Radeon HD 7950 Owner's Thread

[Zargon]

I left the memory stock since temps were fine


do I need to raise the mem clocks?

 

[iCyborg]

Just got mine, and reading this thread again, I see others have seen the same thing:
At stock everything my card idles at 54C with 500/1250 clocks. If I disable one of the monitors, it goes to 300/150 and idle temp 36C, this is the same as my CPU idle temp. My case (Antec 900), although decently ventilated, is rather cramped, but still, this 54C is kinda worrying, I was hoping it'd be in the forties at worst. Bumping up the fan from default ~40% to 50% gets me to 49C, at around 55% the fan is starting to be audible.
Furmark after ~30mins maxed at 80C, though fan was running wild there at almost 80%, not sure I want to OC it any further.
Also, the included miniDP-DP adapter is not working, I'll try to see if they'd be willing to replace it...

On the bright side, it works fine otherwise and S3 SLEEP WORKS with 12.6! (this is important to me, and people reported issues earlier).

 

[Binky]

54C is fine. What makes you think it's not ok?

 

[joe_H]

Just got mine, and reading this thread again, I see others have seen the same thing:
At stock everything my card idles at 54C with 500/1250 clocks. If I disable one of the monitors, it goes to 300/150 and idle temp 36C, this is the same as my CPU idle temp. My case (Antec 900), although decently ventilated, is rather cramped, but still, this 54C is kinda worrying, I was hoping it'd be in the forties at worst. Bumping up the fan from default ~40% to 50% gets me to 49C, at around 55% the fan is starting to be audible.
Furmark after ~30mins maxed at 80C, though fan was running wild there at almost 80%, not sure I want to OC it any further.
Also, the included miniDP-DP adapter is not working, I'll try to see if they'd be willing to replace it...

On the bright side, it works fine otherwise and S3 SLEEP WORKS with 12.6! (this is important to me, and people reported issues earlier).

I had similar temps running dual monitors off this card. Decided to get a cheap HD5450 to run the secondary display. The card idles around 38C now instead of 54C. I also changed out the tim on the card, and dropped my load temps by 5-6C.

 

[Fire&Blood]

Just got mine, and reading this thread again, I see others have seen the same thing:
At stock everything my card idles at 54C with 500/1250 clocks. If I disable one of the monitors, it goes to 300/150 and idle temp 36C, this is the same as my CPU idle temp. My case (Antec 900), although decently ventilated, is rather cramped, but still, this 54C is kinda worrying, I was hoping it'd be in the forties at worst. Bumping up the fan from default ~40% to 50% gets me to 49C, at around 55% the fan is starting to be audible.
Furmark after ~30mins maxed at 80C, though fan was running wild there at almost 80%, not sure I want to OC it any further.
Also, the included miniDP-DP adapter is not working, I'll try to see if they'd be willing to replace it...

On the bright side, it works fine otherwise and S3 SLEEP WORKS with 12.6! (this is important to me, and people reported issues earlier).

You can get some good TIM in there and throw in a fan dedicated to cool the GPU and you will regain a good chunk if not all of that increased idle temp.

On 1100/1575 @1075v: I replaced my exhaust 80mm fan for a 120mm one, added 2 more 120mm to cool the GPU. Had to remove the 80mm because it was noisy but while it was the third fan dedicated to the GPU cooling the card's front, I had idle temps as low as 30C-33C depending on ambient.

 

[iCyborg]

54C is fine. What makes you think it's not ok?

It's almost 20C higher temp just by connecting an additional 1680x1050 monitor which never does any 3D, and this extra heat being dumped into the case does affect the longevity of all components, GPU included.

@Rick
The case itself has 4 fans including a 200mm one, PSU also has its own fan, and GPU itself has 2 fans. I don't think one more will help much. Anyway, my case and its location certainly contribute to the problem, but idling temp (36C) is not bad with single monitor, load temps are also not too bad, but idle with 2 monitors is higher than I expected. Seems these higher temps are common for 8+6 pin versions, even though ref voltage seems low. I can live with it, 24/7 bitcoin miners have it worse after all...

 

[zaydq]

http://www.3dmark.com/3dm11/4141143

New 3dmark score. I got my gpu score to 10 000 finally! But I somehow lost a few hundred points on my physics score. Not sure why?

 

[chimaxi83]

Using same CPU/RAM speeds/timings?

 

[RussianSensation]

zaydq, 10k on 7950! Nice

I am amazed so many of you guys are getting 80%+ Asics. MSI must have really binned these chips.

 

[Fire&Blood]

http://www.3dmark.com/3dm11/4141143

New 3dmark score. I got my gpu score to 10 000 finally! But I somehow lost a few hundred points on my physics score. Not sure why?

Right on! That's $450 GPU territory right there. I wonder how many OCed reference 7970 cards you beat because of better cooling hence clocking high enough to get past them.

Only guessing here, maybe you aren't completely stable, a voltage bump may be the answer.

 

[Silverforce11]

zaydq, 10k on 7950! Nice

I am amazed so many of you guys are getting 80%+ Asics. MSI must have really binned these chips.

So is the ASIC reading in GPUz proven to be accurate reflection of overclocking potential?

 

[RussianSensation]

I get that impression for 7900 series. The higher the #, the less voltage they need on air to reach higher clocks. The lower the Asic #, the better the card is for watercooling and exotic cooling. Mine is just an avg overclocker with Asic of low 70s.

 

[hyrule4927]

http://www.3dmark.com/3dm11/4141143

New 3dmark score. I got my gpu score to 10 000 finally! But I somehow lost a few hundred points on my physics score. Not sure why?

Nice score there! Was that at stock volts on the memory? Just curious because my memory seems to top out pretty low at 1375 before I see a performance loss from errors. I'm not sure what kind of voltage increases the memory can safely handle, so I haven't tried bumping that up at all yet.

I get that impression for 7900 series. The higher the #, the less voltage they need on air to reach higher clocks. The lower the Asic #, the better the card is for watercooling and exotic cooling. Mine is just an avg overclocker with Asic of low 70s.

My ASIC is around 86.3%, for reference.

After a little more stability testing I've gotten 1200/1375 benchmark stable (consistent scores and no issues 3DMark 06, Vantage, and 11) with around 1.15 volts, but I had to increase it to 1.175 to stop Crysis from crashing after 30 minutes or more of gameplay. I still got one crash in Saints Row 3 at that speed and just dropped the clocks a bit. I'm not entirely sure if the crash was OC related, but until I get more testing done, the best completely stable settings I've managed are 1175/1375 at 1.175 volts.

 

[RussianSensation]

Ya, that's good. 1.175V is the stock voltage for a 925mhz 7970. My 7970 can only do 1150mhz at 1.174V. So think about how much binned your 7950 is since normally 7950 is a chip that couldn't pass to be a 7970. The fact that your card can do higher clocks than my 7970 shows that MSI is literally picking the top 7950s for this card.

 

[zaydq]

I just upped vcore I didn't touch memory voltages. My 1200/1800 is stable at. 1168mv. Even more so than my 1200/1600. It runs at full load 74c with a good fan profile. I should try and up my voltage on my cpu. Because I have had a 4ghz run with 7600 points and even that seems too low.

 

[iCyborg]

The lower the Asic #

My ASIC is around 86.3%

ASIC = Application Specific Integrated Circuit. In this context, it's the GPU (i.e. card minus PCB, cooling etc). It's not ASIC number, it's asic quality rating/number. Not that I know how do they come up with it...

That said, my reading is 88.2%.

I tried to organize cables a bit better and moved the case from under the desk to beside the desk, hoping to get a couple C lower - I was surprised to see idle temp dropping 8C (from 56C to 48C) and load temp in furmark from 80C to 70C with fan at 55% instead of ~80%. Boy, what a difference this simple change did.

How do I read stock voltage? Afterburner for some reason doesn't let me read or change voltage, is there anything else besides 'unlock' in the settings? The highest value I've seen in GPU-Z after load testing is 1.006V. Why the heck do these 7950s with boost need 1.25V for 925MHz? Mine runs at ~0.94-0.95V at almost all times.

Putting powertune to 20% and clocks to 1100/1500 works fine, though load temps are getting into high 70s again. I feel like I could easily run this 24/7, but being a casual gamer, I'll be keeping it at stock, maybe change it on a per-game basis.

 

[thujone]

so i've realized all the fans that came installed in my thermaltake case are crap. realized yesterday that both the fan at the front and the top exhaust (both 200mm) put out so little air that i couldn't tell which direction the fan was blowing. replaced the front one with 2x120mm... ordered a new 200mm fan for the top exhaust. i also found a couple old 60mm fans sitting around for the rear VGA exhaust. hooked them up to the built in fan controller and they work great... except one has WAY too much whine. in any event... all this fan reconfiguring dropped my temps a good 10C. and this is without the new 200mm top exhaust.


still waiting on thermaltake to come through with an RMA for the wireless fan on the side panel... but i'm not sure it's going to work much better than any of the other ones. might end up having to replace it with an aftermarket too.

i really love this case otherwise... and the fans on my water 2.0 radiator that came from thermaltake seem to work good... but yeah... beware thermaltakes stock case fans.



also... regarding asic... my asic quality is 89.2 in gpuz. but i can't get past 1150 no matter how much voltage i give the card.

 

[Binky]

Is anybody testing for lowest possible voltages for bitcoin mining? I've managed to get 1.0v at 1100/625 and it's stable for mining at 99% gpu usage, but using the desktop while mining might lead to crashes. It seems to run fine at 1.0v if I leave it alone. I'm stable at about 1.01v though (mining, not gaming) at 1100/625 with any desktop usage while mining. That's pretty impressive.

Voltage needed seems to be very low for me at or below 1100, but it increases rapidly after 1100. I have no need for more than 1100 so I'm good.

 

[thilanliyan]

Is anybody testing for lowest possible voltages for bitcoin mining? I've managed to get 1.0v at 1100/625 and it's stable for mining at 99% gpu usage, but using the desktop while mining might lead to crashes. It seems to run fine at 1.0v if I leave it alone. I'm stable at about 1.01v though (mining, not gaming) at 1100/625 with any desktop usage while mining. That's pretty impressive.

Voltage needed seems to be very low for me at or below 1100, but it increases rapidly after 1100. I have no need for more than 1100 so I'm good.

I have mine at 1100 at 1.037v (stock). For me that is low enough. I have actually not tried to push higher than that at all, but will in time.

My 6950 I have undervolted to 1.0v at 800MHz. Stock was 1.15v at 850MHz but I found it ran too hot and loud at those speeds so backed off.

 

[Rvenger]

I have mine at 1100 at 1.037v (stock). For me that is low enough. I have actually not tried to push higher than that at all, but will in time.

My 6950 I have undervolted to 1.0v at 800MHz. Stock was 1.15v at 850MHz but I found it ran too hot and loud at those speeds so backed off.

Wow you are lucky. My ASIC score is an 87% and I need to pump 1.12v to get to 1100 core.


EDIT: That must be for mining and not gaming.

 

[RussianSensation]

How do I read stock voltage? Afterburner for some reason doesn't let me read or change voltage, is there anything else besides 'unlock' in the settings? The highest value I've seen in GPU-Z after load testing is 1.006V. Why the heck do these 7950s with boost need 1.25V for 925MHz? Mine runs at ~0.94-0.95V at almost all times.

Exactly! A lot of 7900 series owners have noted that some of those chips undervolt from the factory. If you use GPU-Z VDCC and set that to Maximum or Average, chances are it'll be below your MSI Afterburner if you have a good chip. I use HWInfo64 Sensor together with GPU-Z Sensors to compare and they report similar #s. My chip is just average, when I set 1.200V in MSI Afterburner, it shows an average of 1.094V but sometimes spikes to 1.174V. This is why I think some of the current 7950 cards such as the Gigabyte Windforce 3x, 950mhz Sapphire Dual-X and MSI TF3 are actually better than this GPU Boost 7950. They can hit those speeds at 1.0V or less.

There is no reason at all for AMD to apply 1.21-1.25V to get 7950 stable at 925mhz. It's just them being extremely conservative because they have to cover for those bad 50-60 ASIC chips and make sure they can hit those clocks. What happens is that it would penalize the best 7950 chips unless MSI, Gigabyte and Sapphire, etc. customize this new bios themselves. What happened though is now it paints most 7950 in a worse light because AMD is applying this 1.2V+ voltage to poor reference 7950 models that to begin with had low ASIC.

Here is Gigabyte Windforce 3x 7970 with ASIC of 58%. It needs very high "real world" VDDC to overclock as you see below:

1.238V VDCC in GPU-Z to reach 1165mhz

vs. Sapphire Dual-X with ASIC of 85%. It needs much lower "real world" VDCC to overclock:

1.102V VDCC in GPU-Z to reach 1145mhz

Source

This is why so many of you guys are getting such high overclocks with low voltages. You got binned 7950 chips, which seems to be a consistent trademark of MSI TF3 7950.

 

[Attic]

I've found that on my two TFIII's that i don't need much voltage to hit 1100 on the core. I started off anticpating a need for a voltage bump, but I have since found that I can lower voltage from the default (1086 in Afterburner) and still hit 1100c stable on both cards.

Currently in MSI afterbuner I plug in 1056mv to get my cards to 1100c but I actually haven't tried lower. Default one card came at 1086mv other I haven't been able to verify. With MSI afterburner set to 1056mv HWInfo shows actual max reading of 1.031-1.034mv. ASIC quality on cards is 88.6 and 85.2.

Temps for me were outrageous at first, 80c+ at stock settings on load and thermal shutdowns when oc/d. After adding 120mm blower onto cards (not exhaust) my load temps while oc/d dropped ~15c, after reapplying TIM (AS5) temps dropped again another 4-6c at load. The card needs to run at 71c or lower to keep noise levels suitable IMO. So i'm still looking to lower temps a bit more while retaining highest possible o/c. Currently at 1025c on each card i'm hitting 74c/73c load temps on the cards. 1100c gets me to 76c/75c.

The TIM was applied way to liberally before I replaced it. Shame this kind of meaningful oversight made it into the card.

So far i'm cranking Batman AC and Sleeping Dogs.

 

[zaydq]

http://www.3dmark.com/3dm11/4150548

Slightly over 10k... i guess i sort of figured out my cpu score but I would think it should be in the 8000s...

Is it reasonable to try for 1300 core clock on air with voltages above 1.175?

 

[Fire&Blood]

http://www.3dmark.com/3dm11/4150548

Slightly over 10k... i guess i sort of figured out my cpu score but I would think it should be in the 8000s...

Keep in mind that most setups you are compared against are K chips 4.5GHz+, driver differences and memory play a role too.


I just started bitcoin . At stock clocks and 1.012v, it doesn't exceed 68C/58% fan setting (fan set to auto control). I get up to 500 MH's. I never tried it on the 5770 so I can't compare scores. I've been hearing about BTC for a while now so I thought why not.

 

[djsb]

Just wondering, does anyone else's card freeze the display when you OC the core too high for the voltage you're using? I had started thinking my card was a relatively bad one because whenever I went above 1050MHz the display would freeze. Not a hard lock, because the audio would continue and programs would keep running, but I would still have to manually restart the machine. It seemed weird because I expected the driver to just crash and recover, like my older cards did. Anyway, I think my card is actually pretty good because Afterburner's default voltage is 0.993V, but the card is not even close to cool, and it sure heats up fast when I up the memory speed. I'm guessing it'll get better when the season changes (my computer room is pretty terrible... stuffy in the summer, chilly in the winter).