HD4870X2 doesn't like voltages? :(

[AgentGOD]

I pumped 1.31v into both GPU cores via RivaTuner trick (verified by GPU-Z and VT1103 plugin), and set the clock speed to 800 MHz on each core by having ClockCloning turned on. However, after running FurMark for about 5 seconds, the system froze, screen went black, and then shortly after, the 4870 X2's fan started spinning at 100% (was very loud). System was frozen, and didn't respond to "Reset button". I had to shut it off, and turn it back on.

I wanted to try more voltage, because my particular card isn't 100% stable at 800 MHz core at stock voltage (I was using CCC for this). It can handle Crysis for about 10 minutes at the clock, but then system will freeze, then 2 minutes later, screen will go black and reboot.

In each scenario, I had CCC installed, but Overdrive disabled in the first scenario, and enabled in the second and ONLY CCC in the second.

Any help? :)

[NickF]

Back in the day, I had a similar issue with my 6800GT. I would try to give it a but more volts to match the 6800 Ultra, and it would instantly become unstable. I almost bricked my card because I did it via BIOS flashing.

[AgentGOD]

Doesn't solve my problem tho :(

[Santi U]

the 4870 X2's fan started spinning at 100%

What did/do you have your fan set to when running furmark? With my 4870x2 @1.3v and fairly good case airflow my VRM temps were 90c~ @100% fan. I now only use games I play to test for stability, furmark is just too harsh and only shows stability for absolute max thermal conditions imo.



I wanted to try more voltage, because my particular card isn't 100% stable at 800 MHz core at stock voltage

Hmm I would try lower increments from stock voltages than to 1.31v. Test each voltage stepping and see what it nets you in terms of max overclock (if you have done this method than my bad :) ). I would do this with 100% fan to try and eliminate overheating as the cause. Once you get to where you want to be with oc/voltage then lower your fan to see if it's stable or the decibel range is ok for you for that oc.

The fan spinning to 100% usually means (depending on bios specs) max threshold has been met, which is 112c~ for my asus vanilla card. Also sometimes the card will spin to 100% fan or so for me on a cold boot so it could be that it did this on reboot but to be safe and eliminate this variable i would use 100% fan.

You can see my results in the soft mod voltage thread (http://www.xtremesystems.org/forums/showthread.php?t=215521&page=13)

(I apologize if what I said/suggested sounds demeaning as sometimes help thread responses can be as you give not enough information at the time, kinda, heh )

[tool_462]

Have you tried verifying the voltage with a DMM?

[AgentGOD]

Have you tried verifying the voltage with a DMM?
Never have. I assume what the volterra regulator reports should be accurate.


What did/do you have your fan set to when running furmark? With my 4870x2 @1.3v and fairly good case airflow my VRM temps were 90c~ @100% fan. I now only use games I play to test for stability, furmark is just too harsh and only shows stability for absolute max thermal conditions imo.

Fan is at AUTO. And I'm NOT going to run the card at 100% fan 24/7, way too loud.




Hmm I would try lower increments from stock voltages than to 1.31v. Test each voltage stepping and see what it nets you in terms of max overclock (if you have done this method than my bad :) ). I would do this with 100% fan to try and eliminate overheating as the cause. Once you get to where you want to be with oc/voltage then lower your fan to see if it's stable or the decibel range is ok for you for that oc.

The fan spinning to 100% usually means (depending on bios specs) max threshold has been met, which is 112c~ for my asus vanilla card. Also sometimes the card will spin to 100% fan or so for me on a cold boot so it could be that it did this on reboot but to be safe and eliminate this variable i would use 100% fan.

You can see my results in the soft mod voltage thread (http://www.xtremesystems.org/forums/showthread.php?t=215521&page=13)

(I apologize if what I said/suggested sounds demeaning as sometimes help thread responses can be as you give not enough information at the time, kinda, heh )
I started by testing the voltage setting method with 1.27v, and it worked fine in FurMark I believe (at least for the time interval I tested). I did this small baby step just to test and see if setting voltages via RivaTuner actually worked, and it did. So then, I went to 1.3125v on both cores, and then set the GPU clock speeds via RivaTuner, the above happened.

[i found nemo]

what are temps?

[AgentGOD]

what are temps?
Right before it rebooted (second scenario), it shot up to 95C.

[Jor3lBR]

Your card doesn't like 100oC temps not volts mate.

What is that Clockcloning tool? Download link?

[AgentGOD]

Your card doesn't like 100oC temps not volts mate.

What is that Clockcloning tool? Download link?

I meant this:
http://img15.imageshack.us/img15/1091/imeandis.jpg

It's required for CrossFire overclocking in RivaTuner.

And how do you know it's temperature issue?

[BeepBeep2]

If it was 95c on 2D reboot it was most likely around 105-115c on shutdown.
100c is 212f..uhh it's unsafe dude.

These ATI cores don't like being pushed further than 85-90c. 86c would be my ultimate max. "safe".

Try 775 on the core and 1950 (3900) on the memory. If it can't pass FurMark for at least 1 and a half hours, it's not a good overclock.

[AgentGOD]

If it was 95c on 2D reboot it was most likely around 105-115c on shutdown.
100c is 212f..uhh it's unsafe dude.

These ATI cores don't like being pushed further than 85-90c. 86c would be my ultimate max. "safe".
It ran at 92C stock clocks and voltage, and stock fan settings.

Btw, I just manually cranked up my fan speed to 50% in CCC, and now I'm idling at 42C, whereas before I as idling at 65C.
http://img54.imageshack.us/img54/7458/vvvvvvv.jpg

Would it be safe to keep it there? :P

[AgentGOD]

Cranked it up to 70% and ran FurMark for ~2 min:
http://img3.imageshack.us/img3/7436/furmark000000.jpg
Saw it gradually climb from 40C to 89C.

[AgentGOD]

Hmm, apparently he got to 890 MHz core on 55% fan:
http://forums.guru3d.com/showpost.php?p=2840681&postcount=6

[AgentGOD]

Here's a more "real-world" situation:
http://img8.imageshack.us/img8/4397/4870x2oc.th.jpg (http://img8.imageshack.us/my.php?image=4870x2oc.jpg)

Recorded while playing Call of Duty 4 for 2 minutes. Maximum GPU and memory clocks shown, as well as maximum GPU temperature for both cores.

Fan was set to 50% manually.

[Solus Corvus]

You may have hit the limit of your card. You can try benching it with 100% fan to see if it makes any difference. But it also comes down to luck of the draw - some cards are higher clockers then others.

I have the thermals under control with a full cover water block and dedicated loop on my 4870x2. But no matter how high I push the voltage (via rivatuner) it won't go past 840 on the cores. If I try, I get the same lockup that you describe.

[AgentGOD]

You may have hit the limit of your card. You can try benching it with 100% fan to see if it makes any difference. But it also comes down to luck of the draw - some cards are higher clockers then others.

I have the thermals under control with a full cover water block and dedicated loop on my 4870x2. But no matter how high I push the voltage (via rivatuner) it won't go past 840 on the cores. If I try, I get the same lockup that you describe.

Which lockup do you mean?

[AgentGOD]

Seems that 55% fan + 800 MHz core OC in CCC = stable! I was able to play Crysis for a while, and no freeze, no artifacts.

[AgentGOD]

Here's Call of Duty: World at War, played Nazi Zombie for 20 rounds (co-op)! In screen shot, I'm showing maximum GPU temperature, core clock, and memory clock.

Screen shot:
http://img89.imageshack.us/img89/2282/hehload.jpg

[Santi U]

Fan is at AUTO. And I'm NOT going to run the card at 100% fan 24/7, way too loud.

My recommendation was not intended for 24/7

"I would do this with 100% fan to try and eliminate overheating as the cause. Once you get to where you want to be with oc/voltage then lower your fan to see if it's stable or the decibel range is ok for you for that oc."

Working with the less variables the better.


You may have hit the limit of your card. You can try benching it with 100% fan to see if it makes any difference. But it also comes down to luck of the draw - some cards are higher clockers then others.

Exactly :)

Maybe your psu can't handle higher loaded voltage? It seemed like my 750w was maxing out @1.3v with a similarly clocked/volted quad.

[AgentGOD]

Maybe your psu can't handle higher loaded voltage? It seemed like my 750w was maxing out @1.3v with a similarly clocked/volted quad.
That's possible.

[Solus Corvus]

Which lockup do you mean?

The system lockup and blank screen you describe in the first post. My card never goes much over mid 40c while running furmark. But once I push it past the clock limit (~840 core) I get an instant lockup in furmark no matter how high I push the voltage.

As Santi U and I said, you could bench it with the fan at 100% to eliminate heat as a variable. You could also run a fan pointing at the pins on the top of the HS at the same time - the VRMs get quite hot. Obviously this wouldn't be suitable for long term use but it can help you determine how the card behaves at the limit when you eliminate/reduce heat as a factor.

To determine if the power supply is a factor you could run a second power supply to the card only, if you have another power supply lying around or one you could borrow.

[AgentGOD]

The system lockup and blank screen you describe in the first post. My card never goes much over mid 40c while running furmark. But once I push it past the clock limit (~840 core) I get an instant lockup in furmark no matter how high I push the voltage.

As Santi U and I said, you could bench it with the fan at 100% to eliminate heat as a variable. You could also run a fan pointing at the pins on the top of the HS at the same time - the VRMs get quite hot. Obviously this wouldn't be suitable for long term use but it can help you determine how the card behaves at the limit when you eliminate/reduce heat as a factor.

To determine if the power supply is a factor you could run a second power supply to the card only, if you have another power supply lying around or one you could borrow.
I just did some more experimenting. Ran card at 830 MHz core, and voltage, I pushed it to 1.3625v via RivaTuner. I also pushed the fans to 55%. So I decided to try FurMark again, 1920x1200, 8x AA, Stability Testing. And, initial load temp was only ~65C in but then same symptom happened, 4 seconds later, blank screen, GPU fan starts spinning loudly at probably 100%, and system is frozen.

I read somewhere that some 4870/X2's have some sort of "OVP" that kicks in when you set voltage over 1.3v, maybe this is why. In each voltage scenario, I used over 1.3v (initially 1.3125v, now 1.3625v).

Also, at the top (first post), I used 800 MHz, voltage at 1.3125v, same symptoms. But, stock voltage, 800 MHz, with fan at 55%, I seem to run Crysis very stable, whereas before, it would just freeze after 10 min of gameplay. FurMark worked too with 800 MHz core @ stock voltage.

[Solus Corvus]

The rivatuner plugin that lets us control voltages also measures OVP and OCP status. In all my testing it has never shown as active. Though it is certainly possible that OVP is kicking in and the plugin just doesn't properly sense it. We can't even be entirely sure that the voltages being reported are accurate.

The next step for me is to get a DMM and soldering iron. First I want to simply verify that the software control is actually changing the voltages in the expected way. Then I want to try a hard voltmod to see if there is any change in the overclock behavior when adjusting voltages that way.

[Santi U]

I read somewhere that some 4870/X2's have some sort of "OVP" that kicks in when you set voltage over 1.3v, maybe this is why. In each voltage scenario, I used over 1.3v (initially 1.3125v, now 1.3625v

I heard something similar about 1.36v~.

If I have the time today i'll try some psu testing. I don't have a killa-watt or DMM but disabling 2cores and running 1.09v~ stock should be easy enough to see if it's being limited by that.

[i found nemo]

idk what kind of fan you have, but i put an old dell fan infront of my gpu and it's not loud, even with my other fan on 100 %. don't be scared of noise, crank that sucker up!

[BeepBeep2]

Running FurMark for only 2 mins isn't enough. My 4850 is 76c at 2 minutes and 82c at 8 minutes and about 86c at 20 minutes. It stays around (+1-1)88c after that. My core is at 710.

[BeepBeep2]

Hmm, apparently he got to 890 MHz core on 55% fan:
http://forums.guru3d.com/showpost.php?p=2840681&postcount=6

...in no way is that stable, you need high-end water/Dry Ice(!) to get that high (on both cores). He may be lying. Oh, and Visiontek cards aren't good overclockers too. 850 may be possible at high volt's/water but no.

[CedricFP]

I am not sure if the X2's like voltages for extra clocks or not.

At stock, I can bench at 842 on the core - I can even game Crysis Warhead for a couple of hours before it locks up. However, once I move to 847 or above, I always lock up in GT2 of 3DMark06. However, I can still play Crysis for an hour or so.

So I used the rivatuner trick to give my card 1.37v (OVP limits it to 1.371) and I am still only 842 fully benchable. Again, any higher, such as 847 or 852 and I freeze 5 seconds into GT2.

You would think going from 1.26 --> 1.37 would give me at least 5 extra mhz.

On water btw - load temps never exceed 48 on both cores.

[AgentGOD]

I just tried going to 1.28v, and the same reset described above happened. Blank screen, fan very loud, etc. ONLY after running FurMark for a few seconds though.

[AgentGOD]

Hmm.... looks like I might have found the cause:
http://img9.imageshack.us/img9/7025/hmmvqp.jpg
Look at GPU VRM temps. This is at 792 MHz core, stock voltage, 55% fan speed. After running 73 seconds of FurMark at 1920x1200, no AA, full screen. The temps were still climbing too.. GPU temp was hitting 92*C at the time before I exited FurMark.

This is also with all of the Antec 900 fans running at maximum speed. Ambient temp @ 80*F (room temp).

I guess more volts would have pushed the VRMs over the limit, and caused the card's self-protection to kick in during FurMark.

[AgentGOD]

Should I try sticking a 120mm 60cfm fan to the side panel as intake for the Antec 900?

Since it would blow right over the GPU, it might lower GPU/VRM temps?