Sapphire Toxic HD4850 Crash

[defach]

I have a Sapphire Toxic HD 4850 512mb from Newegg. I replaced an Nvidia 7800GT with it. I experience random crashes in a wide variety of games; the screen blanks out, monitor goes to no signal mode, computer freezes and the last half second of audio loops. Only way out of this is a hard reboot. I will detail all of the things I have tried and give my system specs below. I am looking for someone else with a similar problem or any suggestions anyone can make to fix this. I am aware that a new BIOS has been released for the sapphire 4870 cards, but it will not be compatible with my card.

I NEVER had any crashes with my Nvidia card. I am not bashing here, just providing a compare to rule out other parts.

Troubleshooting:
600 watt PSU, tested rails, they push fine.
Multiple memtests, no errors
Scandisk/defragmentation of all hard drives
Disabled antivirus / other scanning programs
Safe mode cleaning of drivers, tried 8.7, 8.8, 8.9 beta and now using 8.9
Profile (the XML file) editing to force clocks to not swap between 2d/3d
GPU/Memory is NOT overheating -- entire system is within temperature ranges
RBE to set all cores at a constant clock speed with the highest voltage
Changed all 3d settings in CCC to every possible combination. I seem to have the best luck with AAA off, VPU recover on (even though it doesn't work) force off vsync, allowing applications to control AA and sampling
Loading GPU with Folding@home to keep it busy (seems to help)
Different resolutions, windowed/fullscreen settings

Games affected:
Bioshock - Unplayable. 2-15 minutes into the game, randomly on any level, black screen freeze
Civilization 4 Colonization - at any point, even sometimes on the menus
World of Warcraft - This game runs the best, no crashes in windowed mode, maximized, can run two instances (of the game, multiboxing) just fine
Call of Duty 4 - First game I played with card, played the whole game without a single crash, still does not crash.

Inconsistancies:
I was once able to play Bioshock for 8 hours with only one crash about halfway through. I have tried to reproduce the settings and can not get a similar experience.
I was also once able to play Civ 4 Colonization for six hours without a single crash. The next day when I resumed playing, I crashed within minutes.
WoW would occasionally crash when using the test realm client in windowed mode without maximizing it (removing the title bar, edges)
Sometimes after changing some 3d settings, I will get a crash when just using applications (not gaming or video intensive). This crash is different -- it will VPU recover itself, and then tell me I need to restart because the display driver stopped responding
I have thrice gotten a BSOD while not gaming that gave a Thread stuck in device driver error. I have attempted all of the troubleshooting related to that (namely, new drivers) and have been unable to fix all of my problems.

System Specs:
Windows XP Pro 32bit SP3
5g ram (i know it's overkill, memtested fine)
Xfi Platinum
Pentium D 2.8

I am at a loss for ideas here. I firmly do not think that the card is 'defective', only that this is a driver or GPU BIOS issue. I will be happy to provide additional information. I apologize for the length and unpretty formatting. Thank you to everyone who reads this.

 

[oli_ramsay]

lol no

 

[Polaris573]

Free bump because the forum anti-spam software blocked your posts and made a mess of your thread. Sorry about that.

 

[erocker]

At any time during any of the crashes did a small red LED turn on, located on the bottom of the card?

 

[defach]

Thanks for all the help getting the post sorted out!

I have not noticed it. I will take a look at it during crashing -- right now there are no LED lights at all on the card. Just have to turn the computer around. I will crash it right now and be back in ~15. Using 8.10 RC2 now, same deal.

 

[defach]

No LEDs at all during or after crash. When I reboot the system, I also do not see any LED lights come on. I am in a dark room with a case window and minimal case lighting. I can see the back (non fan) side of the card easily. I do not see where there would be LEDs to light up.

 

[berp]

Were you able to find a solution? I'm having the same problem and I've tried pretty much the same things you mention without any success.

 

[DonInKansas]

Did you uninstall Nvidia drivers, run Driver Cleaner before installing ATi drivers? The remnants may not be playing nicely together.

 

[spearman914]

More detailed plz . 5GB RAM? How the freak is there a combination that makes 5 GB RAM.
AND WTF::: How the hell you play Cod 4 and Bioshock on this thing:

System Specs:
Windows XP Pro 32bit SP3
5g ram (i know it's overkill, memtested fine)
Xfi Platinum
Pentium D 2.8

 

[defach]

I have resolved my problem. I needed to underclock to stock (not Toxic) speeds using Rivatuner. Underclocking with any other method did not work. I apply the profile on startup, and I have had ZERO crashes since. Looks like there was a problem with the TOXIC cards and they can't run at advertised speed.

Yes I did clean the NVIDIA drivers, and I have zero FPS problems running COD4 or Bioshock. What is so bad about my setup? I realise it's not a Core 2 Duo, but the PD's were great chips at the time, they just run a little bit hot. I run them at 1680x1050 with all options on, too.

I had two 1g chips, bought two 2g chips, one 2g chip failed, replaced it with another 1g. 3x1g + 2g = 5g async.

 

[WarhammerTX]

try just using the 2 x1 gig chips for a total of 2 gigs and put it back at default speeds and see if that fixes the problem.

 

[Muhad]

Maybe your over-clocking your card to much.

 

[Paintface]

I have resolved my problem. I needed to underclock to stock (not Toxic) speeds using Rivatuner. Underclocking with any other method did not work. I apply the profile on startup, and I have had ZERO crashes since. Looks like there was a problem with the TOXIC cards and they can't run at advertised speed.

Yes I did clean the NVIDIA drivers, and I have zero FPS problems running COD4 or Bioshock. What is so bad about my setup? I realise it's not a Core 2 Duo, but the PD's were great chips at the time, they just run a little bit hot. I run them at 1680x1050 with all options on, too.

I had two 1g chips, bought two 2g chips, one 2g chip failed, replaced it with another 1g. 3x1g + 2g = 5g async.

i have the same problem, sadly underclocking does not fix it for me, i suspected my creative soundcard giving issues, did you try playing at toxic clocks without your xifi card pulled out and drivers disabled? this would help me out in trouble shooting.

thanks in advance

 

[[I.R.A]_FBi]

rma card.

 

[blackhood15]

I seem to have the same problem. How do you underclock the card? I also seem to have seen a few references to updating the GPU Bios on a few sites, how do you do that?

My system does not BSOD but does crash and I get a screen of vertical lines. Not always consistent with games. Games which I can't get more than 2 mins into are: bioshock/crysis/cryis warhead/empite total war (inconsistent as to when it crashes)/farcry2/stalker. Convesly I got all the way through medol of honour airbourne and so far so good with COD World at War.

Any help from anyone would be appreciated as i'm about to give up on PC gaming altogether

My system is:
Vista home
2 x OCZ 2GB KIt (2x1GB) DDR2 1066MHz/PC2-8500 Reaper Memory HPC Edition Dual Channel Kit CL5
ASUS M3A 770 Socket AM2+ 8 channel audio ATX Motherboard
AMD Phenom X4 Quad Core 9950 2.6GHz Black Edition Socket AM2+ Retail Boxed Processor
Sapphire HD 4850 Toxic Edition 512MB GDDR3 Dual DVI HDTV Out PCI-E Graphics Card
Creative soundblaster X Fi fatality

 

[SonDa5]

Toxic over heating problem possibly nailed on the head

I'm 99.9% sure I have figured out the root of a huge problem with the Toxic HD4850 crashing.

Simple.

The card gets hot because the fan spins so slow.

I figured this out using FurMark and enabling Xfire for it.
Using Furmark to test for stability and running CCC to underclock the Hd4850 Toxic as close as possible to the reference design VisionTek HD4850.

I'll post some screen shots later on of the temps of each card along with the speeds of the fans all while under load.

Right now I'm going to try to figure out a way to increase the speed of the Toxic HD4850 fan. CCC doesn't do a thing for it.

Fan of Toxic maxes out around 1950 RPM and the VisionTek tops out at over 7860RPM using CCC.


The VisionTek stock heat sink and fan keep it running at 65/69/65 degress stable with the fan at 100% under 99% load from Furmark.

Toxic with GPU clock adjust to 655 mhz and GDDR3 RAM@1100 mhz under Furmark load runs 87/96/88 with only 90% load.


I think the percentage of load able to process is limited to the heat of the GPU. Since the VisionTek GPU is much cooler it can process 99% load while the hot Sapphire Toxic can only handle 90% of full furmark load.

I'm sure overall temps of both card in Xfire effect how efficient it can process.

 

[SonDa5]

Here is a screen shot that covers the situation.

 

[SonDa5]

In response to this locked thread:
http://forums.techpowerup.com/showthread.php?p=953069#post953069




This fan is very quiet.

Big mistake in not making the fan control adjustable.

This card has so much potential to perform but is limited because the fan doesn't spin fast enough to keep it cool.

Where is the OP of that thread? That was the only post made by the OP.

 

[zlimvos]

At any time during any of the crashes did a small red LED turn on, located on the bottom of the card?

Hello there
I have crashes with the red led turning on. I am trying now to find what it means and going through forums. I hope you are still around to tell me cause there is nothing in the 'manual'..
(I suspect temperature)
Sorry for reviving the post, but i figured maybe more googled and ended up in this post for same reason.

cheers

 

[blackhood15]

no problem. After reading various comments on here and also even getting a letter puplished in PCZone mag about the issue, I still kept getting crashes, intended to send card back, but never got around to it, so just put up with it, kept machine well cooled and settings on low, with intention to replace with a dx11 card.
However, i randomly looked to recheck drivers last month and since then the card seems ok, but I haven't really tested it, but certainly no crashes since, but not used with extreme games yet.
I don't remember ever having any red light issues

 

[Widjaja]

I know old resurrected post.

I have seen these issues before with the toxic cards sadly.

Seems they are just toxic waste in some cases and under-clocking is the only answer.

But then again.

My HD4850 does the same thing as the OPs if I run my AC Twin Turbo passive.
The strange thing is.....the temperature read out is below stock cooler temperatures and this will happen which leads me to believe it is not the core but another component of the GPU which is over heating....

I point to the VRMs as the VRM heat sinks supplied with the Twin Turbo are rather small and not directly in contact with the main cooler itself.

So I would say the same thing which happened to my HD4850 when running passive has happening to the Toxic versions.