Overclock.net banner

GTX670 CPU scaling tests [3930K 3.5GHz -> 5.1GHz]

7K views 23 replies 16 participants last post by  AlexNJ 
#1 ·


Introduction

So I had some free time and a GTX670 available so I decided to check out how much CPU frequency actually mattered at my resolution [2560x1440]. After doing some tests I realized I could actually share the results since the 670 is actually a really common GPU with nice price/perf and with the Korean monitors becoming more and more common the setup I had is suddenly a pretty common one.

Enjoy.

Settings & Setup


Setup
  • Intel Core i7 3930K @ 3.5ghz, 4.0ghz, 4.5ghz, 4.8ghz, 5.1ghz
  • Asus Rampage IV Extreme w/EK blocks
  • 2x 4GB Corsair Vengeance 1600MHz 7-8-7-21 1T @ 2133MHz 9-11-9-27 1T
  • Asus GeForce GTX 670 DirectCu II @915Mhz (980Mhz boost)
  • [X25-MG2 80GB, 2x Samsung 1TB Spinpoint F3 in raid0, WD Green 750GB
  • Corsair AX1200
  • Windows 7 64bit Home premium
Cooling
  • CPU: Single stage phase change cooling
  • Motherboard: Watercooling
  • GPU: Stock Air cooling


Games & Benches

  • StarCraft 2: Wings of Liberty [2560x1440]
  • The Elder Scrolls V: Skyrim [2560x1440]
  • Portal 2 [2560x1440]
  • The Witcher 2: Assassins of Kings [2560x1440]
  • DiRT 3 [2560x1440]
  • Crysis [1920x1200]
  • Just Cause 2 [2560x1440]
  • 3dmark 11 [Performance preset]
  • 3dmark Vantage [Performance preset]

Notes
  • Driver version: 301.42
  • Force PCI-e 3.0 patch: yes
  • All games at maxed settings w/ 8xAA
 
See less See more
8
#2 ·
Benchmarks

StarCraft 2: Wings of Liberty

As the bench the last 4 in game minutes of a normal 22 minute 1v1 game was used. PvZ.



Portal 2

As the bench the beginning sequence of the game where Wheatley transports you in your room was used.



The Elder Scrolls V: Skryim

As the bench, the beginning sequence of the game until the moment you reach character creation was used.



The Witcher 2: Assassins of Kings

As the bench, the beginning sequence of the game until the moment you get to the first dialogue option was used.

 
#3 ·
DiRT 3

As the bench the in-game benchmarking tool was used.



Just Cause 2

As benches, the three in-game benchmarking tools were used. The fps readings are averages.



Crysis

The crysis benchmarking tool was used. The resolution was lowered to 1920x1200 due to the tool not having a 2560x1440 option.



3dmark 11

The performance preset was used.



3dmark Vantage

The performance preset was used.

 
#6 ·
Thank you, I have a similar system except I'm under SLI GTX680. CPU @ 4.5Ghz to 4.6Ghz is the sweet spot for performance per watt. Gaining over 4.8Ghz actually doesn't benefits much anymore except for 3D Rendering / Bench (Maya, 3D Max).
 
#7 ·
Quote:
Originally Posted by djriful View Post

Thank you, I have a similar system except I'm under SLI GTX680. CPU @ 4.5Ghz to 4.6Ghz is the sweet spot for performance per watt. Gaining over 4.8Ghz actually doesn't benefits much anymore except for 3D Rendering / Bench (Maya, 3D Max).
Yeah that would be my conclusion as well. Pushing the CPU over 5ghz give you more epeen, is more fun is and is beneficial with F@H so I'll keep mine chugging at a higher frequency.

And I really wish I had a second one to test with, would be interesting to see the results.
 
#10 ·
Less noticeable at high resolutions and more noticeable with SLI. Not that I have any experience of SLI setups with the 600 series, I'd still imagine that they're just as CPU hungry as the 500 series if not more. 580SLI and tri SLI setups for example saw pretty huge benefits from SB.

Also weird to see a post here, just stumbled on my thread again at random
tongue.gif
 
#12 ·
Starcraft II and synthetic benchmarks appear to be the only applications with a palpable difference from CPU overclocking. Not sure how that translates into what percent of games will be sensitive to overclocking, but my guess MMOs would be, and any games with lots of physics and multiple units to keep track of with lots of configurations (such as Starcraft 2).
 
#15 ·
Quote:
Originally Posted by malmental View Post

any changes because of driver updates.?
Lol I have no idea why you would bump this
tongue.gif


The 670 is my brother's GPU, I might test it again some of these days. Maybe even do a 3930K vs. 8320 comparison with that and my titan.
 
#16 ·
Quote:
Originally Posted by Alatar View Post

Quote:
Originally Posted by malmental View Post

any changes because of driver updates.?
Lol I have no idea why you would bump this
tongue.gif


The 670 is my brother's GPU, I might test it again some of these days. Maybe even do a 3930K vs. 8320 comparison with that and my titan.
that's what I'm talking about....
I'm gathering all the info I can because the upgrade bug has me once again..
 
#17 ·
Nice test. Really good to see and some interesting results. The gain over 5GHZ seems non existent or not even a gain but a disadvantage. I hope I can get my 3820 at about 5,2 just for the epeen hahaha and then get it down to 4,9 for 24/7 or max memory bandwith depending on the ram capabilities.
 
#18 ·
Really should test some MMOs or games with high intense physics built into them. Planetside 2 really benefits from clock to clock performance on the CPU, also starcraft and WOW might be a good choice too. overall good results, kinda makes me satisfied with my 4.9ghz on my 2500k right now, seeing the difference between 5ghz and 4.9ghz isn't too significant...
biggrin.gif
 
#23 ·
Yeah thanks for posting......Kind of highlights the need for a balanced system....no point in the majority of cases in ocing the crap out of your cpu if your running more mid range graphics.

What would be nice to see is how this would pan out if you were running one or two high end gpu's that were oc'ed to the limit.....just how many GHz on the cpu are required to feed say 2x heavily oc'ed 780's......interesting stuff
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top