18-11-2008, 04:58 AM
|
#1 |
|
Senior Member
Join Date: Jan 03 2007
Posts: 225
|
Core i7 and Games, Memory Testing 6GB vs 3GB
Back in April 2008 we tested 4GB vs 2GB. Now that the Core i7 memory offerings are different, we did more testing with 6GB vs 3GB. We got very similar results. Vista with 3GB on i7 is good but 6GB is much better.
 APPLICATION NOTE 811 EVA2000, please relocate this if there is a better section for it. 
|
|
|
|
18-11-2008, 04:39 PM
|
#2 |
|
Administrator
|
AN811: Gaming Performance Analysis- 6GB vs. 3GB
AN811: Gaming Performance Analysis- 6GB vs. 3GB Mike Clements, Corsair Memory Inc. Introduction 
With the introduction of the Intel Core i7 processor series, which features an on-die, triple-channel memory controller, consumers now have a choice between 3GB (3 x 1GB) and 6GB (3 x 2GB) system memory densities. This Application Note aims to assess the potential gaming performance advantages of installing 6GB of system memory compared to 3GB of system memory, using the same testing methodology as described in AN804. The results echo those of AN804 and show that 6GB of system memory is essential for gaming PCs based around Core i7 CPUs, with up to a 3x increase in performance compared to an identical PC with 3GB of RAM. How we tested To conduct the testing we built a high-performance Core i7 test system that reflects the type of PC that a gaming enthusiast would buy. By using high-end components we also preclude the possibility of any single item of hardware causing performance bottlenecks, which allows us to accurately measure the effect of increasing the memory density. Memory Configurations
Test Platform
  Testing Methodology Due to the various methods available for benchmarking different games, we used a variety of tools to assess the impact of using 3GB and 6GB memory array sizes. The specific testing methodology for each game is detailed below for each title. In each case we used a resolution of 1680 x 1050 and configured each game to run at the highest detail settings. We ran each test 3 times, rebooting between each test, and recorded the average of the 3 runs. All of the test components were run at stock speeds and all unused peripherals were disabled in the BIOS. In addition, the following BIOS settings were used:
Results World in Conflict World in Conflict is a modern, highly resource-intensive Real-Time Strategy (RTS) game, which was released in September 2007. World in Conflict supports DirectX 10 graphics cards and is a tough test of overall system performance. The game includes a built-in benchmarking tool, which we used to assess the performance of this game with 3GB and 6GB system memory sizes. Each test was run at 1680 x 1050 in DirectX 10 mode. World in Conflict performance The following results from World in Conflict's built-in benchmarking tool show that there are clear performance benefits from using 6GB of system memory compared to 3GB of system memory. The minimum, average and maximum frame rates all increase substantially when using 6GB of system memory. The average frame rates increases by 19%, from 75fps to 89fps while the minimum frame rate increases by 50% from 54fps to 84fps. The increase in minimum frame rate is most significant since the minimum frame rate determines the overall smoothness of the gameplay.  Crysis Warhead Crysis Warhead is one of the most demanding games on the market, requiring high-end graphics processors, CPUs and plenty of memory. We used the latest Crysis title, Crysis Warhead, to examine the performance impact of using 3GB of memory compared to 6GB. To assess the frame rates in this game we used FBWH Tool, which was developed by 'Mr.John' on the InCrysis forums. This allows for detailed adjustment of graphical and system settings, in addition to providing accurate benchmarking options. We used the latest version of FBWH Tool, Version 0.31, which you can download here . The following benchmarks were run at 1680 x 1050 in DirectX 10 mode. We used both the GAMER and PERFORMANCE presets and tested with the 'Cargo Flythrough'; benchmark. The GAMER preset uses high graphical detail settings and is a more resource-intensive test. Crysis Warhead performance The results using the GAMER preset shows that, although the average and maximum frame rates are virtually identical for 3GB and 6GB memory array sizes, the minimum frame rate is increased by 26% when using 6GB of system memory.  The PERFORMANCE preset reduces the graphical complexity of the game resulting in a less 'GPU-bound'; scenario, whereby the video card is not the performance limiter. The results show that the average and minimum frames rates are higher with 6GB of memory, although the gap in minimum frame rate is smaller than when using the GAMER preset, most likely due to reduced memory requirements. The results indicate that using 6GB of memory is highly advisable to run Crysis Warhead at high graphical detail settings. In addition to running pre-scripted benchmarks using FBWH Tool, we also analysed the in-game performance of Crysis Warhead using FRAPS (version 2.9.6, build 7637). To assess actual gameplay performance we played through an easily repeatable scene in the 'All the Fury'; chapter on the airfield. The results show that the in-game memory usage of Crysis Warhead is significantly higher than when running only the benchmark, which resulted in significant performance differences between 3GB and 6GB memory array sizes. Once again the most significant difference was to the minimum frame rate, which dropped to an unplayable 11fps when using 3GB of memory, compared to 33fps with 6GB - a substantial difference of 3x.  Crysis Warhead memory usage analysis The dramatic difference in minimum frame rate between 6GB and 3GB memory array sizes can be partly explained by analysing the memory usage of Crysis Warhead for the two configurations. Crysis Warhead - 3GB memory array size  The above screenshot was taken during the 'All the Fury'; level and shows the memory usage of the Crysis process when 3GB of system RAM was installed. The total amount of memory comitted to the process (Private Bytes) is over 1.4GB, of which only 1GB is stored in physical memory. This difference leads to more regular hard disk access and stuttering in the game. Crysis Warhead - 6GB memory array size  With 6GB of system RAM installed in the PC, the Private Bytes figure has remained roughly the same as with 3GB installed, but the physical memory usage has increased to 1.4GB. Storing more game data in physical RAM reduces the need to load data from the hard disk drive, ensuring a consistantly higher minimum frame rate. Warhammer Online The Massively Multiplayer Online Role Playing Game (MMORPGs) genre is a hugely popular and influential area of PC gaming with titles such as World of Warcraft being played by 11 million people worldwide. Warhammer Online is one of the latest and most sophisticated games in this genre, and while not quite as graphically demanding as Crysis Warhead, it still requires significant amounts of system memory in order to run smoothly. We played though a repeatable section of the game and recorded the average, minimum and maximum frame rates using FRAPs. Warhammer Online performance The results show a clear performance gain from using 6GB of system memory compared to 3GB of system memory. With 6GB of RAM the minimum frame rate was 16fps, 45% higher than when using 3GB of RAM.  Warhammer Online memory usage analysis Examining the memory usage of Warhammer Online during actual gameplay helps to understand the reason for the increased performance when using 6GB of memory compared to 3GB. Warhammer Online - 3GB memory array size  With 3GB of system memory the physical memory usage of the Warhammer Online process during an intense battle scene with 60 combatants reached 1.2GB. Private Bytes peaked at 1.35GB. Warhammer Online - 6GB memory array size  With 6GB of RAM the Private Bytes figure rises to just over 1.4GB and 1.35GB is stored in physical memory. The extra memory headroom of the 6GB configuration leads to smoother performance in the game. Game loading times To assess whether installing 6GB of system memory improved game loading times, we ran two different tests using Crysis Warhead and World in Conflict. For Crysis Warhead we loaded a saved game from the 'Call me Ishmael'; chapter, recording the time taken from clicking on the load button until 'Press Any Key To Continue'; was displayed. We repeated the test three times, rebooting between each test. For World in Conflict we used a section from the 'Behind the Iron Curtain'; campaign. We recorded the time taken from clicking on the 'Continue Campaign'; button until 'Press Any Key To Continue'; was displayed. The test was repeated three times, rebooting between each test. Windows Vista's SuperFetch process was disabled to ensure that the caching function of this process did not affect loading times. Game loading time results The results show that using 6GB of system memory significantly impoved the game loading times of these two titles. The Crysis Warhead saved game loaded 12% faster with 6GB compared to 3GB, while the campaign in World in Conflict loaded 23% faster with 6GB of system memory compared to 3GB of system memory.  Average Game Loading Times  Total Windows Vista x64 memory usage In addition to analyzing the memory usage of each individual game process, we also recorded the total memory usage of the whole system, with no other applications running other than the specified game. The chart below shows that Warhammer Online and Crysis Warhead both clearly require more than 3GB of total system memory in order to run smoothly. World in Conflict uses less memory in total, but the extra headroom that 6GB of system memory provides still resulted in higher performance.  Conclusion These results show that using 6GB of system memory provides significant performance benefits in today's games when using the Intel Core i7 platform. The analysis shows that 3GB of system memory is insufficient to run modern games, such as Warhammer Online and Crysis Warhead, resulting in poor performance. The lack of memory when using 3GB of RAM results in increased hard disk drive access, sometimes called thrashing. This causes in-game stuttering, which reduces the minimum frame rate. The minimum frame rate is crucial for fluid gameplay since this determines the overall 'smoothness'; of a game. In our tests the minimum frame rate in World in Conflict rose by 50% when we upgraded from 3GB of system memory to 6GB of system memory. In Crysis Warhead the jump was far more significant, with the minimum frame rate increasing by a factor of 3x when upgrading from 3GB to 6GB of memory. The Crysis Warhead results also highlight the need for large amounts of system memory when using high graphical detail settings. The results from Warhammer Online are similar to World in Conflict and Crysis Warhead and show a minimum frame rate increase of 45% when upgrading from 3GB to 6GB of memory. The message to enthusiasts who are looking to build a Core i7 system for gaming is clear - installing 6GB of memory will provide significantly higher frame rates and a considerably smoother gaming experience. Summary of results  Additional Resources
|
|
Last edited by eva2000; 18-11-2008 at 05:16 PM. |
|
|
|
|
|
18-11-2008, 05:54 PM
|
#4 |
|
Senior Member
Join Date: Oct 07 2007
Location: London, Canada
Posts: 280
|
playing from the other side of the fence, not everyone is going to be running a pair of GTX280s. i understand the point of the demonstration is to show the difference in memory and to eliminate bottlenecks, but coming to 100% concrete conclusions...
By eva2000
...might be a bit misleading unless backed up with tests on a system using a single GTX280/GTX260 or even a single HD4870. i know, the average i7 early adopter has more money than most so higher end GPU performance will likely be on their table but there are still going to be plenty of guys running single GPUs on i7 and without evidence of 6GB benefiting all setups, the above claim can't really be made. unless appended with, "In our test configuration..." or "With no GPU bottleneck..." etc...
again, the information is great and it is nice to see Corsair continue with providing useful information. i am just stating that from a "review" standpoint and not a "marketing" stand point, blanket statements should be backed up with all encompassing information. |
|
|
|
|
19-11-2008, 02:41 AM
|
#5 |
|
Senior Member
Join Date: Jan 03 2007
Posts: 225
|
By eva2000
Man, my brain still hurts from all of this testing 3 vs 6.
|
|
|
|
|
19-11-2008, 07:34 AM
|
#6 |
|
Senior Member
Join Date: Oct 07 2007
Location: London, Canada
Posts: 280
|
By Yellowbeard
that would be very interesting to see. i should have my EVGA X58 here this week and a 6GB kit from you cats so once i get the reviews and whatnot out of the way, i will try to get some 3GB vs 6GB testing done as well to help verify your results. getting my hands on GPUs of that caliber might be a bit tough for me but i'll see what i can do. again, excellent testing and some very nice information for the high end rigs being built right now
|
|
|
|
|
19-11-2008, 03:26 PM
|
#7 |
|
Administrator
|
By Yellowbeard
Mike nothing panadol won't fix
By 3oh6
yeah i wonder how much even GPU memory size would effect FPS/Gaming - 256MB/512MB/768MB/868MB/1024M in terms of how much of a slice of available system memory is taken as well??
|
|
|
|
|
19-11-2008, 03:38 PM
|
#8 |
|
Senior Member
Join Date: Jan 03 2007
Posts: 225
|
By eva2000
My initial inclination is that the results will be even more pronounced with a single GPU. However, I have not done a lot of GPU/gaming based testing so it may be just hte opposite.
I'm probably not going to run the full test suite but, I'll pick a few and remove a GPU and see what happens. |
|
|
|
 |
| Bookmarks |
| Tags |
| 3gb, 6gb, core, games, memory, testing |
«
Previous Thread
|
Next Thread
»
| Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
| Thread Tools | |
All times are GMT +11. The time now is 03:01 AM.