Good review, I am planning out my C2D system and or, dual Xeon (conroe) based system as my next build. The only question I am trying to get over is if I want SLi or not, seeing that I have not yet found a Dual Xeon motherboard that can run SLI.
Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Very fun game, and nice review. After looking at those numbers I had to go into the game and check what kind of FPS I was getting since I play at 1600x1200 with a second monitor and all settings except shadows at high( I am running a C2D @ 3.15ghz although only a 7900gt vid card) Lo and behold at was sitting at 20FPS without a whole lot on the screen(according to the in game command -- Alt-F11 ) I guess I just have never noticed
As for the crashing you experienced, I saw that too, but only when the SupCom exe would try to use more than 1.5GB of RAM. Apparently it is not able to address memory space above 2GB, and crashes when it tries to. This thread on the GPG Forums shows a simple way to mod the exe so that it can address memory higher than 2GB.
I've used it on my system and have had zero crashes since. It also doesn't interfere with Multiplayer play, since they aren't doing any sort of CRC check on the exe at connection.
Now I just need to get more than 2GB of RAM, since with this mod in place I've seen SupCom use as much as 2.9GB of RAM, which means it's paging like crazy, though it's only noticeable when I zoom in and out rapily from one place on the map to another.
odd choices of settings. my 3800+ with a single 8800 GTS is getting similar settings as the xp single core test at 1600x1200. also i've found setting priority to above normal helps IMMENSLY.
Excellent review. Dugg! The review answers almost all my wonders about how multi core the game really was. Especially the task manager bit!
A note about the task manager bit. According to ExtremeTech "Supreme Commander uses two main threads, one for the "simulation" and one for rendering. If a quad-core CPU or equivalent is detected, it will also spawn threads for audio and graphics driver management."
I myself was hoping it would split multi-cpu enemies (the "Simulation thread") across the cores, i guess we are not quite there yet in RTS games.
If anyone is interested in running their own canned benchmark, you can complete the following steps.
Change the shortcut to add /perf
"C:\Program Files\THQ\Gas Powered Games\Supreme Commander\bin\SupremeCommander.exe" /perf
then it'll run through a 4 cpu scrim.
then when its complete check your /bin folder in the SC dir.
Thank You!!
I have been waiting for solid performance numbers within SupCom to help answer the debate of Quad vs Dual for some time.
As thorough as your benching was, I would make one recommendation.
Most SupCom players such as myself only play multiplayer, and we are the community seeking these benchmarks more then anyone. However, in multiplayer games, there is no 'Computer Player AI' being crunched on other cores.
I state this due to the explanation of how SupCom supports multiple cores; it is not due to it being 100% fully multithreaded, but that it uses multiple sub engines, each of which which can be moved onto any core as a whole:
I would like to have seen some benchmarks of a multiplayer game where 'Computer Player AI' is never being added into the equation. Not many people are playing 6 player games with the AI; the AI is not very good in SupCom, and all but new players to the game could beat all of them single handedly.
This might make the Quad vs Dual core difference even more negligible for us in the SupCom multiplayer community.
Thanks again,
Terrific article! One thing that made me go "Huh?"--I noticed that at 1600x1200 with quad cores under XP you had TR MSAA enabled. At 2560x1600, you were using TR SSAA. How were you able to run a higher-quality transparency AA at a higher resolution? That's weird! Wonder if something in the latest drivers favors widescreen?
That's why you benchmark the replay.
Every game in SupCom is recorded for playback, and GPG even includes the 'replay vault' where you can watch some of the best players duke it out. Recent 6-way FFA's between the top 20 ranked players have had hundreds of downloads each and have become quite popular to watch.
GPG even includes one such multiplayer replay within the retail game for benchmarking - with a built in performance test! You must run the executable with the extension "/map perftest", and this will benchmark your system and spit out an overall score similar to 3D mark. This single-number result is not nearly as thorough as the results you produce however - and I would love to see that.
SupCom also includes a console command, ren_shownetworkstats. This shows the maximum smooth game speed that each person in a multiplayer game is capable of running the current game at that time. I have never seen anyone in multiplayer with a higher score than my 3.6C2D, which makes me wonder if moving to quad in another 6 months is worthwhile (and no this is not an intent to brag or show off e-peen, I built this system for SupCom after slogging through the beta on an athalon 64, and it still gets choppy in multiplayer 6 player games).
Again, you rock and I'm very appreciative of the existing benchmarks you've produced thus far.
I thing i am wondering is, could the choice of 2GB of memory affected the results?.
Its pretty well know that SupCom can used loads of memory. I have seen 2.7GB virtual size and huge private working sets of over 2GB.
Could that have possibly limited performance, especailly on large maps with lots of AI controlled units?
As for the crashing you experienced, I saw that too, but only when the SupCom exe would try to use more than 1.5GB of RAM. Apparently it is not able to address memory space above 2GB, and crashes when it tries to. This thread on the GPG Forums shows a simple way to mod the exe so that it can address memory higher than 2GB.
I've used it on my system and have had zero crashes since. It also doesn't interfere with Multiplayer play, since they aren't doing any sort of CRC check on the exe at connection.
Now I just need to get more than 2GB of RAM, since with this mod in place I've seen SupCom use as much as 2.9GB of RAM, which means it's paging like crazy, though it's only noticable when I zoom in and out rapily from one place on the map to another.
I have Vista, and I have 3 GB installed. I've seen it climb REALLY close (~1911 MB) to using 2 GB of memory, even playing games like Oblivion, R6 Vegas, and STALKER. But I've never seen it use more up to or over 2048 MB.
I'm not too sure if those fps numbers are representative of actual gameplay. I played the whole game at dual 1600x1200 with 4xAA and 16xAF on my 8800gtx and E6400@stock.
I'm also running Windows Vista and have played the game for hours on end with no crashes whatsoever.
I don't doubt you guys got the right fps numbers, but I think you need to redefine what 'playable' is for RTS titles, this isn't an FPS.
Didn't anybody notice the memory usage difference between Vista an XP? I suspect Vista needs 4gig to be efficient.
I've been SOOO angry with all the other sites who have done "performance testing" with SupCom and then using the built in demotest to present their results, a LAME proposition since SupCom doesn't really begin using multicore until WELL into a game against MANY AI's.
lolWill someone test with an AMD duel, I thought that Intel chips arn't really duel core vs Amd duel cores. Intel are two chips slapped together.
I'm not too sure if those fps numbers are representative of actual gameplay. I played the whole game at dual 1600x1200 with 4xAA and 16xAF on my 8800gtx and E6400@stock.
I'm also running Windows Vista and have played the game for hours on end with no crashes whatsoever.
I don't doubt you guys got the right fps numbers, but I think you need to redefine what 'playable' is for RTS titles, this isn't an FPS.
1ST off I would like to post, that I have admired the work that Kyle Bennet and the rest of the staff of [H] have done over the past several years. [H] has been in my book the goto sight for reviews for me, because I feel they are the most honest and ubiased reviewers on the planet.
My only issue with [H] and all of the rest is that the test beds for these systems are almost allways done using high-end components. The average builder (such as myself) will not spend $4000 on a gaming system.
I just built a gaming system for $1200, that I am very statisfied with, but when I want to find a review on like a 8800GTS 320, who is going to spend $300 on a video card, but still have a $950+ processor.
If you are going to test something that is midgrade, also do a test using a midgrade processor.
Test system used in this informative review. (all prices = NewEgg)
Motherboard: EVGA Nvidia nForce 680i SLI $249.99
CPU: Intel Core 2 Quad QX6700 $970.00
Memory: 2GB Corsair XMS2 Dominator CM2X1024-888C4D $253 (I could not find this ram anywhere. do I with with the CM2X2024-640C4D)
Hard Drive: Western Digital 74GB Raptor SATA/150 $159 (ORM)
O/S: Microsoft Vista Ultimate $249.99 (upgrade)
Dual BFG:GeForce 8800GTX 768MB (SLI) $819.99 - $1639.98
Case: ? Lets say $150
PSU: 1000Watt? Thermaltake Toughpower 850W W0131RU NVIDIA QUAD-SLI APPROVED? $259.99
Cost of test system: $3,931.95
Nice.
My only issue with [H] and all of the rest is that the test beds for these systems are almost allways done using high-end components. The average builder (such as myself) will not spend $4000 on a gaming system.
I just built a gaming system for $1200, that I am very statisfied with, but when I want to find a review on like a 8800GTS 320, who is going to spend $300 on a video card, but still have a $950+ processor.
If you are going to test something that is midgrade, also do a test using a midgrade processor.
Kyle, thanks for replying to my post.
Does [H]ardOCP sell any of the systems or components they buy for testing?
Good point. I personally think playable in an RTS would be 25-30 FPS. I might be off here but anything more dosen't really matter in an RTS does it?
Kyle, thanks for replying to my post.
Does [H]ardOCP sell any of the systems or components they buy for testing?
dual monitor mostly rapes your video card, and rapes it good. i've got a friend who ahs it setup and he jsut uses the second for a GIANT minimap as he can't use it for much else without getting 2 frames a second.
er... why would you do that? If destroyed as in opened to peek at the inside then I can understand.
I would be that in the case of things like ES processors and prototype boards, [H]ardOCP probably has some sort of contract that says they have to destroy them when finished. Prototypes often contain technologies that don't make it into the final product, and it wouldn't be cool if those technologies fell into the wrong hands.
er... why would you do that? If destroyed as in opened to peek at the inside then I can understand.
Up until now games have been mostly single-threaded. This means that they are only capable of taking advantage of a single-core on the CPU. If you have dual or quad-core processors the game will not utilize those extra cores efficiently to provide any gameplay experience improvements. Therefore multi-core CPUs have been pointless for gaming.
I would be that in the case of things like ES processors and prototype boards, [H]ardOCP probably has some sort of contract that says they have to destroy them when finished. Prototypes often contain technologies that don't make it into the final product, and it wouldn't be cool if those technologies fell into the wrong hands.
Pointless? Not if the other 3 cores are Folding. This is a pretty surprising oversight coming from the Web site of the #1 Folding team.
He said 'pointless for gaming', I don't think his intended statement could be any clearer.
I have Vista, and I have 3 GB installed. I've seen it climb REALLY close (~1911 MB) to using 2 GB of memory, even playing games like Oblivion, R6 Vegas, and STALKER. But I've never seen it use more up to or over 2048 MB.
I am sure others have pointed out their experience, but I haven't read past this post, yet. Pretty long thread, but good reading.
I have Vista and 3GB of RAM, and I play SupCom. On a 20x20 (medium) map, 500 unit cap (smallish), and 4 players on a multiplayer map that I hosted (one other human player, two AI), I got 2.3GB physical RAM usage and a 2.85GB swapfile. It seems you are not playing the "right" games for >2GB usage.
An 8 player 81Km by 81Km game with a 1000 unit cap per player might raise that 2.3GB number up a bit.
EDIT: I should have noted originally that all I did was crash before I found the Boris thread on the GPG forum for updating the BCD in 32bit Vista for >2GB user address space and using that utility to add the >2GB aware flag to the SupCom .exe. That was an awesome contribution he made. I figured I should give credit where credit was due.
Do you need to reboot after applying the fix?