New computer slower than older computer...

Yep, I first realised after encoding a newly bought CD into mp3's.
So then I downloaded Sisoft Sandra and did some benchmarks.
First some details on the two computers.

The older comp is a one of those mini setups, a Soltek Qbic.
The same as this.
http://www.hardcoreware.net/reviews/review-214-1.htm
It has a P4 2.8GHz (Northwood), 1GB RAM (2 x dual channel DDR400) and
an ATI 9600 Pro. Mothorboard model is SL-B8E-F and uses the Intel 865G
chipset.

The newer comp is based around an Asus P5GDC-V.
CPU is a P4 3.2GHz (Prescott 640), 1GB RAM (2 x dual channel
DDR2) and video card is an ATI 800XL.

Some Sandra benchmarks...

CPU arithmetic benchmark.
Older comp - dhrystone = 8430
whetstone fpu = 3535
whetsone iSSE2 = 6202

New comp - dhrystone = 5176
whetstone fpu = 3440
whetsone iSSE2 = 3786

__________________________________

CPU multimedia benchmark.
Older comp - Integer = 21487
Float = 30684

New comp - Integer = 19046
Float = 18869

___________________________________

Memory bandwidth.
Older comp - RAM bandwidth Int Buff - 4271
- RAM bandwidth float buff - 4271

New comp - RAM bandwidth Int Buff - 4873
- RAM bandwidth float buff - 4887

___________________________________

Cache and memory benchmark.
Older comp - Speed factor = 13.2
New comp - Speed facotr = 8.2

___________________________________


As you can see, the CPU benchmarks slower. I really can't understand
it. Might I have gone wrong somewhere?

 

My first guess would be a setting in the BIOS like the front side bus is
set too low for your CPU and it's not running at 3.2GHz. Does CPU-Z (or
sandra) confirm the CPU is actually running at 3.2GHz?
good luck,
Ed

 

<<snip>>

I would take a look at processor temperatures. When the CPU
hits 70C, it will go into thermal throttle. That is where a
certain percentage of cycles are not used for computing, as
the processor is trying to keep the temperaure down.

If it was my computer...

1) If I could not afford to upgrade the Intel cooler, I would
be using a bit of good thermal paste between the heatsink
and the CPU. Just be very careful not to use so much that
you get it smeared into the LGA775 socket, as that cannot
stand any contamination. The Intel cooler will work, but
will likely be loud while doing it. The Intel cooler speed
is normally controlled by the air temp moving through the
fan. Make sure the fan gets the full 12V (no Qfan nor
rheobus devices in its path), so the fan can run as fast
as it needs to.

2) The case cooling is pretty important. You have to move a
lot of air through the computer case. Sometimes buying a
larger case will give you more vent area, and have more
room for fans. You can use a rheobus on the case fans,
and adjust them for the best compromise between noise
and effectiveness. The reason for doing this, is to get
the "warm cloud" around the processor heatsink pulled to
the outside of the case. I have a case with a 120mm fan
on the back, and the reason it doesn't work too well, is
because the case doesn't have enough vent holes to let
the fan do its work. I had to remove the plastic grillwork
on the front of the case, to get decent airflow. If
removing the side of the computer case lets the case fans
speed up, then you are vent-area limited.

3) A third party heatsink may help. Heatsinks like the
Thermaltake XP-120 and XP-90 family are pretty large.
The Zalman 7000 and 7700 are another family of products.
Some of these products need an adapter kit for LGA775,
so be careful to do your research first. With oversize
heatsinks, usually there is a clearance requirement between
the motherboard top edge and the PSU, in order that the
heatsink does not bump into the PSU. For the Thermaltake
products, you buy a separate fan, and sometimes it can be
hard to find a nice fan with a tachometer output for the
RPM readings.

Get a copy of Asus Probe to start, and have a look at your
temperatures. A copy of Prime95 (using "torture test" option)
can be used to run the CPU at 100%. You will most likely
find you have a high idle temperature to start with, and
then the CPU will be "bumping its head" against the 70C
ceiling.

Paul

 

Thanks Ed. I had a look at that and it all checks out.

 

I think your onto something Paul.
It's a bit odd though because the temp does show as over 70C in the
bios but if I use a util in Windows such as Asus PC Probe it shows
that after being on for about 10 minutes and with pretty much only
Forte Agent running the CPU is at 42C and the MB at 42C.
Strange that the bios reports a different temp.

I actually ordered a CPU cooler last week. It'll probably arrive in
the mail this week.

I do have a spare larger case. I'll see how the new CPU cooler goes
first.

Thanks for your help.

 

With regard to your temps, that means you are idling at 42C.
By measuring current consumption with my ammeter, I've determined
that operating in the BIOS screens is roughly a 50% CPU usage, so
your 70C is happening at say 50% CPU loading. And, at that point,
the CPU is already throttling. That means when you run a benchmark,
the CPU is going to be throttled quite a bit.

Install your new cooler and keep your copy of Asus Probe handy.
Get a copy of Prime95 (mersenne.org) and run the "torture test"
option. With your new cooler, your CPU temperature must stay below
70C, while Prime95 is running, in order to see good benchmarks
after your thermal testing is complete.

In terms of thermal performance, the Intel cooler could be
about 0.33C/W thermal resistance. That means, if the processor
draws 100W of power, the processor will be 33C hotter than the
case air temperature. If room temp is 25C, case air temperature
is 32C, then with that cooler the CPU would be at 65C. Heatsinks
like a Zalman 7000 or 7700, or a Thermaltake XP-120, have a
thermal resistance down around 0.2C/W - generally the rule of
thumb is that you want a heatsink which is larger than the footprint
of the CPU socket. The toughest thing with large heatsinks like
that, is determining whether they will fit or not before you buy
them. If the heatsink is no larger than the Intel heatsink,
it will likely have the same 0.33C/W thermal resistance as well.

A delta of 7C (10F) is a rough target for good case air
ventilation. I.e. The motherboard temp should read 32C, if
the room air temp reads 25C. If you cannot get there (say the
case air is 40C), consider adding ducted air flow down to the CPU
heatsink/fan. That may help a bit. I got a 6-7C reduction in case
temperature, just by making more vent area available on the front
of my case. (Try to balance inlet and exhaust CFM ratings, for
best results. If you put an 80mm inlet fan on the front, put an
80mm exhaust fan on the back as well. This helps control dust,
expecially if you have removed the dust filter to improve air
flow. When counting fans, the PSU fan doesn't count as a fan,
as it is too slow to move appreciable air.)

Another test for case cooling, is to note the CPU temperature with
the computer case side on, then remove the side of the computer
case and measure the CPU and motherboard temperature again. If
there is a big improvement with the side off the case, that means
you need to improve case cooling. After all, forcing air through
the case, should work better than convection cooling with the
side off.

If you still aren't meeting reasonable temperature targets, water
cooling is another alternative. Place the radiator outside the case,
for best results. (I have seen case modding articles, where the
radiator was kept inside the case. Which means the case fans still
have to work hard to keep things cool.)

Paul

 

Thanks again Paul.
I'll report back when I get the CPU cooler.
I ordered one of these...
http://www.phoronix.com/scan.php?page=article&item=114&num=1
http://hi-techreviews.com/nuke/primecooler4/primecooler4_page1.htm

I hope it does the trick. I think the spare larger tower case I have
might also be needed.

 

It seems likely that if both MB and CPU are reported as "42C," then
one of them is incorrect. (I'm not suggesting that the CPU is
necessarily hotter than 42, though, because 42 seems awfully hot for a
MB.) Can you double check the results with Motherboard Monitor 5?

Ron

 

HA ha! how will you explain that to your wife. hehe... nono i'm just
having fun.

Maybe these two computer don't have the same chipset. i would
recommend to get an other one.

Or maybe work around your drivers like you did with your old system.
If you did so then your cpu is slower than the old one nut has the
ability to handle more bandwidth. It like throubered and a barton you
don't have the same speed for the same number like the t-bred at 2400
= a 2600 barton but it has less instruction than the barton.


a+

 

Could it be that your old machine had hyper threading turned off and you
new one has it turned on?

Not being an Intel user I don't know what the impact of hyper threading
is on bench marking.

 

Tony et al:
Greetings and hallucinations from just north of Fantasy Land (Washington,
DC)!
Have you checked that the BIOS is set to actually run the system full out? I
had that problem with a friend's ABIT board a while back. It could not find
the CPU in the BIOS, so it defaulted to a 1.1 GHz machine rather than the
2.4 GHz one that it was supposed to be. A new BIOS helped there.
Peace,
Paul
<snipped>