Nvidia’s “Detonator” 337.50 versus Catalyst 14.2 Beta 3

Nvidia has released their new GeForce Beta driver 337.50 to gain performance especially in CPU-limited situations for DX11 games. Partly, this effort is in response to AMD’s new Mantle API which attempts to gain performance by coding “closer to the metal” at the expense of fragmenting the industry as it will only run on GCN architecture Radeons.  Currently Mantle is being featured in only two games, Thief and Battlefield 4.  By benchmarking 22 games, we are going to compare the AMD driver sets, Catalyst 13.12 and the latest 14.2 Beta 3, to Nvidia’s WHQL 325.23 and Beta 335.70.

Nvidia has claimed some impressive performance increases for DX11 with these drivers for both single and multi-GPU.  Nvidia claims “up to 64%” improvement for single GPU, although the greatest improvement seems to be “up to 71%” with SLI. We are going to test single-GPU configurations by focusing on the GTX 780 Ti, the GTX 760, and the GTX 750 Ti today; and we will follow up with Part II by focusing on SLI. We will naturally compare the GTX 780 Ti to AMD’s R9 290X in Uber mode, as well as compare the GTX 760 and the mildly overclocked PowerColor R9 270X PCS+ as they are direct competitors, approximately in the same $250 price range.

So as not to bottleneck our graphics, we are using Intel’s Haswell platform with our Core i7-4770K overclocked to 4.0GHz which is a good match for Ivy Bridge at 4.5GHz. We also use 16GB of Kingston’s “Beast” DDR3 clocked to 2133MHz, ECS’ flagship Z87 Golden Motherboard and two identical 2TB Toshiba HDDs; one for AMD and one for Nvidia.

Test Configuration – Hardware

• Intel Core i7-4770K (reference 3.5GHz, HyperThreading and Turbo boost is on to 3.7GHz; overclocked to 4.0GHz; DX11 CPU graphics), supplied by Intel.
• ECS GANK Domination Z87H3-A2X motherboard (Intel Z87 chipset, latest BIOS, PCOe 3.0 specification, CrossFire/SLI 8x+8x) supplied by ECS
• Kingston 16 GB HyperX Beast DDR3 (2×8 GB, dual-channel at 2133MHz, supplied by Kingston)
• GeForce GTX 750 Ti, 2GB reference design and clocks, supplied by Nvidia
• Nvidia GTX 760, 2GB reference design and clocks, supplied by Nvidia
• Nvidia GTX 780 Ti, 3 GB reference design and clocks, supplied by Nvidia
• PowerColor R9 270X PCS+, 2GB, at PCS+ speeds (1100/1425MHz), supplied by PowerColor
• PowerColor R9 290X, 4GB at Uber speeds; no throttling
• Two 2TB Toshiba 7200 rpm HDDs
• Thermaltake ToughPower 775W power supply unit supplied by Thermaltake
• Thermaltake Water2.0 Pro watercooler, supplied by Thermaltake
• Onboard Realtek Audio
• Genius SP-D150 speakers, supplied by Genius
• Thermaltake Overseer RX-I full tower case, supplied by Thermaltake
• Philips DVD SATA writer
• HP LP3065 2560×1600 thirty inch LCD

Test Configuration – Software

• Nvidia GeForce 335.23 WHQL and 337.50 drivers. High Quality, prefer maximum performance, single display.
• AMD 13.12 WHQL and 14.2 Beta 3 Catalyst drivers; latest CAPs. High Quality – optimizations off; use application settings.
• Windows 7 64-bit; very latest updates
• Latest DirectX
• All games are patched to their latest versions.
• VSync is off in the control panel.
• AA enabled as noted in games; all in-game settings are specified with 16xAF always applied; 16xAF forced in control panel for Crysis.
• All results show average, minimum and maximum frame rates except as noted.
• Highest quality sound (stereo) used in all games.
• Windows 7 64, all DX10 titles were run under DX10 render paths; DX11 titles under DX11 render paths

The Benchmarks

Synthetic

• 3DMark 11
• Firestrike – Basic & Extreme
• Heaven 4.0

DX9

• The Witcher 2
  
• Borderlands 2
  
• Aliens: Colonial Marines

DX10

• Crysis

DX11

• STALKER, Call of Pripyat
• Civilization V
• Max Payne 3
• the Secret World
• Sleeping Dogs
• Sniper Elite V2
• Hitman: Absolution
• Far Cry 3
• Tomb Raider: 2013
• Crysis 3
• BioShock: Infinite
• Metro: Last Light
• GRID 2
• Splinter Cell: Blacklist
• Total War: Rome II
• Batman: Arkham Origins
• Assassin’s Creed IV: Black Flag
• Thief

Before we look at the performance, let’s check out Nvidia’s release notes:
http://www.nvidia.com/download/driverResults.aspx/74636/en-us

We see some pretty impressive claims for performance increases given by Nvidia and we will test some of these games using the GTX 750 Ti, GTX 760 and the GTX 780 Ti to cover a wide performance range of video cards since the last WHQLs 335.23. Of course, the increases are given as “up to” and are not necessarily applicable to our system which is not generally CPU-limited. We will also look at AMD’s progress with Catalyst 14.2 Beta 3 since we last tested with WHQL 13.12.

The Results and the Conclusion

The Big Picture

First up, we have what we call “the Big Picture” with all of our information on one big chart.  This chart will be sub-divided into smaller charts as it tends to be confusing with a lot of data in one place.

 UPDATE 4/27/14

Sleeping Dogs was able to complete the benchmarks and it was not a driver issue as we originally suspected.  The R9 270 achieved 41.8 fps at 2560×1600 and 76.3 fps at 1920×1080 with Catalyst 14.3 Beta 3.  Using Extreme AA, the R9 290X at Uber speeds recorded 38.1 fps at 2560×1600 and 68.7 fps at 1920×1080 with Catalyst 14.3 Beta 3.

AMD’s 14.2 Beta 3 ‘Mantle Drivers

Let’s focus on the performance increases that AMD has made since WHQL 13.12, last year:

As you can see, performance is all over the place. There are some improvements and also some losses. The Sleeping Dogs benchmark crashed with the latest drivers with both the 270X and the 290X.

Let’s look at Thief and the Mantle pathway included with 14.2 Beta 3. Mantle worked fine for Hawaii’s 290X to give a small performance increase, but it chugged and stuttered using the 270X. At any rate, the GTX 760 is a bit faster generally than the overclocked R9 270X in DX11, while the GTX 780 Ti was solidly faster than even the R9 290X’ Mantle performance.

Now let’s look at Nvidia’s new beta drivers compared with the last WHQL:

We don’t see the huge increases indicated by Nvidia with Rome II, but then we are also not CPU-limited. Also, Rome II game patches have been regularly updated and they have brought their own performance increases which may affect comparisons between the drivers now.

Next, let’s compare the latest AMD latest Catalyst 14.2 beta 3 to Nvidia’s GeForce 337.50 betas:

The GTX 780 Ti continues to gain performance over the R9 290X while the GTX 760 also gains over the mildly overclocked R9 270X.

Conclusion

Nvidia new GeForce 337.50 beta driver has brought some serious performance increases to the GTX 750 Ti, the GTX 760, and to the GTX 780 Ti.  Although the emphasis was placed on eliminating the CPU bottleneck in DX 11 games which we do not generally suffer from, we saw good performance increases nearly across-the-board. They are worthy of an unofficial “detonator” designation as they bring great free performance.  Although they are beta, we encountered no issues and we highly recommend these drivers.

On the other hand, we cannot recommend AMD’s latest Beta drivers unless you are playing Battlefield 4 and need the performance. Thief does not practically benefit by using Mantle for the R9 290X which already has enough performance, and it causes serious issues with our R9 270X.

Nvidia expects that there will be further optimizations for GeForce drivers and we look forward to testing SLI in part two, including GTX 680 SLI and GTX 760 SLI when we receive a new Galaxy video card this week.  We have also been seriously delayed by our now-fixed web site’s issues in publishing the GTC wrap-up, and hope to have it up by this weekend.

Happy Gaming