Upgrading from P3 800Mhz to Celeron 2.6GHz - Is that wise?

  • Thread starter Thread starter *Vanguard*
  • Start date Start date
V

*Vanguard*

Any CPU-only benchmarks around that simply compare that various Intel
CPUs against each other? For example, and with all else being equal,
what would be the gain in performance, if any, between a Celeron 2.6GHZ
400FSB and a Pentium 3 800MHz 100FSB? The mobo will get changed but I'm
just wondering how much of a performance boost I might expect just based
on the CPU change.
 
Any CPU-only benchmarks around that simply compare that various Intel
CPUs against each other? For example, and with all else being equal,
what would be the gain in performance, if any, between a Celeron 2.6GHZ
400FSB and a Pentium 3 800MHz 100FSB? The mobo will get changed but I'm
just wondering how much of a performance boost I might expect just based
on the CPU change.
Click to expand...


It'll vary by app, how much it's processor instead of PCI or disk I/O
or (???) bound, but I'd ballpark it at a 175% average increase while
the CPU is the most significant bottleneck, but ranging from a little
to a lot higher in benchmarks. CPU-only benchmarks are of little use
though, since modern CPUs are now by far the fastest part in a system.
If the motherboard only supports ATA33, AGP 1, of course there are
other significant bottlenecks too, as I'm sure you're aware.

The primary factor is the most demanding and/or most important use of
the machine. It's already seldom you'd wait on the GUI with a 800MHz
CPU but if you're trying to do something like encode video realtime,
play semi-modern games, it'll make all the difference.

Dave
 
*Vanguard* said:
Any CPU-only benchmarks around that simply compare that various Intel
CPUs against each other? For example, and with all else being equal,
what would be the gain in performance, if any, between a Celeron 2.6GHZ
400FSB and a Pentium 3 800MHz 100FSB? The mobo will get changed but I'm
just wondering how much of a performance boost I might expect just based
on the CPU change.
Click to expand...

If you are just changing the mobo/cpu you probably won't see much
"improvement" unless you are doing CPU intencive apps like photo/video
editing. I p3/100 vintage system basically will have nothing useful for a
new system inside it to reuse. Instead of a celron, I'd be looking at
either a P4 2.4C or an AMD system depending on what the application is.
 
*Vanguard* said:
Any CPU-only benchmarks around that simply compare that various Intel
CPUs against each other? For example, and with all else being equal,
what would be the gain in performance, if any, between a Celeron 2.6GHZ
400FSB and a Pentium 3 800MHz 100FSB? The mobo will get changed but I'm
just wondering how much of a performance boost I might expect just based
on the CPU change.
Click to expand...

Acccording to sandra twice the speed but whether this translates into
actual performance increase depends on what apps and other hardware, CPU
intensive apps will show a big improvement such as rendering or graphics.
Match it with a good video card and you'll see a big boost in the newer
games.

Lane
 
*Vanguard* said:
Any CPU-only benchmarks around that simply compare that various Intel
CPUs against each other? For example, and with all else being equal,
what would be the gain in performance, if any, between a Celeron 2.6GHZ
400FSB and a Pentium 3 800MHz 100FSB? The mobo will get changed but I'm
just wondering how much of a performance boost I might expect just based
on the CPU change.
Click to expand...

You could download SiSoftware prgram called Sandra and that shows
different speeds of CPUs.

http://www.sisoftware.co.uk/index.htmldir=dload&
location=sware_dl_x86&langx=en&a=

Nick
 
You could download SiSoftware prgram called Sandra and that shows
different speeds of CPUs.

http://www.sisoftware.co.uk/index.htmldir=dload&
location=sware_dl_x86&langx=en&a=

Nick
Click to expand...

The problem with Sandra is that it's not at all accurate when
comparing CPUs with different sized cache. In particular it will
overrate Durons, Celerons, C3, etc. Perhaps they've improved things
recently but it's been like that for several years.


Dave
 
*Vanguard* said:
Any CPU-only benchmarks around that simply compare that various Intel
CPUs against each other? For example, and with all else being equal,
what would be the gain in performance, if any, between a Celeron
2.6GHZ 400FSB and a Pentium 3 800MHz 100FSB? The mobo will get
changed but I'm just wondering how much of a performance boost I
might expect just based on the CPU change.
Click to expand...

Thanks for the replies. I realize more gains are realized with a better
mobo, higher ATA rates and using RAID for the disks, and such. All the
other components were getting replaced, too. I was just wondering if
the CPU alone when changing from P3 to Celeron but upping the clock
would give me any *potential* boost, if any. Presumably an AMD Athlon
XP 2500+ (333MHz, I think that's the Barton with the doubled L2 cache)
would be better than the Celeron 2.6GHz and they're the same price
($85-$89).

If I wanted to overclock the Athlon 2500+, will I have to do tricks with
the traces (opening/shorting them) on the CPU, or can it simply be
overclocked by changing the BIOS settings? I tend to prefer a stable
system and haven't bothered to overclock any of my boxes yet. I
previously used Pentiums, but it's something I'd think about with the
AMD. Anyone experienced with the AOpen AK77-600 MAX motherboard
(http://snurl.com/36vj)? I like their "watchdog" hardware-based failsafe
(http://snurl.com/36vi) to reset back to detection defauts if you screw
up the overclocking too bad without having to open the case and short
the CMOS clear jumper. MSI is usually a good mobo brand, too, and the
MSI KT600 Delta-LSR looks good (http://snurl.com/36vg). They have their
Core Center utility with its auto testing that increases the FSB until
the system reboots (presumably to whatever was the last good FSB value).
AOpen's watchdog is in hardware. MSI's is in software so I still might
have to clear CMOS via the jumper if the reboot hangs. However, I don't
like that AOpen has me usurp the mic-in and line-in connectors to get
6-channel sound whereas the MSI can use its S-bracket (without the
S-bracket you're stuck doing the same thing as AOpen). Couldn't get the
manual for the Gigabyte boards (download was dead from them), and am
still looking at Soyo and Asus (Abit and Biostar are out of the running
after disqualifying them based on features, but I'm also leery of their
quality).

Previously I was thinking of getting mobos with the nForce2 Ultra400
because the dual-channel access to double the bandwidth sounded good,
but benchmarks at tomshardware.com show the single-channel 400MHz
(DDR400) that's now supported by KT400A and KT600 chipsets is almost
equal to what the nForce2 provides. And with single-channel, I don't
have to worry about getting the memory sticks in pairs to ensure they
exactly match, so I could add one 512MB stick now and get another later
(although I do match the architecture).
 
*Vanguard* wrote:

Previously I was thinking of getting mobos with the nForce2 Ultra400
because the dual-channel access to double the bandwidth sounded good,
but benchmarks at tomshardware.com show the single-channel 400MHz
(DDR400) that's now supported by KT400A and KT600 chipsets is almost
equal to what the nForce2 provides. And with single-channel, I don't
have to worry about getting the memory sticks in pairs to ensure they
exactly match, so I could add one 512MB stick now and get another later
(although I do match the architecture).
Click to expand...

First off, AMD's don't benefit much from dual chanel ram, not like a P4
does. Second DO NOT buy a via chipset board over a nvidia unless you like
dealing with flakey drivers etc. Those nvidia boards have proven themselves
to be stable while Via has proven many times they release buggy software,
hardware and drivers.

What makes you think the KT400 chipset is faster than the nvidia in single
channel mode? And they don't have to be "matched" pairs to work right..
 
stacey said:
*Vanguard* wrote:



First off, AMD's don't benefit much from dual chanel ram, not like a
P4 does. Second DO NOT buy a via chipset board over a nvidia unless
you like dealing with flakey drivers etc. Those nvidia boards have
proven themselves to be stable while Via has proven many times they
release buggy software, hardware and drivers.

What makes you think the KT400 chipset is faster than the nvidia in
single channel mode? And they don't have to be "matched" pairs to
work right..
Click to expand...

Exactly. It isn't like dual-channel on P4 boards where you have to have
exact matched pairs. nForce 2 boards (are great) have two, totally
independant memory controllers. You just put the slower stick in the master
socket and it sets the specs for the other channel to the same.
 
stacey said:
First off, AMD's don't benefit much from dual chanel ram, not like a
P4 does. Second DO NOT buy a via chipset board over a nvidia unless
you like dealing with flakey drivers etc. Those nvidia boards have
proven themselves to be stable while Via has proven many times they
release buggy software, hardware and drivers.

What makes you think the KT400 chipset is faster than the nvidia in
single channel mode? And they don't have to be "matched" pairs to
work right..
Click to expand...


"Whereas dual-channel boards can give you problems if your RAM modules
are less than perfectly identical (which is why manufacturers started
selling matched DIMM pairs several months ago), single-channel systems
are relatively problem-free." (http://snurl.com/376z) The article
discusses nForce2 dual-channel versus VIA single-channel for AMD (it is
not discussing Intel). When looking at http://snurl.com/376y, it
mentions to be careful to not use x4 architectured memory and instead
use x8 or x16, so you also have to be watchful of what type of sticks
you use. Dual-channel still requires more care in picking your memory
than does single-channel. It seems best to populate both slots in the
nForce2 dual-channel mobo with identical memory when you first add
memory (i.e., populate the slots at the same time).

I did NOT say the single-channel KT600 was faster than the dual-channel
nForce2 Ultra400. I said that it was nearly equal (which means the
KT600 is *slower*) but the difference is often insignificant. The
benchmarks show numbers whereas I typically check the percentage of
difference so I can see how much better one is over the other. Also,
some benchmarks are superfluous. Do you really notice when Quake is
playing at 65 fps versus 68 fps? See
http://www6.tomshardware.com/motherboard/20031107/nforce2-400-11.html
for some benchmarks.

My problem was the nForce2 mobos didn't have the most critical feature
that I required: RAID - and on both IDE and SATA ports. I could find
good nForce2 mobos equally priced to the KT600 mobos. But once I
checked against the requirement for RAID, lots of nForce2 mobos got
chopped off the list. Of the nForce2 mobos left that had RAID, once I
checked which ones provided not just RAID on the SATA ports but also on
an IDE3 port then the rest got chopped off and I was left with an empty
list. Regardless of all the hoopla over the wee bit faster memory, the
hard disk is by far the slowest major subsystem so I want to speed that
up by using RAID 0 stripping (I'll use disk images for backup/restore).
My current drives which will get migrated to the new system are IDE-only
so I need RAID that uses IDE ports. I would also like RAID on the SATA
port for later drives upgrades, but the first requirement was RAID on
IDE. As an example, if you choose MSI as the mobo maker, they have only
one nForce2 mobo with RAID (http://snurl.com/376y) yet it only provides
RAID on the SATA ports or by IDE3+SATA (one drive on IDE3 and the other
on SATA1 or SATA2). Both my current drives to be reused are IDE only so
this setup won't work for me.

However, I will have to rethink my mobo selection criteria. While
concentrating on having an IDE port for my current IDE-only drives to
provide RAID, that probably won't help if only one IDE port supports
RAID. For RAID to work well by overlapping read/write across multiple
drives, they would have to be on separate IDE channels, so I'd need 2
IDE ports to successfully execute RAID. DOH! RAID for 2 drives on the
same IDE channel probably won't help much. So I guess I'll be
recompiling my mobo list. I'll have to use the old drives as non-RAID
until I get a SATA drive later. However, the MSI nForce2 mobo gives me
1 extra IDE port and 2 SATA ports whereas the MSI KT600 mobo gives me
the 1 extra IDE port and *4* SATA ports so I can add more drives later
without having to usurp a PCI slot with a controller card.

I'm looking at AMD for a CPU since the price differential from a P4 will
pay for the mobo or a hefty chunk of it. This is for a home computer.
If it were a business computer then I'd probably stick with a P4.

As far as bugginess in drivers, there hasn't been one manufacturer that
someone hasn't complained about. I've had problems with Creative,
Intel, IBM, Logitech, Promise, Adaptec, and so on but I continue to use
their products. The trick is to keep a historical store of all the
drivers you have updated through so you can revert when the newest one
turns out to cause problems for you. I tend to turn a deaf ear to those
claiming flakiness in drivers since most of such reports are from
personal experience rather than a statistical analysis across a large
user population. However, if you have some industry statistical reports
regarding VIA being more flaky than nForce2 then I'd like to read those.
 
Also, whether for the nForce2, KT400, KT400 (and presumably for the
KT600), you get better memory performance by using DDR333 memory (or
setting the memory clock so it is DDR333) rather than DDR400 memory!
See:

http://www.tomshardware.com/motherboard/20030417/index.html

From the article, "The differences that we found in all of the
benchmarks are very small; too small to feel anything at all, even. I
had to run all benchmarks three, four or sometimes even five times to
make sure I got a reliable result rather than a benchmark variance."

So I'm not basing my mobo selection on dual- or single-channel but on
the other features that I want: RAID (a must!), 10/100 LAN, USB 2.0,
serial (preferably 2 of them), parallel, IEEE 1394 would be nice,
overclocking features, PS/2 for keyboard and mouse, onboard audio (will
replace later, if needed), suspend to ram or disk, AGP 4x/8x, would like
AUX-in on mobo besides just CD-in, and so on.
 
*Vanguard* wrote:

As far as bugginess in drivers, there hasn't been one manufacturer that
someone hasn't complained about.
Click to expand...

But Via has had MORE than their share as well as buggy hardware that driver
updates can't fix. And as long as you buy two of the same memory sticks
when you buy your ram, you shouldn't have any problems. Most nforce makers
have "supported" rams sticks as well? And like I said on an AMD system
running the nforce board in single channel mode doesn't hurt performace
enough to be that worried about it.
However, if you have some industry statistical reports
regarding VIA being more flaky than nForce2 then I'd like to read those.
Click to expand...

Do some google searching on Via problems.. BTW the latest Via based IDE
raid controller is a POS, uses a BUNCH of CPU cycles in use and has poor
performance, as well as buggy drivers. Again do some searching about that
as well. We've had several people on this group complaining about them in
the last few weeks.
 
*Vanguard* said:
Also, whether for the nForce2, KT400, KT400 (and presumably for the
KT600), you get better memory performance by using DDR333 memory (or
setting the memory clock so it is DDR333) rather than DDR400 memory!
See:

http://www.tomshardware.com/motherboard/20030417/index.html

From the article, "The differences that we found in all of the
benchmarks are very small; too small to feel anything at all, even.
Click to expand...

Yep, AMD's aren't bottle necked my memory bandwidth like a P4 is. I went
from single chanel to dual chanel ram on my P4 and the difference was VERY
noticable, everything else being equal.
 
Back
Top