Chris,
In many ways I do agree with you, Intel has spent many years designing and
adapting their rackmount systems to be as bombproof as possible, and for my
$$$ intel certainly takes the cake. Primarily because of their depth of
experience when integrating these types of solutions.

I do feel its important to point out the following however
Currently it is a complete fallacy that intel "run cooler" than AMD. As it
happens Intel currently hold the crown as producing the hottest desktop chip
avalible. Thats right, the 3.2 EE processor (and the 3.2c p4) both consume
more power and produce more heat than any of the 32bit AMD offerings. Intel
was the first of the two companies to break the 80w barrier with the 3.06b ,
and with current parts you are kidding yourself if you think that the server
offerings have any less heat production.

This is however, where the greater experience of Intel does show. They build
systems with better heat disapation mechanisms than AMD, who most often
leave it up to external companies to develop competent solutions to the
thermal issues rather than do it themselves. This long experience with rack
systems is the reason I select Intel, as despite their greater heat
production, they are a better product with a longer history behind it.

To the original thread starter.
Dual anything is be my choice especially if you get charged per RU.
Xeons currently over the AthlonMP's, Opterons I'd still hang off for a few
months yet to see if the dev goes well enough to take serious advantage of
the 64bit that the cost is worth it.
However, if you find yourself a real good AMD person/company, the Dual
Athlon can be cheaper and potentially better for HLDS, especially if you use
the "barton" core (512kb l2 cache) AthlonMP cpus.

Matt


----- Original Message -----
From: "Chris Meisinger" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Thursday, January 22, 2004 1:41 PM
Subject: RE: [hlds_linux] Dual Xeons or 2 P4?


> > Maybe so, but they also accomplish the same performance at lower clock
> >speeds.
> >
> >-Simon
>
> This has nothing to do with a server application.   For a server, you want
> it to run as cool as possible, and in todays 1u applications, case space
> comes at a premium.  Less fans means more room for HD's, Gig-E cards and
the
> like.  Let us also not forget what one of the main enemies of solid state
> electronics is.  HEAT.   Heat speeds up the degradation process and is the
> primary factor in electronics failures.  Don't believe me?  Take the Fan
off
> of your AMD and see what happens.  Yeah, that smoke is a direct result of
> things getting to hot.  How do you think the capacitors and diodes on the
> motherboard react to the elevated ambient temperatures over a long period
of
> time?
>
> I know I still have PIII-500's humming away in a datacenter that are
almost
> 5 years old now, and they're running fine.  AMD didn't even launch the
> Duron/T-bird series until (late?)2000, and the first year or so was spent
> mostly exchanging processors for ones that didn't have crushed cores.
Intel
> has proven the test of time.  Not saying today's AMD's won't or can't, but
> they just haven't been around long enough yet.
>
> So yes, while they "accomplish the same feats at the same clock speed",
they
> do it at the cost of added thermal energy.  As someone who's helped to run
a
> very large datacenter, I can tell you that the difference in temperature
> between an Intel and an AMD might not seem like a lot, until you've got
800
> of those machines.
>
> Also, Xeons aren't really that expensive anymore.   Go check Pricewatch
and
> do some comparisons.  Intel is really pushing Itanium as its die-hard
server
> solution, and Xeons are at a mid-point performance wise, with the price
> reflecting that.  I can build a Dual Xeon 3.06 Ghz (Hyperthreaded, of
> course) server for cheaper than I can buy a Single AMD Opteron 248.  (Dual
> Xeons + mobo cost between 900-1100 dollars depending on mobo, Single
Opteron
> + Mobo costs between 1200-1400 dollars depending on the motherboard)
Yeah,
> go look for yourself, Xeons are cheaper.
>
> AMD might have something with Opteron, but right now, it's really too
early
> to tell.   Most Linux distros don't take advantage of it, and 99% of your
> standard binaries and utilities aren't compiled with support for it
either.
> So running an Opteron is more for coolness factor than anything else.  You
> can talk about "What's coming out next week/month/year" all you want, but
> the fact is, its not here yet.  There might be some marginal gains to be
had
> with running an Opteron, but nothing awesome.  This could change with
time,
> but only time will tell. =)
>
> AMD has some good ideas, and some cool applications, but for right now, in
> my opinion, the only way to go for server applications like what we're
doing
> with HLDS is Intel.  Hyperthreading is VERY cool on a Linux based HL
server.
> Like when I can watch top, and I can see my gameservers bouncing between 4
> cpu's and other tasks being offloaded on the fly, not having to wait, or
> fight for CPU time, this makes me and my players happy.
>
> I firmly expect to see the AMD fanboys riled up and give me 23432432
> different reasons as to why AMD is better because they read on xxx review
> site that it just is, but I've run both platforms, and seen the difference
> for myself.  Given the choice between processor x, which is less
expensive,
> has proven track record spanning years of reliability and is fully
supported
> by an OS I put on it, or Processor Y that is slightly more expensive,
about
> as fast as the other processor, runs hotter, and doesn't have ANY
extensive
> datacenter testing yet, I know which one I'd go with.
>
>
> My .02.
>
> Chris
>
>
>
> _______________________________________________
> To unsubscribe, edit your list preferences, or view the list archives,
please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux


_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Reply via email to