Sign in with
Sign up | Sign in

Report: 3GB vs 6GB RAM on Core i7 Benched

By - Source: Tom's Hardware | B 51 comments

Planning to build a brand spanking new Intel Core i7- based gaming system with that fancy new Intel X58 motherboard ? Just make sure you toss 6GB of memory into the mix for the best possible experience, according to Corsair’s recent research.

Results from recent tests performed by Corsair indicate significant performance gains on systems with 6GB of RAM versus 3GB of RAM and Microsoft Windows Vista (x64). Corsair decided to go all out in its testing, using high-end system parts to eliminate any other hardware bottlenecks during the tests. Results indicate that it is essential for enthusiasts to opt for 6 GB, rather than 3 GB, if they are looking for the best possible experience, and of course, to move to a 64-bit operating system to support the large memory configuration.

Corsair utilized its own 3 GB and 6GB 1,600MHz Corsair Dominator DDR3 triple channel kits : (TR3X3G1600C8D , TR3X6G1600C8D). Test system specifications were as follows :

Corsair Test System Specification
Board ASUS P6T Deluxe (BIOS revision 0804)
CPU Intel Core i7-965 Extreme Edition
Video Nvidia SLI 280GTX
Storage 2 x Seagate Barracuda 7200.10 320GB in RAID-0
Display 24-inch TFT / LCD 1680x1050
OS Windows Vista Ultimate 64-bit SP1
Driver Nvidia 180.43 BETA (Required for SLI on ASUS P6T Deluxe)

Corsair Test System Configuration
Memory Manually Configured @ 16000MHz 8-8-8-24, 1T (1.65v)
QPI Frequency Manually Configured to 6400MHz
BClock Manually Configured to 133MHz
Hyper Threading Enabled

Testing took place across several different games using built in benchmarking tools when available or third party, well known benchmarking tools when needed. Each metric was run three times with a reboot in between each run. Results for all three runs were used to calculate the average.

Substantial increases were noted in most cases, and in others, minimal increases. The areas across all games that showed the most increase was on the ‘minimum’ frame rate, which is the most limiting and affecting factor in gaming. The minimum frame rate of a game usually depicts heavy load on your hardware, and in most systems, indicates heavy memory usage – either system or video ram. Disk thrashing usually accompanies this characteristic. The higher your minimum frame rate is, the smoother the over-all experience tends to be.

Corsair’s Game Line-Up :

World in Conflict : Maximum details (Built-in Benchmarking Tool) Crysis Warhead : GAMER & PERFORMANCE settings (FBWH Tool) Warhammer Online : Maximum details (Actual Game Play + FRAPS)

All of the scores below come courtesy of Corsair. For more information on our own Core i7 testing, including number to indicate that adopting three channels won’t buy much extra performance, please read UPDATE : Core i7 : Blazing Fast, O/C Changes.

World in Conflict Tests (Built-in Benchmarking Tool)

Results from these tests showed substantial increases across minimum, maximum, and average frame rates. The average increased by 19 percent. World in Conflict is a highly CPU intensive game similar to Supreme Commander, which explains the substantial increase in performance as heavy CPU load would hinder GPU data steam outputs. Each test was run at 1680 x 1050 in DirectX 10 mode.

Crysis Warhead Tests (FBWH Tool)

Test runs on Warhead did not show much in overall performance gain, but then again, Crysis is easily one of the most demanding titles out there when it comes to system and video resources, however. The minimum frame rate did increase from 17.9 to 22.5 utilizing the GAMER preset and DirectX 10 features. Not a substantial increase, but an increase none the less.

Results for the PERFORMANCE preset showed more increase over GAMER, however, the results for PERFORMANCE were quite a bit as it is, and would provide a very smooth playback no matter what. Both presets were run at 1680 x 1050 in DirectX 10 mode.

During actual game play with Crysis Warhead while using FRAPS, results did show a substantial increase in the minimum frame rate — jumping from 11fps to 33fps, using 3GB and 6GB respectively. This difference alone definitely makes a difference between unplayable and playable. FRAPS game play results were based on the scene All the Fury, repeatedly using the GAMER preset and DirectX 10.

Warhammer Online Tests (Actual Game Play + FRAPS

Warhammer can be pretty demanding on your hardware when played at maximum details especially when you are situated in an area with a lot of activity. It may not be as graphically demanding as Crysis, or other games for that matter, but it is definitely memory intensive.

Tests were run playing through a repeatable section of the game using FRAPS, results from each run were record and used to calculate the average. Results from these tests show increases across minimum, maximum, and the average. The average increased from 48fps to 53fps.

Game loading times were also tested across all three titles, and of course, Crysis again being the king crunch, showed quite a decrease in load times when the system ran 6GB of RAM versus 3GB. From the results we can see that a saved game load in Crysis Warhead showed a 12 percent faster load time with 6GB of system memory, while World in Conflict showed a 23 percent faster load time for the campaign ‘Behind the Iron Curtain’. Those are pretty impressive increases.

And lastly, during the tests, total system memory usage was recorded as well. All titles naturally used more memory when the system was configured with 6GB of RAM as opposed to 3GB. When more memory is available, more of it will generally get used.

This increased usage of memory by each title also decreases disk thrashing or stuttering due to disk access during game play. According to the results we can see that the total Windows memory usage was highest with Warhammer Online when the system was configured with 6GB of RAM.

The end result of Corsair’s test was that overall performance increased noticeably when the system was configured with 6 GB versus 3 GB. Most enthusiasts will do anything to squeeze extra performance out of a system, and this looks like a smart move. As 12 GB kits are not readily available at the moment, its hard to know whether performance would change as much as it does when jumping from 3 GB to 6 GB.

Moving to larger system memory arrays of 6GB will require users to move to the 64-bit platform, as 32-bit platforms do not support over 4 GB of memory.

Display 51 Comments.
This thread is closed for comments
  • -3 Hide
    Pei-chen , November 19, 2008 12:17 PM
    This should be top new next to the 4870 X2 reviews.
  • 0 Hide
    frankscastle , November 19, 2008 12:17 PM
    Anybody else saying "duh". I thought 8gb was the "sweet spot" for vista 64-bit, anyway?,1775.html I will wait to see 12gb bench marks before I get all worked up.
  • 3 Hide
    neiroatopelcc , November 19, 2008 12:35 PM
    Perhaps I just expect more, but I don't see the average improvement to be anything worth twice the money. ddr3 is expensive as it is!
  • -1 Hide
    Anonymous , November 19, 2008 12:42 PM
    I think that they mean 1600 MHz on the ram clock, not 16000 MHz.
  • 1 Hide
    trinix , November 19, 2008 1:06 PM
    6 gigs is better, now I'd love to see a review where we would take one part or the other and see if you are on a budget, what would be best to safe on.

    Get a better processor, memory or vidcard for example. Other than that, thanks, it's nice to know that more is better. Though we already knew that.
  • 0 Hide
    velo116 , November 19, 2008 2:05 PM
    I don't understand why despite having 3GB of ram the games don't max out the 3GB. Why can't the same results be acheived with 4GB as all of the total windows memory usages are under 4gb??
  • 4 Hide
    scook9 , November 19, 2008 2:08 PM

    wow...really.... thats like intel taking a quad core and saying they work better. Why doesnt TOMS do a article on this, not the memory manufacturer.
  • -3 Hide
    kschoche , November 19, 2008 4:16 PM
    Could this have anything to do with the locality of the data in memory? Having to hit a memory stick physically farther away from the onboard memory actually affects performance and having higher density sticks would ideally reduce having to go farther away (stick 2/3 instead of 1) , but I guess not by *this* much??
  • 7 Hide
    nihility , November 19, 2008 4:25 PM
    Anyone else wondering if 4GB would be enough to see the same results? There are some articles about the performance difference between dual and tri-channel and they're really small.
  • 1 Hide
    Anonymous , November 19, 2008 4:50 PM
    Yeah i'd like to know if the same results occur with 4 GB of ram as i'm still rockin my LGA 775 q9550
  • 0 Hide
    Anonymous , November 19, 2008 6:07 PM
    For nihility: with 3x64-wide memory controllers, you want a multiple of 3 memory modules.

    I'm curious how it would behave with 12GB. I *know* that for the GIS processing I'm doing (IBM INtellistation/Opteron/Linux based),
    16GB is _lots_ faster than even 8 GB...
  • 2 Hide
    lopopo , November 19, 2008 7:06 PM
    "Just make sure you toss 6GB of memory into the mix for the best possible experience, according to Corsair’s recent research."

    enough said
  • -4 Hide
    seboj , November 19, 2008 7:17 PM

    wow...really.... thats like intel taking a quad core and saying they work better. Why doesnt TOMS do a article on this, not the memory manufacturer.

    QFT. Seriously, this should be on the footnote page for a test bench that Tom's actually does themselves, not a full article itself. When did we start caring what the manufacturers claim?

    There should be a disclaimer at the top of this page that reads:

  • 3 Hide
    Tjik , November 19, 2008 7:34 PM
    Thankfully I see some have reading skills because I fear that most won't notice that this isn't an independent test by Tom's Hardware.

    The title says: "Report: 3GB vs 6GB RAM on Core i7 Benched", suggesting this is a hardware related result. I however doubt that hardware plays such a big role here, since it just as well could be Vista related. Unfortunately Vista seems to lack ability to scale well in accordance to amount of RAM available. Looking at the results I'm more inclined to say we see an example of bad code.
  • 4 Hide
    tuannguyen , November 19, 2008 7:42 PM
    Hi All -

    Just wanted to chime in a quick note to clarify that this news piece is about a report done by Corsair engineers. It's mentioned in the article that it's not from us. If you find the information useful for buying decisions for Core i7, take advantage of it. If not, that's cool too!

    The report is about 3GB vs. 6GB memory. In some instances, Corsair's report even indicates that there's virtually no difference in performance gains -- in other areas, there are.

    Once you've made up your mind on whether double memory is useful, the best thing to do is do some shopping for your favorite brand and go from there.

    Corsair did not buy this coverage nor did it gift us for it. We just felt it was an interesting study between 3GB and 6GB memory on Intel's latest Core i7 platform -- not about the Corsair brand.

    / Tuan
  • -7 Hide
    Anonymous , November 19, 2008 7:46 PM
    Latest core i7 platform? there was an earlier one? COOL!!! lol
  • -1 Hide
    Shadow703793 , November 19, 2008 8:52 PM
    What I want to know is dose the RAM speed effect performance since it now has a IMC?
  • -2 Hide
    tsnor , November 19, 2008 10:07 PM
    1. Note vista 64 was used even in the 3GB test. Vista 32 would be a better, higher performing 3GB solution
    2. Dual Channel enambed with both the 3GB and 6GB configs? Or were 1/2 the dimms pulled leaving a single channel?
    3. As noted everywhere, this is not a Tom's review.
  • -5 Hide
    JonnyDough , November 19, 2008 11:21 PM
    Just make sure you toss 6GB of memory into the mix for the best possible experience, according to Corsair’s recent research.

    In other news:

    Ford says you should own at least five of Ford vehicles for the best driving experience, the National Dairy Association says you should be drinking 5 cows worth of milk a day for the best health, and Coke says that Pepsi is bad for you. Seriously?

    Why would you take any test results into real consideration that come from a company with a vested interest in selling you more of what they're saying you need?

    Marketing 101. If you believe everything you read please go to school and take it.
  • 0 Hide
    fulle , November 20, 2008 12:06 AM
    For the person that asked if dual channel mode was disabled on the 3GB kit... I would doubt it. Corsair wouldn't want to give the impression that their 3GB kit was inferior in performance to competitors. The only difference seems to be the amount of memory used. Its obvious that the information carries bias, given that Corsair makes memory, but the results are interesting.

    The positive effect on the minimum framerates in all the games seemed substantial. The Crysis Warhead FRAPS test was especially interesting to me.

    Can Toms be troubled to give us an unbiased set of benchmarks on this?
Display more comments