SimHQ originally intended to publish the benchmark results below as a technology article. However, the results aren't really that surprising and it's probably not worth the effort it would take to format such a piece. Still, the numbers are interesting and useful, so they have been posted here as a "mini-article". Please use this thread to comment.
Test SystemCPU: Opteron 148 @ 2200 MHz
Motherboard: Asrock 939Dual-SATAII
GPU: 256MB PCI-E 7900GT using Forceware 91.33
Memory: 1x, 2x, or 4x 512MB 400MHz DDR (Corsair XMS3200 v1.2) (1x512 is a single channel configuration while 2x and 4x512 are dual channel configurations)
Sims tested- IL-2 Forgotten Battles + Pacific Fighters + Aces Expansion Pack 4.04
- Lock On + Flaming Cliffs 1.12a
- Falcon 4: Allied Force 1.07
- Wings Over Vietnam 07.10.06
- rFactor 1150
Test settingsAll sims were tested at 1280x960 (exception: rFactor tested at 1024x768) with 4xAA and 8xAF forced on for all tests. The goal was to place a reailstic but not excessive load on the video card, in order to recreate realistic game settings. RAM was set to two different speed/latency combinations; the first was 266MHz with 3-6-6-15-2T timings (representative of older memory), and the second was 400 MHz with 2.5-3-3-11-2T timings (representing current average DDR settings).
TestsThe following were evaluted for each of the five sims tested:
- Sim load time from the desktop to the sim's user interface
- Mission or track load time from the user interface to the game's 3D environment
- Average/minimum/and maximum framerates during a session, or instantaneous framerate in-game
RESULTS: FB+PF+AEP 4.04 load time from desktop1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
46 seconds1x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
36 seconds2x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
25 seconds4x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
24 secondsRESULTS: FB+PF+AEP 4.04 Black Death track load time1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
26 seconds1x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
26 seconds2x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
11 seconds4x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
11 secondsRESULTS: FB+PF+AEP 4.04 Black Death track average/minimum/maximum framerates1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
23/3/401x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
29/11/482x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
32/12/514x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
37/13/51The IL-2 series sees a lot of improvement from a move from low- to high-speed RAM; there is a significant decrease in game loading time and a significant increase in minimum framerate. A jump can also be seen from 512MB to 1024MB; game load time drops again and this time track load time does as well. Framerates don't change much, however. Moving from 1024MB to 2048MB results in very little change anywhere.
RESULTS: Lock On + Flaming Cliffs 1.12a load time from desktop1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
29 seconds1x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
29 seconds2x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
25 seconds4x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
24 secondsRESULTS: Lock On + Flaming Cliffs 1.12a MiG-29 demo track load time1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
76 seconds1x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
53 seconds2x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
21 seconds4x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
21 secondsRESULTS: Lock On + Flaming Cliffs 1.12a MiG-29 demo average/minimum/maximum framerates1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
23/3/401x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
44/17/862x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
45/24/874x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
47/25/109While Lock On's game load times remain relatively unaffected by memory parameters, massive improvements can be seen in mission load times when increasing the speed of 512MB memory and moving to 1024MB. The same can be said about framerates; minimum framerates increase from 3fps to 17 fps and then to 24 fps. Neglible improvements are seen from a move to 2048MB. Lock On is an extreme example of a game that has a 1024MB "sweet spot"; a 53-second mission load time compared to a 21-second load time is the difference between twiddling your thumbs/rolling your eyes and a reasonable wait.
RESULTS: Falcon 4:Allied Force 1.07 load time from desktop1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
32 seconds1x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
26 seconds2x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
24 seconds4x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
24 secondsRESULTS: Falcon 4:Allied Force 1.07 Day 1 Balkans H-hour +2 SEAD escort mission load time1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
35 seconds1x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
40 seconds2x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
25 seconds4x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
25 secondsRESULTS: Falcon 4:Allied Force 1.07 Day 1 Balkans H-hour +2 SEAD escort mission instantaneous framerate in cockpit on runway1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
201x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
282x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
284x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
29AF shows a slight improvement moving from slow to fast memory, but beyond that not much effect can be seen. The CPU likely is the limiting hardware here.
RESULTS: Wings Over Vietnam 07.10.06 load time from desktop1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
17 seconds1x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
17 seconds2x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
9 seconds4x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
8 secondsRESULTS: Wings Over Vietnam 07.10.06 Operation Bolo single mission load time1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
17 seconds1x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
19 seconds2x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
17 seconds4x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
18 secondsRESULTS: Wings Over Vietnam 07.10.06 Operation Bolo single mission instantaneous framerate in cockpit in flight1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
551x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
672x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
724x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
73A large decrease in load time can be seen moving from 512MB to 1024MB of RAM, and framerates jump when increasing RAM speed as well as moving from 512MB to 1024MB. However, it's doubtful you'll be able to tell the difference between 55fps and 67fps. WOV also had the shortest load time of any sim so an improvement here is not all that important. There is no significant improvement moving from 1024MB to 2048MB.
RESULTS: rFactor 1150 load time from desktop1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
17 seconds1x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
17 seconds2x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
9 seconds4x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
8 secondsRESULTS: rFactor 1150 BMW Sauber practice @ Silverstone (Northamptonshire) track load time1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
17 seconds1x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
19 seconds2x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
17 seconds4x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
18 secondsRESULTS: rFactor 1150 replay of Lupos @ Lienz Alstadt average/minimum/maximum framerates1x512MB (512MB total), 266 MHz, 3-6-6-15-2T:
27/14/431x512MB (512MB total), 400 MHz, 2.5-3-3-11-2T:
27/14/432x512MB (1024MB total), 400 MHz, 2.5-3-3-11-2T:
27/14/434x512MB (2048MB total), 400 MHz, 2.5-3-3-11-2T:
27/14/43Out of the five sims tested, rFactor responds the least to memory changes. Track load times and framerates were unchanged for all memory configurations, and only the game load time improved when moving from 512MB to 1024MB. Just like WOV, this is already a short load time.
CONCLUSIONSignificant improvements can be seen in sims when moving from 512MB to 1024MB of RAM. Very little to no improvement is typically seen when moving from 1024MB to 2048MB of RAM. Although there are no doubt specific situations where 2048MB of RAM is helpful, at this time it should not be considered a "must-have."