http://img706.imageshack.us/img706/3918/xmpexample.jpg
Get cpu-z if you don't already have.

Example of xmp on my machine, swap my numbers for yours. :


xmp-1866
1866 is the mhz I set my ram at in bios

The following will be under ram timing in bios.
This is what I put from looking at my xmp.

Cas 8
Ras to cas 10
Ras precharge 10
Tras 26
trc 33

2T (your bios may call it "CR" or "command rate")

This will be under Dram voltage in your voltage area.
In my case voltage = 1.6v

IF you opt to use the 1T profile you may lose stability, if you don't it's a small gain.

Bump your ram to higher xmp profile if not already.
Upgrade to ssd if not already.

Can't do much with laptop mang.

With that gpu, this might help.

http://www.sinisterswarm.com/forums/showthread.php?t=16086

Turns out the reason my cards were running @ 100/60 was a downclock bug on my second 5870's oc software.

After some serious tweaking I was able to use my crossfire @ 100% to make tera look better and retain a constant fps with minimal stuttering in town, none outside.

http://img841.imageshack.us/img841/5034/settingsb.jpg
http://img59.imageshack.us/img59/9167/teracustom.jpg
(That was my fps low spot before)

Fear.exe is my base xfire profile

Forcing vsync and triple buffer with d3doverride, maintain a constant 60ps (60hz monitor).
Ambient lighting 1, fxaa off, replaced with SweetFX.
enabled Atitexturefilteroptimizationchecks in ini
diabled core parking on my rig
specified CPUNunlogicalprocessors=8 (not 100% sure here if improvment)

Stopped my old stuttering problems in town by disabling framesmoothing and mouse smoothing.
Disabling real time optimization also helped to a lesser degree.

While AA can go higher, reserving the rest of my gpu power for sweetfx additions.

Oh yeah and there's no flickering anymore, no longer using radeonpro for profiles.

The packet loss was a illusion caused by checksum calculation post wireshark.
Capturing traffic locally with network card optimizations enabled.
The fps issue was my 5870 downclocking itself because of bad OC software, also fixed.

In the process I figured out how turn off packet throttling from windows media sharing.
Thanks for the replies, even if they were a lil late .

I'm more a hardware/driver specialist. I've come at the issue of my fps fluctuating 25 every 2 seconds when around other people from all hardware/driver standpoints to no avail.

I tried running wireshark and and every third packet is black with red writing.

Error :
"Header checksum: 0x0000 [incorrect, should be 0x18ea (may be caused by "IP checksum offload?)"

Info :
"54 49798 > scp-config [ack] seq=1252 ack=113975 win=260864 len=0"

I'm probably wrong cause this isn't my field but, does this mean I'm recieving packets in the wrong order?

Research so far is pointing toward :
Error is result of capturing my own traffic locally with network card optimizations enabled.
If this happens after capturing, wireshark is simply reporting it as a error when it really isn't.

Guess back to the drawing board for fps instability only when around other people.

Got another 7-10 fps average and got rid of dips by setting tera.exe affinity to core 4 and 6.
Recommend to anyone using dulldozer/piledriver chips.

(If your running stock and peg your cores assign 2 also)

This thread sparked up my competitive ocing drive. I am not a amd fanboi, I like using whatever works best in given situation. On the flip side I most enjoy doing things people claim can't be done.

So....
I took those 2x5870's in crossfire and built a amd system with higher NB oc capability around them to prove it's software settings/NB oc for amd. 8350+asus crosshair v formula z.

D3d override for triple buffering dx, xfire Afr designated in application setting for tera.

http://imageshack.us/a/img705/796/specsm.jpg

While the benchmark is clearly much lower than say a 3770k or 3930 oced (mainly due to single core cpu physics test), I have no cpu bottleneck holding me down.

My frames are 60-80 with max everything spare this one random spot in velika trade district (dunno). Nexus just can't be done @ 60fps with the way this game is coded imo. Testing without my second card present my fps drops to 40-55.

The point here is a amd crossfire rig can infact play tera as "perfectly" as anything can and on top of that I'm running a Minecraft and UT server.

This chip is by no means oced as far as it could be. I'm running 1.332. vcore, the rated max is 1.55 and there's a handful of people running it at 1.75 with extreme cooling.

Next goal:
Going to order 2x more of my ram chips and try loading tera into a 28gb ramdisk.

I've noticed the more I tweak my software for fps, the more frequently I experience random player models kind of flickering. Not whether to call successful experiment yet. This is definetly a tera and not cpu stability issue.

Final verdict I'm calling this a partial success. The 2570 NB made a huge diff on amd cpu and got me to the 60+ mark, but I was unable to get my crossfire to more than 100/60 load balancing without player characters flickering and this took some pretty janky driver software mods. Amd fine, crossfire bad.

I have my system pretty tweaked for tera.
Custom app settings, D3D override, crossfire functioning well enough.
Went 8350 because I'm hosting multiple servers too.

I get 60 fps everywhere except this spot (obviously excluding nexus).
Does everyone lag here?
http://img688.imageshack.us/img688/4506/laggyspot.jpg

Rig : http://imageshack.us/a/img705/796/specsm.jpg
cards @ 875/1250

Sup with this spot?!

http://www.overclock.net/t/884103/1055t-overclocking-guide

Your ram timings will likely need to be higher than mine unless your using Gskill pi series 1600mhz6cas. You don't need them anyway, I only use them to brag about superpi results.

Your not gonna get 60fps in a crowded area on this rig, but you should at least be able to run 40+ if you ebay another 6950. OR save up and buy a 7970.