7950 overclocking guide/tips?

[Techhog]

So, I want to try OCing my 7950, hopefully getting to 1GHz without having to add voltage. It hits 30-35C idle and 60-65C under load right now. (Actually, it's never gone above 62C to my knowledge, but to be safe...) What's the best way to go about this? What should I use to stress test it and for how long?

 

[nsafreak]

I would recommend reading this thread: http://forums.anandtech.com/showthread.php?t=2259333 , lots of good tips and advice on how to overclock a 7950 series card.

 

[sublime420ec]

Open catalyst control center....go to performance tab then click amd overdrive.....the for the power control setting put it to 20% then slide your core to 1ghz....shou d be able to hit at least 1100 core and 1300 on mem but all u want is 1ghz so just use ccc to o/c

 

[Techhog]

Thanks guys! Now to see if I can push this...

 

[Sohaltang]

I used CCC too. Somereason it works better than the other "hardcore" tweaking utilities. Maybe Im just an idiot. Check the benchmark threads and you will see my setup runs pretty well with the OEM CCC max overclocks

 

[Daxwax]

Couldn't adjust voltages on mine, which makes it hard to fine-tune.
Ended up flashing a different firmware which lowered voltage a bit, and using CCC overdrive as the others mentioned.

 

[Techhog]

Looks like I can't really push this. At 1050/1300, it can't open benchmarks. :/ Trying to figure out why now... Probably a reallybbad chip I guess....

EDIT: yep, can't do 3D on 1050 core at stock voltage. I'm gonna stick to 1000/1300 since that seems to work.

 

[Techhog]

Okay, this is scary. My Valley benchmark scores are extremely low, and the GPU load can only reach 99%. What is going on? Did I break something?

 

[wand3r3r]

The load at 99% is nothing to worry about.

Each time that it crashes due to not handling the OC, you have to go move the clocks in e.g. afterburner and re save them since it reset them after a driver crash due to the OC.

Try MSI afterburner and you make sure you put the power % to the max (+20%). Check what voltage your card is at. Stock voltage is up to 1.25v so feel free to raise it a bit once you find the limits, I wouldn't go over 1.25 or so.

 

[Techhog]

The load at 99% is nothing to worry about.

Each time that it crashes due to not handling the OC, you have to go move the clocks in e.g. afterburner and re save them since it reset them after a driver crash due to the OC.

Try MSI afterburner and you make sure you put the power % to the max (+20%). Check what voltage your card is at. Stock voltage is up to 1.25v so feel free to raise it a bit once you find the limits, I wouldn't go over 1.25 or so.

Actually, my computer just needed a restart to ix the issue lol. Now to see if I can get it a little further...

EDIT: Yeah, 1000 is the limit at stock voltage. Oh well. I'm not gonna bother increasing voltage because I don't feel that it's worth the risk if I got such a bad chip. Really a shame, especially since I have yet to see temps go above 63C.

 

[wand3r3r]

Actually, my computer just needed a restart to ix the issue lol. Now to see if I can get it a little further...

EDIT: Yeah, 1000 is the limit at stock voltage. Oh well. I'm not gonna bother increasing voltage because I don't feel that it's worth the risk if I got such a bad chip. Really a shame, especially since I have yet to see temps go above 63C.

What is your stock voltage? AMD's stock voltage is 1.25v on boost editions which means it's perfectly safe and not expected to be enough to reduce the lifetime or do anything significantly damaging. Personally I don't think there is any risk in going up to the AMD stock voltage. Manufacturers have dropped them as low as 1.0x Volts since they haven't needed 1.25 to run at 800/925 clocks.

 

[Techhog]

What is your stock voltage? AMD's stock voltage is 1.25v on boost editions which means it's perfectly safe and not expected to be enough to reduce the lifetime or do anything significantly damaging. Personally I don't think there is any risk in going up to the AMD stock voltage. Manufacturers have dropped them as low as 1.0x Volts since they haven't needed 1.25 to run at 800/925 clocks.

How do I find my voltage?

EDIT: Wait, I think I found it... Okay, let me try that...

 

[wand3r3r]

How do I find my voltage?

EDIT: Wait, I think I found it... Okay, let me try that...

Try MSI Afterburner.

It should show you the default voltage.

To enable voltage monitoring/control you need to go to Settings/General/Unlock voltage monitoring (and control).

When overclocking you should probably put the power usage to +20% too.

 

[Techhog]

Try MSI Afterburner.

It should show you the default voltage.

To enable voltage monitoring/control you need to go to Settings/General/Unlock voltage monitoring (and control).

When overclocking you should probably put the power usage to +20% too.

Well, I did it. This is how it should look, correct?

It was stable in Valley at those settings. Fan speed topped out at 58%. VRMs were 72/75C max.

 

[schmuckley]

I'd use Trixx..It might take some experimenting to find the right one..
Sometimes you have to use one for voltage and another for clocking .
Do the core 1st..find max for that.
Do the RAM next..find max for that. (hint) when the frames drop off when you go higher..you back down by about 25Mhz)
Then put them together.
I game @ 1025/1350 with a Sapphire.
I think your card might need VRM sinks..not sure though..

 

[wand3r3r]

Well, I did it. This is how it should look, correct?

It was stable in Valley at those settings. Fan speed topped out at 58%. VRMs were 72/75C max.

Yep, just like that.

What was the stock core voltage? It's at 1.25v now, did you set that yourself? If you want you can save that to a profile (there are 5 profiles) and when you press reset it will set it to the cards defaults.

About trixx, I don't think there is any reason to use it, they serve the same purpose.

 

[Techhog]

Yep, just like that.

What was the stock core voltage? It's at 1.25v now, did you set that yourself? If you want you can save that to a profile (there are 5 profiles) and when you press reset it will set it to the cards defaults.

About trixx, I don't think there is any reason to use it, they serve the same purpose.

I set it myself... and I don't think it's stable. I tried OCCT and I lost the video card signal pretty quickly. The stock seems to be 1.031V.

I got a really bad chip. I think I should just leave it at stock.

 

[Techhog]

Just uninstalled MSI Afterburner. I can't afford to risk the card like this, so I'm done. The graphics card shuts down in OCCT when at higher than stock voltage.

 

[MrK6]

Just uninstalled MSI Afterburner. I can't afford to risk the card like this, so I'm done. The graphics card shuts down in OCCT when at higher than stock voltage.

Why are you running OCCT? It's a crap program. All it does is overload your power circuitry and potentially hurt your card. See if you can in any way access your VRM temps (try GPU-Z or ASUS Tweak). What model (manufacturer and specific model) card do you have and what's the cooling solution like? The reason I'm asking is that OCCT is more than likely overloading your VRM's and negating any boost you're getting from increasing voltages. The goal is to figure out how robust your VRM cooling solution is so that you can figure out how much head room you have.

Also, use games to test video cards. Benchmarks generally don't test the full chip and "power virus" programs like OCCT or Furmark just stress the heatsink and do nothing to test stability.

 

[Techhog]

Why are you running OCCT? It's a crap program. All it does is overload your power circuitry and potentially hurt your card. See if you can in any way access your VRM temps (try GPU-Z or ASUS Tweak). What model (manufacturer and specific model) card do you have and what's the cooling solution like? The reason I'm asking is that OCCT is more than likely overloading your VRM's and negating any boost you're getting from increasing voltages. The goal is to figure out how robust your VRM cooling solution is so that you can figure out how much head room you have.

Also, use games to test video cards. Benchmarks generally don't test the full chip and "power virus" programs like OCCT or Furmark just stress the heatsink and do nothing to test stability.

*sigh* Okay, I'll give it one more go. Make is in sig, model number is AX7950 3GBD5-2DHPP.

Also, GPU-Z isn't showing graphics memory usage anymore... I'm really scared that the memory chips are dead...

EDIT: Okay, it's just GPU-Z being weird.

 

[schmuckley]

http://i.imgur.com/pELkBqS.jpg?1?8008

Bare VRMs is your problem.Need a sink/some RAM sinks or something.The best way is kindy risky and nullifies any chance of RMA.
I'd just get the short Enzotechs and stick um on and hope.
The other way is to get thermal adhesive,Take the tape off of the sinks and lap them..then apply a drop to each VRM chip with a toothpick..then smoosh the sinks on so they hold.
Hrmm..possibly an old motherboard VRM sink might would work well..with some good thermal tape.

 

[MrK6]

http://i.imgur.com/pELkBqS.jpg?1?8008

Bare VRMs is your problem.Need a sink/some RAM sinks or something.The best way is kindy risky and nullifies any chance of RMA.
I'd just get the short Enzotechs and stick um on and hope.
The other way is to get thermal adhesive,Take the tape off of the sinks and lap them..then apply a drop to each VRM chip with a toothpick..then smoosh the sinks on so they hold.
Hrmm..possibly an old motherboard VRM sink might would work well..with some good thermal tape.

Yep, there's the problem. I wouldn't even run OCCT at stock on this card with bare heatsinks; I actually think it's criminal that they're allowed to sell the card like that.

Ranting aside, you need to find a way to cool the VRM's, even if you don't want to overclock it (although heck, you might as well since you're going through all of this). I agree that the Enzotech heatsinks are probably the best for this application, buy a few and see what fits. If you can't get the thermal adhesive to hold them on, a tiny drop of super glue in the corner of the heatsink usually does the trick (and I mean very tiny, otherwise you'll have too much holding force and won't be able to remove the heatsinks ever).

Edit: Enzotech heatsinks discussed - http://www.jab-tech.com/motherboard-chipset-cooling/enzontech-mosfet-heatsink-mos-c1/ or http://www.jab-tech.com/motherboard-chipset-cooling/enzotechnology-mosfet-heatsink-mos-c10/

 

[Techhog]

http://i.imgur.com/pELkBqS.jpg?1?8008

Bare VRMs is your problem.Need a sink/some RAM sinks or something.The best way is kindy risky and nullifies any chance of RMA.
I'd just get the short Enzotechs and stick um on and hope.
The other way is to get thermal adhesive,Take the tape off of the sinks and lap them..then apply a drop to each VRM chip with a toothpick..then smoosh the sinks on so they hold.
Hrmm..possibly an old motherboard VRM sink might would work well..with some good thermal tape.

Yep, there's the problem. I wouldn't even run OCCT at stock on this card with bare heatsinks; I actually think it's criminal that they're allowed to sell the card like that.

Ranting aside, you need to find a way to cool the VRM's, even if you don't want to overclock it (although heck, you might as well since you're going through all of this). I agree that the Enzotech heatsinks are probably the best for this application, buy a few and see what fits. If you can't get the thermal adhesive to hold them on, a tiny drop of super glue in the corner of the heatsink usually does the trick (and I mean very tiny, otherwise you'll have too much holding force and won't be able to remove the heatsinks ever).

Edit: Enzotech heatsinks discussed - http://www.jab-tech.com/motherboard-chipset-cooling/enzontech-mosfet-heatsink-mos-c1/ or http://www.jab-tech.com/motherboard-chipset-cooling/enzotechnology-mosfet-heatsink-mos-c10/

Are you really telling me to replace the heatsink? -_-

I'm not hard-modding the card. I'll just keep everything at stock if it's really that bad, and next time I buy a graphics card I'll make sure to buy the the best cooled card I can afford, rather than the best GPU. Okay?

 

[MrK6]

Are you really telling me to replace the heatsink? -_-

I'm not hard-modding the card. I'll just keep everything at stock if it's really that bad, and next time I buy a graphics card I'll make sure to buy the the best cooled card I can afford, rather than the best GPU. Okay?

It's more we're suggesting adding heatsinks than taking any away; I'm suggesting what I believe the problem here is and how you can remedy it. It's your card and you should do whatever you're most comfortable with doing.

 

[Techhog]

It's more we're suggesting adding heatsinks than taking any away; I'm suggesting what I believe the problem here is and how you can remedy it. It's your card and you should do whatever you're most comfortable with doing.

I don't want to disassemble my card, though. :/