Nvidia's GeForce 337.50 Beta driver significantly boosts performance

[Scorpus]

Yesterday, Nvidia posted a new beta driver for their GeForce graphics cards, promising significant performance improvements over previous drivers. Through DirectX optimizations in the new GeForce 337.50 beta driver, Nvidia claims performance gains of up to 64% in single-GPU configurations, and up to 71% in SLI configurations, compared to R334-branch drivers.

In Nvidia's testing of some specific games with a GeForce GTX 780 Ti graphics card, the company found performance in Battlefield 4 was 12% higher under the 337.50 driver, and 16% higher in StarSwarm. Nvidia is boasting that the 780 Ti outperforms the AMD Radeon R9 290X with Mantle enabled in these circumstances, so there's no hiding the motives behind the optimizations.

It appears Nvidia has been reducing CPU overhead as part of the improvements targeting DirectX, but they aren't ready to share specifics on how the gains were achieved. More details can be expected when the driver comes out of beta.

The latest set of performance improvements in the 337.50 beta are leading up to wider gains in DirectX 12. The updated graphics API, which was announced at GDC 2014, is set to bring reduced CPU overhead and low-level improvements to old and new hardware towards the end of 2015.

To grab the latest GeForce beta driver, you can either head to our driver download section, or let the GeForce Experience application update it for you.

Permalink to story.

https://www.techspot.com/news/56300-nvidias-geforce-33750-beta-driver-significantly-boosts-performance.html

 

[Skidmarksdeluxe]

I'll just wait for GFE to update my driver automatically with the WHQL one.

 

[theBest11778]

I'll just wait for GFE to update my driver automatically with the WHQL one.

That's how I usually do it, but hey free performance boost... I'm there!!!

 

[GhostRyder]

Already grabbed this on my laptop, have not really tried it yet beyond a game of League this morning.

 

[Arwed]

Weren't those figured faked like "+% since last NV gen" ? At some point I've heard, that the same card only gets like 3-5% increase. The 65% was meant to be like "the 770 with new stuff is 123% better than 670".. when it was like 118% better anyways.

Warning: These numbers may be faked by marketing"!

 

[Jad Chaar]

It is very impressive how they can increase the performance so much just by tuning drivers. It shows how inefficient some code can be. Most of the numbers are probably inflated though for marketing purposes...

 

[Guest]

It is very impressive how they can increase the performance so much just by tuning drivers. It shows how inefficient some code can be. Most of the numbers are probably inflated though for marketing purposes...

We are talking about 3-5% advantage from that better driver. Its impressive, as it comes from a free software update - but it aint even close to what Mantle/DX12 deliever due to their API nature.

 

[Jad Chaar]

We are talking about 3-5% advantage from that better driver. Its impressive, as it comes from a free software update - but it aint even close to what Mantle/DX12 deliever due to their API nature.

Yeah but in some games like BF4 and Arkham Origins and benchmarks like FireStrike the increases are large. I am referring to the results from >>this<< video if you are curious.

 

[thetawaves90]

Just look at the y-axis scale on the "geforce + dx11 12% faster" graph; we're talking a gaming performance improvement of about 0.2 of a frame per second over the comparative AMD model and less than 0.05 of a frame per second improvement over the pre-existing drivers.

We have no point of reference for the scale of the "SLI performance graph", other than a "percentage increase" with nothing to base those figures on.

A claim of 64% single-gpu performance increase warrants some details on the benchmarking setup so there can be some assurance of validity of results.