Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

Hi,

Hoping for some help. First of all: I'm not looking for a giant
overclock, no games, it is for hi-res pictures and audio editing. Just
want what's possible with the stuff I got now, making improvements
later. In theory I should have some room ... But I can't even get it to
run normal speeds :-/

This is it:

Abit NF7-S v2, the 2 color memory slots version, latest BIOS 27, (got
it with 22)
Athlon XP Mobile 2600+ AXMG2600FQQ4C
2x 3200PC 512 Mb Kingston Value Ram 2,6 volt KVR400X64C3A/512, CAS by
SPD at 200HHz: 8-3-3-3 and 8-3-3-2,5
Aopen H600A miditower with 350 watt power supply. Seems to be very
efficient, should be enough for the 45 watt mobile CPU, Intel says it's
OK for a Pentium 3,6Ghz. But mine has I slightly different model no:
AO350-12AHNF. Sticker says: +3,3V & +5V = 130 watt.
Coolermaster CP5-8JD1F-0L CPU cooler (nothing special, but performs
well and quite silent)
Zalman ZM-NB47J passive northbridge cooler
Samsung ATA harddisks SP4002H (40 Gb) and SP8004H (80 Gb), Toshiba
DVD-ROM
Old Diamond ViperII Z200 32 Mb AGP 4x videocard (first upgrade
candidate, looking for one with passive cooling)
Logitec optical USB mouse with PS2 converter.

Setup:

I did all recommended BIOS settings, disabling: APIC mode, system &
video cache, FSB & AGP spread spectrum, CPU disconnect function. Also
disabled: SATA, serial port 2, parallel port.
Not sure about: AGP fast write capability (default enabled) and CPC (is
that a setting in Win XP?).

After all BIOS changes I installed Win98SE again to make sure (not all
updates yet), at 10x166 Vcore 1,55 and memory 2,6 with only one 512Mb
stick, added a [vcache] entry in C:\windows\system.ini to allow for 1
Gb:
MinFileCache=0
MaxFileCache=262144
Added the second 512MB stick and started testing.

Testing

I always set clocks and voltages in BIOS (not from Windows), CAS until
now always 8-3-3-3 by SPD, FSB and RAM clock are always the same, by
SPD, confirmed by Memtest. I run Memtest86 first and when cleared,
reboot to Windows, first start 8rdavcore to monitor voltages/temps and
continue testing with Prime95, not doing anything else on the PC. I'm
using a X4-LIFE Inspector to measure total power consumption of the
whole box.

Because a lot of slight overclock experiments did not work (mostly
clearing Memtest, but not Prime), I started at 10x200. 2000Mhz is the
normal speed for the Athlon Mobile 2600+; PC3200 memory should do
200MHz. The NF7 reported voltages are always lower than set in BIOS, so
started out at Vcore 1,550 (BIOS) so it does not drop below 1,45
(enough for the mobile 2600+ at 2000MHz). CPU was 45 degrees C max.

Before Win re-install:
10x200 Vcore 1,55, mem 2,6. Memtest OK, Prime: blue screen (not
directly).
10x200 Vcore 1,55, mem 2,7. Memtest 3x OK, Prime: 1 hour 24 min.

After Win98 re-install:
10x200 Vcore 1,55, mem 2,7. Memtest OK, Prime: 2 hour 18 min.
10x200 Vcore 1,60, mem 2,7. Memtest OK, Prime: blue screen (not
directly).
10x195 Vcore 1,60, mem 2,7. Memtest OK, Prime: I was thinking it still
ran after 11 hours, but (never seen that) Prime is still 'red' and the
PC has crashed, freezing the screen. Keys and mouse are dead. Can
reproduce it, does it again within minutes.

So re-installing windows seems not needed after disabling all these
BIOS settings.
At 10x200 Vcore 1,55 was reported 1,472-1,520 so +/- 0,048
At 10x195 Vcore 1,60 was reported 1,536-1,552 so +/- 0,016


WHAT'S WRONG HERE?

Having to do even more underclocking and/or overvolting to get stable?!
Something must be very wrong or broken :-/

PSU? I know it's only 350 watt, but it seems to be efficient enough and
should power a mobile at stock speed (45 watt) with ease. Both a
digital and needle multimeter show a stable 3,40 volt at the +3,3 rail,
idle and heavy load.
These test did reach 110 watt total power use max.

But these Vcore variations still seem large, already at stock speeds,
I've seen up to +/- 0,07 volt with quite normal overclock settings.
Other voltages vary also, but less.
Are these Vcore variations too large? NF7 regulators?

Somehow still a problem with Win98SE? There is that '512MB barrier',
but the [vache] thing should solve that. Other settings? Memtest
cleares a lot more than Prime (running Win98) ...

The old Diamond ViperII Z200 32 Mb AGP 4x videocard? I've read about
'issues' with the S3 Savage4 Pro chipset, this is a S3 Savage200
series. Use the 1999 driver from the CD now, since had these 'screen
freezes', before re-install I had the latest driver.

BIOS ISSUE?

I have flashed to BIOS 27. I did not think much of it, always get
things running again, but there are these strange things happening
after a change in BIOS and saving.
Most of the time it just reboots, videocard notice, beep, etc.
Sometimes it freezes, black screen, also after saving 'clearly
permitted' settings. Always can fire it up again with reset + Insert or
cold boot + Insert.
Seemingly at random I get this red notice "DO NOT POWER DOWN while
writing BIOS", and it reboots again normally. Recently it sometimes
reboots without the beep after saving. The settings are always saved.

I've also heard about BIOSes which corrupt SATA drives, but I don't
have any and disabled SATA. Should I still go for a other BIOS?


Well, what to say ... HELP!

Boyd Noorda

Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

Update

I wrote:

> 10x200 Vcore 1,60, mem 2,7. Memtest OK, Prime: blue screen (not
> directly).
> 10x195 Vcore 1,60, mem 2,7. Memtest OK, Prime: I was thinking it
still
> ran after 11 hours, but (never seen that) Prime is still 'red' and
the
> PC has crashed, freezing the screen. Keys and mouse are dead. Can
> reproduce it, does it again within minutes.

[cut]

> The old Diamond ViperII Z200 32 Mb AGP 4x videocard? I've read about
> 'issues' with the S3 Savage4 Pro chipset, this is a S3 Savage200
> series. Use the 1999 driver from the CD now, since had these 'screen
> freezes', before re-install I had the latest driver.

I've removed the video driver and set video to standard 640x480 16
color:

10x200 Vcore 1,60, mem 2,7. Prime: still blue screen
10x195 Vcore 1,60, mem 2,7. Prime: still running after 4 houres, no
crash with frozen screen.

Something changed for the better, but it's not solving much.
I'll try to get a other video card.

Any suggestions for a simple, passive cooled compatible video card?
No games, most 'demanding' tasks: image editing and showing video's.

Thanks,
Boyd Noorda

Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

Update 2

> 10x195 Vcore 1,60, mem 2,7. Prime: still running after 4 houres, no
> crash with frozen screen.

But Prime stopped after 4 hours 40 min. :-/

Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

> Hoping for some help. First of all: I'm not looking for a giant
> overclock, no games, it is for hi-res pictures and audio editing. Just
> want what's possible with the stuff I got now, making improvements
> later. In theory I should have some room ... But I can't even get it to
> run normal speeds :-/

This may sound a bit dumb, but try taking the HSF off, ispecting the
thermal compound, and reseating the HSF.

steve

Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

"Steve Wolfe" <unt@codon.com> wrote in message news:<3aqsfqF6dhg70U1@individual.net>...

> This may sound a bit dumb, but try taking the HSF off, ispecting the
> thermal compound, and reseating the HSF.

Thanks, good point. But I monitor temps closely. CPU temp hardly
reached 40 degr. C in these lasts tests.

Best,
Boyd Noorda

Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

joburgnewsdesk@ddh.nl wrote in message news:<1112020034.263006.199320@o13g2000cwo.googlegroups.com>...

Update 3

> I've removed the video driver and set video to standard 640x480 16
> color:
>
> 10x200 Vcore 1,60, mem 2,7. Prime: still blue screen
> 10x195 Vcore 1,60, mem 2,7. Prime: still running after 4 houres, no
> crash with frozen screen.

Prime stopped 'normally' at 4 hours 40 min.

> Something changed for the better, but it's not solving much.
> I'll try to get a other video card.

I've put in a videocard with the Nvidia GeForce2 MX400 and installed
the newest drivers. I imagine a Nvidia video should work well with a
Nvidia chipset.

10x195 Vcore 1,6 volt memory 2,7 volt, 8-3-3-3 cleared Memtest, Prime
ran 15 min.

I've now flashed back from BIOS 27 to BIOS NF7S_d1.0 bios using Abit
NF7 FlashMenu 1.20, downloaded here:
http://www.logan.eclipse.co.uk/downloads.htm
Following instructions found here:
http://www.logan.eclipse.co.uk/abit_nf7_(s)_v2_0.htm

10x195 at Vcore 1,6 memory 2,7 8-3-3-3:
Prime ran 12 min. ending in fatal exception blue screen (windows still
ran)
Max. CPU temp was 45 dergees C, max power consumption: 117 watt
Reported voltages:
Vcore: 1,520-1,568
mem: 2,672-2,704
3,3 v: 3,264-3,296
12 v: 11.734-11,856

I've had memory at 2,7 volt until now to make sure in does not drop
below 2,6 volt the Kingstons wants. (The NF7 does undervolt). But I've
noticed ealier with noname PC3200 memory that even a bit of
'overvolting' does not always help, at least at these 'around nomal'
speeds.
So I lowered memory voltage to the stock 2,6 volt for the Kingston
sticks reported voltages being: 2,576-2,608.

Now at 10x195, keeping Vcore at 1,6, Prime runs for 10 hours already
and still running, the longest I ever got.
But I don't think lowering memory voltage is the whole story ...

I'm still worrying about voltage variations. Anyone?

I'm also wondering about the -5 volt rail, I'll start an other thread
about this issue, since the answer should be simple.

Is there a program which checks stability under Windows a bit faster?
I'd like to explore the possibilities faster, later confirming with
Prime.

Thanks,
Boyd Noorda

Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

raise vcore to 1.75, then lower one notch at a time till prime fails
then raise one notch. test in safe mode to rule out driver conflicts, or
use linux.

Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

"M. Bezzel" <feesa@sbd.net> wrote in message news:<114ise0lckse6b2@corp.supernews.com>...

> raise vcore to 1.75, then lower one notch at a time till prime fails
> then raise one notch.

I'll do that later for sure.
Prime is still running this time, 19 hours now But I'm still
underclocking and already overvolting (10x195 @ 1,6 volt) :-/

> test in safe mode to rule out driver conflicts

That's a good tip, but at 640x480 my monitor util does not fit the
screen ...

> or use linux

That would be nice, I've already downloaded the 'Utimate Boot CD',
will burn a CD asap, hoping to find good utils for testing.
Any other suggestions?

Thanks,
Boyd Noorda

P.S. I did not mean to start a new thread, but the Google groups
interface I use seems to have a lot of problems lately.

Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

> > This may sound a bit dumb, but try taking the HSF off, ispecting the
> > thermal compound, and reseating the HSF.
>
> Thanks, good point. But I monitor temps closely. CPU temp hardly
> reached 40 degr. C in these lasts tests.

Because you're measuring from a thermistor underneath the CPU socket,
that's not entirely reliable. I've seen reports of no higher than 40 C when
temps were actually MUCH higher.

steve

Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

"Steve Wolfe" <unt@codon.com> wrote in message news:<3b1ijmF66vnaiU1@individual.net>...

> > > This may sound a bit dumb, but try taking the HSF off, ispecting the
> > > thermal compound, and reseating the HSF.
> >
> > Thanks, good point. But I monitor temps closely. CPU temp hardly
> > reached 40 degr. C in these lasts tests.
>
> Because you're measuring from a thermistor underneath the CPU socket,
> that's not entirely reliable. I've seen reports of no higher than 40 C when
> temps were actually MUCH higher.

Sure, but I guess the difference is the same with all NF7 boards?
Compared to what temps others do on this board, I guess I'm save
Actually also did higher temps, up to 53 degree C, still stable for
hours. (Will not do that daily

Besides a lot of other things, it seems my problem was to low Vcore,
hoping the 1,45 volt mobile would do normal (not overeclocked) speeds
around 1,6.
Now I'm up to 1,75 volt things are better, but temps and PSU will
limit me (around 11.5x197). I guess the mobile may have saved me money
in the PSU and cooling area
Still don't understand why PC3200 Kingston Value Ram (2x512MB) will
not do 200MHz in dual channel mode, run at specified 2,6 volt and CAS.
And also at More volt does not help.

Thanks, Boyd

Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

Joburgnewsdesk wrote:
> "Steve Wolfe" <unt@codon.com> wrote in message
> news:<3b1ijmF66vnaiU1@individual.net>...
>
>>>> This may sound a bit dumb, but try taking the HSF off, ispecting
>>>> the thermal compound, and reseating the HSF.
>>>
>>> Thanks, good point. But I monitor temps closely. CPU temp hardly
>>> reached 40 degr. C in these lasts tests.
>>
>> Because you're measuring from a thermistor underneath the CPU
>> socket, that's not entirely reliable. I've seen reports of no
>> higher than 40 C when temps were actually MUCH higher.
>
> Sure, but I guess the difference is the same with all NF7 boards?

No, not really. Even on-die thermistors vary up to 10 deg C between
identical boards; more if you have different BIOS versions. For most people,
the rule "if it's (Prime95) stable then it's not too hot (and not getting
damaged)" holds. If you're overvolting(*1) by more than 10% or so then it's
a bit more iffy, but that doesn't usually include to many people.

[...]

(*1) Here I mean overvolting the chip type, not the particular chip. For
example, I wouldn't count running a mobile Barton at 1.65V as overvolting at
all, since the Barton core usually operates at 1.65V. A 10% overvolt on a
Barton core would then be 1.815V, so running a Mobile at 1.8V would be about
as high as I'd want to go before discarding the "stable = no damage" rule.

--
Michael Brown
www.emboss.co.nz : OOS/RSI software and more
Add michael@ to emboss.co.nz ---+--- My inbox is always open

Archived from groups: alt.comp.hardware.overclocking.amd (More info?)

"Michael Brown" <see@signature.below> wrote in message news:<424ccd74$1@clarion.carno.net.au>...

> > Sure, but I guess the difference is the same with all NF7 boards?
>
> No, not really. Even on-die thermistors vary up to 10 deg C between
> identical boards; more if you have different BIOS versions. For most people,
> the rule "if it's (Prime95) stable then it's not too hot (and not getting
> damaged)" holds. If you're overvolting(*1) by more than 10% or so then it's
> a bit more iffy, but that doesn't usually include to many people.

Well, I'm using the 'Prime stable, then back off a bit' approach and
aiming for a max. temp around 50 degree C full load. I guess that's
save?

> (*1) Here I mean overvolting the chip type, not the particular chip. For
> example, I wouldn't count running a mobile Barton at 1.65V as overvolting at
> all, since the Barton core usually operates at 1.65V.

That's right, but the Mobile 2600+ officially already should work at
1,45 at the off. speed of 2000MHz. It does not in my case. Of course
that's a higher setting on the NF7, making sure Vcore never goes below
1,45, in my case a 1,55 BIOS setting to get 1,472-1,1520 Vcore
reported (load and idle), doing 10x200.
This does only run Prime a few hours.

I'm thinking this could be memory. I've noticed with other
PC3200/512MB sticks that they do 210MHz in single channel mode, but
can't do 200MHz in dual channel.
Is that normal?

It also could be PSU or NF7 Vcore. Are the Vcore variations within
specs?
It seems my 350 Aopen PSU or my NF7-S tops out at Vcore 1,75 (BIOS)
doing 11 to 12x multiplier.
11,5x198 gives me Vcore 1,648-1,728 Vcore reported, 0,08 variation.
The whole box uses 147 watt max. dropping to around 140 watt over time
(one HD shuts off). CPU 53 degree C max. Prime stopped at 5 hours 40
min.
I guess a more stable and a bit lower (1,7) Vcore would help a lot?

> A 10% overvolt on a
> Barton core would then be 1.815V, so running a Mobile at 1.8V would be about
> as high as I'd want to go before discarding the "stable = no damage" rule.

I will not reach that with this PSU and cooling, maybe later. Just
trying to get the stuff I have now stable.
But I was hoping for 200MHz FSB/RAM at least, since that should not be
a power issue. Memory voltages (2,6 BIOS) always stay within
2,572-2,608.

Thanks, Boyd