P4P800-E

[Johnny]

I've just built a new PC and the performance test results are abysmal.

I've used passmark performance test v5.0 (eval)

New build system is a

3.0GHz P4 Intel Prescott CPU 800FSB
2 x 512Mb Corsair 2-2-2-5 DDR400 supposedly performance memory (located in
the blue slots as per manual)
Asus P4P800-E board

test scores are
cpu passmark = 323
memory mark = 330

Bios settings are as follows
system information shows
processor type = Intel (R) Pentium (R) 4CPU 3.00GHz, speed 3000MHz, count 1
system memory size = 1024GB
further bios info shows cpu is set to 200 x 15 and is registering as 800 FSB

********
my older system is a

2.8GHz P4 Intel Northwood 533FSB
1 x 512Mb Nanya generic PC2700 DDR333
Asus P4PE board

test scores are
cpu passmark = 413
memory mark = 423

WTF!!! somethings not right here surely. Anyone a clue what the heck is
wrong?

 

[Paul]

I've just built a new PC and the performance test results are abysmal.

I've used passmark performance test v5.0 (eval)

New build system is a

3.0GHz P4 Intel Prescott CPU 800FSB
2 x 512Mb Corsair 2-2-2-5 DDR400 supposedly performance memory (located in
the blue slots as per manual)
Asus P4P800-E board

test scores are
cpu passmark = 323
memory mark = 330

Bios settings are as follows
system information shows
processor type = Intel (R) Pentium (R) 4CPU 3.00GHz, speed 3000MHz, count 1
system memory size = 1024GB
further bios info shows cpu is set to 200 x 15 and is registering as 800 FSB

********
my older system is a

2.8GHz P4 Intel Northwood 533FSB
1 x 512Mb Nanya generic PC2700 DDR333
Asus P4PE board

test scores are
cpu passmark = 413
memory mark = 423

WTF!!! somethings not right here surely. Anyone a clue what the heck is
wrong?

If the CPU temperature gets to 70C or higher, the processor goes
into thermal throttle mode. It slows down the rate it computes,
by not using every clock cycle. It does this until the temperature
drops below the magic threshold.

Load up a copy of MBM5 (mbm.livewiredev.com) or Asus Probe (one or
the other, but not both at the same time), to display
the temperatures and fan speeds while the computer is running
an OS. You can then check what is going on when your benchmark
runs. Both programs have a "log to file" option, so you can
even run 3DMark, then later open the file and look at the temp
profile during the 3DMark run.

An Xbitlabs article claims that a P4PE, when run at FSB800
(overclocked) comes damn close to matching the latency of
a newer dual channel board. But since you were at FSB533,
that wouldn't hold true.

With the RAM you've got, you can select any "aggressive"
or "turbo" type options you want, as the RAM is fast enough
to take those kinds of settings. You have to mess around a
bit, to get PAT enabled, and you should use a Windows utility
to verify what settings are actually working. Something like
CPUZ or CTIAW might be useful.

You should also recognize that you are using a Prescott, and
it has something like a 30 stage pipeline, versus 20 stages
on the Northwood. If the flow of execution is changed, then
30 items get thrown away, versus 20 on the older chip. That
is one reason they put a larger cache on it, to compensate
for the pipelining losses. (There is bound to be some
pathological corner case, where your old processor might be
faster than the new one. But, make sure you've done everything
you can to tweak the new one.)

Where the Prescott design should shine, is it should have
allowed much higher clock speed processors to be built.
In other words, it was Intel's plan, to build a Prescott with
a clock speed that no Northwood could touch, but Intel has
apparently given up on making a 4.0GHz Prescott. But a number
of people have overclocked the existing Prescotts to 4GHz,
so it is in fact possible - it just takes decent cooling,
and a motherboard with a good Vcore design.

http://www.cpudatabase.com/CPUdb

Your performance is off far enough (30%), it has to be
temperature related. We've had a couple of people that simply
couldn't manage to get their CPU temperature down, and that
boggles the mind. Maybe when more LGA775 aftermarket coolers
are available, there will be more options for cooling the
Prescott. If your computer case only had one 60mm fan
on the back, I could understand this.

HTH,
Paul

 

[Pete D]

I get the same result with almost the same machine and my machine is
actually running fine. I would think that the problem is in the software not
your system.

 

[Johnny]

If the CPU temperature gets to 70C or higher, the processor goes
into thermal throttle mode. It slows down the rate it computes,
by not using every clock cycle. It does this until the temperature
drops below the magic threshold.

Load up a copy of MBM5 (mbm.livewiredev.com) or Asus Probe (one or
the other, but not both at the same time), to display
the temperatures and fan speeds while the computer is running
an OS. You can then check what is going on when your benchmark
runs. Both programs have a "log to file" option, so you can
even run 3DMark, then later open the file and look at the temp
profile during the 3DMark run.

An Xbitlabs article claims that a P4PE, when run at FSB800
(overclocked) comes damn close to matching the latency of
a newer dual channel board. But since you were at FSB533,
that wouldn't hold true.

With the RAM you've got, you can select any "aggressive"
or "turbo" type options you want, as the RAM is fast enough
to take those kinds of settings. You have to mess around a
bit, to get PAT enabled, and you should use a Windows utility
to verify what settings are actually working. Something like
CPUZ or CTIAW might be useful.

You should also recognize that you are using a Prescott, and
it has something like a 30 stage pipeline, versus 20 stages
on the Northwood. If the flow of execution is changed, then
30 items get thrown away, versus 20 on the older chip. That
is one reason they put a larger cache on it, to compensate
for the pipelining losses. (There is bound to be some
pathological corner case, where your old processor might be
faster than the new one. But, make sure you've done everything
you can to tweak the new one.)

Where the Prescott design should shine, is it should have
allowed much higher clock speed processors to be built.
In other words, it was Intel's plan, to build a Prescott with
a clock speed that no Northwood could touch, but Intel has
apparently given up on making a 4.0GHz Prescott. But a number
of people have overclocked the existing Prescotts to 4GHz,
so it is in fact possible - it just takes decent cooling,
and a motherboard with a good Vcore design.

http://www.cpudatabase.com/CPUdb

Your performance is off far enough (30%), it has to be
temperature related. We've had a couple of people that simply
couldn't manage to get their CPU temperature down, and that
boggles the mind. Maybe when more LGA775 aftermarket coolers
are available, there will be more options for cooling the
Prescott. If your computer case only had one 60mm fan
on the back, I could understand this.

Thank you for the reply - that gives me some things to look at - i'm going
to have to spend some time being methodical and investigating what is wrong
(surely there is something wrong) The cpu fan is the retail one supplied
with the processor - the case side is almost always open so i dont think
case temperature is an issue.
On another note one of the corners of the cpu cradle on the p4pe board that
anchors the cpu fan has broken, is there any possibility of obtaining or
purchasing a replacement?

 

[Johnny]

I get the same result with almost the same machine and my machine is
actually running fine. I would think that the problem is in the
software not your system.

That was my very first thought - the benchmark software simply isn't running
properly with the new hardware. However (and the system is working perfectly
otherwise) other tasks involving processing muscle are noticably slower on
the new machine. I've been building PC's for years (lots of them) so i'm not
exactly a novice (although with new hardware and bios setups its a damned
sight easier and straight forward than it used to be). I'm inclined to hope
it's simply a temperature related issue but need to spend a little more time
investigating when I find some.

 

[Johnny]

<snip>
If the CPU temperature gets to 70C or higher, the processor goes
into thermal throttle mode. It slows down the rate it computes,
by not using every clock cycle. It does this until the temperature
drops below the magic threshold.

I've yet to see the onboard bios monitor register anything much above 50C
after prolonged use. My first thought was that the benchmark software simply
wasn't utilising the new hardware but i've run transcoding programs whose
performance is also well below what i'd expected to see.

Load up a copy of MBM5 (mbm.livewiredev.com) or Asus Probe (one or
the other, but not both at the same time), to display
the temperatures and fan speeds while the computer is running
an OS. You can then check what is going on when your benchmark
runs. Both programs have a "log to file" option, so you can
even run 3DMark, then later open the file and look at the temp
profile during the 3DMark run.

I'm going to do this now thanks.

 

[Wim Zefat]

Either this software doesn't like the Prescott, or something is seriously
wrong somewehere with your system.

My system:
Asus P4P800-E
3.2 GHz P4 Northwood CPU 800 MHz FSB
2 x 512 Mb GEiL 2-3-3-6 memory (dual channel mode, PAT enabled)
no overclocking

CPU passmark = 514
Memory mark = 540

Wim

Op Sat, 6 Nov 2004 11:55:15 -0000 schreef Johnny:

 

[Johnny]

Either this software doesn't like the Prescott, or something is
seriously wrong somewehere with your system.

Yep, that's about where i'm at. The system works fine, the temperatures are
acceptable. 40C idle to 49-50C when crunching.

 

[Wim Zefat]

Op Sun, 7 Nov 2004 13:49:26 -0000 schreef Johnny:

Yep, that's about where i'm at. The system works fine, the temperatures are
acceptable. 40C idle to 49-50C when crunching.

Sounds about normal for a Prescott. Have you tried another benchmark
program for comparison, such as SiSoft Sandra? Although the processor and
memory results for my system are within a few percent of the baseline for a
3.2 P4 in Passmark, I get rather strange results for my graphics card -
which don't jive with the results from other benchmark programs. So maybe
its just the Passmark program..

Wim

 

[Johnny]

Op Sun, 7 Nov 2004 13:49:26 -0000 schreef Johnny:


Sounds about normal for a Prescott. Have you tried another benchmark
program for comparison, such as SiSoft Sandra? Although the processor
and memory results for my system are within a few percent of the
baseline for a
3.2 P4 in Passmark, I get rather strange results for my graphics card
- which don't jive with the results from other benchmark programs. So
maybe its just the Passmark program..

Well I just selected the turbo setting in BIOS and the board appears to have
died so I'm now in the process of dismantling it.

 

[Pete D]

That was my very first thought - the benchmark software simply isn't
running
properly with the new hardware. However (and the system is working
perfectly
otherwise) other tasks involving processing muscle are noticably slower on
the new machine. I've been building PC's for years (lots of them) so i'm
not
exactly a novice (although with new hardware and bios setups its a damned
sight easier and straight forward than it used to be). I'm inclined to
hope
it's simply a temperature related issue but need to spend a little more
time
investigating when I find some.

Mine does work well, rips a DVD in under 20 mins, plays Doom3 at high
settings at top frame rate, best of luck.