GPU adds to instability, say it ain't so?

Speed addicts anonymous.

Moderator: Starfalcon

GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 6:42 am

Well, more trial and tribulations on my new build (really no time to put it in production, so still on the bench testing).

Again, here's the config:
E2160
Asus P5K-VM
2x1GiB OCZ Platinum Rev2 DDR2-800 (speced 4-4-4-15 @2.1V)
Antec NeoHE 430

Everything tested fine with the IGP, even at 333x9 with the RAM running at DDR2-800 (using the ratios of course).

Borrowed a friend's old BFG 6800GT OC, and it has never been stable for too long. Since I tested the other components for weeks with no trouble, I thought just plugging in the video card with such an ample PSU would be enough (still not breaking 200W draw at the wall, according to my Kill-A-Watt). Long story short, there is an overheating problem which I seem to have resolved with a ghetto-mounted fan blowing the back of the card and the NB heatsink (stupid new ForceWare forced the fan to spin at 53% always I turn it back up to 100% with RivaTuner :o). Between BSODs, random hangs/shutdown, and OS reinstalls (it corrupted things that bad), I finally caved and ran memtest and sure enough it failed, which explained everything (or not?). I've backed things down to DDR2-667 and it seems to be fine after about 6 hours of memtest so far (CPU still at 333x9).

So I ask, is this possible? I did test all the other components with the IGP enabled. PCIe has always been manually locked at 100. Assuming my PSU is good (RTHDRIBL+Orthos test run that was stable showed <200W of draw), how can this be? Bad video card or the NB being iffy at handling the new x16 load while overclocked? I suppose if my board has those NB tweaking option I could try but I guess I'm stuck now.

That said, I would eventually get another card, so this means I have to retest the whole thing? Ouch. :-?
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24141
Joined: Mon May 24, 2004 1:19 am

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 12:15 pm

I'd try bumping up the memory's voltage to 2.2v from 2.1v at DDR2-800 speeds to see if that fixes the memtest. If this is your ram, then you're at the very end of the recommended spectrum and it may just require a little bit more power. 2.15v to 2.2v at least, in my opinion.
Image
Nitrodist
Grand Gerbil Poohbah
 
Posts: 3280
Joined: Wed Jul 19, 2006 12:51 am
Location: Minnesota

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 12:22 pm

Instability can stem from a bad PCIe socket, bad contacts, bad video card, voltage dip from the added video card, or any number of other things. But yeah adding a discreet video card can and will cause instability once in a long while. Which also sucks cause you have to test everything to find out what is causing the instability. Get another video card and if it's still unstable then its either the mobo, the PSU or bad PCIe contacts.

my money is on the mobo
CB5000
Gerbil XP
 
Posts: 365
Joined: Wed Mar 26, 2008 3:46 pm
Location: NW region

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 12:37 pm

Nitrodist wrote:I'd try bumping up the memory's voltage to 2.2v from 2.1v at DDR2-800 speeds to see if that fixes the memtest. If this is your ram, then you're at the very end of the recommended spectrum and it may just require a little bit more power. 2.15v to 2.2v at least, in my opinion.
Unfortunately the BIOS ends at 2.1V. :( However, DDR2-800 4-4-4-15 and 5-5-5-15 at 2.1V were both tested to be fine with the IGP only. Given the memory controller is on the northbride, how would the video card affect it?

CB5000 wrote:Instability can stem from a bad PCIe socket, bad contacts, bad video card, voltage dip from the added video card, or any number of other things. But yeah adding a discreet video card can and will cause instability once in a long while. Which also sucks cause you have to test everything to find out what is causing the instability. Get another video card and if it's still unstable then its either the mobo, the PSU or bad PCIe contacts.

my money is on the mobo
The thing that sort of contradicts those theories is that the card (finally) tested fine at stock (200x9/ddr2-667) and overclocked (300x9/ddr2-600 and 333x9/ddr2-667) settings, once I put the extra fan on the video card and turn up the GPU fan to 100%. Plus the PCIe has been locked from day 1 to 100 even at stock. So PCIe would seem to be fine?

That's why my theory goes to the iffy Northbridge overclocked. May be those "termination voltages" and all that can be of use, though my board doesn't have those.

This card is a loaner and will not be the final card I put in. However at this rate it means I will have to run the same set of stability tests whenever I put the real card in (hopefully by Starcraft 2). Using the IGP has been rock solid for weeks, except that it can't run RTHDRIBL so I can't subject the setup to the harshest test (Orthos/Folding+RTHDRIBL, have been doing Orthos/Folding fine).

Sucks when things don't go as smoothly as imagine. :-?
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24141
Joined: Mon May 24, 2004 1:19 am

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 12:50 pm

It sounds to me like the Northbridge can't handle a 16x PCI connection plus dual-channel RAM.

Have you tried underclocking the PCIe bus, if that's possible? Alternately, you could try running it with 1 RAM module, or (best approach) buying a 3rd-party HSF bigger than what is on your NB and applying it yourself (or just reapplying the current one, if the thermal compound has dried out). Ideally that would be with Arctic Silver (or similar) but if it is just the stick-on type, then just stick it on. Of course it could already be cooked. Last time I had a video card that was failing under load for thermal reasons, addressing those problems only delayed the process by a couple months - then it started causing problems all the time even in 2D mode, and eventually would not work at all.
Saber Cherry
Gerbil XP
 
Posts: 303
Joined: Fri Mar 14, 2008 2:41 am
Location: Crystal Tokyo

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 12:55 pm

Saber Cherry wrote:It sounds to me like the Northbridge can't handle a 16x PCI connection plus dual-channel RAM.
How does that explain the thing actually working with stock and the other overclocked settings.

Saber Cherry wrote:Have you tried underclocking the PCIe bus, if that's possible?
Unfortunately 100 is the minimum setting. :(

Saber Cherry wrote:Alternately, you could try running it with 1 RAM module, or (best approach) buying a 3rd-party HSF bigger than what is on your NB and applying it yourself (or just reapplying the current one, if the thermal compound has dried out). Ideally that would be with Arctic Silver (or similar) but if it is just the stick-on type, then just stick it on. Of course it could already be cooked. Last time I had a video card that was failing under load for thermal reasons, addressing those problems only delayed the process by a couple months - then it started causing problems all the time even in 2D mode, and eventually would not work at all.
Are you talking about the NB or the GPU? I'm confused.
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24141
Joined: Mon May 24, 2004 1:19 am

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 1:01 pm

Flying Fox wrote:
Saber Cherry wrote:It sounds to me like the Northbridge can't handle a 16x PCI connection plus dual-channel RAM.
How does that explain the thing actually working with stock and the other overclocked settings.


It works at stock and overclocked when it is using the IGP, but not a discrete graphics card, right? Or at least, that is what I understand. The IGP doesn't use PCIe 16x. And I would imagine that it goes through the SB, not the NB, but I don't know.

Are you talking about the NB or the GPU? I'm confused.

I'm talking about my GPU, but that's my experience with thermal failure of an IC. Replacing the cooler with a much better one completely fixed the problem for a while, but then the symptoms returned and worsened. As though the failure was progressing anyway... and the thermal problem was just making it worse. So if your NB is in the same situation as my GPU was, then I would venture to guess that the fix I implemented might turn out the same way.
Last edited by Saber Cherry on Thu May 01, 2008 1:03 pm, edited 1 time in total.
Saber Cherry
Gerbil XP
 
Posts: 303
Joined: Fri Mar 14, 2008 2:41 am
Location: Crystal Tokyo

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 1:03 pm

Extra heat from the video card pushing the RAM or northbridge into a marginal condition perhaps? Do you have any way of monitoring the temperatures while running Memtest? If you're literally bench testing (i.e. not in an enclosure), perhaps try adding more local airflow to the northbridge heatsink. Any chance of trying a cooler-running, lower-power graphics card?

Seems very unlikely it would be a power supply issue as such, but perhaps the actual RAM supply voltage drops a bit with the video card installed? Perhaps it's not just 780G micro-ATX boards that have problems with voltage regulation... :wink:
Q9550@3.4 ~ Asus P5E ~ 4GB XMS2-6400C5DHX ~ Gigabyte Windforce 3X 7950 3GB
Xigmatek S1283 ~ LG DVDRW ~ Samsung F3 1TB ~ WD Caviar SE 640GB ~ Asus Xonar DX ~ Corsair HX 520W
Antec P182 ~ Logitech G7/G15/G25 ~ Sennheiser HD555 ~ BenQ FP241W
MrJP
Gerbil Elite
 
Posts: 826
Joined: Wed Nov 20, 2002 4:04 pm
Location: Surrey, UK

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 1:08 pm

I guess people can't read without tables/lists. So here it is (with a bit more info on the configs that I have tested so far):

With IGP:
- 200x9, ddr2-667 = good
- 300x9, ddr2-600 = good
- 333x9, ddr2-667 = good
- 333x9, ddr2-800 = good (I think I tested both 4-4-4-15 and 5-5-5-15 at 2.1V)

With borrowed 6800GT OC (that's after I realize the overheating problems while running 3d benches):
- 200x9, ddr2-667 = good
- 300x9, ddr2-600 = good
- 333x9, ddr2-667 = good
- 333x9, ddr2-800 = bad

Except stock, I think most of these "good" configs have been through at least 12 hours of memtest, Orthos and Folding at least one 2605 SMP WU. RTHDRIBL refuses to run on the G33 so I can't combine the CPU stress with GPU stress. However, the 2 overclocked good configs with the video card has gone through an overnight (8+ hours) session of Orthos+RTHDRIBL. At that config Kill-A-Watt is reporting 190-something watts, so let's call it <200W.

So if it is just that, I can most certainly live with 667 (after all I bought the P5K-VM mainly for its 1:1 ratio). Just wondering which component is not holding up. If it is the NB, then I wonder if those fancy Vtt and others on fancier motherboards is the ticket.

One note: I did try 400x7 and I couldn't seem to get that to work yet on the few times I tried.
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24141
Joined: Mon May 24, 2004 1:19 am

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 1:15 pm

MrJP wrote:Extra heat from the video card pushing the RAM or northbridge into a marginal condition perhaps? Do you have any way of monitoring the temperatures while running Memtest? If you're literally bench testing (i.e. not in an enclosure), perhaps try adding more local airflow to the northbridge heatsink. Any chance of trying a cooler-running, lower-power graphics card?
I have the fan lined up to blow in from a "southeast" direction (if you put the back I/O ports up at the "north" position). So the extra airflow will touch the RAM, CPU and NB HSF, and then to the video card back plate. Finger tests showed that the air arriving at the video card back plate is still quite cool. Ambient according to the thermostat 3 metres away showed mostly 21C.

MrJP wrote:Seems very unlikely it would be a power supply issue as such, but perhaps the actual RAM supply voltage drops a bit with the video card installed? Perhaps it's not just 780G micro-ATX boards that have problems with voltage regulation... :wink:
Hmm... that may be a plausible theory too. It's good to keep the discussion alive.
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24141
Joined: Mon May 24, 2004 1:19 am

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 1:18 pm

Ah, I see. Well, that just makes it more likely to be a NB heat problem, since the failure only occurs when both an 800MHz memory bus and 16x PCIe stress it.

Another possibility is that it is a memory problem in the block of memory normally used by the IGP, which would not matter (just cause a tiny graphics blip), but when the IGP is disabled that memory could be used by the OS and cause corruption. Pretty unlikely, but the way to test it would be to swap the module locations and run MemTest at 800MHz with the IGP. If it is in dual-channel mode, of course, you'd first have to put both modules in the same channel and then try them swapped and unswapped. If you think the RAM is good there's no reason to bother.
Saber Cherry
Gerbil XP
 
Posts: 303
Joined: Fri Mar 14, 2008 2:41 am
Location: Crystal Tokyo

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 2:21 pm

Flying Fox wrote:
MrJP wrote:Seems very unlikely it would be a power supply issue as such, but perhaps the actual RAM supply voltage drops a bit with the video card installed? Perhaps it's not just 780G micro-ATX boards that have problems with voltage regulation... :wink:
Hmm... that may be a plausible theory too. It's good to keep the discussion alive.


I have an asus M2A-VM-HDMI that have terrible voltage regulation and it's a u-ATX form factor. Setting CPU voltage to 1.5 leads to 1.435 volts under load if it's overclocked. I upped the RAM voltages too just to make it more stable. It's also possible the the northbridge isn't optimized to handle that kind of load on this mobo... as from the specs it looks like a mainstream board. It's somewhat likely that the G33 chip isn't getting enough power when it's under stress, if some of the other components are experiencing large voltage droops. Get a multi-meter and test it... never trust what the software tells you... though some voltages, you can't measure.
CB5000
Gerbil XP
 
Posts: 365
Joined: Wed Mar 26, 2008 3:46 pm
Location: NW region

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 4:15 pm

Oh well, even if you can't get 800MHz memory to work with the video card, that's still some pretty nice overclocking. Have you had to bump up Vcore much to get 3GHz stable?
Q9550@3.4 ~ Asus P5E ~ 4GB XMS2-6400C5DHX ~ Gigabyte Windforce 3X 7950 3GB
Xigmatek S1283 ~ LG DVDRW ~ Samsung F3 1TB ~ WD Caviar SE 640GB ~ Asus Xonar DX ~ Corsair HX 520W
Antec P182 ~ Logitech G7/G15/G25 ~ Sennheiser HD555 ~ BenQ FP241W
MrJP
Gerbil Elite
 
Posts: 826
Joined: Wed Nov 20, 2002 4:04 pm
Location: Surrey, UK

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 4:42 pm

MrJP wrote:Oh well, even if you can't get 800MHz memory to work with the video card, that's still some pretty nice overclocking. Have you had to bump up Vcore much to get 3GHz stable?

I have the biggest Vdroop I have seen so far, as reported by CPU-Z (take it for what it is worth). My last stable setting (with IGP, still memtesting 333x9 with video card for now) takes me to 1.4375V in the BIOS and CPU-Z reports as 1.328V/1.344V with Orthos/Prime95MT running. :o

Temps are getting to <27C delta(Tjmax), which is a little too hot, I'll be getting an aftermarket cooler later for sure. My hands are also tied a bit (I treat it as a challenge) by the Asus TM210 case that I am going to put it in. So worst case scenario is to drop back down to 300x9 (Vcore=Auto :D).

The P5K-VM is supposed to be one of the best G33 boards for overclocking. Of course, I'm happy just to stay at 333x9 even if this particular northbridge may be a bit iffy.

I'm getting very close to the point of having more than enough testing and tweaking. I would just like to slap the aftermarket cooler on and put this damn thing into production (minus the video card, since it is a loaner, my next game will be Starcraft 2 so I have lots of time to wait)!
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24141
Joined: Mon May 24, 2004 1:19 am

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 5:09 pm

I'd like to see pics of this ghetto rigged rig.
*edit i used the antec spot cool to help out with a deadzone i had in my case. Works great.
Looking for Knowledge wrote:When drunk.....
I want to have sex, but find I am more likely to be shot down than when I am sober.
Heiwashin
Grand Gerbil Poohbah
 
Posts: 3027
Joined: Wed Dec 13, 2006 12:21 pm
Location: Denham Springs, LA

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 8:51 pm

Heiwashin wrote:I'd like to see pics of this ghetto rigged rig.
Here you go. :D
Image
The purple-ish fan shroud/mount came from my trusty old SX1030 case. The fan is some 80mm variety. You can see I put some scotch tape on that "tongue" onto a greyish surface. That came from some crappy old PSU I have lying around, strictly to elevate the fan to the motherboard's level. The fans were running when I took this, the shutter was faster. As I described before the extra airflow will go through the RAM, the CPU HS, the NB HS, and then to the video card backplate. Notice the thin edges on the fan? At first I just tried to tape to those thin edges, it worked for a while and once it fell off the setup prompt hung. That's how I proved the video card was indeed suffering from some overheating. GPU overheating + (possible) stress of the NB gives me enough variations to run around trying to nail things down. :(

Heiwashin wrote:*edit i used the antec spot cool to help out with a deadzone i had in my case. Works great.
It's weird, I thought this Spotcool thing is a good idea. Surprisingly it is not selling like hot cakes and not a lot of stores carry it anymore. :-?
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24141
Joined: Mon May 24, 2004 1:19 am

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 9:12 pm

You people and your top quality camera's *shakes fist*
Yea, only prob with the spotcool is the led's are brighter than my entire case. I had to lean it facing slightly away from the window.
Looking for Knowledge wrote:When drunk.....
I want to have sex, but find I am more likely to be shot down than when I am sober.
Heiwashin
Grand Gerbil Poohbah
 
Posts: 3027
Joined: Wed Dec 13, 2006 12:21 pm
Location: Denham Springs, LA

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 01, 2008 9:18 pm

Heiwashin wrote:You people and your top quality camera's *shakes fist*
I guess I'll take that as a compliment. This was taken with my aging Canon A40. If you really want to feel picky it's actually slightly overexposed, plus I could definitely use more than 2-point focusing (the camera were only able to do 2 but it was capable of 3, those new fancy 9-point AiAF would be better). It was just a quick snap. ;)

Heiwashin wrote:Yea, only prob with the spotcool is the led's are brighter than my entire case. I had to lean it facing slightly away from the window.
I don't have a window so this won't be a problem for me. However, my smallish Asus TM-210 may be too small anyway.

Edit: finally got through enough System Restore and Repair Installation to get the corrupted test OS back. Running Orthos and RTHDRIBL now at 333x9 with DDR2-667 4-4-4-15. A couple of minutes so far it seems to be fine. Kill-A-Watt finally goes over 200W to like 205W max. Vdroop is now reported to be 1.320V from CPU-Z. We will see how this goes.
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24141
Joined: Mon May 24, 2004 1:19 am

Re: GPU adds to instability, say it ain't so?

Postposted on Thu May 15, 2008 5:25 am

I like to point out that there are sometimes boards that are broken. It is easy to spot one that does not boot. Spotting one that does not work quite right is a bit harder. If you think you know a lot about tweaking, cooling and over-clocking you may never see one (no offence).

I have got an Asus board with a strange BIOS. When I set the memory timings to 5-5-5-15 it configures the memory as 5-5-5-13. When I set it to 5-5-5-17 it then shows as 5-5-5-15 (in applications like CPUZ and AMD's OverDrive). Only Asus will know where those 2T have gone to (they are probably off to lunch).
sdack
Gerbil
 
Posts: 66
Joined: Mon Apr 21, 2008 3:47 am
Location: In another thread, having a nice talk


Return to Overclocking, Tweaking, & Cooling

Who is online

Users browsing this forum: No registered users and 3 guests