VOGONS


First post, by RETROKOMODO

User metadata
Rank Member
Rank
Member

Hello fellow Vogons.

This is an odd one. It's popped up here and there all over the place for years and i've been looking into it for a while. Different theories abound, but i'm yet to find a solid fix. So - here's the problem:

When running the Advanced Pixel Shader test in 3DMark2001 I see maybe a half second of the test run? - and then the whole PC restarts. This is the config I currently have running:

CPU: Athlon XP 3200+
GPU: AGP HD 3850
HDD: 74GB WD Raptor
MOBO: ABIT NF7-S v2.0 (has a dedicated CPU power socket)
RAM: 2GB Corsair Value PC3200
PSU: Corsair HX 520 I think - but more than enough
O/S: XP SP3
GPU Driver: Cat 13.4

I have tried multiple combinations of 3 different socket 462 motherboards, 3 different PSU's with enough on the 5v, 2 different AGP HD 3850's, 2 sets of PC3200 ram, probably 4 or 5 HDD's, AGP Fast Writes Enabled and Disabled, AGP speed set to x4 and x8, and to top it off - 7 different BIOS flashes of my 3850.

I previously had an ASROCK 775DUAL-VSTA setup with the same GPU, RAM and HDD's, and it ran through without restarting - so it isn't the driver version, or the GPU, or RAM. The big difference was it running an Intel 775 CPU.
I also tried the current motherboard with an Athlon XP 1700+ installed - and it ran through without restarting - so it isn't this motherboard..

Whisky commented in another thread 'sometimes when i drop frequency on them to 100-133 FSB 3D Marks runs OK, but not always 😀 I tried over 30 modified and official bios, 20 different ram sticks, and instant restart is always in the same time on both NF7.'

So at lunch today that is what I tried - I dropped the speed of the 3200+ in the BIOS to 2500+ speeds

IMG_20220819_132713.jpg
Filename
IMG_20220819_132713.jpg
File size
211.14 KiB
Views
1410 views
File license
CC-BY-4.0

and the test ran and completed!!

IMG_20220819_132616.jpg
Filename
IMG_20220819_132616.jpg
File size
102.5 KiB
Views
1410 views
File license
CC-BY-4.0

I then popped it back up to regular 3200+

IMG_20220819_132847.jpg
Filename
IMG_20220819_132847.jpg
File size
166.02 KiB
Views
1410 views
File license
CC-BY-4.0

and just as before - instant restart when the test started - no fish.

So things that we know work are - well - everything. It runs fine on all my kit (and I needn't have bothered buying 2 extra motherboards and 2 extra CPU's) but it only runs if I cripple the CPU speed.. What might be going on here do we think? I initially thought my CPU was somehow slightly knackered somehow, but it isn't that. Are the 3000+ & 3200+ just tooooo powerful to handle? As I said my current board has a dedicated CPU socket

IMG_20220819_132941.jpg
Filename
IMG_20220819_132941.jpg
File size
149.64 KiB
Views
1410 views
File license
CC-BY-4.0

Should I maybe try with my spare 750w PSU? Seems a tad overkill for an XP 3200+ with a 77w TDP..

This might be why on a Google search of 'xp 3200+ agp hd 3850' running a 3DMark2001 test, I find nothing with a combo of both CPU and GPU, and I would have though that with both parts being the fastest options for Socket A, there'd be more people trying the combination.

Any suggestions on what to try next? Other than downclocking the CPU? I'm willing to try almost anything at this point. The alternative is to flog my Socket A gear which I don't particularly want to do and go Intel..

Reply 1 of 45, by weedeewee

User metadata
Rank l33t
Rank
l33t

if you have hardware to spare & test... try fully recapping a mainboard & gpu.

Right to repair is fundamental. You own it, you're allowed to fix it.
How To Ask Questions The Smart Way
Do not ask Why !
https://www.vogonswiki.com/index.php/Serial_port

Reply 2 of 45, by pixel_workbench

User metadata
Rank Member
Rank
Member

Try setting CPU Interface = Enabled in bios. On my Abit AN7 I remember having instability if it was disabled and base clock was > 166mhz.

My Videos | Website
P2 400 unlocked / Asus P3B-F / Voodoo3 3k / MX300 + YMF718

Reply 5 of 45, by RETROKOMODO

User metadata
Rank Member
Rank
Member
The Serpent Rider wrote on 2022-08-19, 17:49:

Just add voltage to CPU and RAM.

Good morning! I've been trying voltage changes. RAM is at 2.7 at the mo and I have climbed incrementally to 1.8 for the CPU. Max temp is 52 degrees midway through running Cinebench R10 and all seems stable enough.. It still restarts at the APS test though.. Any suggestions for maximum voltages I should try?

Alternatively, would lowing the FSB to 166 again, then cranking up the multiplier to 3200+ levels work? I've turned it off and gone to a BBQ so can't test right now 😉

Reply 6 of 45, by RETROKOMODO

User metadata
Rank Member
Rank
Member
quicknick wrote on 2022-08-19, 22:06:

Happens a lot to me with NF2 motherboards + ATi videocards...
Haven't found a fix yet.

Hmm. I also had the exact same on my Asus A7N8X-X, but also my Matsonic MS8177C with VIA KT600 & 8237 Chipsets. What other chipsets could I try?! I'm kinda all out of disposable income right now having dropped £100+ on the latest motherboard & 3200+. Further purchases for testing may have to wait until next month 🙁

Reply 7 of 45, by RETROKOMODO

User metadata
Rank Member
Rank
Member
weedeewee wrote on 2022-08-19, 15:31:

if you have hardware to spare & test... try fully recapping a mainboard & gpu.

Alas - no spare parts or soldering kit to hand. The latest board i'm using is running Cinebench nice and stable at increased voltage, so VRM and caps are hopefully fine? Nothing swollen or burst on a cap inspection anyway

Reply 8 of 45, by RETROKOMODO

User metadata
Rank Member
Rank
Member
pixel_workbench wrote on 2022-08-19, 17:21:

Try setting CPU Interface = Enabled in bios. On my Abit AN7 I remember having instability if it was disabled and base clock was > 166mhz.

Tried enabling but seems not to have made any difference i'm afraid. Although this is with me running at 200 x 11. I may drop to 166 and increase multiplier to match as per https://www.cpu-world.com/CPUs/K7/AMD-Athlon% … A3200DKV4D.html

Reply 9 of 45, by RETROKOMODO

User metadata
Rank Member
Rank
Member
RETROKOMODO wrote on 2022-08-20, 09:44:
pixel_workbench wrote on 2022-08-19, 17:21:

Try setting CPU Interface = Enabled in bios. On my Abit AN7 I remember having instability if it was disabled and base clock was > 166mhz.

Tried enabling but seems not to have made any difference i'm afraid. Although this is with me running at 200 x 11. I may drop to 166 and increase multiplier to match as per https://www.cpu-world.com/CPUs/K7/AMD-Athlon% … A3200DKV4D.html

Well that didn't work.. Multiple BIOS resets later and back at square 1

Reply 10 of 45, by AlexZ

User metadata
Rank Member
Rank
Member

My guess there would be the motherboard can't handle 200Mhz FSB reliably anymore.

Pentium III 900E, ECS P6BXT-A+, 384MB RAM, NVIDIA GeForce FX 5600 128MB, Voodoo 2 12MB, 80GB HDD, Yamaha SM718 ISA, 19" AOC 9GlrA
Athlon 64 3400+, MSI K8T Neo V, 1GB RAM, NVIDIA GeForce 7600GT 512MB, 250GB HDD, Sound Blaster Audigy 2 ZS

Reply 11 of 45, by rasz_pl

User metadata
Rank l33t
Rank
l33t
RETROKOMODO wrote on 2022-08-21, 09:37:
RETROKOMODO wrote on 2022-08-20, 09:44:

I may drop to 166 and increase multiplier to match as per https://www.cpu-world.com/CPUs/K7/AMD-Athlon% … A3200DKV4D.html

Well that didn't work.. Multiple BIOS resets later and back at square 1

what exactly didnt work? changing multiplier or running 3dmark on 13x166?
Fastest diagnosis would come from using oscilloscope on Vcore and AGP Vsupply. That would exclude motherboard and leave CPU as the source of problems.

Open Source AT&T Globalyst/NCR/FIC 486-GAC-2 proprietary Cache Module reproduction

Reply 12 of 45, by agent_x007

User metadata
Rank Oldbie
Rank
Oldbie

Possible fixes/causes :
1) Get a fan blowing into nForce north bridge chipset (if it helps, you need better chipset cooler/better airflow over Rialto or AGP power)
Possible cause : AGP power circuits are at the limit.
2) Try changing GPU drivers (possible versions to test : 9.1, 10.2, 11.4, 12.2 [since you have 13.4])
Possible cause : Badly compiled code in newest ATI driver causes test to fail.

Additional options :
1) Increase AGP voltage (may kill bridge on card)
2) Increase NB voltage (may not help or will make this worse)
3) Get a new PSU (I'm assuming PSU just isn't coping well enough with that perticular load because of it's age)

If this fails, I would assume a software glitch which makes this card incompatible with your motherboard (or a bug with DirectX/Driver/3DMark installation). You could always swap card for X1950 Pro 512MB AGP.

157143230295.png

Reply 13 of 45, by RETROKOMODO

User metadata
Rank Member
Rank
Member
rasz_pl wrote on 2022-08-21, 12:44:
RETROKOMODO wrote on 2022-08-21, 09:37:
RETROKOMODO wrote on 2022-08-20, 09:44:

I may drop to 166 and increase multiplier to match as per https://www.cpu-world.com/CPUs/K7/AMD-Athlon% … A3200DKV4D.html

Well that didn't work.. Multiple BIOS resets later and back at square 1

what exactly didnt work? changing multiplier or running 3dmark on 13x166?
Fastest diagnosis would come from using oscilloscope on Vcore and AGP Vsupply. That would exclude motherboard and leave CPU as the source of problems.

Well I changed the bus to 166 and the multiplier to 14, and on a save nothing happened - blank screen. I had to clear the cmos to get back in 🙁 I don't have an oscilloscope or volt meter for testing either, so i'm kinda hoping I can just set stuff and it'll work!

Reply 14 of 45, by rasz_pl

User metadata
Rank l33t
Rank
l33t
RETROKOMODO wrote on 2022-08-22, 15:01:

Well I changed the bus to 166 and the multiplier to 14, and on a save nothing happened - blank screen. I had to clear the cmos to get back in 🙁 I don't have an oscilloscope or volt meter for testing either, so i'm kinda hoping I can just set stuff and it'll work!

14x166 != 11x200

is your CPU unlocked at all? as in can you go above 11? try 12x166

Open Source AT&T Globalyst/NCR/FIC 486-GAC-2 proprietary Cache Module reproduction

Reply 15 of 45, by RETROKOMODO

User metadata
Rank Member
Rank
Member
AlexZ wrote on 2022-08-21, 10:40:

My guess there would be the motherboard can't handle 200Mhz FSB reliably anymore.

and would be the case for all 3 in that case! Someone suggested replacing the capacitors - would that set things straight? They all game just fine with the 3200+, it's just this one annoying thing they won't do..

Reply 17 of 45, by RETROKOMODO

User metadata
Rank Member
Rank
Member
rasz_pl wrote on 2022-08-22, 15:17:
RETROKOMODO wrote on 2022-08-22, 15:01:

Well I changed the bus to 166 and the multiplier to 14, and on a save nothing happened - blank screen. I had to clear the cmos to get back in 🙁 I don't have an oscilloscope or volt meter for testing either, so i'm kinda hoping I can just set stuff and it'll work!

14x166 != 11x200

is your CPU unlocked at all? as in can you go above 11? try 12x166

OK - gone in at 12x166 - windows boots just fine, CPU-Z gives it:

Core: 2004.69
Multiplier: x 12
Bus Speed: 167.06
Rated FSB: 334.11

Aaaand trips on the test

Reply 18 of 45, by RETROKOMODO

User metadata
Rank Member
Rank
Member
rasz_pl wrote on 2022-08-22, 15:44:

try something like 5x200MHz, at least will tell us if its FSB related

On trying 5x200 and saving the bios it reboots to an unresponsive blank screen. On a power cycle it starts saying 'CPU is unworkable or has been changed. Please recheck CPU soft menu' but gives an F1 to continue. I continue in and CPU-Z gives me:

Core: 1202.94
Multiplier: x 12
Bus Speed: 100.25
Rated FSB: 200.49

Sooooo theres that.

went back into the bios and set it to 11x200 a 2200mhz CPU.

Core: 2205.20
Multiplier: x 11
Bus Speed: 200.47
Rated FSB: 400.94

Reply 19 of 45, by RETROKOMODO

User metadata
Rank Member
Rank
Member

For the record, these are my two 3200+ chips

This is the one i've had for about 20 years

IMG_20220822_171240.jpg
Filename
IMG_20220822_171240.jpg
File size
697.65 KiB
Views
1031 views
File license
CC-BY-4.0

This is the one I got last week

IMG_20220822_171245.jpg
Filename
IMG_20220822_171245.jpg
File size
789.88 KiB
Views
1031 views
File license
CC-BY-4.0

There seem to be some fairly big differences between the two! I currently have the recent one in. I know there is a chip off one corner by someone who didn't pay attention when fitting the cooler - but it makes no difference which I use!

The GPU is fine from what I can tell. No funky caps, now running far cooler since I switched coolers (shameless plug - https://www.youtube.com/watch?v=8ueKUWoBu9I) and has a decent amount of airflow going over the Rialto chip. Also having exactly the same on another AGP HD 3850 that as far as I can tell has not been used! or at least - not much. They both completed the test when using a different Intel setup ad the same 13.4 drivers, so i'm not concerned wit them as a point of failure.
Memtest complete without errors on 3 different kits of PC3200 so pretty sure they're all ok too!