[H] users 7950 Overclock Results

[Gumbi10]

Sure. I will do that now.

With regards to the memory, I blue-screen, green-screen, w/e after 30 seconds of running Heaven at 1350 :/ I will run for 2 mins at 1200 mhz and then screenshot (just to let the VRMs heat up a bit, tbh, they probably take a bit longer to stabilise fully, but whatever). I can still fire up my side fan to 800RPM roughly, it's only running at about 500 - 600 right now. I'll do that later though,

 

[Gumbi10]

Here is 10 mins of Heaven. Max temps of 78 degrees on VRM 1, not bad. As you can see it was pretty stable at that temp (as in it didn't change for a certain period of time, unlikely for it to get any warmer). Just to give you an idea of my ambients (heatwave in Ireland right now) so as warm as my ambients will get, although not as warm as they were last week, my GPU idles at 38 degrees. That's a nice comparison for you as we seem to have similar cooling. 59 degrees max on core.

http://imgur.com/RfidLCH

 

[Insanitic]

so my VRM 1 runs 3 degrees cooler than yours with the fujipoly pads. Unfortunately, not too much of a difference and my ambient idle temperature is 30 degrees (AC in fullblast, I too have a heatwave here in Canada ). However, I'm also running at a higher voltage than you AND I ran it for 20mins as opposed to 10min so that might have affected the VRM temps somewhat.
And I see that your gpu usage remains consistent. Mine fluctuates from 96% and above. Its weird. I've tried everything, power limit, oc my CPU and nothing changes.

 

[Gumbi10]

so my VRM 1 runs 3 degrees cooler than yours with the fujipoly pads. Unfortunately, not too much of a difference and my ambient idle temperature is 30 degrees (AC in fullblast, I too have a heatwave here in Canada ). However, I'm also running at a higher voltage than you AND I ran it for 20mins as opposed to 10min so that might have affected the VRM temps somewhat.
And I see that your gpu usage remains consistent. Mine fluctuates from 96% and above. Its weird. I've tried everything, power limit, oc my CPU and nothing changes.

Are you sure you aren't clicking on other windows? Is the bench the only "active" window while benching?

 

[Insanitic]

Are you sure you aren't clicking on other windows? Is the bench the only "active" window while benching?

Yes, the benchmark is the only window active, nothing else is open. The thing is it doesn't always fluctuate from 96% the whole benchmark, but only in some sequences. It mostly stays at 99%

 

[Gumbi10]

Yes, the benchmark is the only window active, nothing else is open. The thing is it doesn't always fluctuate from 96% the whole benchmark, but only in some sequences. It mostly stays at 99%

Throttling? Dunno tbh.

 

[Insanitic]

Throttling? Dunno tbh.

Meh, I guess I'll figure it out. Thanks for the advice though.
If your pads arrive, can you do 20 min run of Heaven at max of course, at your 1200mhz speeds? I'm curious of what your VRM temperatures will be.

 

[Gumbi10]

Meh, I guess I'll figure it out. Thanks for the advice though.
If your pads arrive, can you do 20 min run of Heaven at max of course, at your 1200mhz speeds? I'm curious of what your VRM temperatures will be.

Sure They have already arrived (I work up the country during the week and head home for the weekend, college work placement 'til next year), I'm heading home today and will be back Sunday nioght or some testing

 

[someone11]

Just overclocked mine running at 1.25V (stock voltage)

1150/1550

This was a quick overclock, so it can probably go a little higher but started getting artifacts at 1175.


Heaven 4.0 results @ 720p (don't have a 1080p monitor yet) with highest possible settings
FPS: 58.4
Score: 1470
Min FPS: 22.8
Max FPS: 161.4



Ran burn test with FurMark for 10 minutes and highest GPU temp was 69 and VRM temp was 68.

 

[Insanitic]

Just overclocked mine running at 1.25V (stock voltage)

1150/1550

This was a quick overclock, so it can probably go a little higher but started getting artifacts at 1175.


Heaven 4.0 results @ 720p (don't have a 1080p monitor yet) with highest possible settings
FPS: 58.4
Score: 1470
Min FPS: 22.8
Max FPS: 161.4



Ran burn test with FurMark for 10 minutes and highest GPU temp was 69 and VRM temp was 68.

What make of card? That VRM and core temperature are pretty good.

 

[someone11]

What make of card? That VRM and core temperature are pretty good.

It's a SAPPHIRE HD 7950 3GB with boost running 12.104 drivers. The reason my temperatures are good is because I'm running a custom auto fan control.

 

[Deleted whining member 223597]

Quick question, how do you guys stress test your OC?

I tried OCCT but that gave me errors on stock clocks, so don't know how useful that is.


And heaven I got blue and red flashes(little dots spread throughout) after a measly 50mhz OC(stock is 1000/1250 @ 1.25V). Is my Windforce really just that shitty?

 

[tonyftw]

Quick question, how do you guys stress test your OC?

I tried OCCT but that gave me errors on stock clocks, so don't know how useful that is.


And heaven I got blue and red flashes(little dots spread throughout) after a measly 50mhz OC(stock is 1000/1250 @ 1.25V). Is my Windforce really just that shitty?

I play games to test my oc. Assuming you got the rev 2.0 windforce 7950, the majority of those are very crappy. The one I had could only do 1100/1300 stable. Anything above that, it would begin to artifact.

 

[Deleted whining member 223597]

I play games to test my oc. Assuming you got the rev 2.0 windforce 7950, the majority of those are very crappy. The one I had could only do 1100/1300 stable. Anything above that, it would begin to artifact.

Yep, I guess I'm outta luck here, it artifacts at 1075 already. Maybe I'll throw in my Twin Frozr and see how that goes(Can't run both due to the fail of the coolers, they scrape on each other on the gene-z)

 

[Gumbi10]

Has anybody here used the Arctic 7970 Hybrid cooler? The all-in-one water solution? Apparently it performs fantastically at cooling the card. Does anybody have any number for how it cools the VRMs?

 

[Insanitic]

Has anybody here used the Arctic 7970 Hybrid cooler? The all-in-one water solution? Apparently it performs fantastically at cooling the card. Does anybody have any number for how it cools the VRMs?

I'm wondering about this too. It seems like NOBODY has used this product on a 7900 series and reported on VRM temps. I am curious too.

EDIT: Never mind, found a user review on Newegg concerning the VRMs, although he did mix up the VRM numbers (VRM #1 is the long strip near the power connectors, and VRM #2 is near the crossfire connectors)

Here is it copy and pasted:
Pros: Cools GPU down to 65 full load (from 80C stock).
Very very quiet even at 100% fan.

Cons: Does not have cooling (airflow) on the #1 VRM, temperatures shot up to 90-100C before I pulled the plug. 7900 cards have VRM #1 near the crossfire port and VRM #2 near the power connectors. The 7970 arctic cooler does not blow air over the #1 VRM at all, even though they tell you to put a heatsink on it.

If you are going to pay $160 for a cooler, it better sure as hell cool the ENTIRE card.

This may be ok for medium gaming, but high VRM temps can heat up other components and wear down the life of the card (although they can handle up to 120C). You cannot overclock with this cooler and keep the card safe.

It is a design flaw in the product, because the previous version (non-7970 arctic hybrid) did not have to take this into consideration. This version of the cooler is pretty much identical to its predecessor. 7900 cards VRMs run a lot hotter also because of the greater electricity requirement, but Accelero did not take this into consideration.

I took an 80mm fan and placed it blowing over VRM #1 position and cooled it by nearly 7C, something they should have considered in the design.

The install can be difficult (although I've done 5 at this point) for the new guy.

Other Thoughts: If you are only concerned about GPU temperatures and not the rest of the PCB, this cooler will keep your GPU @ 60-70C max full load. However, the 7900 cards VRMs heat up a TON and should not be ignored. These can be checked in GPU-Z.

I feel like the Accelero Xtreme is better. Oddly enough, based on his results, my Accelero Xtreme even cools the core better. My core usually runs around 55 degreess max.

 

[Gumbi10]

Indeed. According to reviews, though, it does actually outperform the air cooler slightly in both noise and core temps. What a terrible design flaw that is, though, not to have air washing over VRM1.

Ah well, the Accelero is compatible with the 8000 series cards, I wonder is it compatible with the 9000 ones? Having this cooler might justify the cost in upgrading to a new in October. Hopefully it cools the VRMs better on those cards I think I need more fujipoly padding, which I might just get if I end up purchasing an 8000/9000 card.

 

[Insanitic]

Indeed. According to reviews, though, it does actually outperform the air cooler slightly in both noise and core temps. What a terrible design flaw that is, though, not to have air washing over VRM1.

Ah well, the Accelero is compatible with the 8000 series cards, I wonder is it compatible with the 9000 ones? Having this cooler might justify the cost in upgrading to a new in October. Hopefully it cools the VRMs better on those cards I think I need more fujipoly padding, which I might just get if I end up purchasing an 8000/9000 card.

There are currently no news at all about what the gpu/ socket that the gpu would look like in the 9000 series. So can't say at this point if the Accelero can fit for or not.

Would you mind posting a 3dmark 11 performance score with your highest possible overclock? I feel like my 7950 is underperforming. Can you also do it the new cat. 13.8 drivers?

 

[yakev724]

Very happy with my 7950, early MSI BE with the 7970 pcb.

1215/1575 at 1.12V

GPU stays below 80C and VRM below 90C...got to 1250 core with higher voltage but had VRM temps above 100C...


3DMark11 - P10315 (11183 gpu)
Heaven - 1134
Valley - 1539




I ran a new Gigabyte 770 OC in my system (stock speed) and got the following:

3dMark11 - P10609 (11635 gpu)
Heaven - 1195
Valley - 2347


So, pretty happy that I paid about half price for a card that performs so closely to the 770, which was perfect in (single monitor) gaming. This has to be just about the most performance available for the money?

 

[Gumbi10]

Fantastic results man!

 

[45454asasa]

80c on the core? Jesus man...

Nice results though, mine would never be stable at 80c.

btw, what is an MSI BE? You mean TF? Or XFX BE?

 

[Gumbi10]

80c on the core? Jesus man...

Nice results though, mine would never be stable at 80c.

btw, what is an MSI BE? You mean TF? Or XFX BE?

He said under 80. The 7970 PCB probably helps a lot in this regard (don't reference 7970s go over 80 degrees at stock settings anyway?).

My 7950 is stable at 1215mhz at 1.12v too! Your card would be great under water, too. With so much voltage headroom, you could get some crazy high clock rate with 1.25/1.3v. For some reason I have dud memory. It crashes at 1350mhz :/ I actually have a "perfect" 7950 in that I have access to all sensors, both the core AND memory voltage is unlocked and modifiable (after a mod) etc. So I can actually vary my mem voltage from 1.4 - 1.8v. Stock is 1.6v. I could push to 1.7 and see what kind of stability that gives me, but to be honest, I don't think it helps too much with performance. Is pushing the memory to 1.7v detrimental?

I have an Accelero cooler which means 60 degrees MAX of core at 1215mhz (VRMs go higher though, left on stock heatsinks as they are apparently better than the accelero ones; got some fulipoly thermal padding, but haven't used it yet).

 

[yakev724]

80c on the core? Jesus man...

Nice results though, mine would never be stable at 80c.

btw, what is an MSI BE? You mean TF? Or XFX BE?

Core goes to 78-79 only after a few rounds of benching. Stays around 71-72 during gaming.

It's an OC Boost Edition with the TF cooler. It's an earlier model, built on the 7970 PCB (different part no. and power connectors to the 7950 PCB cards).

I'd pull it apart and get some copper shims/compound on the VRM but it's still under warranty and probably not worth the cost/effort anyway.

 

[Gumbi10]

Very nice. I'm disappointed in my memory overclocking, but love my card otherwise. Does anyone know if running my memory is safe or not? As I said, cooling isn;t a problem, VRMs at 80 degrees or less at 1215-25 mhz, and core is under 60.

 

[yakev724]

He said under 80. The 7970 PCB probably helps a lot in this regard (don't reference 7970s go over 80 degrees at stock settings anyway?).

My 7950 is stable at 1215mhz at 1.12v too! Your card would be great under water, too. With so much voltage headroom, you could get some crazy high clock rate with 1.25/1.3v. For some reason I have dud memory. It crashes at 1350mhz :/ I actually have a "perfect" 7950 in that I have access to all sensors, both the core AND memory voltage is unlocked and modifiable (after a mod) etc. So I can actually vary my mem voltage from 1.4 - 1.8v. Stock is 1.6v. I could push to 1.7 and see what kind of stability that gives me, but to be honest, I don't think it helps too much with performance. Is pushing the memory to 1.7v detrimental?

I have an Accelero cooler which means 60 degrees MAX of core at 1215mhz (VRMs go higher though, left on stock heatsinks as they are apparently better than the accelero ones; got some fulipoly thermal padding, but haven't used it yet).

Nice! I'm going to keep mine stock (don't have plans for water in my system), possibly xfire if a pair would be more cost efficient than upgrading to a quicker single card (which may actually be the case with these!).

To be honest I don't think much additional time/tweaking is gonna yield that much more performance. I'll monitor temps and maybe adjust by another 5-10MHz, but pretty much done here...as I said, no plans to pull the warranty sticker and remove the stock HSF.

 

[Gumbi10]

There's a warranty sticker? Ah, that's a pain. I'll just stick on my 7950 stock Heatsink if I run into any trouble.


Picking up a second 7950 second hand or something for 200 bean will give you some serious horsepower. Clock em both at ~1100mhz or more depending on case airflow and you're golden. 1440p gaming easily. New drivers help a lot too

What do you think of pushing my memory voltage to 1.7v?

 

[45454asasa]

He said under 80. The 7970 PCB probably helps a lot in this regard (don't reference 7970s go over 80 degrees at stock settings anyway?).

Oh, all I meant was that I cant believe it doesnt crash. My 7950 is.. similar [not quite as nice ]

Does 1200 @ 1.138v. However temps have to be <65 or so, then it artifacts. So it was hard to imagine nearly 80c and completely stable.

Core goes to 78-79 only after a few rounds of benching. Stays around 71-72 during gaming.

It's an OC Boost Edition with the TF cooler. It's an earlier model, built on the 7970 PCB (different part no. and power connectors to the 7950 PCB cards).

I'd pull it apart and get some copper shims/compound on the VRM but it's still under warranty and probably not worth the cost/effort anyway.

Ah, boost edition.. I see. Does MSI have those damn void warranty stickers on the screws?

Very nice. I'm disappointed in my memory overclocking, but love my card otherwise. Does anyone know if running my memory is safe or not? As I said, cooling isn;t a problem, VRMs at 80 degrees or less at 1215-25 mhz, and core is under 60.

You probably have Elpida 1250, so its not going to go anywhere. I think all 7950s come with that by now. I was lucky enough to get Hynix 1250.

I'd also like to know what a safe memory voltage is. Mine is stable 1775 gaming or 1850+ in benches, @ ~1.612v. However I've read both 1.5v & 1.6v as stock.

 

[Gumbi10]

Is that 1.612 in as reported in GPUZ? You can just take that as 1.6v I think. 1.5v is apparently for the undervolted memory in the VaporX. 1.6v is typical.

Yeah, my memory is a complete dud. Just pushed it to 1.7v and had green lines across screen (memory crash) at 1368mhz (random slider push ). So yeah. That's going nowhere. It's surprising really, because I got this card almost a year ago, and it's one the best 7950s (access to all sensors, memory voltage, core voltage, etc etc), custom 7950 PCB (that accepts 7970 waterblocks) etc. And 1.12v at 1225mhz is damn good.

 

[45454asasa]

No, I set it to 1.612 [It has 0.006v increments] - that extra .012v actually did help. I would get random color screens without it once the card heated up, with that slight overvolt it never does. When you say 1.12v you are reporting what you set or what the sensors say?

That is damn good core clocks you got and be happy with em. I bet I get less than 10% more framerate from 1800MHz to 1250MHz memory.

Unfortunately my card is non-reference so im stuck with a universal waterblock. I was excited about seeing where mine would go and got all my stuff for my first loop last week, but the dummies at EK had labeled their VGA supremacy things wrong, so I got the wrong one -_-. So now I wait 2 weeks until FrozenCPU gets correct stock.

 

[Gumbi10]

1.12v is what I set GPUZ reports the voltage applied AFAIK, (it varies slightly) so it's always lower just under 1v IIRC (am on laptop now ). I have read that this value can be inaccurate, as nothing can accurately report vdroop. I dunno. Either way, 1.12v is what's applied in Trixx And yes, I'm very happy with my OC. I can maintain a 1250 mhz overclock (not sure of exact voltage, slightly higher than 1.14 AFAIK), the Accelero cooler cools the core amazingly well, am at 65 degrees usually at 1250mhz core. VRMs get hot though, almost hit 90 degrees under stress. This is safe though AFAIK. I could improve them a but via fujipoly thermal padding, but it's not worth the effort IMO; also, I'm considering getting a 9000 series card if my Accelero is compatible with them, (will it? Arctic says it's compatible with the 8000 series anyways, and 9000 is non-OEM right? So it should still support them?). so am saving the thermal padding for that card if that's the case )))

 

[yakev724]

What do you think of pushing my memory voltage to 1.7v?

Really not experienced in the matter. Never bothered too much with memory due to low return/risk but might make sense in your case.

Oh, all I meant was that I cant believe it doesnt crash. My 7950 is.. similar [not quite as nice ]

Does 1200 @ 1.138v. However temps have to be <65 or so, then it artifacts. So it was hard to imagine nearly 80c and completely stable.

That seems strange...I think at stock speeds my card gets up to around 60-65C. Rising temps (within normal operating limits) shouldn't lead to crashes/artifacts...

Ah, boost edition.. I see. Does MSI have those damn void warranty stickers on the screws?

Yep, and looks like I have a single large plate covering the VRM though, so don't really see how I'd cool them if I kept the stock cooler installed...

I'd also like to know what a safe memory voltage is. Mine is stable 1775 gaming or 1850+ in benches, @ ~1.612v. However I've read both 1.5v & 1.6v as stock.

Mine's at 1.55 (also adjustable in Afterburner, though I haven't bothered)

 

[Gumbi10]

Applied 1.7v (stock 1.6). (It applies in 0.06v increments so whatever 1.7 applied rounds to) and set 1362mhz memory. It crashed almost instantly. So ya. Memory is a total fail.

By the way, what is the ASIC of your MSI 7950, my guess is it's more than 85, seeing as you have such low voltage for that high overclock.

 

[J Macker]

80c on the core? Jesus man...

Nice results though, mine would never be stable at 80c.

btw, what is an MSI BE? You mean TF? Or XFX BE?

I haven't read much of this thread, but I was under the impression that generally 85C is okay.

Is this not true?

I have 18, yes, eighteen MSI TF3 7950's that I use for mining. Granted, I'm using them a little different by trying to reduce the voltage and maintain stability instead of pushing the fastest core/mem speeds like you guys do.
Most of them have been running at 80-85C for the past 3 months 24/7. They're perfectly stable at 85C.
There's a few of the cards (about 6) that only hit around 75C, probably due to a slightly different airflow configuration or an extra 0.5 inch spacing between the gpus.

I have cgminer set to auto-throttle when it goes above 85C, or kill the process if it hits 92c.

They're set between
1020-1050 mV
1000-1020mhz core
1250 mem

So, I can safely say that 85C is just fine for 24/7 operation with those settings.

 

[J Macker]

Ah, boost edition.. I see. Does MSI have those damn void warranty stickers on the screws?

Yes, MSI cards have a single warranty sticker on one of the screws.

 

[45454asasa]

I haven't read much of this thread, but I was under the impression that generally 85C is okay.

Is this not true?

I have 18, yes, eighteen MSI TF3 7950's that I use for mining. Granted, I'm using them a little different by trying to reduce the voltage and maintain stability instead of pushing the fastest core/mem speeds like you guys do.
Most of them have been running at 80-85C for the past 3 months 24/7. They're perfectly stable at 85C.
There's a few of the cards (about 6) that only hit around 75C, probably due to a slightly different airflow configuration or an extra 0.5 inch spacing between the gpus.

I have cgminer set to auto-throttle when it goes above 85C, or kill the process if it hits 92c.

They're set between
1020-1050 mV
1000-1020mhz core
1250 mem

So, I can safely say that 85C is just fine for 24/7 operation with those settings.

I think I confused people with what I said before. All I meant was I cant imagine running mine at 80C simply because in every profile I have, that results in crashing. But I suppose with a more stable/higher voltage, maybe that wouldnt be an issue for me.

I also used my card for mining for awhile. I found it could run 1200MHz at just under 1.05v IIRC. Even then it was struggling to stay under 70. And after 4 months, voltage requirements rose about 0.010v for all my profiles. Since I havent mined for so long, I think its since recovered.

Applied 1.7v (stock 1.6). (It applies in 0.06v increments so whatever 1.7 applied rounds to) and set 1362mhz memory. It crashed almost instantly. So ya. Memory is a total fail.

By the way, what is the ASIC of your MSI 7950, my guess is it's more than 85, seeing as you have such low voltage for that high overclock.

Dang, 1.7v really netted you nothing at all . Did you try lowering the core to make sure its not a issue with both being higher?

My ASIC is 86.8 - good guess

That seems strange...I think at stock speeds my card gets up to around 60-65C. Rising temps (within normal operating limits) shouldn't lead to crashes/artifacts...

Well, its with a heavy OC. I mean, at stock, I dont know how high it will go before crashing. But at basically any OC setting I have [where I used the lowest stable voltage] once temps creep up, it artifacts and a few degrees after artifacts, display driver crashes.

Yep, and looks like I have a single large plate covering the VRM though, so don't really see how I'd cool them if I kept the stock cooler installed...

Could always try to contact MSI and say your card is overheating and ask if you can reapply TIM or something.

Mine's at 1.55 (also adjustable in Afterburner, though I haven't bothered)

How does yours OC?

 

[Gumbi10]

1200mhz at 1.05v? That sounds insanely low to me. My stock voltage is really low, and even thern I need to up that to get 1200mhz. My stock is 1.012v for 950mhz, and it will do 1000mhz (maybe a tad more, haven't actually tested the limit) at that voltage.

 

[yakev724]

How does yours OC?

1215/1575 at 1.12V

GPU stays below 80C and VRM below 90C...got to 1250 core with higher voltage but had VRM temps above 100C...


3DMark11 - P10315 (11183 gpu)
Heaven - 1134
Valley - 1539

asic is 87.8

As mentioned, I think my card may be a great candidate for liquid cooling, but I'm a practical OCer and have no plans to spend that sort of money on something short of a faster card...might make reselling it a bit easier though!

 

[45454asasa]

1200mhz at 1.05v? That sounds insanely low to me. My stock voltage is really low, and even thern I need to up that to get 1200mhz. My stock is 1.012v for 950mhz, and it will do 1000mhz (maybe a tad more, haven't actually tested the limit) at that voltage.

This is only for miners and some other rare programs. Otherwise it takes 1.138v.

It might be more than 1.05v, I cant test now though. Vertical orientation right now [FT-02] and miner + 1200 = too much heat for half the heatpipes. I used to run the case with front panels on the floor, but since I have the CPU watercooled and external, I cant do that.

I can run it for a few seconds though.. lol, then it gets too hot.

@1.082v it droops to 0.95v but still gives good results in cgminer.

asic is 87.8

As mentioned, I think my card may be a great candidate for liquid cooling, but I'm a practical OCer and have no plans to spend that sort of money on something short of a faster card...might make reselling it a bit easier though!

Ah, missed it. I agree its not the best use of cash, but for me, it means I'll never spend an hour with my finger on checkout with an aftermarket air cooler ever again.

 

[SolidBladez]

FYI, you can remove the stock heatsink on MSI cards. It doesn't invalidate the warranty. They put it there so that in case there is damage and the sticker was removed then they can blame the customer. So as long as you don't damage anything you will be okay.

 

[45454asasa]

FYI, you can remove the stock heatsink on MSI cards. It doesn't invalidate the warranty. They put it there so that in case there is damage and the sticker was removed then they can blame the customer. So as long as you don't damage anything you will be okay.

I dont understand.

It doesnt invalidate the warranty, unless the card is damaged...

But... that means if it dies later on, the card is damaged, and it could be blamed on the customer.

Therefore, the warranty is invalidated..?