*** Official Radeon HD5800 Series Overclocking Thread ***

[zalbard]

I doubt a new BIOS would help.

[Hawkeye4077]

Nice charts.I am sitting on 875/1150 on 1.1v, but what the chart needs is a Brand column. See which cards get what. Mine is an XFX 5850 non XXX

[AndrewZorn]

okay, that is a lot of pages to read through...
- how do you guys get CCC to increase the maximums of the sliders in overdrive? mine is maxed at 775/1125... i saw in the original post that if you dont want to use atitool (which i do not) you need to flash... so is it just MSI owners that get the 1500 max?
- does increasing the GPU voltage increase the VRM temps? my liquid-cooled gpu runs very cool, but the vrms already get hotter than i would like them to.

[Intratech]

Sweetness!

http://www.techpowerup.com/gpuz/28ar9/





All on air baby.

[AndrewZorn]

VDDCs 38*C under load on air? how is this even possible?

[snoro]

fan at 100% speed really help alot for vrm temps

[Intratech]

VDDCs 38*C under load on air? how is this even possible?

Look at fan speed... that and open window with cool winter air.

Anyway, I used Furmark to test that setting further and it locked up in about a minute. I checked the VDDC current and it shot up to 68Amps which was like double what I've ever observed from any HD 5870 during operation. Either Furmark is one hella crazy program or GPU-Z ain't reading that right.

[lowdog]

Look at fan speed... that and open window with cool winter air.

Anyway, I used Furmark to test that setting further and it locked up in about a minute. I checked the VDDC current and it shot up to 68Amps which was like double what I've ever observed from any HD 5870 during operation. Either Furmark is one hella crazy program or GPU-Z ain't reading that right.

It's Furmark, it just canes the crap out of the vid card. Personally I never use it for more than 1 minute, just use it to see the volts and VDDC Phase temps briefly and thats it.

[Frogeye]

Look at fan speed... that and open window with cool winter air.

Anyway, I used Furmark to test that setting further and it locked up in about a minute. I checked the VDDC current and it shot up to 68Amps which was like double what I've ever observed from any HD 5870 during operation. Either Furmark is one hella crazy program or GPU-Z ain't reading that right.

Wow! great clocks. you were able to get that using only 1.16 volts? I'm using 1.18 for stable gaming 900/1140. I haven't really pushed to see if it will do 1k stable.

[Intratech]

Thanks.

Yes, look at the screenshot for the GPU load and the time it was running.. it was at default voltage. It was stable for everything else apart from Furmark which caused a lockup in about a minute. Someone told me not to use Furmark to test for stability so I'm going to loop 3dmark05 or 06 next time.

[dejanh]

Has anyone noticed that more voltage actually had a significantly detrimental effect on their overclock of HD 5850? For example...

1.125V/900/1250 - stable
1.137V/900/1250 - slight artifacting
1.149V/900/1250 - significant artifacting but then...

1.187V/975/1250 - no artifacting, stable

It's like the voltage has to scale with clock speed, i.e., pumping even slightly more volts at certain clock actually has a negative effect and makes the card more unstable. This is not a one off thing, I can pretty much produce it at will.

[Intratech]

I might test out that theory with the Asus HD 5870 actually. Something similar was puzzling me about it, at times it was stable at a certain clock but at other times it would just artifact like mad so maybe the vGPU I was using might be causing it. I'll check that out over the weekend.

[dejanh]

I might test out that theory with the Asus HD 5870 actually. Something similar was puzzling me about it, at times it was stable at a certain clock but at other times it would just artifact like mad so maybe the vGPU I was using might be causing it. I'll check that out over the weekend.

Yeah, it would definitely be interesting to know. I know that 5870 vs 5850 is a bit of an apples to oranges, maybe more like granny smith to red delicious

Anyway, I'm pretty sure this is not a fluke but a lot of people don't look at this kind of stuff. It is puzzling though as it definitely behaves as though voltage scales with core clock and more volts is certainly not always better despite the cooling. It's like the cards have certain voltage "sweet spots" for a given clock. I'm still waiting on my second HD 5850 to test this theory out on as well.

[GregUMR]

Yeah, it would definitely be interesting to know. I know that 5870 vs 5850 is a bit of an apples to oranges, maybe more like granny smith to red delicious

Anyway, I'm pretty sure this is not a fluke but a lot of people don't look at this kind of stuff. It is puzzling though as it definitely behaves as though voltage scales with core clock and more volts is certainly not always better despite the cooling. It's like the cards have certain voltage "sweet spots" for a given clock. I'm still waiting on my second HD 5850 to test this theory out on as well.

That'd be perfect for just about every overclocker with a 5850 that isn't satisfied with his clock speeds. However I am pretty skeptical of it. I guess there could be some interference except at certain voltages, but that's far stretched I think. I'm going to test this idea with my visiontek 5850.

I've been hoping for 1050/1300 stable but I can only manage 960/1250, not a big difference but still.

[dejanh]

That'd be perfect for just about every overclocker with a 5850 that isn't satisfied with his clock speeds. However I am pretty skeptical of it. I guess there could be some interference except at certain voltages, but that's far stretched I think. I'm going to test this idea with my visiontek 5850.

I've been hoping for 1050/1300 stable but I can only manage 960/1250, not a big difference but still.

Not sure how this would be perfect for people overclocking...

Anyway, I did some more tests and definitely my card behaves this way. I tested 1000MHz core at 1.225V and then at 1.237V and while there were no artifacts at 1.225V at 1.237V the artifacts started within the first 2 minutes of test. Then I increased the core frequency to 1005MHz and I managed to go 10 minutes without artifacts. Then I increased to 1010MHz and I went 20 minutes without artifacts.

So very, very strange. It could be some unbelievably weird coincidence that I keep on hitting every time, but I doubt it. It's not the first time I was able to reproduce this either.

[ReverendMaynard]

Not sure how this would be perfect for people overclocking...

Anyway, I did some more tests and definitely my card behaves this way. I tested 1000MHz core at 1.225V and then at 1.237V and while there were no artifacts at 1.225V at 1.237V the artifacts started within the first 2 minutes of test. Then I increased the core frequency to 1005MHz and I managed to go 10 minutes without artifacts. Then I increased to 1010MHz and I went 20 minutes without artifacts.

So very, very strange. It could be some unbelievably weird coincidence that I keep on hitting every time, but I doubt it. It's not the first time I was able to reproduce this either.

you running furmark for your tests?

[dejanh]

you running furmark for your tests?

Running Furmark 1.8.0, ATITool 0.27 beta 4, and Kombustor for general stability tests, then Crysis Benchmark, 3DM06 and Vantage for gaming stability. Basically the whole suite. I observed the same type of behavior with each tool.

Edit: I don't necessarily see this as a problem, just as an interesting quirk. I just did ~1h of ATITool with no artifacts at 1.225V and 1000/1250, bump to the voltage and artifacts start right away, bump down and artifact disappear. In Kombustor this instability will show as flickering of the MSI fur generally, sometimes actual blocks will appear.

[emertX]

My new BABY!

[ChaosAD]

Watercooled emert, right?

[emertX]

Yes!
EK Plexi Nickel!
It's a dream though...but more isn't possible!
I don't know why? I gave him 1,4V and 1150 on core sucks! 1410 on mem either!

[ChaosAD]

Yea, it so obvious that you need more gpu power, lol!

[malkiewicz]

Is this a good score with 2 cards... I can't really OC much without it crashing...

[emertX]

My 5890!

[AndrewZorn]

i have some noob questions:

- can i use any [reference] brand bios on my Visiontek 5850? i have the Asus bios on there now, but what if I use MSI, in order to be able to use Afterburner (and will I be able to?)? i know the safe answer is just to modify the Asus bios i already have on it... since it already works and is already unlocked (unlike the Visiontek)... but if the MSI is the same as all the others but has Afterburner, mind as well...
- i do not have any 'selectable voltages' in RBE. do i just type in the voltage i want for the main/high/3D clock section ("Clock info 00") and leave everything else the way it is if i do not want to mess with 2D and such clocks?
- if i someday figure out my exact desired overclock, can i just put it in the Clock info 00 spot directly... and never have to even use catalyst control center? since it is in the BIOS, i wouldn't need to install CCC to see working powerplay and overclock?

[cirthix]

i have some noob questions:

- can i use any [reference] brand bios on my Visiontek 5850? i have the Asus bios on there now, but what if I use MSI, in order to be able to use Afterburner (and will I be able to?)? i know the safe answer is just to modify the Asus bios i already have on it... since it already works and is already unlocked (unlike the Visiontek)... but if the MSI is the same as all the others but has Afterburner, mind as well...
- i do not have any 'selectable voltages' in RBE. do i just type in the voltage i want for the main/high/3D clock section ("Clock info 00") and leave everything else the way it is if i do not want to mess with 2D and such clocks?
- if i someday figure out my exact desired overclock, can i just put it in the Clock info 00 spot directly... and never have to even use catalyst control center? since it is in the BIOS, i wouldn't need to install CCC to see working powerplay and overclock?

1) yes, but afterburner works on any 5850
2) input vgpu into gpu register box, DO NOT change in the clock boxes or in the overclock wizard dialog box. You'll break powerplay if you do.
3) use the clock wizard rather than editing the boxes directly.