Tested: Nvidia's Performance-Boosting GeForce 337.50 Driver

Nvidia has posted a new driver - the GeForce 337.50 Beta driver. Normally we wouldn't give driver releases all that much attention, though this one is a little different.

Nvidia claimed that this driver will bring some very big performance improvements in a whole number of games. Some games will experience performance improvements of up to 64 percent on single GPUs, while SLI performance can be boosted by as much as 71 percent. The optimizations have been done through optimizations with DirectX, specifically reducing CPU overhead in certain games.

With such claims though, we wanted to have a crack at the numbers ourselves to see what the real-life effects are, so we cleaned our system, installed the drivers, and started the benchmarks. We weren't running the exact same hardware platform as Nvidia was running, but this would only serve to make things a little more interesting.

Test Platform:

Swipe to scroll horizontally
Test Hardware
ProcessorIntel Core i7-4770K @ Stock Clocks
MotherboardAsus Maximus VI Impact
GraphicsReference Nvidia GTX 780 Ti 3GB
MemoryCorsair Vengeance Pro 16 GB (2 x 8GB) DDR3-1866, 1.5 V
System DriveSamsung 830 256 GB SATA3 (6Gb/s)
Operating SystemWindows 8.1 Pro x64

The biggest difference between our test system and Nvidia's test system is the CPU. Nvidia is running a monstrously powerful Intel Core i7-3960X while we're 'only' running an Intel Core i7-4770K. Of course, the i7-4770K is still a very power processor, and it is also something that you're a lot more likely to encounter in a gaming system than an i7-3960X.

1080p Performance:

We actually got some very interesting results for our 1080p performance tests. Tomb Raider and Battlefield 4's performance went up with similar amounts to Nvidia's claims. Unigine's Heaven Benchmark actually ran an average of 0.2 FPS lower over three test runs, though that falls within a margin of error.

The biggest surprise was Star Swarm, as we got a very big performance jump (almost 60 percent) -- much higher than Nvidia's claim of "up to 21 percent". Because I wasn't sure that this was correct, I actually reverted back to the old driver and re-ran the Star Swarm tests another three times. This changed nothing... The Nvidia GeForce 337.50 Beta driver really gives Star Swarm a significant boost! Without more testing, we wouldn't be able to say exactly why we got a much bigger performance jump than Nvidia advertised, though one possibility is that that it was because we had a less powerful CPU installed, and the reduced CPU overhead thus made an even bigger impact. In the past, folks have referred to Star Swarm as "An AMD marketing tool" -- that's probably over now.

1440p Performance:

1440p performance followed the same trends as the 1080p performance, giving both Tomb Raider and Battlefield 4 a nice little performance boost. Unigine's Heaven benchmark again ran just the way it did before, which is interesting in, well, how uninteresting it is.

Unfortunately, we didn't have another GTX 780 Ti at our immediate disposal, so it was a bit difficult to run SLI benchmarks. That said, we'd imagine that the performance gains would, like Nvidia claims, be even higher than on single-card configurations, as you're more likely to run into CPU bottlenecks when running more graphics cards.

All in all this is one of the most impressive driver releases we've seen in a long time. Nvidia's point to all this is to show that you don't need a new API and developer support in order to be able to get performance gains, but that time well spent optimizing drivers can make a big difference too. That said, competition from AMD's Mantle API likely pushed Nvidia to develop these optimizations. In the end, it's the gamers who win.

You can find the drivers available for download here on Nvidia's website, where you'll also find the complete release highlights.

Niels Broekhuijsen

Niels Broekhuijsen is a Contributing Writer for Tom's Hardware US. He reviews cases, water cooling and pc builds.

  • Arkive
    Was doge really a necessary way to start the article?
  • bucknutty
    Does this fix the SLI Profile in TitanFall?
  • Zepid
    Does this fix the SLI Profile in TitanFall?
    This is what I wanna know. TitanFall only uses one of my two 570s in SLi.
  • ykki
    "4770k is still a very power processor"umm... what?
    toms needs proofreading.
  • Zeroplanetz
    There site says OS 32bits. I'm new to all of this so if I have 8.1 64bit I shouldn't do this beta drive?
  • mauller07
    time to wait for these to drop and actual performance analysis to roll in since people know nvidia have fixed the AMD performance results, considering the highest settings in Thief using Mantle at 1080p, a stock 290 (not even the x) gets 68 fps avg on beta 14.3.i dont mind them bigging up performance improvements, but the outright lying is such a joke.
  • coupe
    Nvidia's revolutionary marketing driver. Guaranteed 150% better marketing to combat AMD Mantle hype.
  • spazoid
    Was doge really a necessary way to start the article?
    Maybe not necessary, but very much appreciated.
  • clonazepam
    Was doge really a necessary way to start the article?
    I'm gonna say yeah cuz I learned something new! =)
  • CaptainTom
    Am I the only one who finds Nvidia's marketing slides hilarious? They make it look like they are massively beating AMD when really all they are showing is that with their "Wonder Drivers" their $700 card beats AMD's $560 card by 12%. Wow what a victory Nvidia! /s