Differences between AMD and Intel CPUs [was: Re: Dragonfly and Hyperthreading....]

[EM1897 at aol.com]

In a message dated 3/9/2005 1:19:29 PM Eastern Standard Time, Freddie Cash <fcash-ml at xxxxxxxxxx> writes:

>On March 9, 2005 09:18 am, EM1897 at xxxxxxx wrote:
>> I'm sorry Matt, but you are just plain wrong. Assuming
>> PCI-X is a requirement, a 3.2Ghz P4 + MB is $360. and
>> to get equivalent performance you're looking at $700.
>> to $900 for a MB with an AMD processor and I'd argue that
>> you need more than that. If you need a 1U appliance,
>> you're looking at $950. base vs $440. base. There is
>> simply no justification for it currently.
>
>You can't compare a P4 system with an Opteron system.  One's a desktop 
>processor, the other is a server processor.  You would compare the P4 
>system with an Athlon64 system, and find the Athlon64 system to be less 
>expensive, quieter, cooler, and "just as fast".
>
>You have to compare a P4 Xeon system with an Opteron system.  And once 
>you do that, you find that the Xeon system will always be more 
>expensive than the Opteron system.  (It is up here in Canada, anyway.)  
>It's nigh impossible to find just the Xeon CPU for less than $500, let 
>alone a CPU and mobo combo.
>
>You also have to be sure you are comparing the correct speed bins of the 
>two processors.  In all the benchmarks out there (on Ars Technica, 
>Tom's Hardware, Anandtech, etc), the Opteron (at a much lower clock 
>speed) is either faster than, the same as, or within 15% slower than a 
>P4 / Xeon system running up to 2 GHz faster.  Once you start comparing 
>SMP systems (especially 4-way systems) you find the Opterons will 
>almost always come out on top.  Which one is the better value then?  A 
>super-high clocked system, running smoking hot, or a lower clocked but 
>just as fast performing system that doesn't sound like a jet engine?
>
>We have two IBM 2U servers here, each with dual 3.0 GHz Xeon processors 
>and 6 GB of RAM each.  These things sound like jets when idle.  Once 
>the systems get loaded, they sound worse.  There are 10 fans in these 
>2U boxes.  During testing, these were in our office, and nearly drove 
>us insane.  I haven't checked the temps recently, but they were always 
>pushing 80 C with the case top off during testing.
>
>We also have two whitebox servers here, each with dual Opteron 242s (1.8 
>GHz) and 8 GB of RAM each.  These things are whisper quiet, even under 
>load.  There's only 4 fans in each of these.  The loudest devices are 
>the SCSI drives.  The temps under load push 60 C.
>
>The price difference between the two sets of servers above is almost 
>$5,000 CDN each, with the Opterons being cheaper.  I haven't run any 
>benchmarks on these systems as they run a mix of Fedora Core (64-bit), 
>RedHat Enterprise Linux (32-bit), and FreeBSD (32-bit).  If we didn't 
>have a support contract with IBM for the hardware and software on those 
>servers, it'd be very hard to justify using them.

First of all, yes I can. Secondly, there is no difference
between a P4 and a Xeon performance-wise, so the distinction
between "server" and "workstation" is really just 
semantics when you are talking uniprocessor, and I've 
already rejected BSD SMP implementatations. If you pay more for a xeon then you are just a fool, unless you need DP or MP. There was a time when you couldn't get PCI-X for a P4, but those days are over. So paying for uni-xeon system is stupid.

If you paid that much for an SMP system running FreeBSD,
then you're flying blind for certain. You've overpaid 
because you don't know enough about hardware to know
what you're doing. Your xeon systems "sound like jets"
because you've bought something that sounds like a jet.
A P4 system with a single blower (which is what we use)
will likely outperform what you've bought for 1/5th the
cost. (my current 5.3R benchmarks show that FreeBSD
with 2 3.06Ghz xeons is actually slower than a single
P4 processor system. They've done a helleva job!).

If you are arguing that its possible to pay more for
a xeon system than an opteron system, then point taken.
But considering that you admit that you have  no idea
what the performance is of either system, I don't see
the point you're making, except to also admit that you've
paid WAY more then you need to pay. 

My assumption was that whoever is doing the purchasing
actually knows what he's buying, which doesn't seem to
be the case for you. Nothing personal, but you got 
fleeced rather badly.