Wednesday, April 23, 2008

Detailed AoC hardware and performance analysis.

Hello all. There are quite a few questions here about performance and hardware. Hopefully this thread will help answer them.

After finding out that I got into the PvP beta, I quickly called my friends to tell them to check their e-mails. After finding out that we ALL got into the beta, we hightailed it to a remote cabin for some fishing, drinking and AoC.
Here are the rigs used in these tests.

Rig I – Me
Q6600 @ 3.6 GHz.
GeForce 9800 GX2 @ stock. (Later OC’d to 675 to compare to the Dual Core, then taken back down to 600 because I don’t want to use my computer as a heater.)
Vista 64 bit.

Rig 2 – Mike
E8400 @ 4.0 GHz
GeForce 9800 GX2 @ 675 MHz
Vista 64 bit.

Rig 3 – Stev-o
Pentium 4 @ 3.6 GHz
GeForce 6600 XT @350 MHz

Rig 4 – Meng
Phenom 9850 @ 3.0 GHz
Radeon HD 3870 @ 900 MHz
Vista 64 bit

Rig 5 (Laptop) Stev-o
T9300 @ 2.8GHz
GeForce 8600M @ stock.

First. There IS no DX 10 support at the moment. None whatsoever! You are looking at DX 9 results!

Let’s start with the install.
The Q6600 finished first, followed closely by the E8400, the 9850 a bit later, followed by the T9300 and finally the mighty Pentium 4!
All in all, the install went smoothly on all machines.

Settings (in the beginning tutorial jungle area.)
Mike and I ran the game at 1600 by 1200 with EVERYTHING maxed. (Filtering, Shadows, Bloom). The game ran flawlessly and never dropped below 60 FPS! If we would up the resolution a bit but keep the settings as they were we would both experience some stiffness (character movement was a bit off, seemed like commands took a second to respond, and FPS drops). Toning down Bloom and Shadows we were able to max the resolution and still have FPS in the 60+ area. The game looks and plays beautifully. Since we wanted to play at above 60 FPS, with everything cranked up that is the resolution we stayed at.

Meng had some really weird problems with Shadows. Whenever he would have them on, he would get weird stretch marks across the screen and some weird shapes would pop up all over the screen in yellowish and or green colors. The fix was to turn the Shadows completely. After he did that, the game ran great. He was also running it at 1600 by 1200, although on average at around 52 ish FPS. Never dropped below 50, never went above 55.
We spoke to a few Phenom and 3870 users that said they were having the same problems with shadows. So heads up. If you have a Phenom and a 3870, you might be in for a nasty surprise. When we swapped his 3870 for my GX2, the problem went away. When we plopped his 3870 into my rig, I did not have the problem either. Seems it is a combination of the Phenom and the 3870 that causes this. Hopefully Funcom fixes this soon… Of course it could also be a driver issue…

Steve’s laptop did a great job running the game as well. He ran it at 1680 by 1050 with everything maxed but the shadows! His FPS hovered around the same mark as Meng’s.

Now here is where we all got a nice surprise! Steve’s P4 machine managed to handle the game incredibly well! At 1024 by 768 without Bloom and with low-medium settings he never went below 40 FPS! However Bloom and higher quality shadows really killed the system, making the game virtually unplayable.
But we decided to turn that stuff off and see what we can do (Shadows at low, Bloom off completely. AA at first setting). We managed to get it playable at 1280 by 800 and at 40+ FPS (without any of that “eye candy” very cool considering the system is ancient.) With further tweaking and turning things off we saw pretty awesome potential in this engine. You could probably get it to play at even higher settings if you want to go into the 30 or so FPS range… We didn’t feel like it, since this machine was here just so we can see how it plays on old hardware.
But if you disable shadows completely you will have a HUGE FPS gain, especially on older machines. Turn off Bloom and you can up the resolution to some crazy number… Filtering should remain on as the game has some jaggies without it, although they are much less noticeable at higher resolutions.

Settings (in PvP).
Currently there is no large scale PvP… Disappointing, I know. There is no AV type map where you have hundreds of people clashing into each other with a ton of spells going off at the same time. As it is, the FPS in PvP seems to be close to the PvE portion of the game. However like I sad, at the moment PvP is limited to just a few groups so I don’t know how this will hold up in sieges. Surprisingly even with multiple people throwing spells at us; our FPS stayed the same as it did during the “single player” portion. Very good job Funcom.
So as you can see the engine is very scalable. Able to play on low and high end systems.

Dual Core VS. Quad core.

The Phenom is disqualified from this because it performed sub par to my Q6600, even when we swapped videocards... My Q6600 spanked it silly (10+ fps at the same settings, using the same videocard. The performance gain of my Q6600 increased as we upped the resolution, filtering and quality. In the end the Q6600 took the lead by 17 FPS in the starter area at max settings and high resolution.)
So let’s begin!

In the left corner is the mighty Q6600 overclocked to 3.6 GHz, in the right corner is the E8400 clocked at a beefy 4.0GHz.

At resolutions prior to and including 1600 by 1200 there is no difference whatsoever. We had the same fps. Taking it to higher resolutions, I was hovering closer to 54 ish FPS where as Mike was really close to 60. So I OC’d my video card to match his and voila! The FPS was once again the same! This is where we said “To hell with it. Time to crank up the resolution to the max setting possible and go wade into the jungle! 1920 by 1200 here we come!” and so we did.
Finally his E8400 showed its strength. On average at this resolution he would constantly beat me by a good 5-8 FPS. We were both still above 30 but you could notice the difference in playing at 60+ and then dropping down to 40ish. However if we had not seen the game at 60 FPS, than 40, and even 30 FPS would have been more then enough to satisfy us.

It looks like MHz mean more to the game than multiple cores (if you have at least 2). My guess is that it is NOT in fact optimized for Quad Core machines. My 3’rd and 4’th cores were NEVER used above 20%. Honestly this is a big disappointment. The game IS made for multi core computing, but it seems to be optimized for Dual Cores at this point and time.

64 BIT OS VS. 32 BIT OS.

Since the Phenom and the T9300 performed so close to each other, we decided to do some torture tests. We lowered the Phenom to a lower clock speed than the T9300 (because Phenom is a Quad, we clocked it 300 MHz lower then the T9300 to do a good comparison) and dug up an old GeForce 8600 GT. (Although this had more RAM and a slightly higher clock speed then the M version. So we messed with the bios of the GT to make it identical to the M version and then flashed it. Tada!)

Now we took the game to a neutral resolution of 1024 by 768 and then upped it as we saw fit. Surprisingly the extra RAM and the 64 bit version of Windows did not do anything. Once we reached the higher resolutions, the FPS started to steadily drop on both machines, at pretty much equal increments! (1-2 FPS differences at most).
So all in all, the game does NOT take advantage of more then 4GB of RAM and 64bit computing.

So there you have it. 2 Funcom myths debunked. The game does NOT run faster on 64 bit machines, and the game does NOT run faster on Quad Cores then on Dual Cores.
On the plus side surprisingly the game does an excellent job of running on low end hardware and still looks great.

If you have any requests on resolutions and rigs you want us to try, post here and I will try to oblige. We all do have work tomorrow so we will be calling it quits in a few hours.

For now we are going back for more PvP. I will be writing up a review of the game some time this week as well. So far it looks great. Yes I am disappointed in a few things (no soul corruption or spell weaving, no bounty hunting and you can’t move while comboing… Making you a sitting duck for ranged classes… Etc, etc.) But the game is still just plain fun. Hopefully a few patches will make it even better. I mean hell they fixed class balance in pretty much 1 day! So I’m sure they can pull this off. Expect a full review some time this week.

Beta Testers, please list your rig and the settings you used and I will also update this post to match your findings. This way we have a 1 stop shop for hardware info.

If you enjoyed this post, make sure you subscribe to my RSS feed!

0 kommentarer:

Post a Comment

Age of Conan Cheats © 2009