Best configuration for game server: Nwn - Bioware

Hello all readers!

I have a great struggle to take decision between Athlon and PIV processor.

My needs is to have a really good "server" without graphic performance for a game called Neverwinter Nights.

I want to run my server on a Linux operating system like Red Hat.

Have 24/24 hours, 7/7 days open.
With at least 40 IP connections.

I want to best processors to have speed in for process many I/O request.

(in hope to put 2 Raid ScsII hard drive mode II)

And have speedy ram like PC2700 or RDRAM...

So, I want to have equivalent of the PIV 2000 MHZ power.

If people with knowledge can give me informations about my needs, I will be very happy to read it.

For my expressions, my mother language is French, so I ask you to be less severe...

In hope to know the best products for my needs,

Arnok Ciel Argenté
25 answers Last reply
More about best configuration game server bioware
  1. Have you thought of using FreeBSD instead of Linux?
    It would suit your needs better than Linux and it is also more secure than Linux. Not that I dont like Linux, Linux is my favorite OS, but you have to choose the OS that suits your needs best which a *BSD probably would. (BTW, before any *BSD Vs Linux Flames come in - this site runs FreeBSD ;)... well it used to last time I checked a week or 2 ago)
    Also, I would suggest that you use ECC RAM if you want a rock solid server, the negligable performance drop is a small price to pay for a solud system.
    Also, if you dont mind paying the extra, I would be inclined to get a P4, only due to its superior thermal protection.
  2. No, I don't know about FreeBSD and I think Linux facilities are good to do many thing...

    What will be the good and the bad things about FreeBSD. I don't even know if Neverwinter Nights will support it.
  3. I dont really want to go into *BSD VS Linux - It has a habit of starting flamewars :). Howvever, FreeBSD is very similar to Linux, all I can suggest is that you do a bit of research on FreeBSD and make your own mind up. I will say that FreeBSD is more secure than Linux, it is one of the most secure OS's out and it makes an exceptional server OS that is as solid as a rock (Yahoo run FreeBSD).
    But if the game doesnt support it (I have a feeling it will run it though ) then there is nothing that you can do about it anyway.
    /d4n13l
  4. Hmmm...

    For Linux/UNIX/BSD, the Pentium IV is probably not such a good choice. Most of the apps will not be P4-optimized. Is the Linux-based server for Neverwinter Nights P4-optimized?

    If it is, the P4 might have a shot at performing better. If not...go with either a P3 or an AthlonXP solution. If you can custom-compile NWN server from source code (doubtful) <i>and</i> get a copy of Intel's optimizing compiler for Linux, you could conceivably produce a P4-optimized server. Then, of course, there's the difficulty of getting dual P4 Xeons--they're rumored to be in <A HREF="http://www.theinquirer.net/09110101.htm" target="_new">very short supply</A>, especially at 2.0GHz.

    Just to give you an idea how horribly the P4 performs under Linux, <A HREF="http://www6.tomshardware.com/cpu/01q4/011031/xpvsp4-12.html" target="_new">check this out</A>.

    With all that in mind, I would personally recommend something different--and really, is the CPU the bottleneck here? I've never set up an NWN server myself, so I don't know. I would expect the network connection to be the real bottleneck though, especially if you're serving over the Internet.

    If I really needed CPU power, I would go for dual AthlonMP's myself.

    I can't say if FreeBSD or Linux is better for this; I lean towards Linux, since lately, it's very solid and seems a bit faster than FreeBSD. Many Linux-ELF apps can run under FreeBSD though, so keep an open mind.

    Kelledin
    <A HREF="http://www.linuxfromscratch.org/" target="_new">LFS</A>: "You don't eat or sleep or mow the lawn; you just hack your distro all day long."
  5. Thanks for the idea...

    I have posted a request to know if NWN will be P4 optimized for Linux.

    If somebody can give me much more informations about RAID for the serveur and others things, I will be very interest.
  6. Quote:
    Just to give you an idea how horribly the P4 performs under Linux, <A HREF="http:// www6.tomshardware.com/cpu/01q4/011031/xpvsp4-12.html" target="_new">check this out</A>.

    I think this is guilt by association. You have to ask your self, does presents of poor compilation numbers necessarily imply poor performance based on the operating system?
  7. Considering that the compile job is based on a lot of what the system does, it's a fairly good indicator. A compile job is a lot of integer work, some file copying, some floating point, and heavy memory usage, and a lot of it is the same stuff a system does on a regular basis. It calls on a lot of different binaries for different tasks--not just gcc components for compiling. It tests stability better than almost anything else. The only thing it doesn't stress is the network layer, which is barely CPU dependent at all (from the standpoint of Intel vs. AMD).

    Of course, it would be nice if we had some decent Apache benchmarks or the like, but <A HREF="http://www.linuxhardware.org/article.pl?sid=01/10/09/1514233&mode=thread" target="_new">here</A> is the best I could find. There's another article <A HREF="http://www.linuxhardware.org/article.pl?sid=01/10/15/1443234&mode=thread" target="_new">here</A> from the same site. I dislike both articles for not including 2GHz P4's, and the latter for not including dual Xeons (cheap bastards).

    Kelledin
    <A HREF="http://www.linuxfromscratch.org/" target="_new">LFS</A>: "You don't eat or sleep or mow the lawn; you just hack your distro all day long."
  8. I think you misread my logic or I misread yours. Whether or not the CPU performs well during a set task, has nothing to do with its overall performance or the OS. Could the same difference be seen on a Windows platform during a separate complication test? I would say that compilation is quite different that the work done by the OS. In general an OS does very little computation and a lot of house cleaning by moving memory, reloading page tables, setting up a DMA channels, keeping track of open streams, calling interrupt hooked code, etc. On the other hand a compiler does a lot of computation, parsing lexicals, building tables, searching tables, loading code libraries, spitting out object code and then linking it all together. I would say that the reason the AMD performs better during compilation is due to the amount of unpredictable branches in the task.

    It is fair to say that the performance of an Apache web server could be related to a compile benchmark. I think your logic was just a bit reaching by saying the OS’s performance may have a relation to a compile benchmark. Otherwise, I’m defending the Linux OS more than the P4’s performance.
  9. An honest question: anyone want to explain without laughing why Win2K Server or Advanced server wouldn't do the job? I have no preferences, really, though if I wanted to do something similar, I have access to free copies of W2K Advanced Server
  10. "Win2K Server or Advanced server wouldn't do the job"

    Sorry sdausmus but I don't want to know performance of server, I want to know the best configuration hardware to run a Neverwinter Nights Server with Linux on it.

    All:

    Please someone, I have one good point for AMD cause Linux is not PIV optimized...

    Anyone who have other information for this?

    I add that the NWN server, is a game server that will take "all" the machine to do is job. This new game technology will take a load of server's ressources...
  11. You need bandwidth for best performance on game servers. The slowest P4 crushes the best AMD in memory bandwidth by far.

    Second factor is stability, AMD just lost by large margin.

    Hard drive speed is important and not platform dependant. RAID for best performance, but prob not needed.
  12. Hmmm, let's see, I've posted three Linux benchmarks so far. In those three benchmarks, the Pentium IV performed miserably, and the Athlon+760MP got a ten out of ten for stability.

    The facts apparently speak for themselves. It would be nice if you had some links to back up your claims...

    Kelledin
    <A HREF="http://www.linuxfromscratch.org/" target="_new">LFS</A>: "You don't eat or sleep or mow the lawn; you just hack your distro all day long."
  13. Those benchmarks are meaningless as they all use generic kernels. That makes it a very unfair comparison as you always rebuild your kernel to take the upmost advantage of whatever hardware you use. A far better comparison would be to benchmark properly optimised kernels, specifically for each CPU. As we all know the P4 relies heavily on optimised code. As a consequence, you will see benchmarks using a generic kernel favour the Athlon. I suspect that the gap in performance will be shortened if optimised kernels were used. But, the Athlon would take first place all the same, just with a smaller margin.
  14. I dont know about you, but I havent heard about any p4 optimised linux kernals, do you have links to show there are any, or perhaps you have coded your own?

    "The Cash Left In My Pocket,The BEST Benchmark"
    No Overclock+stock hsf=GOOD!
  15. There are some SSE optimizations in the Linux kernel, and there is a selection for "Pentium-4" in the kernel config's processor selection. I don't know how good the Pentium-4 optimization is.

    You could perhaps optionally use Intel's Linux C/C++ compiler to compile the kernel, but that's risky. Using an unsupported compiler version (or even an unsupported optimization level) can cause problems. YMMV.

    Then, of course, there's all the libraries in a Linux system, like the GNU C libraries, openssl, etc. etc. All these libraries will affect the performance of any apps which use them. They normally default to no higher than i586 (Pentium classic) optimizations, and usually even lower. You can usually get the source code for this stuff and compile it all yourself (like I do), but that could take days. Not to mention which, some of it is just as compiler-picky as the kernel.

    Kelledin
    <A HREF="http://www.linuxfromscratch.org/" target="_new">LFS</A>: "You don't eat or sleep or mow the lawn; you just hack your distro all day long."
  16. Matisaro,
    I dont know whether I detected a hint of sarcasm in your last post(?), hehe. No, I havent coded my own P4 kernel, mainly because I dont have a P4 - they cost too much and Athlons are faster :).
    My point was that the benchmarks were meaningless as they used generic kernels which could be run on a lowly 386, hardly a real-life analogy of a P4 Vs Athlon system. I would be far more interested in benchmarks of fully optimised systems. The kernel may use a bit of SSE, but what about SSE-2? It will only be a matter of time before we begin to see entire distros that take full advantage of the P4. After all, this is Linux.
    I am not defending the P4, merely defending fair benchmarking.
  17. Kruguer talk about:

    "You need bandwidth for best performance on game servers. The slowest P4 crushes the best AMD in memory bandwidth by far."

    With DDR 2700, It seems that RDRAM is not so far in front of DDR...

    "Hmmm, let's see, I've posted three Linux benchmarks so far. In those three benchmarks, the Pentium IV performed miserably, and the Athlon+760MP got a ten out of ten for stability."

    I don't see some stability test on Athlon... Do you have some links, I am curious to have some comparaison test of it. But I supposed the motherboard to be important in stability.

    Much more speedy RAM and greater stability! I think is the main concern for a game server.

    1 point: Linux not optimized for PIV. Athlon
    1 point: For much more speedy RAM. PIV
    1 point that float in the air for stability...

    With all the last read that I done over the subject, PIV is much more better for a gaming computer with multimedia alogrithme and Athlon is much more better to be a server with a lot of RAM.

    It seems that latency of RDRAM is greater when a high number of megs is in use.

    In my server, I want 1 gig or 2 gigs of RAM.
  18. Quote:
    I dont know whether I detected a hint of sarcasm in your last post(?), hehe. No, I havent coded my own P4 kernel, mainly because I dont have a P4 - they cost too much and Athlons are faster :).


    NONONONO. I meant that alot of linux fans do that kind of stuff, not attacking you or anything.

    "The Cash Left In My Pocket,The BEST Benchmark"
    No Overclock+stock hsf=GOOD!
  19. Latest benchmarks put the athoon xp 1900+ infront of the fastest p4 in almost EVERYthing.


    "The Cash Left In My Pocket,The BEST Benchmark"
    No Overclock+stock hsf=GOOD!
  20. Latest benchmarks put the athon xp 1900+ infront of the fastest p4 in almost EVERYthing.


    "The Cash Left In My Pocket,The BEST Benchmark"
    No Overclock+stock hsf=GOOD!
  21. I'm just wondrering how much sse2 optimization would benefit a Linux server?I mean how much can a server utilize simd instructions?

    And FUGGER,you idiot.It's the bandwidth between the server and outside world that matters ,not the "netburst"(or whatever) between cpu and ram.
  22. "And FUGGER,you idiot.It's the bandwidth between the server and outside world that matters ,not the "netburst"(or whatever) between cpu and ram."

    I think that you cannot judge someone on his idea with yours.

    The speed of the RAM beetween CPU and RAM is essential to don't create a lag of information.

    Nwn server permit me to create a zone for 10 megs in RAM. If I have 2000 MEG, I will have more than 160 zones accessible by 40 players.

    If I have 40 zones for 40 players (1 player on 1 zone), it is sure that the server cannot give 400 MEG of information to process in the "same time"... this example is a hard one, but If I want to have more than 3000 PNJ and 160 zones it is clear that the best Memory is important.

    I read an article on RAM over there, and it seems that DDR is much more faster than RDRAM over this benchmark.
    http://www.hardwaresecrets.com/morememory.html

    This article seems to say different thing. Anyone on the subject seems to say different thing. It is politicaly problem. I want fact, where I can find fact????

    Please, this is not a place to insult other, but to have discussion over the fact.

    Thanks to participate,

    Arnok Ciel Argenté
  23. Sorry, I posted these two a little further up:

    <A HREF="http://www.linuxhardware.org/article.pl?sid=01/10/15/1443234&mode=thread" target="_new">AthlonMP tested</A>--this is where it gets rated 10/10 for perfect stability.
    <A HREF="http://www.linuxhardware.org/article.pl?sid=01/10/09/1514233&mode=thread" target="_new">AthlonXP tested</A>--here, it actually gets 8/10 for stability, simply because there's an increased possibility that you'll get a bad board.

    d4n13l: remember what I said about apps being compiler picky. Also note that when Mandrake optimized their distro for i586, a lot of people griped about it.

    I'd be perfectly willing to try building a P4-optimized distro, if someone would loan me a P4 box and a legal copy of Intel's Linux C/C++ compiler. (I'm actually serious on this point, though I can't promise anything.)

    Kelledin
    <A HREF="http://www.linuxfromscratch.org/" target="_new">LFS</A>: "You don't eat or sleep or mow the lawn; you just hack your distro all day long."

    UPDATE: I just got this from Intel's web site:

    "You can use familiar Linux development tools such as emacs, make and gdb with the Intel Compiler to produce application-oriented executables that are substantially object code compatible with gcc (note: Intel C++ for Linux is designed for application development and will not currently build the Linux kernel)."

    The compiler is free for non-commercial usage. It may or may not work for me (I don't run a supported distro).
  24. To toss in a few more options:

    1. There is Solaris for X86 (instead of Linux/FreeBSD)
    2. Ebay has really cheap SUN Servers that would run true
    Solaris as well.
    Both of these options deserve consideration as there is no
    more stable and performing environment, tested and proven on the internet for years.
  25. I don't know if NWN will support OS for Risc structure... It could be interesting, but we need some benchmark to compare performance between our new PC tech vs. the old server tech...

    I have a benchmark over there:
    http://www.linuxhardware.org/article.pl?sid=01/10/09/1514233&mode=thread

    I think It have been done on Linux, so it could be much more interesting than benchmark on Windows. Windows 95 have some detection OS mode to put breakdown on Athlon. It seems that's was not the case for other version of Windows. This is a test done by a friend who have install Win95 with a Pentium and put AMD k6 on it after win install. Compare the setting between install win direct over ADM processors... and the first tweak option have been much more speedy than the second...

    So, I want to know how can we consider a benchmark a real fact or a politicaly benchmark???

    For now, with the last bench, the news over the DDR333 and PC2700 tech, I decide that Athlon is much more appropriate for a Linux NWN server than a PIV... PIV is optimized for graphic and multimedia thing SSEII, but I need strong Integer Calcul for the server part!

    Thank's the Kelledin and others who take time to give me some way of searching...

    Arnok Ciel Argenté
Ask a new question

Read More

CPUs Servers