P5V800 stability with Pentium D 920

Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

Threaded View
Time to chuck this mobo/CPU and start over, or can this be made to

I put together this setup a couple of years ago, based on an ASUS
P5V800-MX board.  Last year I upgraded the memory to 1.5 gig Corsair
DDR400 and got a Corsair HX520W power supply.  I also added a 500 gig
Seagate HD planning to replace the old 160 gig (wound up leaving the
old one in and switch booting from one to the other as needed).  I
installed the lastest BIOS.  Xp home has all the updates.

I have the larger drive setup to boot Linux (Ubuntu) or Xp, and I use
them both.  The old drive boots Xp.

The PC's been running okay with a Celeron D 2.8, although I've never
been thrilled with the performance.  I just got a used dual core
Pentium D 920 cheap on eBay and installed it.  As expected, it works a
lot faster.  I like it.  However, it will do a spontaneous reboot if
stressed.  Unacceptable.  This is not an Xp problem because it will do
the same thing in Linux -- load up a few resource-hungry applications
and start doing things and it will fail.  Both cores are working and
show activity under Linux and Xp.  Good, big Zalman CPU heat sink/fan
-- heat doesn't seem to be a prob.

Since I've only had the new CPU for a day or so, I'm not ready to give
up yet but I can't afford to spend much more time with it.  I also
can't justify spending much money on it either.  I figure if I spent
$200 on a new mobo/CPU/Memory combination and sold off the Celeron,
the D 920, mobo, and memory for $75 or so, I might justify going out-
of-pocket $125.  That's what I'm trying to decide right now.

Anyone here have an idea about how to stabilize the system?  BIOS
tweaks, or other tweaks/patches?  I need to decide quickly if this can
be made to work or not.

Thanks in Advance, Alan.

I've copied below output from PC Wizard, in case that gives you any

Voltage, Temperature and Fans
 Hardware Monitoring :    ITE IT8712F
Voltage CPU :    1.21 V
DIMM :    3.28 V
+3.3V Voltage :    2.94 V
+5V Voltage :    6.85 V
+12V Voltage :    11.19 V
Chassis Fan :    2657 rpm
Processor Temperature :    40C
Mainboard Temperature :    36C
Hard Disk Monitoring :    S.M.A.R.T
Hard Disk Temperature ST3500630AS :    48C
Hard Disk Temperature ST3160023A :    45C

System Summary
 Mainboard :    Asus P5V800-MX
Chipset :    VIA P4M800CE
Processor :    Intel Pentium XE 920 @ 2800 MHz
Physical Memory :    1536 MB
Video Card :    ATI Technologies Inc Radeon 9600 XT (V350)
Hard Disk :    ST3160023A (160 GB)
Hard Disk :    ST3500630AS (500 GB)
CD-Rom Drive :    PLEXTOR CD-R   PX-W1210A
DVD-Rom Drive :    TOSHIBA DVD-ROM SD-M1302
Monitor Type :    ViewSonic PF815 - 20 inches
Network Card :    VT82C570 MV IDE Controller VT6102 Rhine II Fast
Ethernet Adapter
Operating System :    Microsoft Windows XP Home Edition 5.01.2600 Service
Pack 2
DirectX :    Version 9.0c

LPC bus :    Yes
PCI Bus :    Yes
Bus PCI-Express :    Yes
USB Bus :    Yes
SMBus/i2c Bus :    Yes
Bus HyperTransport :    Yes
Bus CardBus :    Yes
Bus FireWire :    No

 General Information :
Type :    Intel Pentium XE
Internal Specification :    Intel(R) Pentium(R) D CPU 2.80GHz
Model Number :    920
Codename :    Presler
Revision :    B1
Technology :    0.065=B5
CPU ID :    F.6.2
CPU IDEx :    F.6.2
Microcode :    MU0F620F

 Instructions :
IA-64 Technology :    No
Intel64 (EM64T) :    Yes
FPU128 :    No
SSE5 :    No
SSE4a :    No
SSE4.2 :    No
SSE4.1 :    No
S-SSE3 :    No
SSE3 :    Yes
SSE2 :    Yes
SSE :    Yes
Extended 3DNow! Technology :    No
3DNow! Technology :    No
3DNOW Prefetch :    No
3DNow! Pro Technology :    Yes
AMD MMX Technology :    No
MMX Technology :    Yes
Cyrix MMX Technology :    No
CLF - Cache Line Flush :    Yes
CX8 - CMPXCHG8B :    Yes
CX16 - CMPXCHG16B :    Yes
CMOV - Conditionnal Move Inst. :    Yes
MON - Monitor/Mwait :    Yes
POPCNT :    No
RDTSCP :    No
SEP - Fast System Call :    Yes

 Miscellaneous :
XD - No-execute Page :    Yes
VT - Vanderpool Technology :    Yes
TXT - Trusted Execution Techn...    No
SVM - Secure Virtual Machine :    No
FPU - Co-processor Built-in :    Yes
FXSR - Fast Float Save & Rest...    Yes
xTPR - Send Task Priority :    Yes
DAZ - Denormals Are Zero :    Yes
FFXSR :    No
ExtApicSpace :    No
3DNow! Technology :    No
PBE - Pend. Brk. EN. :    No
LAHF - LAHF/SAHF Inst. :    No
ABM :    No
MASSE - Misaligned SSE :    No
OSVW - OS Visible Workaround :    No
IBS :    No
P1GB - 1GB Page Size :    No
WDT - Watchdog Timer :    No

 Features :
VME - Virtual Mode Ext. :    Yes
DE - Debugging Extension :    Yes
PSE - Page Size Extension :    Yes
TSC - Time Stamp Counter :    Yes
MSR - Model Specific Registers :    Yes
PAE - Physical Address Extens...    Yes
MCE - Machine Check Exception :    Yes
APIC - Local APIC Built-in :    Yes
MTRR - Memory Type Range ...    Yes
PGE - Page Global Enable :    Yes
MCA - Machine Check Architec...    Yes
PAT - Page Attribute Table :    Yes
PSE36 - 36-bit Page Size Exten...    Yes
PSN - Unique Serial Number :    No
DS - Debug Trace & EMON Sto...    Yes
SS - Self Snoop :    Yes
ACPI - Software Clock Control :    Yes
TM - Thermal Monitor :    Yes
TM2 - Thermal Monitor 2 :    No
EST - Enhanced SpeedStep T...    No
HTT - Hyper-Threading :    Yes
SBF - Signal Break on FERR :    Yes
DSCPL - CPL qualified Debug S...    Yes
CID - Context ID :    Yes
LT - LaGrande Technology :    No
PDCM :    Yes
DCA - Direct Cache Access :    No
EPS - Enhanced PowerSaver :    No

 Features Hyper-Threading :
Technology :    Yes   -   Disabled

 Features Multi-Core :
Physical Processor #1 (Core #1) :    Apic ID 0
Physical Processor #1 (Core #2) :    Apic ID 1

 Power Status :
Digital Thermometer :    No
Operating Point Protection :    No
Dynamic FSB :    No
Dynamic Acceleration :    No

 Addressing Information :
Physical Addressing max. :    36-bit
Linear Addressing max. :    48-bit

 Advanced Settings :
In Order Queue Depth :    12
Fast-Strings :    Yes
x87 FPU Opcode :    No
Thermal Monitor :    No
Split-Lock :    Yes
Performance Monitoring :    Yes
Prefetch Queue :    Yes
Branch Trace Storage :    Yes
C1E - Enhanced Halt State :    No
Temperature Control :    66C

 Mainboard Upgradeability :
Socket/Slot :    LGA 775
Upgrade interface :    Unspecified
Supported Speed :    2800 MHz (or more)
Supported Voltage :    1.2V

Re: P5V800 stability with Pentium D 920

Quoted text here. Click to load it

This sounds like a PSU problem.  The Pentium D is a considerably
greater load than the Celeron it replaced. Looking at the voltages
seems to confirm this - none of the key voltages are within the 5%
tolerance called for by the spec.  Whilst even a 10% variation is
usually here nor there, your 5V supply is way above even that.
For 5V to be that high suggests that the PSU is overloaded at 3.3V
which is consistent with the greater load of the Pentium.  Try
replacing the PSU with something beefier and see if that resolves
the problem.

Andrew Smallshaw

Re: P5V800 stability with Pentium D 920

Quoted text here. Click to load it
Thanks, Andrew, for the thoughts.  I don't think you are correct, but
it did help get me going to solve this -- not quite solved though.  I
have more information ...

I called Corsair tech support.  The guy was skeptical of the voltages
reported by the Windows program.  He said the only way to be
completely confident with the voltages reported was with a volt
meter.  However, he also said to look at what it said in the BIOS.  I
checked there, and the voltages reported were all fine -- very close
to spec.  5v was 4.94 and so on.  He said the 520 was plenty of power
for what I had in my system.  They would do the RMA if I wanted but he
really didn't think that was the problem.

He also suggested running memtest.  It turns out that I also have
Corsair memory, and I did that.  I had run memtest overnight (fine)
when I last upgraded the memory a year or so ago.  So, I ran it just
now and got failures (multiple failures at same address).  I pulled
out one module (the 512) and ran it again (with the remaining 1 gig
module) and found no errors with about 8 minutes of testing.  So, I
got an RMA on that memory module, and was expecting that it would have
solved the prob.  Not quite.

The system seemed to run better without the offending 512 module, and
I was able to load it up with quite a few programs running at once,
but I was eventually able to make it fail.  The tech support guy also
suggested running orthos in safe mode to stress the system.  That ran
for about 10 minutes before the system failed.  So, it's running
better, but not totally reliably.

I'll run memtest overnight on the 1 gig module.  I'm open to
suggestions, but leaning toward using it for now (without stressing it
too much), and then replacing it (mobo/mem/cpu) soon.

Thanks, Alan

Re: P5V800 stability with Pentium D 920

On Fri, 14 Mar 2008 16:21:30 -0700 (PDT), dechert@gmail.com

Quoted text here. Click to load it

Agreed, that PSU is more than enough for the system as
described.  It is not reliable to take a 3rd party software
application and assume it is even reading the right *things*
to determine voltage level, only after comparison to a
multimeter reading (keeping in mind the software tends to
report lower than actual values unless there is an
artificial offset used and then the actual reading may be
closer to correct but the margin for error actually
increases in some cases), OR if others have already
determined whether the specific software correctly works
with the board, then one has found a software suitable for
later monitoring purposes.

However, that a model of PSU should be enough doesn't rule
out that the particular specimen could be faulty but based
on what you wrote below it seems not to be the problem or at
least not clearly so.

Quoted text here. Click to load it

The remaining question is do those addresses ALWAYS fail in
every kind of test (every memtest test #).  If they always
fail it indicates bad addresses, but if not it tends to
implicate general memory subsystem instability which is not
necessarily damning of the memory because when you installed
the faster processor it most likely used a faster FSB &
memory bus than the Celeron did.  One potential resolution
to this is trying a different board bios to increase memory
compatiblity.  Another is to try manually setting the memory
speed and timings to more conservative values.

Sometimes memory is ok, that it meets JEDEC specs per it's
rating, but after the windows of opportunity for returning
it to the seller is gone the manual timings option would be
needed, and if you RMA and receive same memory again it
could similarly have the problem if it were the board that
was to blame.

Quoted text here. Click to load it

If you pull one and then no errors, it doesn't implicate the
other yet.  Next you should have put the other module in and
pulled out the one currently installed to see if errors
remain.  Even then, if the board is to blame it can be that
having only one or the other modules installed would be ok
but having two modules causes errors.  There are a few other
tweaks that can sometimes improve memory stability including
raising memory voltage in bios, raising MCH voltage (either
of these should only be raised a small amount if any), and
relaxing the memory timings a bit.  Reducing the memory bus
speed should be the last resort especially if it had to be a
ratio change resulting in a large reduction in bus speed
because then you aren't quite getting the performance you
paid for.

Quoted text here. Click to load it

Orthos will fail from CPU instability, including CPU
overheat, OR from memory and/or northbridge instability.  
Before booting windows you should run memtest86+ for many
hours, like overnight, and only if that has zero errors then
proceed to boot windows and run other tests.  Personally I
like to ensure some margin, to manually set the memory speed
(or FSB speed when a ratio) to a few percent, perhaps 5%,
higher speed for the test, as this will tend to help find
borderline stability issues and I do want any system to have
some margin for stability not just barely stable as systems
do tend to degrade slightly over time and sometimes
memtest86+ will not catch errors if a system were barely
instable rather than grossly so.

By increasing memory speed (keeping timings set to manual
but not changing them yet, keeping set to manual just helps
to ensure the bios didn't apply different timings when you
increased the bus speed) you tend to find more errors that
would otherwise be unfound.  Granted, you could even
introduce errors that wouldn't have occurred when raising
the bus speed but IMO any system should have that 5% or so
margin in how it's running and of course you would return
the system to stock speed after the memtest86+ test passed
with no errors.

Quoted text here. Click to load it

Inspect the motherboard for failing capacitors, particularly
the large ones around the CPU socket.  Vented or domed tops
or leaky residue around top or bottom are common indicators.

Try manually lowering the FSB speed to what it was when the
Celeron was running, only temporarily as a test.  Next run
the Memtest86+ overnight and Orthos.  I would not "lean
towards using it for now" if you have any errors at all,
assuming you have the Celeron still it can be reinstalled so
you have a stable system until you get a chance to resolve
this in your free time - but I can't know how important or
unimportant your time or the stability is, different people
have different needs per system.  It is just troubling to
think that you could corrupt data in your windows
installation while running a combo that isn't quite stable.

Especially, do not defrag the hard drive until you are
certain, after a period of time, that the system is fully
stable, because if you did corrupt files including windows
OS files it becomes much more of a PITA to find the root
cause of the error with multiple things going wrong.

You did not mention your CPU temperature.  If the heatsink
isn't mounted good or the case has poor ventilation these
can cause errors, particularly after having ran for awhile
to heat up but it doesn't take long for CPU to heat up
running Orthos, or if it had been running for awhile before
starting the Memtest86+ testing.

You haven't mentioned what settings your board bios
supports, but check on a bios update regarding memory
compatiblity and if there is one put the Celeron back in,
confirm stability through testing, then flash the newer
bios.  If there is no newer bios then if the current bios
supports reducing memory bus speed that is what I would try
first, remembering that it doesn't necessarily indicate
faulty memory unless always the same memtest86+ addresses
are in error.

You could be one of the rare people who simultaneously has
more than one problem, possibly a combination of memory
instability and CPU overheating, had you ever previously ran
Orthos to see how much your temps rose?  I mean prior to
installing the new CPU?

Re: P5V800 stability with Pentium D 920

Quoted text here. Click to load it
Okay, case closed.  All is well and my system appears to be crash-
proof.  Thanks to all for ideas. Kony, your hunches were close, mainly
about memory timings.  It looks like there are no hardware problems at
all.  I conclude that there were some memory timing issues that only
surfaced with the faster processor installed.  Even with slower memory
timings I have now set, the overall performance is still much greater
than the Celeron -- most things seem twice as fast.  Benchmarks seem
to be on par with other dual core systems.

After *much screwing around* this was resolved by loosening memory
timings.  Here are timings that worked:

cas 2.5 -> increased to 3.0
Trp 4T -> increased to 5T
Tras 7T increased to 8T
Trcd 4T increased to 5T

Slowing the memory from 400 to 333 did not help.  I wound up putting
it back to DDR400.

There never was any heat problem. As I mentioned, I have a big Zalman

Orthos cannot crash the system running in safe mode by itself, or
running as one of many programs in a normal Windows session. Linux is
running fine, too.  Memtest is running without errors.

Thanks, Alan.

Site Timeline