Problems with Epox EP-8K7A+, Radeon 9000 Pro, and AGP 4x?
Gary Funck
Radon 9000 "Evil" Pro video card. The system has a 1.2G Athlon, and 266 MHz
FSB running 512 M of generic DDR with ECC enabled (but no scrubbing).
Nothing overclocked, but the memory timings were tweaked a little; two hard
drives running in RAID-1 (mirror) mode. Had been running fine for over a
year, very stable. After adding the video card in its default mode, 4x AGP,
128 M AGP aperture, the system would generally boot fine, but eventually
would just hang, where the system wouldn't accept mouse or keyboard input,
but some things kept running (like sound playback from an MP3). The system
could be forced to fail predictably and quickly by doing something like a
DVD playback while ripping an audio CD at the same time (on a separate CD-R
drive). After some experimentation, I found that AGP had to be run in 1X/2X
mode, *and* "System Performance" had to be set to "Normal" (instead of
"Fast"). This performance setting seemed to revert the SDRAM timings back to
"Auto", among other things (thus making the memory timing more
conservative). I tried various other combinations of AGP settings, but it
seemed that only the combination of AGP 1x/2x and conservative memory
timings worked (just running with conservative memory timings alone wasn't
sufficient).
The user's manual says the following about the board's AGP support:
"The AGP interface that is in the AMD 761TM, it supports a subset of 3.3V,
66MHz components, 3.3V 66/133 MHz AGP 2.0 compliant components, and the new
1.5V 66/266 MHz components."
While trying to resolve the problem, I updated the BIOS to the 5/29/2002
(latest) rev., and the latest AGP miniport drivers (vers. 5.33), but this
had no effect on eliminating the system hangs at AGP 4x.
I also added the Win2000 registry patch suggested on the Epox web site,
http://tinyurl.com/33em
which has a pointer to this Microsoft KB article:
http://support.microsoft.com/default.aspx?scid=KB;EN-US;Q270715& which
states:
A Windows 2000-based computer may stop responding (hang) when you use an
Accelerated Graphics Port (AGP) program such as Ziff Davis 3D WinBench 2000.
This behavior may occur more frequently depending on the computer's
processor, memory, and graphics configuration. This problem is known to
occur with Nvidia GeForce 256 and Matrox G400 video adapters on computers
with AMD Athol processors. This issue is not specific to any particular
video adapter type.
I'm writing this note to make other users aware of the potential problems
and to ask if anyone has had similar experiences, or can suggest a
combination of parameters that might let me run the system at AGP 4x? Has
anyone gotten a Radon 9000 Pro to run 4x AGP on an EP-87KA+ system? What
sort of system timing parameters are you using? (I suppose I could try
dropping the FSB speed back to 200Mhz, but didn't really want to sacrifice
the overall system performance for a small boost in video performance).
Separate question: how much performance does 4x AGP buy over 2x?
Derek Wildstar
2) AGP 4X : 2X has a minor, almost unmeasurable difference in performance at
your specific configuration.
3) Boost AGP driving to EA (instead of DA)
4) Shut off FW, SBA and Bootstrap SBA
5) Disable ECC
6) Suspect PSU robustness
7) Run MEMTEST86 to investigate ram pettern fidelity
8) Swap video card for potentially uncompromised card.
"Gary Funck" <ga...@intrepid.com> wrote in message
news:_AMF9.134221$Yb1.1...@sea-read.news.verio.net...
Gary Funck
"Derek Wildstar" <ab...@mindspring.net> wrote in message
news:SMRF9.177651$QZ.28270@sccrnsc02...
> 1) Drop AGP aperture to 64M
>
I tried aperture sizes, 32, 64, 128. Didn't seem to have an effect on fixing
the failure mode. What are the pros/cons of aperture size, in terms of
performance and system stability?
> 2) AGP 4X : 2X has a minor, almost unmeasurable difference in performance
at
> your specific configuration.
Thanks. From the reading I've done, apparently 4x is a win only in certain
(bandwidth limited?) situations. But, 4x runs at a lower voltage; are there
implications relating to demands on PSU in terms of need for greater
stability, or increased current flow?
>
> 3) Boost AGP driving to EA (instead of DA)
>
Thanks, that was next on my list of things to try.
> 4) Shut off FW, SBA and Bootstrap SBA
>
They're off.
> 5) Disable ECC
>
Yeah, sigh. I tend to prefer system stability versus raw performance, and my
comfort level is higher with ECC on. Does this interact badly with things
like the video card?
> 6) Suspect PSU robustness
>
Maybe, but I did make an effort to buy a good quality PSU with more than
adequate power rating. Not at that computer at the moment, so don't remember
the exact specs.
> 7) Run MEMTEST86 to investigate ram pettern fidelity
>
Good idea.
> 8) Swap video card for potentially uncompromised card.
>
Okay. Thanks again.
Doughnut
I have a mate who bought a radeon 9000 128mb card, All the system ever
does is hang when playing games, sometimes an hour sometimes 10 minutes. I
have read other post relating to lock up on Radeon 9000, It does sound that
it could be the psu but the drivers for this card are still early releases.
Just my little bit....lol
"Gary Funck" <ga...@intrepid.com> wrote in message
news:%oSF9.134237$Yb1.1...@sea-read.news.verio.net...
Gary Funck
"Doughnut" <ab...@dsl.pipex.com> wrote in message
news:3de7ff57$0$9445$cc9e...@news.dial.pipex.com...
> Hi,
> I have a mate who bought a radeon 9000 128mb card, All the system ever
> does is hang when playing games, sometimes an hour sometimes 10 minutes.
I
> have read other post relating to lock up on Radeon 9000, It does sound
that
> it could be the psu but the drivers for this card are still early
releases.
> Just my little bit....lol
>
Well, I might add that I did pull my unit out from among 20/so boxes with
Fry's return tags on them. I bought the new-in-the-box Pro model instead of
the regular models in the return heap, but maybe I should've taken the hint?
.... LOL
PS: if you see your buddy, ask him if he tried running in 2x mode,
conservative system timings and ask if that helped?
PPS: In this case, I *doubt* it is the PSU. Yes, I've read the posts about
5v being pulled down hard on this EP-8k7A+ board. But this seems a little
far-fetched. From what I recall the PSU was a 350w unit, and it is fed by a
new UPS that is running at 50% of its power rating.
Nick M.V.Salmon
> Well, I might add that I did pull my unit out from among 20/so boxes with
> Fry's return tags on them. I bought the new-in-the-box Pro model instead
of
> the regular models in the return heap, but maybe I should've taken the
hint?
> .... LOL
>
> PS: if you see your buddy, ask him if he tried running in 2x mode,
> conservative system timings and ask if that helped?
>
> PPS: In this case, I *doubt* it is the PSU. Yes, I've read the posts about
> 5v being pulled down hard on this EP-8k7A+ board. But this seems a little
> far-fetched. From what I recall the PSU was a 350w unit, and it is fed by
a
> new UPS that is running at 50% of its power rating.
Just because the PSU is capable of supplying the mobo doesn't mean the mobo
can handle the more modern VidCard's power draw - push the O/C & Voltage
right up on an 8K3A and both the 5V + indicated CPU core Voltage in MBM5 can
start leaping all over the place BUT, read the Voltage at the ATX connector
and it will likely be a solid 5V or slightly less - undersized mobo traces
causes it I guess.
Ciao...
[UK]_Nick...
--
Nick M.V.Salmon Master Mariner MN(Retd.)
My four init...@btinternet.com to reply direct
Steven
> 1) Drop AGP aperture to 64M
>
> 2) AGP 4X : 2X has a minor, almost unmeasurable difference in performance
> at your specific configuration.
>
> 3) Boost AGP driving to EA (instead of DA)
>
Can u explain a bit in detail what this DA/EA thingy is doing - have seen
this in the bios but never touched it yet. There is no documentation on this
setting available.
Derek Wildstar
"Steven" <weas...@gmx.net> wrote in message
news:as9ukv$p6rr1$2...@ID-97755.news.dfncis.de...
There is a good reference to this and other less than obvious BIOS settings,
and you are wise not to fiddle with what you don't understand. This is one
of those BIOS settings than can physically damaage your system.
Here's the text:
AGP Driving Value
Options : 00 to FF (Hex numbers)
This option is slaved to the AGP Driving Control BIOS function. If you set
the AGP Driving Control to Auto, then the value you set here won't have any
effect. In order for this BIOS option to work, you need to set the AGP
Driving Control to Manual.
The AGP Driving Value determines the signal strength of the AGP bus. The
higher the value, the stronger the signal. The range of Hex values (00 to
FF) translates into 0 to 255 in decimal values. By default, the AGP Driving
Value is set to DA (218) but if you are using an AGP card based on the
NVIDIA GeForce2 line of GPUs, then it's recommended that you set the AGP
Driving Value to the higher value of EA (234).
Due to the nature of this BIOS option, it's possible to use it as an aid in
overclocking the AGP bus. The AGP bus is sensitive to overclocking,
especially in AGP4X mode and with sidebanding enabled. As such, a higher AGP
Driving Value may be just what you need to overclock the AGP higher than
normally possible. By raising the signal strength of the AGP bus, you can
improve its stability at overclocked speeds.
But be very, very circumspect when you increase the AGP Driving Value on an
overclocked AGP bus as your AGP card may be irreversibly damaged in the
process!
BTW, contrary to some reports, increasing the AGP Driving Value won't
improve the performance of the AGP bus. It is not a performance enhancing
option so you shouldn't increase the value unless you need to.
Derek Wildstar
> I tried aperture sizes, 32, 64, 128. Didn't seem to have an effect on
fixing
> the failure mode. What are the pros/cons of aperture size, in terms of
> performance and system stability?
This feature gets more attention than it deserves, and as time goes on, it
will become moot.
The only real difference is that AGP Aperture of <16M forces AGP 1X with
resultant loss of AGP functionality (features, not just capacity).
However, some configurations experience error when the Size is set to 128 or
256 due to the OS (VIA the GART) {pun intended} reserving certain parts of
memory as being "AGP-able".
It's not that the memory is actually reserved, but the subtle housekeeping
involved causes error. I've experienced systems, notably Epox ones, that
refused to behave when set to 128M, but any other setting was fine.
Gary Funck
>
> "Steven" <weas...@gmx.net> wrote in message
> news:as9ukv$p6rr1$2...@ID-97755.news.dfncis.de...
> > Derek Wildstar <ab...@mindspring.net> wrote:
> > > 1) Drop AGP aperture to 64M
> > >
> > > 2) AGP 4X : 2X has a minor, almost unmeasurable difference in
> performance
> > > at your specific configuration.
> > >
> > > 3) Boost AGP driving to EA (instead of DA)
Is this setting in the Epox 8K7A+ BIOS? I didn't find it under advanced
peripherals, and on various other menus that I looked at. Is there some
graphics tweak program that has to be run to boost the AGP driving?
rb
"Nick M.V.Salmon" wrote:
Actually it's due to 5V rail being used on the 8K7A to get 3.3V through
voltage regulators. This is why the 5V has such a load on it with the
8K7A.