Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 04 Feb 2005 21:25:19 -0700
From:      Scott Long <scottl@freebsd.org>
To:        Astrodog <astrodog@gmail.com>
Cc:        Nathan Vidican <nvidican@wmptl.com>
Subject:   Re: Intel EMT64 Xeon vs AMD Opteron
Message-ID:  <42044AAF.1010002@freebsd.org>
In-Reply-To: <2fd864e05020419382a5e21b3@mail.gmail.com>
References:  <000001c50a3c$50f2eba0$6800000a@r3140ca> <20050204103708.21608.qmail@web26801.mail.ukl.yahoo.com> <2fd864e05020419382a5e21b3@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Astrodog wrote:

> On Fri, 4 Feb 2005 11:37:08 +0100 (CET), Claus Guttesen
> <cguttesen@yahoo.dk> wrote:
> 
>>>Cost wise, AMD Opteron 246 is roughly the same cost
>>>as a 3.0Ghz Xeon ... But
>>>how do they compare performance wise; specifically
>>>related to FreeBSD?
>>
>>We have a dual xeon (nocona) @ 3.2 GHz and a dual
>>opteron @ 2 GHz, both with 4 GB RAM and running the
>>amd64-port. My impression is that the opteron performs
>>*slightly* better than it's Intel-cousin.
>>
>>regards
>>Claus
>>
>>_______________________________________________
>>freebsd-amd64@freebsd.org mailing list
>>http://lists.freebsd.org/mailman/listinfo/freebsd-amd64
>>To unsubscribe, send any mail to "freebsd-amd64-unsubscribe@freebsd.org"
>>
> 
>  From what I understand, EM64T is essentally an extention to x86, so
> it will understand the AMD64 instructions, much the same way an
> Athlon64 does. Opteron, once again, from what I've read on the topic
> is "Actual" 64-bit, not an emulated version. Generally, I find Opteron
> to be the best "Bang for your Buck", though what motherboard, and what
> features you need there may also play a role there. AMD, so far, has
> implied that the dual core opterons will be Socket 940, If that pans
> out, the 940-based solution will be significantly more expandable,
> since there's little to no chance of Intel continuing to use their
> current Xeon socket when their Dual Core offerings come out, and I
> suspect it would be technically impossible, given the Memory
> Controller issues that its bound to create. Since AMD put the memory
> controller on-die, they can resolve this issue in the core, and not
> involve the chipsets of the motherboard itself.
>  Remember, Hyperthreading isn't dual core, its kinda like adding
> another "Lane" to the processing pipeline of a single processor, so
> that when something stalls, other things can still happen.
>  Hypertransport, on the other hand is AMD's method of connecting SMP
> CPUs to eachother, memory, and devices on the motherboard.
> 
> Sorry about the Hypertransport/Hyperthreading thing, but there seems
> to be a great deal of confusion about what each are, and what's
> good/bad about them, and they relate to the AMD/Intel decsion you're
> making pretty explicitly.
> 
> Personally, I say go with the Opteron. Worst case, performance and
> reliability are the same, and you're supporting the underdog. Best
> case, it blows your socks off, and in a year, you can go dual core.
> Either way, you can't loose.
> 

Both the AMD and Intel offering are just extensions to the ia32 design. 
  Opteron is no more 'true' 64-bit than Nacona is.  There are differences
in features; Opteron and Athlon64 have dropped some legacy features,
EM64T doesn't (yet) have NX page protection support, etc, etc.  Beyond
that, they operate in pretty much identical ways.  Where Opteron has
the advantage is that it doesn't have the long instruction pipeline of
Nacona, and it has HyperTransport and an embedded memory controller
instead of Intel MCH bottleneck.  Nacona has Hyperthreading, which can
be a benefit, but it's mostly a toss-up.

As far as a real-world comparison, I just did a 6-CURRENT buildworld on
the following machines:

2x Nacona 2.8GHz, HTT enabled
4GB RAM
FreeBSD 6-CURRENT/i386 (32 bit mode)
ICH5 SATA, Maxtor

2x Opteron 246
2GB RAM
FreeBSD 6-CURRENT/amd64 (64 bit mode)
Adaptec U160 SCSI, Seagate

The time to build world was almost identical at 31 minutes each. 
Granted, the buildworld test isn't a very good overall test as it's 
often I/O bound, but it gives a rough estimate.  I haven't spent much
time running the Nacona in 64 bit mode, but what I did didn't suggest
that it would perform all that much better.

Scott



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?42044AAF.1010002>