[ut3] Linux Server and AMD Athlon 2600 dual processor
Joel Wiramu Pauling
aenertia at aenertia.net
Mon Nov 19 00:06:56 EST 2007
Generally. While enhanced instruction sets are useful for specific
transforms, they are not some sort of GLOBAL program optimisation.
Read : http://freshmeat.net/articles/view/730/
Reading comments essential. This is presuming that Ryan is using GCC to
compile. Choice of compiler is going to greatly change the profiling,
and the profiling of the code is where you get your optimisations,
rather than just flicking some instruction set use flags.
On Hā, 2007-11-17 at 13:51 +0000, Alex Boag-Munroe wrote:
> While "ricer" flags make a marginal difference, I was under the
> impression that being able to utilise these instruction sets is
> actually a decent boost during runtime...no?
>
> On 17/11/2007, Joel Wiramu Pauling <aenertia at aenertia.net> wrote:
> I think this is the point Ryan was making. While I will be the
> first to admit I am a ricer (someone who wants to squeeze
> every marginal gain out of my hardware). I also know that
> those gains due to compiler opt flags are very Marginal.
>
> 64bit advantage is however appreciable on servers with really
> large maps in a rotation, as the processes (even ut2004) can
> climb to above 4gb memory usage (presuming you have a system
> with more than 4gb) In situations where you have a big map
> like AS_MOTHERSHIP, and another big map like AS_ASSAULT
> loading in the background.
>
> If ryan says that ut3 will primarily be gfx bound I believe
> him. I know that this was deinately the case on an opteron
> 248, with 8gb of ram. For ut2004. UT2004 was a very clean
> scalable engine and would run on p3's fine. So long as you
> crippled fx options and rez.
>
> So... while I asked about arch optimized binaries, I also know
> how imperceivable difference it makes.
>
> Cheers.
>
> JoelW
>
>
> On 17/11/2007, Dominic Lepiane <archangel at nibble.bz> wrote:
> Yeah, I would like to know how much this pandering to
> out-dated CPUs is going
> to affect the rest of us. I think sse2 was a
> reasonable requirement.
>
> <edit>Pretend I had a huge long rant/flambait about
> people with old
> processors</edit>
>
> On Thursday 15 November 2007 19:39:25 Kris Kersey
> (Augustus) wrote:
> > I agree. I know that you need to support the lowest
> common denominator
> > but those of us with higher-end chips would hate to
> lose performance.
> > Would it be a big inconvenience to have at least two
> binaries? I know you
> > have to draw the line somewhere. Although, one for
> sse, sse2, sse3, and
> > sse4 doesn't sound like a bad idea. ;-) I'd be
> happy to do the
> > benchmarking to test how much it actually makes a
> difference.
> >
> > Thanks,
> > Kris Kersey (Augustus)
> > LinuxHardware.org Site Manager
> > augustus at linuxhardware.org
> > Gentoo Linux AMD64 Developer
> > augustus at gentoo.org
> > AIM: Augustus22
> >
> > On Fri, 16 Nov 2007, Joel Wiramu Pauling wrote:
> > > Any chance of having a wrapper script that calls
> an optimized binary for
> > > the particular cpu/arch flags?
> > >
> > > I know this increases the deployment payload but,
> I am a CFLAG opt nut,
> > > and disc space is cheap.
> > >
> > > (Yes you could do install time magic also,
> but...)
> > >
> > > I know that with ut2004 there were issues with
> getting amd64 binaries
> > > in-line with i386 and vs versa. So perhaps having
> one unified binary
> > > wrapper would be the way to go.
> > >
> > > On 16/11/2007, Rick Page
> <mpcacrucesalus at gmail.com> wrote:
> > >> Thanks for the update Arne,
> > >> The Celeron is Pentium 4 level hardware, but I am
> seeing better
> > >> performance at a lower cost with the AMD. I have
> been comparing the
> > >> two machines with different games and wanted to
> try UT3 demo before
> > >> the retail release. I think I have seen enough
> good things with the
> > >> AMD machine that I am going to stick with it.
> Please, this is not to
> > >> start an Intel vs AMD war. It is just what I have
> observed. And yes I
> > >> know that the AMD processor in this case is
> better than the slower
> > >> Celeron, but price is also a factor here.
> > >> Rick
> > >>
> > >> On Nov 15, 2007 2:30 PM, Arne Brodowski
> <arne at planet-rcs.de> wrote:
> > >>> Hi Rick,
> > >>>
> > >>> to quote Ryan from the 14th October:
> > >>>
> > >>> ---
> > >>> It's built with -march=pentium4 -mtune=pentium4
> -msse2 -mfpmath=sse ...
> > >>> so it will use generic instructions that were
> only available on
> > >>> P4-level hardware, and sse2 for all math
> operations.
> > >>>
> > >>> If you're still on Pentium III level hardware,
> it's time to upgrade.
> > >>> ---
> > >>>
> > >>> AMD Athlon 2600MP CPUs don't have the SSE2
> Instruction-Set, so you
> > >>> can't run the UT3 binary.
> > >>>
> > >>> best regards,
> > >>> Arne
> > >>>
> > >>> Rick Page wrote:
> > >>>> Hello everyone,
> > >>>> I have been running the demo for some time on a
> 2.1GHz Celeron without
> > >>>> too much problems. I am now trying to get it to
> run on an AMD Athlon
> > >>>> 2600 MP. I am not sure if it is a problem with
> the betademo or if it
> > >>>> is a problem with my setup. The Celeron is
> running Red Hat 9 and the
> > >>>> AMD is running CentOS 5. Yes I know that there
> really isn't much a
> > >>>> difference between the two OS's. When I try
> running ./ut3demo server
> > >>>> DM-ShangriLa?........., I get an Illegal
> instruction. Nothing else
> > >>>> comes up, just Illegal instruction. If I just
> try ./ut3demo, I get the
> > >>>> same message. Yes, I know, not very helpful is
> it. That is why I am
> > >>>> leaning a little more towards my setup, but
> just wanted to check to
> > >>>> see if anyone else had this problem. Any
> helpful suggestions on where
> > >>>> to start to look would be greatly appreciated.
> > >>>> Rick
> > >>>>
> > >>>>
> > >>>> ---
> > >>>> To unsubscribe, send a blank email to
> ut3-unsubscribe at icculus.org
> > >>>> Mailing list archives:
> http://icculus.org/cgi-bin/ezmlm/ezmlm-cgi?64
> > >>>
> > >>> ---
> > >>> To unsubscribe, send a blank email to
> ut3-unsubscribe at icculus.org
> > >>> Mailing list archives:
> http://icculus.org/cgi-bin/ezmlm/ezmlm-cgi?64
> > >>
> > >> ---
> > >> To unsubscribe, send a blank email to
> ut3-unsubscribe at icculus.org
> > >> Mailing list archives:
> http://icculus.org/cgi-bin/ezmlm/ezmlm-cgi?64
> >
> > ---
> > To unsubscribe, send a blank email to
> ut3-unsubscribe at icculus.org
> > Mailing list archives:
> http://icculus.org/cgi-bin/ezmlm/ezmlm-cgi?64
>
>
>
> --
> Dominic Lepiane
>
> "Of all the things I've lost, I miss my mind the
> most."
> - Ozzy Osbourne
>
> ---
> To unsubscribe, send a blank email to
> ut3-unsubscribe at icculus.org
> Mailing list archives:
> http://icculus.org/cgi-bin/ezmlm/ezmlm-cgi?64
>
>
>
>
>
>
>
> --
> Alex Boag-Munroe
>
> Lack of planning on your part does not constitute an emergency on
> mine.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://icculus.org/pipermail/ut3/attachments/20071119/8e3d540f/attachment.pgp>
More information about the ut3
mailing list