The WAR on snake-oil tweaks

A PC enthusiast with a repressed penchant for Macs, Matt Butrovich joins TR as a guest blogger. He’ll be sharing his thoughts on all things tech, from digital photography to gaming.

Gamers will do anything to squeeze a few extra frames per second from their computers. With the recent releases of several high-profile PC games, the echo chamber effect for what I would call snake oil fixes has increased dramatically. I’ve been playing a good amount of Warhammer Online: Age of Reckoning (WAR from here on out) lately, and like most new PC games, it’s got some room for optimization. When I began playing on what I would consider a high-end machine (at least for an MMO), I consistently saw frame rates drop to nigh-unplayable single digits during heavy combat, but my attempts to locate the bottleneck went unsolved. CPU usage wasn’t being maxed out on any of my four cores, and there was still plenty of physical memory left available. Lowering my visual settings and throwing a modest overclock at my video card did practically nothing, which led me to believe graphics performance wasn’t the root of the problem, either.

Not willing to give up, I decided to hit some forums to see what other players were saying about performance issues. Unsurprisingly, I discovered I was not alone, and I found hundreds of posts dedicated to slowdown and stuttering problems without a reasonable explanation in sight. Now, I had no problem accepting that the game just needed more optimizing from the developers—after all, MMOs are notorious for the incomplete that they ship in. “Just be patient. Give it a few weeks. Wait for Mythic to roll out some more patches.” That’s what I told myself to get by. But then, I started seeing posts entitled, “increase your fps by 20+!” and “Nvidia choppy/low fps fix!!!” I’ll be honest: curiosity got the better of me.

I went into the discussion threads expecting very little, which turned out to be a wise approach, since the “solutions” did nothing but make my head spin. Almost every single one was a snake-oil fix, with questionable advice and unverified gains. Gamers were digging through the Windows Control Panel changing just about any option they could get their hands on, then coming back with reports of frame-rate or latency gains without any quantifications. I skipped past anything that had a link to Black Viper or otherwise recommended disabling system services, and I ignored anyone who advised disabling something but couldn’t actually tell you what it did. One of my favorite tips involved disabling your network adapter’s TCP checksum offload capability, supposedly in order to decrease latency and disconnects. For those users whose NICs didn’t have that capability in the first place, a handy registry key was provided for them to disable it anyway—that way, they’ll be doubly sure no checksum offloading will be going on!

Attacks on SpeedStep and Cool’n’Quiet were rampant, as well, with gamers telling each other not to let energy-saving profiles steal their precious CPU cycles, even though those technologies have minimal (if any) impacts on system performance. I’m sorry, but I’m not about to run my 140W TDP processor at full speed 24/7 just because some dude who apparently doesn’t have a shift key told me to via an online forum. Windows Vista’s User Account Control was another sacrificial lamb, not so much for performance gains, but because WAR requires elevated rights to run (presumably because of PunkBuster). The overriding consensus was to disable UAC with no thought given to the consequences this might entail for the largely uninformed users following these directions.

Moving past the remaining “fixes” that I deemed downright dangerous or damaging (flashing your video card BIOS with directions provided by XxLegolasxX), I landed on the slightly more rational but still out-there tips. To give you an idea of what I mean, I found one thread with thousands of views that said Nvidia users could create a driver profile to dramatically increase WAR frame rates. I was already forcing triple buffering, vsync, 8x anisotropic filtering, and 4x antialiasing in my global settings whenever I played WAR, which worked just fine despite the thread saying WAR didn’t respect those settings. How would a custom profile generate performance benefits over identical global settings? Curious, I went ahead and copied my global settings into a new profile for war.exe and fired up FRAPS. Guess what? There was no frame rate change. I’m not about to call the dozens of posters that are reporting benefits liars, but it certainly did nothing for me, and I haven’t seen any evidence supporting the argument other than, “It runs better for me!”

After an evening of entertainment provided by snake oil fixes, I decided to call it a night and admit that the game simply had room for optimization. Not all hope is lost though. Since I began playing Warhammer Online about six weeks ago, I have actually managed to increase my frame rates a significant amount. Though many of you likely practice these methods, I’ll fill you in on my secrets anyway:

  • Keep Windows up to date via Windows Update. This is sort of a no-brainer: Microsoft is continually rolling out patches that increase the stability and performance of Windows. If you don’t have Automatic Updates on, at least remember to check every month or so.
  • Keep your device drivers up to date, especially those that relate to gaming like graphics-card and sound-card drivers. Nvidia released new WHQL-certified drivers not long ago, and those drivers managed to improve my frame rates and stability in WAR significantly. I have to imagine more improvements like those are on the way.
  • Keep your games up to date. The majority of WAR‘s almost-daily patches over the last several weeks have been designed to improve client stability and performance, and in my experience, users are seeing benefits. This also means reinstalling with the retail version of WAR instead of the updated beta client. While the code-base should be the same, some users report a performance benefit after switching to the retail client. I honestly believe that any performance changes are caused by the game files being less fragmented after a clean install, but if you’ve got the 20 minutes it takes to make the switch, it can’t hurt.
  • Keep your computer well maintained. That means watching out for malware, keeping your hard drives defragmented, and keeping non-critical processes (like instant messaging apps, media players, etc.) to a minimum while gaming. Also remember to pop your system’s case open every once in a while and make sure all components are clean and working properly. A dusty computer leads to increased temperatures, and hotter components may throttle themselves down to slower speeds in order to remain stable.
  • Make sure your machine is configured properly, particularly if you assembled it yourself. This includes verifying that the CPU, RAM, FSB, and video card are running at their proper speeds and storage devices are running in their fastest mode (DMA for parallel ATA devices and AHCI for SATA ones if your hardware supports it).
  • Upgrade your machine. Be realistic about what to expect with regard to the minimum/recommended system requirements, and see what other users are reporting with their computers. Upgrading is easier said than done for some people, but there are several inexpensive ways (like adding more RAM if necessary) to improve your computer’s performance without breaking the bank.

With all of that said, the only real solution to improving game performance without throwing money at the problem is patience, particularly if you’re looking for fixes for a recent release. Spore will continue to get patched, as will Crysis Warhead and WAR. Until then, keep tweaking your BIOS, keep tinkering with .cfg files, but most of all, keep playing. That’s what being a PC gamer is all about.

Comments closed
    • Shining Arcanine
    • 11 years ago

    Matt Butrovich, slowness in a multithreaded application when all cores are not being fully utilized seems to indicate a problem in the synchronization of the multithreaded algorithm. In particular, either one thread is locking a resource other threads need and it is not unlocking it fast enough or far, far too much locking/unlocking is being done.

    Multithreaded algorithms needs to be designed so that the individual threads can operate independently of each other for significant periods of time, rather than insignificant periods of time, as multithreading is a coarse form of parallelism that does not allow much interaction between the individual threads.

    Regardless of how your system is configured, I assume that the developers did a very poor job of multithreading their algorithms and if I had to guess what they poorly multithreaded (not being a games programmer myself, but having a professor who is), I would say that it is the collision detection.

      • sdack
      • 11 years ago

      I once saw how somebody started a thread for each object in its application, and ended up with thousands of threads.

      Half of all programmers only have knowledge of software techniques but none over hardware architectures. Only because the hardware is fast enough do these programmers never care about it at all.

      Eventually you will find an application that sticks out of the mass and you think, “Hey! There is an entire team of competent programmers.”, or maybe it was written by just one or two people who know exactly what they are doing.

      Everyone then wants to by like these “gurus”, who in other industries you would only call “skilled” and “talented”. You find a lot of these wannabes in the game industry where almost every new game needs to have its own 3D engine but all other parts of the game come from all around and are being knit together carelessly.

      Typical “over-performers” (that is what I call them) then throw in expensive sound tracks and video scenes, delay the deadlines by years and when a game is finally being released (and for a record breaking price) have completely forgotten to offer modding tools. Some game studios close after many years of what seemed to be a success because they have never made any profit.

      In the end, software developers have become an overrated, overpaid and little caring group of workers, and with incomes continuing to fall.

    • bcronce
    • 11 years ago

    “@cd “C:\Program Files\Bethesda Softworks\Oblivion”
    @start “Oblivion” /b /realtime “C:\Program Files\Bethesda Softworks\Oblivion\Oblivion.exe””

    I’ve had issues before when a program wasn’t coded correctly and while running realtime it would hog all the cpu and causeda catch 22. It ate so much cpu the network stopped working, but it was spooling for a response from the network drivers.

    In most multi core systems, this wouldn’t be an issue since the network driver would probably be offloaded to a cpu with more “time” on it.

    Defragging only helps load times, unless you’re low on memory and constantly reading the HD

    • Jigar
    • 11 years ago

    Welcome to TR, Matt. 🙂

    It’s good to see you here. BTW thanks for the tips. 🙂

    • swaaye
    • 11 years ago

    Good tips. Some game forums have people that know what’s going on (and know how to actually solve problems) and others are loaded with what you ran into. 🙂

    Just a little note: AHCI isn’t going to bring performance gains in game rigs, but it could bring headaches. I wouldn’t recommend people mess around with AHCI unless they need hotplug support (if you use eSATA, say). It also brings native command queuing, but that is useless for single-user workloads.

    AHCI can be trouble with some combinations of OS, chipset, motherboard and/or drives. It usually adds boot time because the BIOS will detect the drives in a different manner. And, it could even result in slower drive performance if the drives don’t support NCQ properly. This is one of those instances where keeping it simple (by sticking with SATA’s legacy mode) is almost always the way to go.

      • sdack
      • 11 years ago

      I would like to disagree on the subject of command queuing (CQ). There is no such thing as a single-user load these days any more. People have multiple applications running and lots of smaller background tasks at the same time. They often come with auto-save features, disk caching, file logging, etc. and are multi-threading. The I/O pattern of the single user is changing into a pattern similar to that of multiple users. Even now while I am writing this am I seeing disk I/O of ~10kB/s on my single user desktop machine. CQ does not give a huge boost but instead was just simple to implement. It got worse when some hard disk makers implemented it badly, gathering data to make CQ more effective, delaying it and causing a performance drop in some situations. If implemented right does CQ not cause any delays but leaves the disk device a chance to reorder operations according to the head movements. The higher the disk I/O gets the more often CQ has a chance of improving the performance. The reason why CQ still gets recommended for multiple users only is because multiple users cause more disk I/O. Everyone should get disks with CQ, really. The technique is not to blame but instead the makers who tried to sell it as the next best feature and messed it up in the making.

        • swaaye
        • 11 years ago

        I stand by what I said unless you can conclusively prove me wrong with some test results. NCQ is definitely one of those technologies that people say they can “feel” but are instead feeling the power of placebo. The gains to be had are very minimal (if there are any) with desktop workloads. On the other hand, the potential problems and complications with AHCI are definitely real.

        Thankfully, we are almost done with needing “performance” from mechanical drives with the coming cheap SSD solutions.

          • sdack
          • 11 years ago

          Before one can talk about placebo effects one needs to understand them. It is a known fact that one can only notice a change when the difference is 15% or more. I.e. when you are driving a car at 50mph and drive the same road again with 57mph will you be able to tell the speed difference without looking on the meter. These 15% apply to a lot of situations and only trained people can notice smaller differences.

          The placebo effect can be measured and predicted. Overclock your computer by 5% and some people will tell you it is faster. They however only know this by measuring the difference but not by feeling it. Sometimes a gain smaller than 15% can be noticed because of side effects. I.e. when stuttering in a game disappears. You will not be able to tell the difference just by looking at the frames but by noticing the missing side effect. The same is true for the example with the car. Most people will not notice if the car is going only 10% faster. But when they cross rail tracks they can notice a difference in the sound it makes as the ears provide additional information to what they see and thereby can tell the car is driving faster. Once the sound is gone it again becomes too difficult to tell.

          Now most new features that come with new hardware bring often as little as 1%-5% of an improvement. Each one on its own cannot be felt. Install a memory module into your computer with low latencies and you will not be able to notice it as the gain is often not more than 5%. Does this mean its not worth it or it is a placebo? Yes, for some it is. However, what matters is the sum of improvements you put into a system and once you get above 15% it can be felt. As such is CQ about as important as all those other little features you can invest in.

    • NeronetFi
    • 11 years ago

    2 words……”Memory Leaks” been seeing this in alot of the new games.

    • sdack
    • 11 years ago

    After reading the report I felt just like the author – I had read another report on performance problems (with this one here being written by a professional).

    So here are some tipps that work for me under Windows XP (I am still new to Vista but it should help, too):

    r[http://www.tweakguides.com/.< ]§ These do not work all, of course, but it is a great source of information about what can be tweaked in games via their .ini-files for instance. Another site §[<http://www.speedyvista.com/services.php< ]§ explains the Vista background services, which I still need to learn about, but it appears to be a good place to start. One more and a useful read in case you need a checklist: §[<http://www.pcstats.com/articleview.cfm?articleID=2238<]§ There is however no single switch that does it all-in-one, instead tweaking is like garden work.

      • Meadows
      • 11 years ago

      It’s called placebo effect. Read up on it.

        • sdack
        • 11 years ago

        Oh, really? It is also called experience. Keep reading up on it or start experimenting on your own (you probably are too lazy for it).

    • Anomymous Gerbil
    • 11 years ago

    Defrag, really? A lot of people would argue that makes zero difference.

      • Meadows
      • 11 years ago

      MMOs with a huge world and a voluminous player base are bound to use a lot of memory. With it comes a lot of caching, paging, or both – much more than you’d see with the average game. Defragging probably doesn’t help the framerates, but it helps lessen the spikes a little.

    • MadManOriginal
    • 11 years ago

    Hi, welcome to TR. Ignore Meadows and the other resident flamebaiting gotta-win-opinion-arguments-on-the-internet folks.

      • CasbahBoy
      • 11 years ago

      I don’t consider Meadows so much flamebait as is /[

        • eitje
        • 11 years ago

        I can’t think of a time that he’s provided something constructive in a front page thread. It’s always shooting down someone else’s idea; never putting forth his own, stand-alone opinion.

          • Meadows
          • 11 years ago

          Now now, that’s not true and you know it.

      • PerfectCr
      • 11 years ago

      Damage has shown his willingness to deal with front page trolls in the past. I have no doubt if he continues down this path he’ll be dealt with accordingly.

      Oh and welcome, great post! 😀

      • Meadows
      • 11 years ago

      I like you too!

    • grantmeaname
    • 11 years ago

    Hi Matt!

    Adisor might try to bring out your hidden mac lover. Don’t say you weren’t warned ;-).

      • Meadows
      • 11 years ago

      You somehow reminded me of the South Park episode about “bi-curiosity”.
      Hmm.

    • eitje
    • 11 years ago

    i wonder if it could be a network issue, and not graphics or CPU at all. maybe the lag is because they have their scene rendering attached to their networking, such that “never get out of sync” is more important than “render 40 PCs and 80 NPCs in real time”

      • ssidbroadcast
      • 11 years ago

      I doubt it.

      • Meadows
      • 11 years ago

      You can’t have rendering attached to networking. Can you even imagine the amount of information your computer plays with every second to render a world for you?
      If you were to handle this with network cards, /[

        • ztrand
        • 11 years ago

        Of course you can have rendering attached to network info. It’s just not optimal. A simple rule like “wait for 64-byte global server-state udp package before rendering next frame” or “never render more than 200 frames without update from server” might be good/needed in certain scenarios but will cause frame-rate lag in others.

          • Saribro
          • 11 years ago

          Wouldn’t something like “render attack-action, then wait ’till server tells you what happened” be more sensible than do it in X amount of frames?

            • ztrand
            • 11 years ago

            i’m sure it would. I was just making up stupid examples to illustrate a concept.

        • eitje
        • 11 years ago

        i think you took me too literally. ztrand has the idea.

    • steelcity_ballin
    • 11 years ago

    Playing with max settings 4xaa 16x af, and adaptive AA with latest cat drivers on a 4850 with no issues. I don’t think I’ve ever dipped below 30 fps save for massive keep sieges and 100+ pc on the screen.

    I used to get bad lag in Morkeins template, haven’t in a long time, running smooth in Tor Anroc now.

      • Steba
      • 11 years ago

      i’m playing with similar (slightly higher) settings and at 1920 x 1080p and i rarely have a problem either. i do admit the scenarios lagged horribly at first. However, after 1.0.1, they became smooth as silk for me. I use a 9800 GTX SSC.

      as pete stated, the only time i see any lag now is during heavy open world RvR with many many players such as keep sieges. I’m sure there working on improving this. but i don’t care how optimized your code or network is. when theres that many people running around on the screen at once spamming spells and killing each other, theres going to be some issues.

      but then again meadows knows everything “Problem is, WHO is no longer a beta but it sucks technologically”. I think certain people (cough) have been stuck in beta their whole life.

      and i’m sure he’s basing that on his own personal expierences with the latest patches. but whatever.

        • steelcity_ballin
        • 11 years ago

        You can try turning off playability effects. This is the puffs of smoke, the blast radius looking stuff from siege weapons, basically any voluminous effect that comes from anywhere. I turn it off for everyone, helped a lot IMO, mostly noticeable in crowded areas. I also turn off all name tags for everyone.

    • herothezero
    • 11 years ago

    As noted here earlier on TR in another article, I’ve always found the “disable services for which you don’t know their purpose” method of “optimization” to be pretty friggin’ stupid.

    • zimpdagreene
    • 11 years ago

    Nice. Well said. I think most of us up has tried the snake oil more than twice .

    • DrDillyBar
    • 11 years ago

    Never played WOW, and lost interest in WAR when I found out it was going to be based on Fantasy rather then 40K….

    Good to see a blog on the subject though. Good stuff.

    • Tamale
    • 11 years ago

    Welcome Matt, nice post!

    It certainly seems like your advice is sound considering the horrible problems that plagued WoW in its earliest days vs. the rock-solid stability of it now.

      • Meadows
      • 11 years ago

      I don’t recall it ever being unstable, and I’ve been playing since 3 years now. I do know that I’ve upgraded through several hardware generations in this time, so my experience about speed was altered somewhat.

        • Usacomp2k3
        • 11 years ago

        So you missed the first 25% of it’s existence which all falls under ‘early’ I think.

          • Meadows
          • 11 years ago

          Haven’t been a beta tester if that’s what you’re shooting at, and I don’t remember if the game was accurate with its debut date here, but as for the beta portion, that’s not meant to be stable and carved in stone as such.

          Problem is, WHO is no longer a beta but it sucks technologically.

            • Usacomp2k3
            • 11 years ago

            I was more referring to the fact that the game was launched November 2004, which is just a month shy of 4 years ago. So if you’ve been playing for the last 3, then that least a year of tweaking that you were able to take advantage of.

        • David
        • 11 years ago

        Clearly you didn’t play at launch. WoW was a decent release as far as MMO’s go, but it was stilly ridiculously slow and unstable.

        Granted, a lot of connection issues came from the games unexpected popularity.

      • jackaroon
      • 11 years ago

      If you want a more realistic (and disappointing) comparison of what to expect at launch vs. “after a few patches” . . . check out Mythic’s other MMO, DAOC, and watch it stutter like porky pig. If you’ve got any modern gaming equipment, the -average- FPS will probably be through the roof, but only until something interesting happens.

    • Scrotos
    • 11 years ago

    How hard is it to right-click the icon and “run as administrator” for the privileges that PB requires? That’s what I do when Enemy Territory gives me lip. And that pain in the ass UAC can remain on to torment me. Everyone wins!

    • Usacomp2k3
    • 11 years ago

    Welcome!

      • RickyTick
      • 11 years ago

      Welcome Matt. Interesting read.

    • Meadows
    • 11 years ago

    Hi there.

    So, effectively, you’re giving irrelevant tips that many gamers are aware of, after stating that nothing helps WHO anyway? Curious.

    Warhammer Online will be the next Crysis, we can yell “unoptimized!” at both now.

      • Scrotos
      • 11 years ago

      Actually, I think if you comprehend his post, you’ll realize that many MORE gamers are NOT aware of those tips. Hence, why so many “snake oil” fps cures are put forth and why so many people chime in that they work.

      You can do better than that. If you want to be vitriolic towards the new guy, at least make it interesting!

        • Meadows
        • 11 years ago

        I agree that the WHO forum inhabitants are not the sharpest knives in the drawer, but why doesn’t he help them, then? The users of this particular site tend to be bright on average, except a few outliers who I will generally speak against anyway.

          • Firestarter
          • 11 years ago

          The advice that he gives has been given to those who seek the ‘snake-oil’ tweaks. They either already took those advices to heart and still want the tweaks, or they ignored them.

          • Fighthouse
          • 11 years ago

          l[<...but why doesn't he help them, then?<]l Just because he didn't say so doesn't mean he didn't.

          • Steba
          • 11 years ago

          You must have a tendancy to speak out against your self quite often then.

    • Damage
    • 11 years ago

    Welcome, Matt. Good to have you.

    And let me be the first to say that disabling Vista’s UAC can sometimes lead to real performance increases. It’s been an issue in my performance testing. Dunno whether it affects WAR, though.

    Also, I’ve seen bad interactions between Cool’n’Quiet, multi-core processors, and Windows XP in certain cases in the past. That’s one of the reasons AMD released its dual-core optimization tool. Again, no idea wether WAR is affected. Seems unlikely it would be with a well-patched OS and new drivers these days.

    But someone was gonna say it. Might as well be me. 😉

Pin It on Pinterest

Share This