Opteron Clock speed

[Dan Irwin]

hi,

I was wondering, which opteron is equal in power to which model of
Intel processors (Aka P4\Xeon)? Does anybody know of an artical
online that compares P4 and Opterons? I know their not exactly normaly
considered on par, but for a workstation they just might be. Kinda
sorta

thx again,

dan

 

[Ed]

hi,

I was wondering, which opteron is equal in power to which model of
Intel processors (Aka P4\Xeon)? Does anybody know of an artical
online that compares P4 and Opterons? I know their not exactly normaly
considered on par, but for a workstation they just might be. Kinda
sorta

thx again,

dan

Just something to look at while you wait for (better) replays. ;p
http://reviews.zdnet.co.uk/0,39023100,39145079-2,00.htm
http://www.techreport.com/reviews/2004q2/dually-opterons/index.x?pg=1

 

[Rob Stow]

hi,

I was wondering, which opteron is equal in power to which model of
Intel processors (Aka P4\Xeon)?

There will always be some apps where the raw MHz of P4/Xeon will
give it an occasional win over AMD64. The varying caches sizes for
the Xeons also complicate the picture. However, here's a guideline
that I find does the job for *most* tasks:

For single processor systems, P4 needs 50% higher clock than AMD64
eg., 2 GHz Athlon 64 vs 3 GHz P4
For dual processors systems, Xeon needs 70% higher clock than AMD64
eg., 1.8 GHz Opty vs 3 GHz Xeon
For quad processors systems, Xeon needs 90% higher clock than AMD64
eg., 1.4 GHz Opty vs 2.6 GHz Xeon

For single and dual cpu systems, my personal experience probably
gives me an excessive bias in Opteron's favour. I have yet to
have a chance to play with a 4-way Opty system, so for the 4-way
guideline above I have merely generalized from the rare 4-way
benchmarks I have seen.

Does anybody know of an artical online that compares P4 and Opterons?

There are hundreds, if not thousands, of articles out there
that have lots AMD64 vs (P4 or Xeon) benchmarks. Google.

I know their not exactly normaly
considered on par, but for a workstation they just might be. Kinda
sorta

Depends on what you are going to do with your workstation.
Some apps run much better on a P4 or Xeon because of bad
programming that stops them from using SSE2 on AMD64 processors.
Dirty tricks by Intel compilers can also unnecessarily
slow down apps that are run on non-Intel processors.

 

[Tony Hill]

I was wondering, which opteron is equal in power to which model of
Intel processors (Aka P4\Xeon)? Does anybody know of an artical
online that compares P4 and Opterons? I know their not exactly normaly
considered on par, but for a workstation they just might be. Kinda
sorta

It's really quite tough to compare the two because it varies quite
significantly from one application to another. For some applications
the Opteron 140 (running at 1.4GHz) could easily beat out a Xeon
3.2GHz. On other applications an Opteron 148 (running at 2.2GHz)
could be beaten by a Xeon 2.0GHz chip.

Doing a quick search on Google turned on a few comparisons.

http://www.extremetech.com/article2/0,1558,1402989,00.asp
http://www.sudhian.com/showdocs.cfm?aid=487
http://www.aceshardware.com/read.jsp?id=65000288
http://tech-report.com/reviews/2003q4/opteron-x48/index.x?pg=1


I'm sure that there are dozens of others, just do a bit of searching
on Google and you should be able to turn up some info more specific to
any applications you might be interested in.

 

[Nate Edel]

I was wondering, which opteron is equal in power to which model of
Intel processors (Aka P4\Xeon)? Does anybody know of an artical
online that compares P4 and Opterons? I know their not exactly normaly
considered on par, but for a workstation they just might be. Kinda
sorta

There have been several articles -- search on Tom's hardware for their
Xeon/Opteron comparison, for example.

VERY rough rule of thumb, for raw integer performance, consider an AMD
Opteron/Athlon 64FX or the Athlon 64 3200+/3400+ (2ghz/2.2ghz 1mb cache) as
equal to an 800FSB Pentium 4 processor with a 50% higher megahertz rating.
For example, any of the 2ghz/1mb cache models should perform at least as
well as the 3ghz Pentium 4.

 

[Nate Edel]

There will always be some apps where the raw MHz of P4/Xeon will
give it an occasional win over AMD64.

Vector SSE2 code, basically. Is there anything else?

For single and dual cpu systems, my personal experience probably
gives me an excessive bias in Opteron's favour.

I doubt it's excessive, actually. Between the much better way of doing
duals, and the combination of a big cache and a much better integer
capability, it seems deserved to me.

 

[Dan Irwin]

now, my next question is, am i better off with two 240s or one 146 and
30 more memory(i can't imagine that'll get me much but...)

 

[Nate Edel]

now, my next question is, am i better off with two 240s or one 146 and
30 more memory(i can't imagine that'll get me much but...)

Depends on for what workloads. If you're running applications that
parallelize well, you'll do MUCH better with the two 240s. In the worst
case, which is one and only one cpu-bound completely single-threaded
application, you'd do no better with two 240s than with one 240 (and
possibly a little worse.)

For software development work, if you're doing big compilation jobs and you
have a halfway sensible IDE/make system, duals help a lot. Ditto for CAD --
I think every major package is pretty well multithreaded these days.

For gaming, the dual 240s will help a LITTLE as compared to a single 240,
but probably not enough to beat a single 144, let alone a 146.

 

[Tony Hill]

now, my next question is, am i better off with two 240s or one 146 and
30 more memory(i can't imagine that'll get me much but...)

The answer to that is a definite "it depends".

Really it could go either way depending entirely on what applications
you use and where you feel you need the more performance. If you just
run mainly one single-threaded application at a time, a pair of
processors aren't going to help you much at all, the faster single
processor will be better. If you're applications are multithreaded
then the dual processor system should be faster.

The real advantage of dual-processor workstations though is not so
much that they are faster at finishing a single task, but rather that
they make the machine much more responsive for multitasking. So even
if your applications are not multithreaded and they would actually run
a bit slower on the dual-processor machine than the faster
single-processor setup, the dual still might be better. Since only
one of the two processors would be used for running your application
in the background, you've still got another processor free to continue
working on something else in the foreground. Even if it's just
something simple like word processing and e-mail, having a
dual-processor machine can be better.

As for the memory question issue, again it depends. More memory only
buys your performance until you're no longer swapping. If all your
applications run entirely from memory with 1GB of memory, going to 2GB
of memory buys your nothing in terms of performance. On the other
hand, if you are swapping then adding more memory HUGELY improves
performance.

 

[Dan Irwin]

What about the gap in clock speed? Would 6 whatevers make the 146
better even for multi tasking?

 

[Tony Hill]

What about the gap in clock speed? Would 6 whatevers make the 146
better even for multi tasking?

In a word, no. In fact, some would argue that even a single processor
that is twice as fast does not feel as "responsive" as a half-speed
dual-processor setup. I don't know if this is true and it is
extremely difficult to test this since a lot of it is more qualitative
than quantitative.

However, even for the quantitative testing of multitasking a dual
Opteron 240 (1.4GHz) should be better than a single Opteron 146
(2.0GHz).

 

[Paul Gunson]

In a word, no. In fact, some would argue that even a single processor
that is twice as fast does not feel as "responsive" as a half-speed
dual-processor setup. I don't know if this is true and it is
extremely difficult to test this since a lot of it is more qualitative
than quantitative.

It is sorta true IME. i have an old dual PIII setup (550MHz) which was
pretty responsive when switching windows etc, i now have an Athlon64
3200+. its getting to the stage with PCs that because the A64 is so
fast i can barely notice the difference, but on things like dialing with
a software modem (lol) it lags for maybe a second, it wouldn't lag at
all on the old dual P3 box because there's 50% CPU resources left.

happy not to have bought dual Optys tho, 'cause the cost of the procs +
board would've been several times my current system.

 

[Nate Edel]

What about the gap in clock speed? Would 6 whatevers make the 146
better even for multi tasking?

146 = 2ghz
240 = 1.4ghz

Close to a 50% boost, but context switches are expensive, and the total
cache between the two processors is 2X the cache of the single. It depends
on how much you really multitask. _IF_ you were building a machine for
multiple users, I'd say definitely go for the dual 240s.

 

[George Macdonald]

It is sorta true IME. i have an old dual PIII setup (550MHz) which was
pretty responsive when switching windows etc, i now have an Athlon64
3200+. its getting to the stage with PCs that because the A64 is so
fast i can barely notice the difference, but on things like dialing with
a software modem (lol) it lags for maybe a second, it wouldn't lag at
all on the old dual P3 box because there's 50% CPU resources left.

happy not to have bought dual Optys tho, 'cause the cost of the procs +
board would've been several times my current system.

Which mbrd for your Athlon64? I'm ready to take the plunge but wondering
about waiting for the new nForce3/250(?) chipset. Personally I've never
had a problem with VIA chipsets in the past but my experience with nForce2s
has been so smooth that I'd like to wait, if it's not going to be too long.

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

 

[Paul Gunson]

Which mbrd for your Athlon64? I'm ready to take the plunge but wondering
about waiting for the new nForce3/250(?) chipset. Personally I've never
had a problem with VIA chipsets in the past but my experience with nForce2s
has been so smooth that I'd like to wait, if it's not going to be too long.

i'm using an Albatron (VIA) K8X800 ProII. its awesome... no probs
yet.... and no additional noisy chipset fan on this board.

i'm not sure if i would've gone with VIA before the A64 line...

an intersting point to note dual procs does not alway mean dual
performance in multi-threaded apps. a test 3D render we've done in maya
took 58 seconds on my A64, whereas a friend tested the same scene on a
dual Opteron 248 with a score of 36 seconds. i would've thought it would
be at least twice as fast considering the 248 itself is 0.2 GHz faster
than the 3200+ .... mind u he was using a Tyan board.. not sure if that
made any difference or not.

 

[Grumble]

On other applications an Opteron 148 (running at 2.2GHz)
could be beaten by a Xeon 2.0GHz chip.

You'd have to look quite hard to find such an application...

Especially if we only consider Xeons with 1 MB of L2 cache.

 

[Tony Hill]

You'd have to look quite hard to find such an application...

Especially if we only consider Xeons with 1 MB of L2 cache.

Perhaps, but they certainly do exist. One situation that jumps to
mind are applications that are tuned to run from 512KB of L2 cache.
While this would run entirely from cache on both processors, the
2.0GHz Xeon has roughly twice the bandwidth of the Opteron. If your
just doing something like a Linpack-style work and running your entire
dataset out of L2, chances are that you're going to be mostly
bandwidth-limited by your L2 cache.

There are other, much more esoteric situations that you could easily
encounter. The simple fact of the matter though is that the P4/Xeon
and their "Netburst" core is quite different from the Opteron. While
the latter is almost always faster clock for clock, that won't always
be the case, so buy whatever processor is best for the software you
need to run.