Intel Application Accelerator 2.2 and 2.3 Beta have a nonpaged pool memory leak
Dan Harkless
hardware/software combinations (like mine) Intel Application
Accelerator 2.2 and 2.3 Beta have a bad nonpaged pool memory leak that
renders your system useless within a matter of days, requring frequent
reboots.
My system was built from components within the last several weeks, and
includes a Gigabyte GA-8IRXP motherboard (running the latest BIOS, F7)
with a P4 1.6A CPU, Matrox G550 graphics card, and Creative Labs
Audigy eX sound card. It is currently running Win2K Pro SP3, but the
memory leak also occurred under SP2. There are two DVD-ROM drives on
the secondary IDE channel and nothing on the primary IDE channel.
System drive is a RAID 1 array connected to the onboard Promise 20276
controller.
I was having problems where after the machine was up for a few days,
various pieces of software would start getting weird failures. I
found this message:
The server was unable to allocate from the system nonpaged pool
because the pool was empty.
in the System Log, so I looked for memory leaks. All the applications
appeared to be fine, which meant a driver memory leak.
I installed the Windows 2000 Support Tools from the \Support directory
on the Win2K install CD. I ran poolmon.exe and found that tag SpDN
was constantly allocating 1920 bytes (sometimes more), and was never
giving them back. For instance, yesterday at 4:42 PM, shortly after
rebooting, the poolmon line for SpDN was:
Tag Type Allocs Frees Diff Bytes Per Alloc
SpDN Nonp 3501 (40) 1768 (20) 1733 166368 (1920) 96
By today at 1:56 AM, it was up to:
SpDN Nonp 269357 (40) 134696 (20) 134661 12927456 (1920) 96
To find out which driver SpDN was associated with, I went to the
C:\WINNT\system32 directory in a Cygwin bash window and typed:
strings --print-file-name *.[sS][yY][sS] drivers/*.[sS][yY][sS] \
2>/dev/null | fgrep SpDN
and found that the only driver containing that string was
IdeChnDr.sys, which a little searching turned up as being part of the
Intel Application Accelerator 2.2 I had installed.
I found a Usenet post mentioning that 2.3 Beta was out, so I
uninstalled 2.2. Prior to installing 2.3 Beta, I verified with
perfmon that Pool Nonpaged Bytes was no longer constantly rising. I
then installed 2.3 Beta. Unfortunately, it does not fix the memory
leak (not too surprising since there were no memory leaks mentioned in
its Release Notes' Issues Resolved section).
I had been running Win2K SP2 (plus all Windows Updates), so I tried
applying SP3, but this didn't help either.
I have reported this bug via the form at
<http://support.intel.com/support/chipsets/iaa/feedback.htm>, so
hopefully Intel will be able to fix this problem prior to the final
2.3 release.
I'm slightly surprised I haven't seen any reports of other people
experiencing this leak. Wonder what it is about my setup that's
triggering it...
--
Dan Harkless
use...@harkless.org
http://harkless.org/dan/
RAV
"Dan Harkless" <use...@harkless.org> wrote in message
news:4189894b.02081...@posting.google.com...
Pete Ulrich
"RAV" <r...@nospamplease.com> wrote in message
news:ulvk434...@corp.supernews.com...
> Sorry for being a newbie. What does this software do?
Intel calls it a 'Storage Controller' and it speeds up data transfer.
However, as with this poster, some have reported problems. Intel claims
that none of the problems are the fault of their software.
RAV
did. Sounds like there is no good reason for the install if it is
susceptable to problems.
"Pete Ulrich" <pmul...@netnitco.net> wrote in message
news:3d60...@news.netnitco.net...
Jeff Cochrane
maybe you :)
It is not required at all, so why bother installing it?
Jeff C.
"Dan Harkless" <use...@harkless.org> wrote in message
news:4189894b.02081...@posting.google.com...
---
Bhgtbvat znvy vf pregvsvrq Ivehf Serr.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.381 / Virus Database: 214 - Release Date: 02/08/2002
PDRR_Engineer
One of the machines I have here is a Dell system, P4 1.7, using the i850
chipset. I have had the IAA v2.2 installed for some time, but cannot say
that I have seen any application anomalies indicative of a memory leak.
Your post is intriguing however. Maybe I'll install the poolmon utility and
monitor the output for a while.
Art...
"Dan Harkless" <use...@harkless.org> wrote in message
news:4189894b.02081...@posting.google.com...
John
I installed the application accelerator on my 8IRXP - it comes on the
motherboard CD so I assumed many people would have it installed - its just
there in the menu along with all the other drivers as I recall - I just
thought we were supposed to install it along with all the other stuff on the
CD
"Jeff Cochrane" <spam...@aol.com> wrote in message
news:ajpjlc$1c8jbh$1...@ID-135764.news.dfncis.de...
MC
I built my PC, no RAID though.
Intel 850EMV w/ onboard LAN and sound, 2MHz, 256MB,
2 80G Maxtor's on IDE1, 2 cd devices on IDE2, XPpro.
Mark
MC
have a very nice IDE info applet showing channel and device
properties (ie, exactly what mode(s) the IDE devices support
and are running at (and more info).
Plus, Intel says this:
1. Intel® Application Accelerator Driver - This technology increases the performance of the Input/Output subsystem transfer rate,
greatly enhancing the system speed.
2. Intel® Advanced Pre-Fetch Module - With this technology, available for Pentium® 4 processor based systems running Microsoft*
Windows* 2000 Professional and Windows XP only, an overall additional performance enhancement is realized for Windows 2000 and
Windows XP - based applications.
Again, no problems here with IA 2.2, XPpro, Intel 850EMVL (no raid).
"Jeff Cochrane" <spam...@aol.com> wrote in message news:ajpjlc$1c8jbh$1...@ID-135764.news.dfncis.de...
John
Have you checked for memory leaks?
"MC" <please...@news.group> wrote in message
news:lW789.51$864....@dca1-nnrp2.news.algx.net...
T Thome
He might be on to something though... don't dismiss it out of hand.
I recently upgraded two PC's both with Intel chipsets. Installed IAA 2.2 and
after about 3-4 days of run time, they would be locked up with an exception,
usually something in user.exe...
1) Soyo P4-IS2 w/P4-1700 Win98SE
2) Intel D815EPEA2 w/P3-866 Win98SE
They both have ICH2 IDE controllers...
I've got a couple of BX boards with P3's, no problems - similar config,
different chipset.
I've backed out the IAA drivers and will run with the MS Drivers for a few
days and see if this makes any difference...
tim
Eric Gisin
did fix some bugs the MS Win2K driver had.
Is it possible to rip out just the driver from IAA? It's just an INF and two
SYS files.
"Jeff Cochrane" <spam...@aol.com> wrote in message
news:ajpjlc$1c8jbh$1...@ID-135764.news.dfncis.de...
Dan Harkless
> The reason that no-one has reported it is because nobody uses it........Except
> maybe you :)
An 'intel "application accelerator" OR iaa' Google Groups search turns
up 2,260 posts, so I don't think it's just me. Anyway, I suspect that
if it's not used that much, it's generally because people don't know
about it, not because they have something against it.
> It is not required at all, so why bother installing it?
Because one of the things I'm going to be using this machine for is
uncompressed digital video capture and editing, which requires lots of
IDE bandwidth and speed.
The one thing I'm not sure about is whether the IAA drivers affect IDE
RAID drive accesses (I have my system drives in a RAID 1 off of the
builtin Promise controller and the data drivers in RAID 0 off of a
3ware card).
I'm guessing probably not -- probably only the Promise and 3ware
drivers would be involved in those cases, which would mean IAA would
only be helping I/O to and from my DVD drives, and the bump in speed
there would likely be unnoticeable (though I've read posts where
people said IAA fixed jumpy DVD playback).
Eric Gisin
|
| Because one of the things I'm going to be using this machine for is
| uncompressed digital video capture and editing, which requires lots of
| IDE bandwidth and speed.
|
People do occasionaly report being stuck in PIO mode with the MS drivers.
| The one thing I'm not sure about is whether the IAA drivers affect IDE
| RAID drive accesses (I have my system drives in a RAID 1 off of the
| builtin Promise controller and the data drivers in RAID 0 off of a
| 3ware card).
|
| I'm guessing probably not -- probably only the Promise and 3ware
| drivers would be involved in those cases, which would mean IAA would
| only be helping I/O to and from my DVD drives, and the bump in speed
| there would likely be unnoticeable (though I've read posts where
| people said IAA fixed jumpy DVD playback).
|
The App Accellerator is a layer well above the device driver, and does
prefetch during EXE/DLL loading.
Windows XP has something similar, so it is probably redundant for that OS.
Dan Harkless
> | The one thing I'm not sure about is whether the IAA drivers affect IDE
> | RAID drive accesses (I have my system drives in a RAID 1 off of the
> | builtin Promise controller and the data drivers in RAID 0 off of a
> | 3ware card).
> |
> | I'm guessing probably not -- probably only the Promise and 3ware
> | drivers would be involved in those cases, which would mean IAA would
> | only be helping I/O to and from my DVD drives, and the bump in speed
> | there would likely be unnoticeable (though I've read posts where
> | people said IAA fixed jumpy DVD playback).
>
> The App Accellerator is a layer well above the device driver,
I see. I got the impression IAA was a replacement ATA driver
specifically for Intel's chipsets (vs. Microsoft's general-purpose
one). Reinforcing this impression was Intel's statement "Replacing
the Intel® Ultra ATA Storage Driver, the Intel Application Accelerator
supports..." and the fact that if you look in the Device Manager, the
IAA drivers get applied to the ATA controller.
If IAA is indeed at a higher level than the normal ATA drivers, I'd be
tempted to think it might indeed speed up my Promise and 3ware RAID
drive accesses, but again, the fact that the driver is tied to the
motherboard's primary ATA controller would seem to discount this
possibility.
Either way, Intel could certainly be more clear in their documentation
as to exactly what this thing does, on a technical level.
> and does prefetch during EXE/DLL loading.
Intel's <http://www.intel.com/support/chipsets/iaa/> site seems to
indicate that it accelerates general-purpose disk access, not just DLL
and EXE loading.
Arge
ago, and have not had any problems what so ever.
Bob
"RAV" <r...@nospamplease.com> wrote in message
news:um0978...@corp.supernews.com...
SilverBack
Simple example for me to look at was running a common benchmark for my video
card. I ran 3D Mark 2001 without the Application Accelerator, score 7981.
Installed it and reran the test, 13053. Big difference.
"RAV" <r...@nospamplease.com> wrote in message
news:um0978...@corp.supernews.com...
SilverBack
The Intel® Application Accelerator is a new performance software package for
Intel desktop PCs.
Intel Application Accelerator also provides:
Faster Boot Time via Accelerated Operating System Load Time
Accelerated Disk I/O for Games, Graphics Applications, Disk Utilities, and
Media Authoring Applications
Performance-enhancing Data Pre-fetcher for Intel® Pentium® 4 processor-based
systems
Support for 137GB and larger hard drives
Requires:
a.. Must have previously installed the Intel® Chipset Software Installation
Utility on your system
a.. Windows* Millennium Edition (Me), Windows Second Edition (SE), Windows
98, Windows 2000, Windows NT* 4.0, or Windows XP Operating System
a.. A motherboard with a supported chipset
a.. An Intel® Pentium® 4, Intel Pentium III, Intel Pentium II, Intel
Celeron®, Intel XeonT, Intel Pentium III Xeon, or Intel Pentium II Xeon
processor
a.. Ultra ATA/66 or Ultra ATA/100 compatible logic either on the
motherboard, or on an Ultra DMA PCI adapter card
a.. Ultra DMA compatible BIOS
a.. Ultra ATA/66 or Ultra ATA/100-compatible IDE device such as a hard drive
or CD-ROM
a.. A 40-pin, 80-conductor IDE cable. (A 40-pin 40-conductor IDE cable can
be used with the Intel® Application Accelerator, but the 40-pin 40-conductor
cable will limit the device to a maximum transfer rate of Ultra ATA/33)
Supported Chipsets
The Intel® Application Accelerator supports the following Intel® chipsets:
a.. Intel® 810 Chipset
b.. Intel® 810E Chipset
c.. Intel® 810E2 Chipset
d.. Intel® 815 Chipset
e.. Intel® 815E Chipset
f.. Intel® 815EP Chipset
g.. Intel® 815P Chipset
h.. Intel® 820 Chipset
i.. Intel® 820E Chipset
j.. Intel® 840 Chipset
k.. Intel® 845 Chipset
l.. Intel® 850 Chipset
m.. Intel® 860 Chipset
"Dan Harkless" <use...@harkless.org> wrote in message
news:4189894b.02081...@posting.google.com...
anodos
IdeChnDr.sys (I think I got that right).
Anyway, I believe it is either related to reading/writing near the end
of the disk/partition, or just happens after a certain amount of use.
I know that when my drive is nearly full, I got the BSOD quite often.
It will occur almost every time I attempt a full system backup or a
disk defrag.
use...@harkless.org (Dan Harkless) wrote in message news:<4189894b.02081...@posting.google.com>...
Anony
Intel Application Accelerator
http://www.intel.com/support/chipsets/iaa/
"anodos" <sir_a...@yahoo.com> wrote in message
news:c69843c.02082...@posting.google.com...
geno_...@tin.it
Prior to IAA2.2.2 I never had such an issue where the system seemed to freeze, almost dead-locked
and lost drag&drop functionality when recovered thru ctrl-alt-del switching to task manager or
logging off and relogging in (but the drag&drop remains killed forever until you reboot then).
It happened with Win2K SP2SRP1 and Win2K SP3, both installed from scratch.
Geez! I lost 3 days and I've got a lot of work to do. Sigh!
Now, IAA 2.2 seems to work flawlessly under 98SE because I can even capture 50mins+ AVIs at 6MB/s
with the StudioDC10+ card and no more than 1 frame lost during capture.
However under Win2K I was already experiencing something a bit weird regarding mouse movements.
Something that never happened on another system based on ASUS P4B266 i845D and P4 1.60A up to IAA2.1
at least.
But that behaviour was on the P4B533 i845E P4 2.26B system with IAA 2.2 and Win2k SP2SRP1.
I think I'll revert back to IAA2.1 which seemed to be more stable on the P4B266.
I've done installation images just to be sure to recover in a short time if anything goes wrong.
geno_...@tin.it
application errors appearing sometimes--however 98SE it's surely not stable per se).
IAA 2.2.2 under Win2K SP2SRP1 as well SP3 almost causes a deadlock while IAA 2.2 caused weird mouse
behaviour and I was assuming that Logitech drivers were the cause but it wasn't the case.
I think I'll revert back to IAA 2.1. Without IAA at all the system is way slower than what it could
really do. It's unfortunate that Microsoft OSes don't come with Intel chipsets optimizations in
bundle, not even on additional ServicePacks.
All of this happening on a ASUS P4B533 i845E ICH4 P4 2.26B system.
A P4B266 i845D ICH2 P4 1.60A system with IAA 2.1 never showed any symptoms like this.
Because on P4B533 the com ports I've connected the logitech mouse to are interfaced with the Intel
LPC controller then I suspect that IAA2.2 already does something weird under Win2K more than 98SE
(where other than some applications crashing maybe sometimes more than the usual rate..but not by
much because DC10+ can capture at 6MB/s for 50mins+ with no troubles and no lost frames) BUT IAA
2.2.2 it's surely heavily buggy.
I hope Intel can solve these bugs and make their IAA reliable because it's surely not right now.