5870 gray screen crash.

Whoa wait what? 1350mV? Mine is set to 1149mV and I believe stock is 1175mv?

Yeah, I rebooted (GSODed at the desktop after several days of uptime) and now it's at 1125.

1350 worked like a charm for gaming, though. I assumed that was what it wanted when 1349 was causing all the crap...

FWIW, I've never touched it (except to force it from 1349 to 1350). So, if it's still not right, then I don't know why or how it got to where it is. Not my doing.

But the 1349 is why my gaming was all f-ed up, presumably since 10.3s... After installation of 10.4s is the first time I noticed the 1349... I just realized "49" sounded too goofy... And since I can't seem to force it there anymore now, I doubt I did it even accidentally...
 
Whoa wait what? 1350mV? Mine is set to 1149mV and I believe stock is 1175mv?

Okay, I need someone who knows their stuff to respond here. (don't care who, just need to get this figured out...)

in the past when my fan's not worked, a reboot had fixed it...

Well, now it's crystal what's happening...

When I reboot, apparently my card's core voltage just goes to whatever setting it likes, for who knows what reason...

Last two reboots today it's booted to 1125... Last week (last previous reboot, yes the system's been up for a week...) it booted to 1349...

Now, here are the facts:

Didn't play a game 'til Thu (started TDU over from the start...) fan was not working, was about to reboot when I noticed the odd "1349", put it it 1350, everything worked perfect...

So, past week's core voltage settings:

1349: when playing TDU things are fine, but fan will stop working, will go down to ~60%, and sometimes down to 0%... card begins to overheat, alt-tab to desktop, fan speeds up when Powerplay kicks in and drops core clock to 400, affecting core voltage setting, so fan works again... So, no GSODs, etc, just fan wouldn't work correctly...

1350: everything worked perfect... Since Thu have played 18 hours of TDU, not a single problem... So, if 1350 is wrong, well, it appears to work...

1125: (where it's set itself to for last two reboots...) Start TDU, immediately primary screen glitches, and 2nd screen (dual monitors, 2nd monitors is for monitoring whatever) major glitches... fan works fine, but I GSOD within five minutes...


So, two issues:

1) If 1125 sounds correct, why do I GSOD within minutes?

2) I never thought to monitor this, but why can my core voltage go to whatever the hell setting it likes after every reboot?

Apparently #2 is my problem. I never thought to monitor what my card auto-sets itself to after a reboot... Any ideas why it is doing this?

And is 1350 fine? It work flawlessly the past week, with 18 hours spent on TDU, sometimes playing for four hours straight without a single glitch...
 
^Default voltage is a bit tricky since I use MSI Afterburner and the software does not reset the voltage to the exact interval. However, I have owned five HD 5870s and the voltage before adjustment is indeed 1.149V. Sometimes, Afterburner sets it to 1.125V or 1.163V if I reset my overclock profile. Therefore, Afterburner does not choose a consistent interval. I've never encountered the GSOD during stock settings, however, I have simulated the issue during overclocking. The most reliable cause is memory corruption of if the memory clock is set above limitations. I would consistently push the cards above their max to find the ceiling for overclocking. Typically, these cards artifact with easy above 1275+ memory clock. A few golden cards can push above 1300+. Once the wall is reached, it cannot be solved regardless of voltage. However, I have discovered an aggressive fan profile sometimes sustain their maximum setting longer.

What I'm trying to say is lower your memory clock or increase your fan setting and voltage. A stock HD 5870 should run 850/1200 without voltage adjustment. If you're using 1350mV to run stock setting then you've received a bad card. In comparison, most of the cards I tested can push 1000/1300 at 1350mV. But GSOD almost always occur above the 1300 ceiling. It should not happen at 1200. The issue may reflect that the manufacture yield is still not good enough to hit 1200 default on every card. Consequently, ATi has released custom memory settings for cards that do not perform at factory specs.
 
A stock HD 5870 should run 850/1200 without voltage adjustment. If you're using 1350mV to run stock setting then you've received a bad card. In comparison, most of the cards I tested can push 1000/1300 at 1350mV.

I have an XFX XXX, so it's manuf OCed to 875/1300. Is this why 1350 worked so well?

FWIW, "1349" CV caused the CV to go to 1.265 during gaming... Like I said, fan didn't want to work right. But at 1350 card was fine leaving fan speed set at auto.

This worked for 18 hours, not a single glitch at all.

Do you recommend I keep it at 1350?

Or, do I just have a bad card?

Either way, I think this sucks. Bad card that acts fine if I manually adjust settings, or bad settings from ATi/XFX. My expectations if for these things to work out of the box, and especially not for me to have to manually adjust something for possibly every reboot.

Or are you suggesting that MSIA is auto-adjusting this itself, and getting it wrong?

(FWIW - I'm inclined to stick to 1350... it works...)

And if this is the solution, it kind of stinks that it took nearly half a year to resolve... Although gaming initially on it was fine, it was the desktop that was unstable...

But I don't monitor such settings, so I wasn't watching for it...



ps I believe Kyle mentioned months ago that ATi was "quietly" upping the voltage on future cards?
 
There is no harm with sticking to 1350mV if it works for you. Since you mentioned you own the XXX version, I am surprised ATi decided to upclock to 1300 without fan adjustment or voltage adjustment. As I said before, most of the cards I tested cannot hit 1300 clock stable; this means that only two of the cards I had can run ATi Tool for 3+ hours at 1300 but with voltage at 1350mV. My observation is memory stability does not benefit directly from voltage because the same cards can hit 1275 at only 1225mV. Apparently, there is a wall or a ceiling at a certain point.

Point is: I'm speculating ATi were overoptimistic about the memory capacities of the stock cards because not many of them can run at 1300 regardless of voltage. On the otherhand, the core clock has shown increasing stability with better voltage, although I have hit certain walls as well. If you find your setting works best at 1350 then keep it. However, if I had your card, I would downclock the memory and upclock the core, which scales much better. FYI, I do not keep my cards at 1000/1300 for 24/7 use, but rather at 950/1275/1225mV since the heat builds up pretty fast in crossfire.

Lastly, your voltage should not change without adjustment because the XXX settings is in the BIOs. You voltage monitoring during your gameplay and if it still happens ask XFX. They may or may not ask to reflash your BIOs. However, it could be the ULPS setting working since it does flux the settings. You may wish to disable it if you're not worried about energy savings.
 
Last edited:
If you find your setting works best at 1350 then keep it. However, if I had your card, I would downclock the memory and upclock the core, which scales much better. FYI, I do not keep my cards at 1000/1300 for 24/7 use, but rather at 950/1275/1225mV since the heat builds up pretty fast in crossfire.

Dropping my memory from 1300 to even just 1299 causes PowerPlay to act up. At the desktop (dual monitors), it will kick the clock from 400 down to 157, which causes glitches on my screens. Anything below 1300 causes this. Even in games I will get screen glitching with anything below 1300.


Lastly, your voltage should not change without adjustment because the XXX settings is in the BIOs. You voltage monitoring during your gameplay and if it still happens ask XFX. They may or may not ask to reflash your BIOs. However, it could be the ULPS setting working since it does flux the settings. You may wish to disable it if you're not worried about energy savings.

Well, apparently my core voltage does, as in the past rebooting fixed the problem before I understood what was going on. I'm positive I was not fudging with it, and as said, now I can't even force the core voltage to 1349 if I wanted to. 1349 will set it to 1350.

I'm going to Google it, but what is ULPS? How do I disable it? It sounds like it is Powerplay? If so, do you think this can be part of my culprit?

Thanks for the help here. I admit I expected it to work out of the box, so I haven't been paying attention to where these settings are ending up. I didn't feel I had any need or reason to monitor where my core voltage was ending up every rebooot.

With 9.12/10.1/10.2, games were fine, but the desktop was really unstable. (My assumption was powerplay was going bonkers...) 10.3/10.4, that flip-flopped. Desktop's been pretty stable (can last over a week...), but games can tend to be FUBARed,

FWIW - I'm still having issues with TDU. Two weeks ago when I played the game from scratch (yes, I restarted the game from the beginning twice in the last week, just to clarify), both times I had issues with TDU coincidentally (?) right around the 18-hr mark. Last time about a half-dozen GSOD/BSOD/reboots later, the game went back to rock-solid stable. Otherwise it wouldn't last 5 minutes (like now), or one time even 5 seconds.

In the past hour or so I've experienced a GSOD or driver crash in TDU with the core at 1125 and at 1350. So, neither is a fix. But, TDU was solid at 1350 for the previous 18 hours.

FYI: I've never seen the voltage change during a session. Only upon initial boot may I see whatever number... It'll stay put, but it'll boot to whatever core voltage it feels like...

Let me know if I should talk to XFX/ATi, or if I should just RMA.
 
I found [HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Control\Cl ass\{4D36E968-E325-11CE-BFC1-08002BE10318}]

Under 0004, I have ENABLEULPS = 1. (The others I found are already at 0). Do I need to change this 0004 one to zero to disable ULPS?
 
That's what I did. Make sure the 0000.. keys are set to 0 under EnableUlps. However, I should mention I had to do this for Crossfire. I'm not sure if this matters for single configurations. In Crossfire, the ULPS will hardlock any overclocking and blue screen immediately. However, removing this impass let's me overclock normally for both cards. I'm not familiar with PowerPlay, it could be a software exclusive to XXX owners. The other issue is the card normally drops to 157mHz in 2D applications. It's a ridiculous but power saving feature. If you're experiencing aftifacts it means you need to upclock the 2D profile by creating a profile in CCC and manually editing the values in notepad. I've made a thread here about it once so check my history.
 
I'm not familiar with PowerPlay, it could be a software exclusive to XXX owners.

Powerplay (to my understanding) is the power saving feature that downclocks the card. Sounds like you are calling it ULPS. Neither of us had heard of it by the other name? (Ultra Low Power Settings? Yes, we're talking about the same thing, different names...)

The other issue is the card normally drops to 157mHz in 2D applications. It's a ridiculous but power saving feature. If you're experiencing aftifacts it means you need to upclock the 2D profile by creating a profile in CCC and manually editing the values in notepad. I've made a thread here about it once so check my history.

I tried that once. It resulted in a Windows session that would GSOD, driver resets, GSOD, driver resets, etc etc etc. I had to safemode to remove the CCC profile.

Wouldn't ULPS=0 keep the card at 875 the entire time?
 
Sigh, I've gotten the GSOD a couple of times in The Saboteur. Strangely, that's the only title that I've experienced it in so far since having my 5870 for a few weeks now. Running the 10.3's and recently restored the OC back to my PC (had backed down to stock for a few weeks for no real reason).

Time to dig through the thread again and look at possible fixes...was really hoping I wouldn't run into this as some people said it only affected a small number of cards (right). If this keeps up, I might have to go with a GTX 480 although I hear it's having driver problems of its own right now.

Honestly I kind of miss my 260s in SLI...wanted to go back to a single GPU but if the GSOD persists suffice it to say I'll be annoyed.
 
I dont understand why ppl think that if your videocard is GSOD'ing that lowering the mem or this and that will fix it, point being is that it shouldnt crash at all on stock settings period.
 
I dont understand why ppl think that if your videocard is GSOD'ing that lowering the mem or this and that will fix it, point being is that it shouldnt crash at all on stock settings period.

In another words - RMA it?

Today my card was a major PITA, but changing it to 850/1250 resolved it. (First time I've ever adjust core/mem like that...)
 
Just got a new 5870 today. Diamond HD 5870. $300 brand new sealed in box from a guy on craigslist. Out of the box I got gsod on wow! WTF! I remember reading about it here. So I Manually increased the Core Voltage to 1150 from 1125. Fixed the issue. Ran bench marks perfectly. 0 issues thus far been using it a few hours.

Upgraded from a 5850. Couldn't pass up the deal. Man this thing runs hot tho.... about 10c more during 100% load. 33c idle 70c @100%
 
i lowered my oc on my cpu 50 mhz it also lowered my ram a bit as well.. gsod stopped !!
 
I'd like to add my two cents in for my experience...

My desktop PC's 5850 has run flawlessly.

My laptop's 5870 (which is really more akin to a downclocked desktop 5770) has the GSOD problem with Furmark with newer drivers. With the old 9.12 drivers, the card works flawlessly. Other G73JH laptop owners have reported the exact same thing. I myself attribute it to minor incompatibilities between the generic mobility Radeon drivers and the specific hardware in my laptop...since it works fine on the 9.12 drivers I'm sticking with those.
 
The GSOD issue on G73JH units has been a hot topic over at NBR.com, and I wanted to link you over there for some tips. You can see the links in my sig there for background info on how we came to the possible working setup.
 
I hadn't had a crash in a LONG LONG time.. I started this thread lol. So the other day I decided to play with gadgets in Win7. Then went into a game just to get a GSOD 5 minutes into the game. Went back and turned off gadgets and have gamed many hours since without issue. It could have been totally random this happened. Worth a shot for some if you use gadgets. I dont even see how this would really be an issue when in a game *shrug*.
 
I hadn't had a crash in a LONG LONG time.. I started this thread lol. So the other day I decided to play with gadgets in Win7. Then went into a game just to get a GSOD 5 minutes into the game. Went back and turned off gadgets and have gamed many hours since without issue. It could have been totally random this happened. Worth a shot for some if you use gadgets. I dont even see how this would really be an issue when in a game *shrug*.

I had that, and it was only inside of a game that the fan would stop working. Turn off gadgets and monitoring and everything's fine. For me, it's not random at all. Over-monitoring my system would cause the Vcore voltage to go wonky, causing things to not work.
 
I had that, and it was only inside of a game that the fan would stop working. Turn off gadgets and monitoring and everything's fine. For me, it's not random at all. Over-monitoring my system would cause the Vcore voltage to go wonky, causing things to not work.

Sweet, guess I was on the right track with that one. I have been overclocked and running great for a long time now. Sold my setup I got constant GSOD's with and the guy popped in a 5850 overclocked and never had an issue. Also I kept the video card from that machien for my self and it never had a GSOD on this rig until the other day with the gadget problem.
 
Last edited:
Back
Top