XFX HD5850 Gray Screen in 3D mode

[tkin]

I recently purchased a XFX HD5850(model no: HD-585x-ZAFV||This model with free AVP game: http://www.newegg.com/Product/Product.aspx?Item=N82E16814150477&cm_re=14-150-_-14-150-477-_-Product)

and facing problems with it, at first it was running fine, then I tried testing its stability, it crashed with MSI Kombustor(1680x1050, post fx , unlock power draw on), then with OCCT(1680x1050, shader complexity 8) and furmark(1680x1050, 8xmsaa, xtreme burning mode on).

The crash came as flickers first, followed by a striped screen, the colour of the screen varied, with OCCT it was red, with msi kombustor it was gray and furmark it was blue, the striped screen sometime go away for a short while and return to normal, then lock up again.

Details:

Catalyst 10.5

Win 7(32bit)

Hardware details:

E8400 @ stock speeds

2x2GB Kingston 800MHZ DDR2 RAM

Asus P5Q mobo with latest bios.

Corsair HX620w

System properly cooled with multiple fans



Now about the crash, it would immediately commence if I kept the fan profile to default, which only ramps up at 80c and my card maxes out at 90c(all three temps in GPU-Z maxes out @ 90c), but manually ramping the fan upto 70% prevents the crash for sometime(about 5mins), also while I'm gaming the card never crashed although GPU temp reached 87c(crysis, metro 2033, Just cause 2, borderlands), but when running furmark or kombustor or occt the card would crash when the GPU crosses 72c.



Does this mean the card is faulty? Should I RMA it? I never tried to oc it as it a non reference model and lacks the voltage tuning option, also my room ambient is high(30c+)



PS: I tested my system with my earlier 9800GTX+(and my friend's GTX460) and it was rock stable even at 95c(GTX460) with kombustor/furmark/occt and for my 9800GTX+ it was stable at 76c.

Also note that I live in India and getting a RMA here is an unbelievable pain so its the last option for me.

 

[Maziar]

When your card crashes with benchmarks,there must be a problem with it.
How much time did you play the games ? try running a game(like crysis) for about 1hour and check if it crashes or not

 

[tkin]

Well, now it crashes with crysis, getting the rma today, crap card from a crappier company, should've stayed with nVidia.

 

[Tamz_msc]

You can try to run it with older drivers, like 10.4 or 10.3 catalyst.

 

[Maziar]

Well, now it crashes with crysis, getting the rma today, crap card from a crappier company, should've stayed with nVidia.

Try older drivers like Tamz_msc said
Also it could have been happened if you bought a Nvidia card too.

 

[tkin]

Yes, it could've happened to nVidia cards, but it would be a general problem(artifacts in games and such) but this seems a widespread problem, a lot of people complained about this.

PS: I tried the 10.5 and 10.7 drivers and the crashes happened in both, at first it wasn't crashing under games and later it started to crash in crysis and some artifacts started to appear, I an GPU-Z along with it and saw that during the gray screen my HD5850 clock speeds remained constant(not dropped) but GPU load was 0%, so it wasn't throttling and this bug wasn't happening in 2d mode.

I think this is a hardware bug because when I got it I ran a maxed out furmark and it passed with flying colours, it also passed a 30min OCCT no problem, the next day it crashed under MSI kombustor, then it also crashed under furmark and OCCT in a similiar pattern(flickers at first then the gray/blue/red screen with stripes), so this indicates that the card is degrading because if this was a driver issue it would crash under the first run of furmark, occt not later.
On the next day it crashed under Crysis, the crashes does not happen immediately after starting the game, it takes some time for the GPU to heat up and then the crashes start, so basically it indicates a overheating component maybe a mem chip or a vrm.

If I put my fan speed @ 80% all the three temp sensors in GPU-Z shows below 75c but the card crashes under load, so I think its a component that is not monitored which is heating up and crashing.

Anyway I rma'd it, rma is a pain here in India as I said, it takes a month to get the card back, the distro here is saying I'll get a new card in return, that's what gets me worried, if this new card is a dump I'm done for.

Does anyone know the defect rate of XFX ati cards? XFX is the only vendor here and there's no refund policy here.

 

[bystander]

You said your case is "properly cooled", but that doesn't always mean the same for everyone. The thing is, the 58XX cards are rather prone to overheating, especially the memory, which isn't well cooled to start with and on the edge at stock settings.

Even average cooling is often not enough to perform, and the temps you are getting are rather high for stock settings.

Make sure you have your fans configured in a push pull configuration, and there is free air flow. Fans not configured well, or loose cables in the way may cause problems. Ambiant tempuratures can also cause problems. If you play in a room that is already hot, it's harder to keep the cards cool.

Are you doing any kind of overclocking? I've found the memory on these cards overheat easily at stock settings. I personally have the memory underclocked a little (by 50), which greatly improves stability.

 

[tkin]

This is my system, its not great but given that I don't OC it remains pretty cool, in idle my CPU stays 49c(on both cores), my hdds idle at 35c, 9800GTX+ idles at 61c(it does not reduce frequencies in idle and has only one performance state, like other G92s) so its descent given my ambient stays 30c+(no ac).

I've also noticed that the XFX fan profile has issues, the fan remains at 23% and idles the card @ 50c, when loading the fan starts to ramp up @ 80c and goes to 37% at 90c and stays there, it looks like that the BIOS fan profile tries to keep the card below 90c at any cost but also tries to keep the system silent.

PS: I tried to reduce the frequencies to 700/900 and it crashed but took a bit longer, also I tried upping my ram volts to 1.9 as some were saying, it didn't help, I can't increase the voltage of my XFX HD5850 as it is a non reference model, in load it goes to 1.088v and stays there, now I'm baffled as to what could've actually been the issue?

 

[bystander]

That definately is the issue, your ambient temperatures are very high. All I can suggest you do is open up the case and point a fan at it.

As far as the fan settings go, you can turn it up manually if you wish. MSI afterburner also has the ability to make your own profile that will increase fan speeds more than the default one does.

 

[tkin]

That definately is the issue, your ambient temperatures are very high. All I can suggest you do is open up the case and point a fan at it.

As far as the fan settings go, you can turn it up manually if you wish. MSI afterburner also has the ability to make your own profile that will increase fan speeds more than the default one does.

Wait, 30c ambient is so high that it fried my card? but it does not fry my 9800GTX+ or GTX460, also in India the ambient is high like this and some people here run HD5850s like anything? I'll get a new cabby later.

 

[bystander]

I am not saying it's "frying" your card, just that these cards have an issue with memory running hot without you realizing it. You essentially have 2 options here. You can RMA the card, or underclock the memory 50-100 under stock.

 

[tkin]

I am not saying it's "frying" your card, just that these cards have an issue with memory running hot without you realizing it. You essentially have 2 options here. You can RMA the card, or underclock the memory 50-100 under stock.

Well, I downclocked to 700/900 and its still crashing, now with almost every demanding 3d app. Now can I conclude that the hardware is faulty? I did not oc even a bit.

 

[bystander]

Probably. Your ambients will cause higher than typical load and idle temps, but they shouldn't cause that much of a problem.

 

[tkin]

Probably. Your ambients will cause higher than typical load and idle temps, but they shouldn't cause that much of a problem.

If that's true than it will also ramp up my fan, but my fan stays at 37% at 90c, it does not ramp up, and also my CPU idle is(50,44c, E8400 stock, intel HSF), so my idle and load will be high but why is it crashing and artifacting?

What are the chances that my next replacement after I rma the card will also have this issue?

 

[sunjun]

If that's true than it will also ramp up my fan, but my fan stays at 37% at 90c, it does not ramp up, and also my CPU idle is(50,44c, E8400 stock, intel HSF), so my idle and load will be high but why is it crashing and artifacting?

What are the chances that my next replacement after I rma the card will also have this issue?

u can try NVIDIA GTX 470 i heard that their drivers r much stable n better than ati .. ....btw i recently bought an ati 5850(FX HD-585X-ZAFC)..only 3 days have passed..... overscan prob.. i only played "nfs shift" ,"dirt2" , "under cover" ... "X men wolverin" ....i will try crysis today and i will reply u what will happen ... in bangladesh we r not permitted to change the company .. like i cant go for nvidia ..... i have to stick wid that .... btw what r u using vga dvi or hdmi?

 

[sunjun]

I recently purchased a XFX HD5850(model no: HD-585x-ZAFV||This model with free AVP game: http://www.newegg.com/Product/Product.aspx?Item=N82E16814150477&cm_re=14-150-_-14-150-477-_-Product)

and facing problems with it, at first it was running fine, then I tried testing its stability, it crashed with MSI Kombustor(1680x1050, post fx , unlock power draw on), then with OCCT(1680x1050, shader complexity 8) and furmark(1680x1050, 8xmsaa, xtreme burning mode on).

The crash came as flickers first, followed by a striped screen, the colour of the screen varied, with OCCT it was red, with msi kombustor it was gray and furmark it was blue, the striped screen sometime go away for a short while and return to normal, then lock up again.

Details:

Catalyst 10.5

Win 7(32bit)

Hardware details:

E8400 @ stock speeds

2x2GB Kingston 800MHZ DDR2 RAM

Asus P5Q mobo with latest bios.

Corsair HX620w

System properly cooled with multiple fans



Now about the crash, it would immediately commence if I kept the fan profile to default, which only ramps up at 80c and my card maxes out at 90c(all three temps in GPU-Z maxes out @ 90c), but manually ramping the fan upto 70% prevents the crash for sometime(about 5mins), also while I'm gaming the card never crashed although GPU temp reached 87c(crysis, metro 2033, Just cause 2, borderlands), but when running furmark or kombustor or occt the card would crash when the GPU crosses 72c.



Does this mean the card is faulty? Should I RMA it? I never tried to oc it as it a non reference model and lacks the voltage tuning option, also my room ambient is high(30c+)



PS: I tested my system with my earlier 9800GTX+(and my friend's GTX460) and it was rock stable even at 95c(GTX460) with kombustor/furmark/occt and for my 9800GTX+ it was stable at 76c.

Also note that I live in India and getting a RMA here is an unbelievable pain so its the last option for me.

i think u should go for windows 7 64 bit as u have 4gb of ram...32bit os limits u at 3gb of ram although u have 4gb of ram or hiegher...

 

[bystander]

If that's true than it will also ramp up my fan, but my fan stays at 37% at 90c, it does not ramp up, and also my CPU idle is(50,44c, E8400 stock, intel HSF), so my idle and load will be high but why is it crashing and artifacting?

What are the chances that my next replacement after I rma the card will also have this issue?

90c is quite hot mind you. While the GPU core may be able to function at 90c (I have experienced these problems at temps over 80c), the video memory seems to be the weak point of the GPU. The sensors don't read the temps of your memory, which may be getting hotter than the GPU core.

You could attempt to turn the fan higher manually, which might help, but I'm not sure how much the fan speed effects the memory temps with the way they are placed in the card.

 

[tkin]

u can try NVIDIA GTX 470 i heard that their drivers r much stable n better than ati .. ....btw i recently bought an ati 5850(FX HD-585X-ZAFC)..only 3 days have passed..... overscan prob.. i only played "nfs shift" ,"dirt2" , "under cover" ... "X men wolverin" ....i will try crysis today and i will reply u what will happen ... in bangladesh we r not permitted to change the company .. like i cant go for nvidia ..... i have to stick wid that .... btw what r u using vga dvi or hdmi?

Sorry to post so late, went out with a attack of acute pancreatitis, anyways I rma'd my card, new card running fine, didn't try furmark though, played crysis warhead and borderlands for sometime, no crash or anything, but I'm not touching Furmark or any other stress tool.

I'm in the same water as you are, i can't change my card either, not that I want to though, GTX480 and GTX470 reaches 95c with load, here in India(kolkata) with 35c+ ambient it will hit 105c and start to throttle, if it does not manage to burn out the vrms by then though, my earlier 9800GTX+(EVGA) could plow through furmark, kombustor or whatever without any issue(my card was factory oc'ed) for hours and never crossed 80c but this XFX crap card is hitting 87c with crysis, fan profile of this card sucks huge, xfx is one of the worst companies ever.

I'm holding out now but planning to sell this card and get a new one(this time, looking for reference HD6xxx series or a GF104 with full 384 shaders enabled when it launches).

Local xfx distro said new cards coming in december, if HD68xx series runs cooler and fixes the GSOD issue, I'm gonna get that first thing, but my first priority is getting a nVidia card, specially from EVGA, the GTX470 and GTX465 from evga is available here but the load temperature has me worried.

 

[tkin]

90c is quite hot mind you. While the GPU core may be able to function at 90c (I have experienced these problems at temps over 80c), the video memory seems to be the weak point of the GPU. The sensors don't read the temps of your memory, which may be getting hotter than the GPU core.

You could attempt to turn the fan higher manually, which might help, but I'm not sure how much the fan speed effects the memory temps with the way they are placed in the card.

Mine looked like a vrm issue, otherwise it would crash immediately in crysis.

 

[razzor03]

Hey, i am having the same issue with my xfx ati 5850. I hv sent it for repairing about 5 times and they even replaced it once. I have a nforce 630a mobo which works perfectly without my gpu while they are saying that my gpu is fine. Xfx suck man. It takes 1 whole week to fix them and 3 week to replace it !