I7 VS Core 2 Quad
High resolution gaming performance analysis
Page 1 / 8 07/11/09
INTRODUCTION
This article is a follow up to our recent GTX 285 TRI SLI Scaling Analysis article. If you haven't already read that, I would highly recommend that you take a look and get familiarized with the background for this article. Even if you don't feel like doing that, I will try my best to make ihs article clear and to thoroughly convey the methodology, analysis and results. Here goes...
In our SLI scaling article we investigated the performance of GTX 285 TRI SLI in twenty three game titles. Some were DX 10, while others were older DX 9 titles. The goal was to see how well SLI performs in modern and semi modern gaming titles. As most of you might know, high end multi GPU systems can easily get starved of CPU horsepower. To fulfill this immense CPU demand, we ran the benchmarks at a resolution of 2560 X 1600, with the highest in game image quality settings, resulting in GPU dependent scenarios. Moreover, we overclocked our QX 9650 to a blistering 4.2 GHZ in an attempt to alleviate most of the CPU bottleneck. For the most part, the results were quite satisfying and SLI scaled very well. However, there were certain titles where we believed that our CPU was holding us back ... yes even at 4.2 GHZ !! Far Cry 2, World in Conflict, Fall Out 3 and a few others came in this league.
Fortunately, we now have a Core I7 setup to redo some of those benchmarks and see what kind of performance gains if any do we see.
TEST BED AND BENCHMARKS
The following system was used for the original article:
EVGA 780i SLI
QX9650 @ 4.2 GHZ
4 GB GEIL Esoteria @ 875 5-4-4-12-2T
GPU (3) X EVGA GTX 285 3-SLI @ 648 / 1470 / 2480 (Forceware 182.08)
Gateway 30" XHD
OS Vista X64
For this article, we our using a brand new I7 920 Platform:
EVGA X58 SLI LE
I7 920 D0 @ 3.8 GHZ HT enabled
4 GB Patriot Viper DDR3 @ 1444 MH 9-9-9-25-1T
GPU (3) X EVGA GTX 285 3-SLI @ 648 / 1470 / 2480 (Forceware 182.08)
Gateway 30" XHD monitor
OS Vista X64
The benchmarks that I used for this article are :
Lost Planet Demo (DX10)
Call of Juarez Demo (DX10)
World in Conflict Demo (DX10)
Crysis Demo (DX10)
Race Driver: GRID (DX9)
Far Cry 2 (DX10)
Stalker: Clear Sky Demo (DX10)
FallOut 3 (DX10)
Tom Clancy's HAWX (DX10)
Cryostasis: Sleep of Reason Demo (DX10)
Note that I7 920 is clocked at 3.8 GHZ which is as far as it could go on the stock cooling.
BENCHMARKING METHODOLOGY
Synthetic tests such as 3DMARK06 are easy to perform as all you have to do is run a test and the results are conveniently output as an overall score and average frame rates. Games however may or may not be so easy to benchmark and there are multiple ways to benchmark a game. For this analysis, I have used two methods :
-Built in Demo : A lot of games these days come with a pre-recorded demo which the manufacturer deliberately includes for gamers to benchmark the game and gauge the performance of their systems. It gives you an idea of how well your PC will be able to run the game. This sort of benchmark is good since the exact same scene is run each time you play it. This gives you consistent Frame rate results. The downside is that it may not necessarily be comparable to the actual gameplay performance in the game for a variety of reasons. For example Artificial Intelligence (AI), or sound might be disabled in the demo but real world gaming will obviously have that.
-Real world Benching : A lot of benchers prefer real world benching. They use a program called FRAPS which runs in the background and monitors your frame rates as you play the game . This is useful since you get an idea of the actual gameplay performance. Another benefit is that you get frame rates recorded with time so you can clearly see when your fps dipped below average and what not. You can also very easily get the min, max and average frame rates. However, your results will not be consistent and totally accurate, you will not be able to make a full comparison between your results and someone else's and its tedious !
Besides Race Driver: GRID and Fall Out 3 which were tested using FRAPS, all others have a built in demo. I am only including the results for the TRI SLI case, since a single card showed no change going from QX 9650 to I7 920. As for dual SLI, I did not have enough data (no results for 3.8 GHZ QX 9650) to do an apples to apples comparison. Having said that, TRI SLI scaling should give us a fairly decent idea of how Dual SLI would scale, especially if these results are compared with the original articles.
Ideally we should be comparing both minimum and average FPS, but since due to various reasons the original article compared only the average FPS, we will be going the same route here.
Let us take a look at the results now. Note that you can click on any of the game titles to go to the corresponding image gallery page.
| NEXT |
|---|