How fast is a server to a Desktop

[Sanny]

I have a old Pentium 2.5 Ghz Machine I use for surfing Net. I am just
curious to know how much faster are the Server Machines are.

Say they are 8 Core/ 16 Core??? What is the Max cores Modern Server
has. And How much fast they are than my Single Core Machine.

I need this information because currently I run my code on a Client
Machine. I want to know how fast It will become if I put the code on
Servver which uses Threads for Multi Tasking and Just outputs the
results from the Server.

Will it be 10-20 Times faster than Ordinary Client Machines?

Bye
Sanny

 

[Toby A Inkster]

Will it be 10-20 Times faster than Ordinary Client Machines?

Sometimes yes, sometimes no.

My server is a Dual 1GHz Pentium III, 512 MB RAM. Fast enough for most
purposes.

--
Toby A Inkster BSc (Hons) ARCS
[Geek of HTML/SQL/Perl/PHP/Python/Apache/Linux]
[OS: Linux 2.6.17.14-mm-desktop-9mdvsmp, up 22 days, 16:37.]

Bottled Water
http://tobyinkster.co.uk/blog/2008/02/18/bottled-water/

 

[RedGrittyBrick]

I have a old Pentium 2.5 Ghz Machine I use for surfing Net. I am just
curious to know how much faster are the Server Machines are.

I'd benchmark them to find out.

Say they are 8 Core/ 16 Core??? What is the Max cores Modern Server
has.

There's no maximum. I'm sure there must be servers with hundreds or
thousands of cores (in large numbers of CPUs).

And How much fast they are than my Single Core Machine.

A $10,000,000 server might be slower than your machine, if the server is
very heavily loaded.

A $500 server might be 500 x faster than your machine if your machine is
infested with viruses, trojans, other malware and only has 32 MB RAM.

I need this information because currently

If I needed it, I'd benchmark actual machines in typical loading.

I run my code on a Client
Machine. I want to know how fast It will become if I put the code on
Servver which uses Threads for Multi Tasking and Just outputs the
results from the Server.

I think it futile to ask usenet readers to guess the answer.

Will it be 10-20 Times faster than Ordinary Client Machines?

Maybe but I wouldn't make any important decisions based on that sort of
guess. Why guess when you can measure?

 

[Jerry Stuckle]

I have a old Pentium 2.5 Ghz Machine I use for surfing Net. I am just
curious to know how much faster are the Server Machines are.

Say they are 8 Core/ 16 Core??? What is the Max cores Modern Server
has. And How much fast they are than my Single Core Machine.

I need this information because currently I run my code on a Client
Machine. I want to know how fast It will become if I put the code on
Servver which uses Threads for Multi Tasking and Just outputs the
results from the Server.

Will it be 10-20 Times faster than Ordinary Client Machines?

Bye
Sanny

I'd suggest you ask this in a hardware related newsgroup. Nothing in
your question is about either java or php programming.

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
(e-mail address removed)
==================

 

[The Natural Philosopher]

I have a old Pentium 2.5 Ghz Machine I use for surfing Net. I am just
curious to know how much faster are the Server Machines are.

Say they are 8 Core/ 16 Core??? What is the Max cores Modern Server
has. And How much fast they are than my Single Core Machine.

I need this information because currently I run my code on a Client
Machine. I want to know how fast It will become if I put the code on
Servver which uses Threads for Multi Tasking and Just outputs the
results from the Server.

Will it be 10-20 Times faster than Ordinary Client Machines?

It really depends on many things.

I found that of the two headless linux servers I have built in the last
couple of years, the biggets difference was having a SATA disk on the
newer one.

Neither run into memory or CPU issues to serve small amounts of pages .
Both have around 512Bytes RAM only.


Fast processors are popular because people run BIG graphically intense
desktops.

They are probably a waste of money in a headless server.

 

[C. (http://symcbean.blogspot.com/)]

I have a old Pentium 2.5 Ghz Machine

???!!! old?

I use for surfing Net. I am just
curious to know how much faster are the Server Machines are.

Methinks you're confused about what a server is. Being a server does
not convey special powers to a PC - in fact a server build will oten
have a cheapo graphics card which will make some applications run
slower than a lower spec machine with a better graphics card.

Will it be 10-20 Times faster than Ordinary Client Machines?

It very much depends on lots of factors. If you are currently having
performance problems then say so - and explain what you've done to
isolate the bottlenecks.

C.

 

[Nigel Wade]

I have a old Pentium 2.5 Ghz Machine I use for surfing Net. I am just
curious to know how much faster are the Server Machines are.

It depends...

Say they are 8 Core/ 16 Core??? What is the Max cores Modern Server
has.

Lots, no lots and lots. Actually even more than that, lots and lots and lots.
Personally the most processors I've had access to in a single machine is 128.
But given that those processors were rather aged (by modern terms) MIPSIII
running in the region of 200MHz, your system might run your code faster unless
you are an expert at parallel programming.

And How much fast they are than my Single Core Machine.

The processors? Probably slower on an individual basis.

I need this information because currently I run my code on a Client
Machine. I want to know how fast It will become if I put the code on
Servver which uses Threads for Multi Tasking and Just outputs the
results from the Server.

Will it be 10-20 Times faster than Ordinary Client Machines?

Well, that all depends. It depends on:

a) the number of processors
b) the speed of those processors
c) how many of those processors your code is able to utilize simultaneously
d) the OS, and whether it is able to schedule the threads in your code across
the available processors/cores.
e) how many other processes are running on the server, competing with your code.
f) how long the piece of string is you are currently measuring.

 

[Patricia Shanahan]

It depends...


Lots, no lots and lots. Actually even more than that, lots and lots and lots.
Personally the most processors I've had access to in a single machine is 128.
But given that those processors were rather aged (by modern terms) MIPSIII
running in the region of 200MHz, your system might run your code faster unless
you are an expert at parallel programming.

....

Even with the same processor speed, the single thread performance may
not be as good on a large shared memory multiprocessor as on a small system.

The problem is that as the number of processors and memory modules
increases, it takes more switching to connect each processor to each
memory. It also takes more work to keep processor caches coherent.

Patricia

 

[Roedy Green]

I have a old Pentium 2.5 Ghz Machine I use for surfing Net. I am just
curious to know how much faster are the Server Machines are.

How big is your wallet? Some years ago we ran a multithreaded app we
had written on a machine with 256 CPUs. I think though that the trend
is to use machines not all that much faster than a desktop. Servers
typically have tons of RAM and very fast durable SCSI discs and
multiple ethernet pipes. CPU is not the limiting factor.

I have posted some of the specs on my server at
http://mindprod.com/contact/equipment.html#SERVER

The big concern is the cost of AC power and the cost of cooling a
giant server farm. So now people like Google look at the TOTAL cost
to serve each transaction, including the cost of the server, the
software, the AC power and the cooling. This is tending toward large
numbers of smaller computers, usually rack mounted.

Sun's new Niagara servers give more bang per watt. This approach was
a wise decision. With oil over $100 a barrel and electricity costs
scheduled to rise, it will pay to replace monster energy pig servers
with energy efficient ones.

 

[Roedy Green]

Methinks you're confused about what a server is. Being a server does
not convey special powers to a PC - in fact a server build will oten
have a cheapo graphics card which will make some applications run
slower than a lower spec machine with a better graphics card.

In fact most servers in farms don't even have a dedicated mouse,
keyboard or screen.

 

[Lew]

....

Even with the same processor speed, the single thread performance may
not be as good on a large shared memory multiprocessor as on a small
system.

The problem is that as the number of processors and memory modules
increases, it takes more switching to connect each processor to each
memory. It also takes more work to keep processor caches coherent.

Another factor affecting speed is program correctness. Code not written to
exploit multiple processors, or not to do so correctly, could run slower than
on a single-processor platform, or not at all (at least, not correctly).

 

[Arne Vajhøj]

I have a old Pentium 2.5 Ghz Machine I use for surfing Net. I am just
curious to know how much faster are the Server Machines are.

Say they are 8 Core/ 16 Core??? What is the Max cores Modern Server
has. And How much fast they are than my Single Core Machine.

I need this information because currently I run my code on a Client
Machine. I want to know how fast It will become if I put the code on
Servver which uses Threads for Multi Tasking and Just outputs the
results from the Server.

Will it be 10-20 Times faster than Ordinary Client Machines?

There are servers that are more than 20 times faster than
your P4 2.5 (for multithreaded usage).

You can also find servers slower than your PC.

The fastest single node computer (not counting clusters)
are supposed to be this 2048 core beast:
http://blogs.zdnet.com/Ou/?p=864

Arne

 

[Toby A Inkster]

In fact most servers in farms don't even have a dedicated mouse,
keyboard or screen.

And these days even VGA and PS/2 ports are becoming increasingly rare.
(Blade servers tend not to have them, rack servers usually do though.)

--
Toby A Inkster BSc (Hons) ARCS
[Geek of HTML/SQL/Perl/PHP/Python/Apache/Linux]
[OS: Linux 2.6.17.14-mm-desktop-9mdvsmp, up 24 days, 21:08.]

Bottled Water
http://tobyinkster.co.uk/blog/2008/02/18/bottled-water/

 

[Toby A Inkster]

With oil over $100 a barrel and electricity costs scheduled to rise, it
will pay to replace monster energy pig servers with energy efficient
ones.

A lot of non-database servers could probably stand to use flash memory
instead of conventional magnetic disk drives.

--
Toby A Inkster BSc (Hons) ARCS
[Geek of HTML/SQL/Perl/PHP/Python/Apache/Linux]
[OS: Linux 2.6.17.14-mm-desktop-9mdvsmp, up 24 days, 21:10.]

Bottled Water
http://tobyinkster.co.uk/blog/2008/02/18/bottled-water/

 

[Nigel Wade]

...

Even with the same processor speed, the single thread performance may
not be as good on a large shared memory multiprocessor as on a small system.

The problem is that as the number of processors and memory modules
increases, it takes more switching to connect each processor to each
memory. It also takes more work to keep processor caches coherent.

Patricia

I know. CPU to main memory is a matter of motherboard design and if the designer
got it wrong there's not much the programmer can do about it. But cache
coherency is very much of concern to the programmer.

The first part of the parallel programming course I did on that system was to
demonstrate that very fact. Selecting a suitable stride size when stepping
through a parallelized loop is rather important.

 

[Patricia Shanahan]

Patricia Shanahan wrote: ....

I know. CPU to main memory is a matter of motherboard design and if the designer
got it wrong there's not much the programmer can do about it. But cache
coherency is very much of concern to the programmer.

"Motherboard" seems to assume all processors fit on one board. The
interconnect between processors and memory modules becomes much more
challenging when that is not the case. It is not so much a matter of
getting it wrong than of limits such as wire lengths and chip pin counts.

The first part of the parallel programming course I did on that system was to
demonstrate that very fact. Selecting a suitable stride size when stepping
through a parallelized loop is rather important.

Also, the less sharing of data that is being modified the better. It is
usually better to have each processor work on blocks of contiguous data.

Patricia