Is it worth upgrading a 3500 single core for a 4400 dual c..

Thanks for the rapid response.

I had thought that since the single core I have and the dual core I was
contemplating run off 2.2GHz and 2.3GHz respectively, that there would not
be a great deal of difference between the two.

Does that mean that if there is only one application running, that both
cores would be used to process it, or just the one. The other core,
presumably would handle windows background tasks only?

Thanks

Martin


"Conor" <conor_turton RemoveThis @hotmail.com> wrote in message
news:62apojF22gibuU1@mid.individual.net...
> On Sat, 23 Feb 2008 13:49:29 +0000, Martin C wrote:
>
>> A few years ago, I built a PC using an ASUS A8N-SLI Premium motherboard
>> and an AMD Athlon 64 3500 Venice single core CPU.
>>
>> As this motherboard is an older 939 socket type, I thought about
>> upgrading to a dual core CPU on a 939 socket. I found an AMD Athlon 64
>> X2 4400+ / 2.3 GHz processor at a reasonable price and though about
>> getting it as the motherboard supports Athlon X2 processors.
>>
>> However, I ask myself if it is really worth getting it. I don't do a
>> great deal in the way of multi-tasking, although some of my newer games
>> may make better use of it. Effectively, as I understand it, each
>> processor runs at 2.3 GHz and both running together give the 4400
>> rating. However, as most applications just use the one core, I am
>> effectively losing out?
>>
>> Basically the question I have is : Would it be worth the cost to upgrade
>> from a 3500 single core to a 4400 dual core.
>>
>> TIA
>>
>> Martin
>
> Absolutely worth it. Whilst many apps will only use one core, Windows
> will use both, shunting processes onto the core not being used by games.
> Most newer games are now SMP aware and will utilise both cores.
>
> --
> Conor
>
> I only please one person per day. Today is not your day. Tomorrow isn't
> looking good either. - Scott Adams
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

On Sat, 23 Feb 2008 13:49:29 GMT, "Martin C"
<martin.TakeThisOut@invalid.com> wrote:

>A few years ago, I built a PC using an ASUS A8N-SLI Premium motherboard and
>an AMD Athlon 64 3500 Venice single core CPU.

Venice is fairly overclockable - do that. It should easily
hit 2.7 to 3.1GHz which is the most you can reasonably get
out of that platlform before replacing motherboard, memory
and processor.


>
>As this motherboard is an older 939 socket type, I thought about upgrading
>to a dual core CPU on a 939 socket. I found an AMD Athlon 64 X2 4400+ / 2.3
>GHz processor at a reasonable price and though about getting it as the
>motherboard supports Athlon X2 processors.
>
>However, I ask myself if it is really worth getting it. I don't do a great
>deal in the way of multi-tasking, although some of my newer games may make
>better use of it.

You won't benefit then, those who claim even windows will
benefit keep forgetting that actually looking at task
manager on a dual core system in your scenarios shows that
the second core is barely used at all, like 2% utilization
on average if that. IOW, a single core at 3% higher
clockspeed would about cover that but overclocking your
present CPU will yield much higher performance than buying
the new dual core and running it at stock speed.



>Effectively, as I understand it, each processor runs at
>2.3 GHz and both running together give the 4400 rating. However, as most
>applications just use the one core, I am effectively losing out?

Yes, exactly. Those who see benchmarks of new apps
optimized for dual cores will only benefit if using
similarly modern optimized apps. In your case a single core
at highest clockspeed reasonably possible is the best
performance you can squeeze out of skt 939. A larger L2
cache wouldn't hurt either but at this time it wouldn't be a
cost-effective upgrade either.


>
>Basically the question I have is : Would it be worth the cost to upgrade
>from a 3500 single core to a 4400 dual core.

No it'd be money mostly thrown away, though if in the future
your uses changed then it might be of more benefit but in
the future you would find your parts all that much older and
even moreso feel a bigger upgrade was in order.

In short, I suggest keeping what you have, mildly
overclocking it if the board allows, enjoying that benefit
while you take awhile to let Intel's 45nm processors drop in
price some then move to LGA755 based upgrade in a few
months or longer. However you are thinking about upgrading
NOW so the real question is in which apps you find the
performance lacking? You haven't given a comprehensive
description of the system but given how cheap DDR2 memory is
right now, I'm inclinded to suggest you not ungrade anything
that couldn't be reasonably transferred to a new system
build. IOW, a hard drive would be but DDR(1) memory
wouldn't.
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

On Sat, 23 Feb 2008 15:14:03 GMT, "Martin C"
<martin DeleteThis @invalid.com> wrote:

>Thanks for the rapid response.
>
>I had thought that since the single core I have and the dual core I was
>contemplating run off 2.2GHz and 2.3GHz respectively, that there would not
>be a great deal of difference between the two.

There won't be. Even multi-threaded apps have to be fairly
well optimized to benefit much from dual core processors.


>
>Does that mean that if there is only one application running, that both
>cores would be used to process it, or just the one. The other core,
>presumably would handle windows background tasks only?

It depends on how the application is designed. Also, one
application versus "windows background tasks" is not only
two threads, it is many of them... far more than # of cores
on any modern processor. You should investigate your most
used or most demanding application (including version
number, don't rely on benchmarks of a newer version than you
use) you use or plan on using in the near future.
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

"Martin C" <martin.RemoveThis@invalid.com> wrote...
>A few years ago, I built a PC using an ASUS A8N-SLI Premium motherboard and an
>AMD Athlon 64 3500 Venice single core CPU.
>
> As this motherboard is an older 939 socket type, I thought about upgrading to
> a dual core CPU on a 939 socket. I found an AMD Athlon 64 X2 4400+ / 2.3 GHz
> processor at a reasonable price and though about getting it as the motherboard
> supports Athlon X2 processors.
>
> However, I ask myself if it is really worth getting it. I don't do a great
> deal in the way of multi-tasking, although some of my newer games may make
> better use of it. Effectively, as I understand it, each processor runs at 2.3
> GHz and both running together give the 4400 rating. However, as most
> applications just use the one core, I am effectively losing out?
>
> Basically the question I have is : Would it be worth the cost to upgrade from
> a 3500 single core to a 4400 dual core.

First, ensure your MoBo will support the 90nm process CPUs, since both the 4400+
CPUs are 90nm versions.

Second, in the AMD chart here
(http://products.amd.com/en-us/DesktopCPUSideBySide.aspx?id=69&id=70&id=100),
both 4400+ versions are 2.2 GHz CPUs (though the socket AM2 version may be 2.3
GHz). The difference is that the version with the OPN beginning with ADV
instead of ADA is an 89 watt version, vs 110 watts for the ADA. So, you should
get the ADV version if you can find it, to reduce thermal loads on your system.

Third, I believe the dual-core CPU is worth it if you can afford it. It should
be affordable, since Socket 939 is an older technology, and the CPUs should be
relatively cheap, though specific models may be hard to find. Remember that
even when "not multitasking," your computer is likely running numerous
background processes such as antivirus, firewall, etc. Keeping a CPU core
available for the foreground task will increase performance somewhat. AMD's
rating system suggests you should see an overall 25% performance increase,
though that would likely occur mostly under high CPU loads.

Bottom line: Get one if you can find one.
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

On 2008-02-23, Martin C <martin.TakeThisOut@invalid.com> wrote:
> Thanks for the rapid response.
>
> I had thought that since the single core I have and the dual core I was
> contemplating run off 2.2GHz and 2.3GHz respectively, that there would not
> be a great deal of difference between the two.
>
> Does that mean that if there is only one application running, that both
> cores would be used to process it, or just the one. The other core,
> presumably would handle windows background tasks only?

It isn't as straightforward as that. In reality your processes
(more accurately with Windows, threads) will constantly switch
between the two cores more or less at random as different threads
become runnable and go back to sleep again. An individual thread
is not tied to a particular core for any length of time.

As to how much benefit you will derive from going dual core, it
depends on your applications and usage patterns. Most applications
these days _are_ multithreaded so there is some scope for parallelism,
but the question is really whether the computationally intensive
portions of those applications can be broken into separate threads
for distribution over the available cores. A few basic observations
of the system can also give much useful info - are you regularly
running at 100% CPU? If not the CPU isn't really your bottleneck
anyway.

My own opinion is that it _is_ worth going for additional cores
when upgrading. How much use of them you'll make right now depends
on your usage patterns, but as multi-core machines become ubiquitous
software will be written to take advantage of them to the point
where everyone will use multi-core machines effectively. However,
I wouldn't go multi-core right now just for that - it is rarely a
good idea to have more than half an eye on the future when upgrading,
since when the future arrives you'll often find that technology
has changed and/or prices have dropped.

Should you upgrade? It's up to you but personally I wouldn't
bother. I'm occasionally accused of being a Luddite when it comes
to upgrades or indeed replacing machines, but in my view it usually
doesn't make sense to upgrade to something less than twice as fast
as what you have now, and if you have simple requirements you can
save money by pushing that multiple far higher.

The upgrade you propose is less than twice as fast and in real
world conditions I doubt you will notice much subjective benefit
from the upgrade. Bear in mind that the CPU is only one component
of the system and many of the noticeable delays in computing are
not caused by the processor - e.g. application start up is often
disk bound and web pages are of course network bound.

--
Andrew Smallshaw
andrews.TakeThisOut@sdf.lonestar.org
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

On Sat, 23 Feb 2008 21:46:01 +0100 (CET), Andrew Smallshaw
<andrews.DeleteThis@sdf.lonestar.org> wrote:


> Most applications
>these days _are_ multithreaded so there is some scope for parallelism,

The interesting thing is that most people state this as if
someone were going to buy all new applications instead of
using what they had... back when they were running a single
core system. IMO, most people do not buy hundreds to
thousands of dollars worth of new software when they upgrade
a CPU on the cheap, so "these days" is not really relevant
to the topic.
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

On 2008-02-24, kony <spam.DeleteThis@spam.com> wrote:
> On Sat, 23 Feb 2008 21:46:01 +0100 (CET), Andrew Smallshaw
><andrews.DeleteThis@sdf.lonestar.org> wrote:
>
>> Most applications
>>these days _are_ multithreaded so there is some scope for parallelism,
>
> The interesting thing is that most people state this as if
> someone were going to buy all new applications instead of
> using what they had... back when they were running a single
> core system. IMO, most people do not buy hundreds to
> thousands of dollars worth of new software when they upgrade
> a CPU on the cheap, so "these days" is not really relevant
> to the topic.

I did specifically address the issue of new vs existing software
in my original post. The fact is that most apps out there in the
wild _are_ multithreaded and have been since at least the time of
Windows 95. Take a simple everyday example - Internet Explorer.
There are several threads working. Several may be managing downloads
of various elements of a web page. Another may be redrawing the
screen as you scroll down the page. The Windows flag may be waving
in the top right hand corner to show that it is working - that
needs a thread too. Many apps (not necessarily Internet Explorer)
will have additional truly invisible threads running things like
garbage collection or rearranging data structures.

These separate threads may be scheduled independently and can run
simultaneously on different cores is the system permits. On Windows
threads have been managed by the kernel and independently scheduled
for a long time. This is one area where it was actually slightly
ahead of Unix as many Unixes traditionally delegated this to the
application rather than take it into the kernel, although has by
and large changed by now.

However, the threads discussed so far have been mainly for non-core
ancillary functions. This is work that needs doing and can profitably
be shifted to another core, but the time used by such tasks is
limited. To effectively use a multi core machine to the best
extent possible you need to carve up the innermost, time
critical loops into separate threads. This is what is usually
meant when new releases of applications are described as multithreaded
in non-technical parlance. This ensures that the time-consuming
work is distributed more evenly among the available cores and gets
the most out of a multi-core setup. However, this is not to say
that traditional multi-threaded apps can't benefit from a multi-core
system.

--
Andrew Smallshaw
andrews.DeleteThis@sdf.lonestar.org
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

On Sun, 24 Feb 2008 21:35:41 +0100 (CET), Andrew Smallshaw
<andrews DeleteThis @sdf.lonestar.org> wrote:

>On 2008-02-24, kony <spam DeleteThis @spam.com> wrote:
>> On Sat, 23 Feb 2008 21:46:01 +0100 (CET), Andrew Smallshaw
>><andrews DeleteThis @sdf.lonestar.org> wrote:
>>
>>> Most applications
>>>these days _are_ multithreaded so there is some scope for parallelism,
>>
>> The interesting thing is that most people state this as if
>> someone were going to buy all new applications instead of
>> using what they had... back when they were running a single
>> core system. IMO, most people do not buy hundreds to
>> thousands of dollars worth of new software when they upgrade
>> a CPU on the cheap, so "these days" is not really relevant
>> to the topic.
>
>I did specifically address the issue of new vs existing software
>in my original post. The fact is that most apps out there in the
>wild _are_ multithreaded and have been since at least the time of
>Windows 95.

No, most apps are not multithreaded, at least not
effectively so. The major applications you'd pay big bucks
for have been for much longer, but these are a minority of
what most people use on a PC.


>Take a simple everyday example - Internet Explorer.
>There are several threads working. Several may be managing downloads
>of various elements of a web page. Another may be redrawing the
>screen as you scroll down the page. The Windows flag may be waving
>in the top right hand corner to show that it is working - that
>needs a thread too. Many apps (not necessarily Internet Explorer)
>will have additional truly invisible threads running things like
>garbage collection or rearranging data structures.

Not evenly divided workloads, and a small load at that. It
makes it fairly insignificant as we see in that the user
experience running IE on a multicore system is not
different than on a single core, or more accurately that the
difference is inperceivable.


>
>These separate threads may be scheduled independently and can run
>simultaneously on different cores is the system permits. On Windows
>threads have been managed by the kernel and independently scheduled
>for a long time. This is one area where it was actually slightly
>ahead of Unix as many Unixes traditionally delegated this to the
>application rather than take it into the kernel, although has by
>and large changed by now.
>
>However, the threads discussed so far have been mainly for non-core
>ancillary functions. This is work that needs doing and can profitably
>be shifted to another core, but the time used by such tasks is
>limited. To effectively use a multi core machine to the best
>extent possible you need to carve up the innermost, time
>critical loops into separate threads. This is what is usually
>meant when new releases of applications are described as multithreaded
>in non-technical parlance. This ensures that the time-consuming
>work is distributed more evenly among the available cores and gets
>the most out of a multi-core setup. However, this is not to say
>that traditional multi-threaded apps can't benefit from a multi-core
>system.

Perhaps but benefit in a hypothetical sense has to be
weighted against benefit in a real usability sense. For
example most single apps of the past few years will perform
as well or better with higher clocked core(s) than more of
them.
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

"kony" <spam RemoveThis @spam.com> wrote...
>
>>Take a simple everyday example - Internet Explorer.
>>There are several threads working. Several may be managing downloads
>>of various elements of a web page. Another may be redrawing the
>>screen as you scroll down the page. The Windows flag may be waving
>>in the top right hand corner to show that it is working - that
>>needs a thread too. Many apps (not necessarily Internet Explorer)
>>will have additional truly invisible threads running things like
>>garbage collection or rearranging data structures.
>
> Not evenly divided workloads, and a small load at that. It
> makes it fairly insignificant as we see in that the user
> experience running IE on a multicore system is not
> different than on a single core, or more accurately that the
> difference is inperceivable.

OTOH, take Outlook Express as another everyday example...

When processing a large number of Newsgroup messages or when cleaning up or
compacting a large message store, it can hog the CPU from all other processes
for several minutes. On a multi-core machine, processes and single-threaded
apps running on the other core are unaffected.
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

On Sun, 24 Feb 2008 16:51:12 -0800, "John Weiss"
<jrweiss98155NOSPAM DeleteThis @NOSPAM.comcast.net> wrote:

>"kony" <spam DeleteThis @spam.com> wrote...
>>
>>>Take a simple everyday example - Internet Explorer.
>>>There are several threads working. Several may be managing downloads
>>>of various elements of a web page. Another may be redrawing the
>>>screen as you scroll down the page. The Windows flag may be waving
>>>in the top right hand corner to show that it is working - that
>>>needs a thread too. Many apps (not necessarily Internet Explorer)
>>>will have additional truly invisible threads running things like
>>>garbage collection or rearranging data structures.
>>
>> Not evenly divided workloads, and a small load at that. It
>> makes it fairly insignificant as we see in that the user
>> experience running IE on a multicore system is not
>> different than on a single core, or more accurately that the
>> difference is inperceivable.
>
>OTOH, take Outlook Express as another everyday example...
>
>When processing a large number of Newsgroup messages or when cleaning up or
>compacting a large message store, it can hog the CPU from all other processes
>for several minutes. On a multi-core machine, processes and single-threaded
>apps running on the other core are unaffected.
>

I have never found this a problem, are you sure the system
exhibiting this problem doesn't have other problems?

Don't get me wrong, I do advocate dual core, but for the
right reasons...
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

"kony" <spam DeleteThis @spam.com> wrote...
>>
>>OTOH, take Outlook Express as another everyday example...
>>
>>When processing a large number of Newsgroup messages or when cleaning up or
>>compacting a large message store, it can hog the CPU from all other processes
>>for several minutes. On a multi-core machine, processes and single-threaded
>>apps running on the other core are unaffected.
>>
>
> I have never found this a problem, are you sure the system
> exhibiting this problem doesn't have other problems?

Yep! Repeatable on 3 systems.
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

On 24 Feb, 21:07, kony <s....TakeThisOut@spam.com> wrote:
> On Sun, 24 Feb 2008 21:35:41 +0100 (CET), Andrew Smallshaw
>
> >I did specifically address the issue of new vs existing software
> >in my original post. The fact is that most apps out there in the
> >wild _are_ multithreaded and have been since at least the time of
> >Windows 95.
>
> No, most apps are not multithreaded, at least not
> effectively so. The major applications you'd pay big bucks
> for have been for much longer, but these are a minority of
> what most people use on a PC.

Just curious. I've been following this discussion and you both seem
to set apart processes and threads without comment. What is the
distinction? I thought they were pretty much the same thing.
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c.. 

On Thu, 28 Feb 2008 09:47:57 -0800 (PST), laider DeleteThis @live.com
wrote:

>On 24 Feb, 21:07, kony <s... DeleteThis @spam.com> wrote:
>> On Sun, 24 Feb 2008 21:35:41 +0100 (CET), Andrew Smallshaw
>>
>> >I did specifically address the issue of new vs existing software
>> >in my original post. The fact is that most apps out there in the
>> >wild _are_ multithreaded and have been since at least the time of
>> >Windows 95.
>>
>> No, most apps are not multithreaded, at least not
>> effectively so. The major applications you'd pay big bucks
>> for have been for much longer, but these are a minority of
>> what most people use on a PC.
>
>Just curious. I've been following this discussion and you both seem
>to set apart processes and threads without comment. What is the
>distinction? I thought they were pretty much the same thing.


A process is a task (or whole app) which may contain (at
least one but...) multiple threads it executes. Threads can
be processed by different cores in a multi-core processor
based system, but the effectiveness, efficiency with which
the work is divided varies greatly. Often a single thread
consumes most of the processing time so addt'l CPU cores are
mostly left idle or processing different threads which may
or may not also be consuming a lot of processing time or
mostly idle.
 >> Stay informed about: Is it worth upgrading a 3500 single core for a 4400 dual c..