (12 is looking promising) integrate similar technology, both of which are found on PCs, consoles, and mobile devices.

It should, but I don't expect it to have so much positive impact like in the above results that Nvidia is showing. Also Mantle could have (I don't remember to be honest) a more meaningful impact on frame rate than it will have with a 3930K (that's why Nvidia is hiding the DirectX results in the Thief chart).
Now DirectX12 looks like and maybe it is Mantle, but with 1.5 years more time for extra testing. Nvidia will use Mantle then, when it will be with a name that will not make the marketing department to jump out the windows. They could use Mantle today, but they don't. I wouldn't either if I where in their position.

I would certainly hope not. Who, spending that amount of money, would build such an asymmetrical system? There would be no point buying a 780 to throw into an i3 system, as any benefit over say a 770, would be less due to bottlenecking. Best buy an i5/770 combo, for example.

I already answered that. Most people are not like us spending hours everyday on line reading specs and charts. They will have a specific budget and they will go and build a PC based on that budget. I have seen many buy a $400 cpu and use the on board graphics to try to play games and then when they see that they can't they go out and buy a GT630 with 4GBs of RAM. Or others buy the strongest graphics card they can, and then go for a 4GHz quad core cpu - that sounds OK doesn't it? Quad core. 4GHz. Or in other words call me 750K for example. Others go out and buy a $300 cpu a $300 gpu and then a 800W NONAME PSU to save $50. Not to mention systems that are pre build from big OEMS or shops. You can read many specs there and laugh and do many faceplams.

 

[ViperXTR]

My 3570k is paired with a gtx 660 am i one of those cases?

 

[Naito]

No one? Many. You thing that anybody out there have the money to upgrade both the gpu and cpu? You think that everyone out there knows how to build a PC that it is balanced. How many out there are with an FX or a Phenom II or an i3 or an i5 and a 290X or a 780/TI? Much more than you think. MUCH MUCH MORE.

Care to share your proof? Looking at the steam surveys, the majority of list topping GPUs, excluding IGPs, are mid-range in nature. As far as CPUs go, seems it's almost 50/50 split between 2 and 4 physical CPUs (pretty sure it is excluding SMT).

 

[pr0n Inspector]

No one? Many. You thing that anybody out there have the money to upgrade both the gpu and cpu? You think that everyone out there knows how to build a PC that it is balanced. How many out there are with an FX or a Phenom II or an i3 or an i5 and a 290X or a 780/TI? Much more than you think. MUCH MUCH MORE.

Flawless logic.

 

[techy1]

No one? Many. You thing that anybody out there have the money to upgrade both the gpu and cpu? You think that everyone out there knows how to build a PC that it is balanced. How many out there are with an FX or a Phenom II or an i3 or an i5 and a 290X or a 780/TI? Much more than you think. MUCH MUCH MORE.

about people who "out there are with an FX or a Phenom II or an i3 or an i5 and a 290X or a 780/TI" ... well ... I am sure - they have some other gifts, than brain... and Jesus loves them anyway...
but I doubt that those people are able or ready to spend 600+ on GPU (if they invested in their PC less than 100$ in a year) .... I would love to see some examples, tough >. we are talking here about Ferrari fuel consumption... and many people are so stupid to buy Ferrari, but can not afford gas (just as unlikely scenario)... and you say that Ferrari should do something about that damn fuel conpumption - cuz there is MUCH MUCH MORE - of those people than I think....

 

[john_]

Care to share your proof? Looking at the steam surveys, the majority of list topping GPUs, excluding IGPs, are mid-range in nature. As far as CPUs go, seems it's almost 50/50 split between 2 and 4 physical CPUs (pretty sure it is excluding SMT).

I added a few stuff in my previous post. And Steam surveys? Really? Well I am not an NBA fan and don't live in USA but if you go and do a gallop in Chicago about the most favorite NBA team what results do you expect to get?

 

[pr0n Inspector]

Yeah with that line of thinking I'm not surprised you're defending the stupid and their stupid choices.

 

[HumanSmoke]

No one? Many. You thing that anybody out there have the money to upgrade both the gpu and cpu? You think that everyone out there knows how to build a PC that it is balanced. How many out there are with an FX or a Phenom II or an i3 or an i5 and a 290X or a 780/TI? Much more than you think. MUCH MUCH MORE.

+1
High comedy. Thanks for the laugh

They could use Mantle today, but they don't.

NO.
Mantle is proprietary at this time.
There is no public SDK or documentation available.
They say a picture is worth a thousand words, if so, then here's a picture. Also note the bullet point "Selected access only"

 

[Naito]

I added a few stuff in my previous post. And Steam surveys? Really? Well I am not an NBA fan and don't live in USA but if you go and do a gallop in Chicago about the most favorite NBA team what results do you expect to get?

At least I provided some sort of evidence. I figured Steam surveys were the largest/most reliable source of this kind of information. I mean, the user base is certainly big enough...
Also, your statement doesn't make much sense, but from what I think you're saying; the Steam equivalent to that would be if I went around asking everyone who uses a PC what their favourite game client is (Steam, Origin, UPlay, Battle.net, etc). Most peoples answer will probably be Steam...

 

[Enterprise24]

If rome II get massive boost 60-70% then I will sell my 290 @ 290X and switch to NV cause rome II is my main game and seriously bottleneck by CPU even from 4670K @ 4.9Ghz Uncore 4.7Ghz and RAM 2800Mhz.

 

[burebista]

It's out now

Win 8/Win 7/Win Vista - International
x32
x64
Release notes

English only:

x32
x64

Notebook:
x32
x64


XP:
x32
x64

 

[Fluffmeister]

Free performance, yay!

 

[burebista]

If rome II get massive boost 60-70% then I will sell my 290 @ 290X and switch to NV cause rome II is my main game and seriously bottleneck by CPU even from 4670K @ 4.9Ghz Uncore 4.7Ghz and RAM 2800Mhz.

Dunno if it's 70% increase but a guy on other forum just posted this:
Rome 2 runs great now. Tested some Campaign and Karthago battle. Also gpu-load is now really nice.

He has in signature:
Processor:AMD X4 B50 3,3GhZ (oc)
Videocard:GigaByte GTX 660 oc / 2gb
OS:Win 8.1 (x64)

LE: OK, the guy posted a benchmark.
TW Rome II:
=======================
335.23: 53,8 avg
337.50: 55,4 avg #smoother in battles, gained fps on campaign map

 

[tokyoduong]

I think this is just the usual selective NVIDIA slides again. If a game gained 60+% from a driver update, then it either means you were selective on how you tested it or it was broken before.

 

[Ed_1]

I wouldn't expect much more than 10% but if it increases 10% and also smooths out the min ,avg etc then its good .

Those slides are deceptive, many are SLI performance and many are compared to 290x not 335-337 , like BF4 one .

But whatever they can get by optimizing the driver, its free .

 

[burebista]

Some user benchmarks.

 

[Fluffmeister]

I'm definitely seeing some nice improvements too, just ran a quick StarSwarm test, standard follow/extreme run:

Stock GTX 670/ i7 920.

335.23
== Results ================================================
Test Duration: 360 Seconds
Total Frames: 9615

Average FPS: 26.70
Average Unit Count: 3962
Maximum Unit Count: 5538
Average Batches/MS: 492.75
Maximum Batches/MS: 1413.87
Average Batch Count: 18609
Maximum Batch Count: 120381
===========================================================

337.50
== Results ================================================
Test Duration: 360 Seconds
Total Frames: 14781

Average FPS: 41.06
Average Unit Count: 4269
Maximum Unit Count: 5617
Average Batches/MS: 864.40
Maximum Batches/MS: 2656.92
Average Batch Count: 22857
Maximum Batch Count: 93661
===========================================================

Not the best benchmark IMO, but the with these drivers the FPS definitely didn't tank as bad and overall it was no doubt faster.