-Post pics of your computer-

[Roadhog]

I agree with the furmark part, he shoulda hit a bit more stability testing, but what's wrong with 65C full load? It's not GREAT for water, but it's hella better then stock air cooling.

Well, the 65c part isn't full load is what I meant to say. Furmark would be full load. That's like owning a race car that you can only drive 50mph because it will overheat any faster.

[shadow001]

lol, i've already told him those radiators aren't enough. Which they aren't if you are hitting 65c+ load on water. lol I'd love to see his temps running furmark for a while.

At 925Mhz core/ 2100 Mhz memory and the max voltage that Nvidia inspector allows by default with my cards, and it's still 20~25*c lower than the stock cooler at stock voltages and clocks....You forgot those parts, as usual.


At stock, the cards run at 45*C under load, and i know you love to whip out the furmark issue, but the fact is that it isn't a game and you should also bust both AMD's and Nvidia ass on why their drivers have application detection and when furmark is being run, the cards are downclocked automatically...What, their stock cooling isn't enough for the worst case scenario according to you(furmark).

[Roadhog]

At 925Mhz core/ 2100 Mhz memory and the max voltage that Nvidia inspector allows by default with my cards, and it's still 20~25*c lower than the stock cooler at stock voltages and clocks....You forgot those parts, as usual.


At stock, the cards run at 45*C under load, and i know you love to whip out the furmark issue, but the fact is that it isn't a game and you should also bust both AMD's and Nvidia ass on why their drivers have application detection and when furmark is being run, the cards are downclocked automatically...What, their stock cooling isn't enough for the worst case scenario according to you(furmark).

It's pretty easy, turn off OCP, run furmark for 20 mins. A lot of people want to see the temps too, not just me.

My 6970's don't downclock in furmark. :|

EDIT: Prime95/IBT isn't a game, why use those to check your cpu temps then?

[moshpit]

It's pretty easy, turn off OCP, run furmark for 20 mins. A lot of people want to see the temps too, not just me.

My 6970's don't downclock in furmark. :|

EDIT: Prime95/IBT isn't a game, why use those to check your cpu temps then?

To Hap who just asked me yesterday or today in a different thread - "Why do you bother with all that long OCCT/Prime95 testing when all you do is game?"

This is why. So I'm not fending off these kinds of questions.

Edit: When people ask me "But are you REALLY running stable at that speed?" my answer can invariably be "I did every accepted stability test AND went further with combinations of those tests run together, HELL yeah it's bloody well stable!!!"

[Roadhog]

To Hap who just asked me yesterday or today in a different thread - "Why do you bother with all that long OCCT/Prime95 testing when all you do is game?"

This is why. So I'm not fending off these kinds of questions.

Must be just me, but I like to see if my system can handle worst case scenario and be stable.

[shadow001]

It's pretty easy, turn off OCP, run furmark for 20 mins. A lot of people want to see the temps too, not just me.

My 6970's don't downclock in furmark. :|

EDIT: Prime95/IBT isn't a game, why use those to check your cpu temps then?

Ahem, from the original Anandtech review when the HD6970 was launched:

Although we see both AMD and NVIDIA surpass their official TDP on FurMark, it’s never by very much. After all TDP defines the thermal limits of a system, so if you regularly surpass those limits it can lead to overwhelming the cooling and ultimately risking system damage. It’s because of FurMark and other scenarios that AMD claims that they have to set their products’ performance lower than they’d like. Call of Duty, Crysis, The Sims 3, and other games aren’t necessarily causing video cards to draw power in excess of their TDP, but the need to cover the edge cases like FurMark does. As a result AMD has to plan around applications and games that cause a high level of power draw, setting their performance levels low enough that these edge cases don’t lead to the GPU regularly surpassing its TDP.

http://www.anandtech.com/show/4061/amds-radeon-hd-6970-radeon-hd-6950/7

Have a nice day...

[moshpit]

Must be just me, but I like to see if my system can handle worst case scenario and be stable.

It's not just you. It's the invariable question. That you asked it is no slam on you, it's a FAIR question, and one I'll never be making excuses on to wiggle out of. I'll tell ya straight up, my 850 GPU clocks are rock solid furmark tested, left running overnight while I slept. My CPU clocks are brutally tested, WAY beyond what any forum I may discuss them on will require of me before claiming stability (mainly thinking XS on that, they get MEAN about stability, and I LOVE that!!!).

Edit: Plus, even just on this forum, Caveman_Jim would SHRED me a new one for claiming hard core stability on my overclocks if I didn't do heavy stability testing before making the claim.

Edit #2: See shadows immediate post above for example of the alternative.

[shadow001]

It's not just you. It's the invariable question. That you asked it is no slam on you, it's a FAIR question, and one I'll never be making excuses on to wiggle out of. I'll tell ya straight up, my 850 GPU clocks are rock solid furmark tested, left running overnight while I slept. My CPU clocks are brutally tested, WAY beyond what any forum I may discuss them on will require of me before claiming stability (mainly thinking XS on that, they get MEAN about stability, and I LOVE that!!!).

Edit: Plus, even just on this forum, Caveman_Jim would SHRED me a new one for claiming hard core stability on my overclocks if I didn't do heavy stability testing before making the claim.

You'd also have to turn off OCP on your cards and run furmark overnight at 850 Mhz then...It's the only way to satisfy roadhog...:P

[Roadhog]

Setting powertune to +20% pretty much removes any furmark cap.

[Roadhog]

It's not just you. It's the invariable question. That you asked it is no slam on you, it's a FAIR question, and one I'll never be making excuses on to wiggle out of. I'll tell ya straight up, my 850 GPU clocks are rock solid furmark tested, left running overnight while I slept. My CPU clocks are brutally tested, WAY beyond what any forum I may discuss them on will require of me before claiming stability (mainly thinking XS on that, they get MEAN about stability, and I LOVE that!!!).

Edit: Plus, even just on this forum, Caveman_Jim would SHRED me a new one for claiming hard core stability on my overclocks if I didn't do heavy stability testing before making the claim.

Edit #2: See shadows immediate post above for example of the alternative.

Well it's not so much that, I don't think it's fair to other users to post a watercooling solution for 3, 580's that can only handle basic games, but in reality can't handle 3, 580's at a true 100% load. What if they see that and then buy a similar setup to fold on only to find out it can't handle the load.

He is welcome to prove me wrong by simply turning off OCP and running furmark for a while, but he keeps avoiding doing it because I'm sure he knows in his own head that it won't handle it.

[shadow001]

Setting powertune to +20% pretty much removes any furmark cap.

More from the article:

PowerTune is a power containment technology, designed to allow AMD to contain the power consumption of their GPUs to a pre-determined value. In essence it’s Turbo in reverse: instead of having a low base clockspeed and higher turbo multipliers, AMD is setting a high base clockspeed and letting PowerTune cap GPU performance when it exceeds AMD’s TDP. The net result is that AMD can reduce the dynamic power range of their GPUs by setting high clockspeeds at high voltages to maximize performance, and then letting PowerTune cap GPU performance for the edge cases that cause GPU power consumption to exceed AMD’s preset value.

And in the next page, if you scroll down a bit there's a nice chart about what it throttled down and what isn't, and it looks like furmark is throttled down to 600 Mhz core speeds to maintain the same TDP...

http://www.anandtech.com/show/4061/amds-radeon-hd-6970-radeon-hd-6950/8

However, by setting the powertune to a 20% increase like you said, does remove the cap, but makes the card hit 90*C and consume 300 watts rather than the usual 250 watts it does....not a good idea if you want that card to last in the longer term.

[Roadhog]

More from the article:





And in the next page, if you scroll down a bit there's a nice chart about what it throttled down and what isn't, and it looks like furmark is throttled down to 600 Mhz core speeds to maintain the same TDP...


http://www.anandtech.com/show/4061/a...deon-hd-6950/8

You should really read the whole article, or own an ATI card with powertune before making false claims. At 0% it can throttle, at +20%, it won't throttle. I already said that.

http://www.anandtech.com/show/4061/amds-radeon-hd-6970-radeon-hd-6950/9

As expected, power and temperature both increase with FurMark with PowerTune at 300W. At this point FurMark is no longer constrained by PowerTune and our 6970 runs at 880MHz throughout the test. Overall our power consumption measured at the wall increased by 60W, while the core clock for FurMark is 46.6% faster. It was under this scenario that we also “uncapped” PowerTune for Metro, when we found that even though Metro was being throttled at times, the performance impact was impossibly small.

[Roadhog]

However, by setting the powertune to a 20% increase like you said, does remove the cap, but makes the card hit 90*C and consume 300 watts rather than the usual 250 watts it does....not a good idea if you want that card to last in the longer term.

3 of mine run about 97c load all the time... I don't see them dying prematurely.

[shadow001]

3 of mine run about 97c load all the time... I don't see them dying prematurely.

And did you also notice in the article how varying the powertune levels did very little to nothing, in terms of impacting the performance levels in actual games, so why make the cards run that hard for little to no gains?...Same applies to OCP when the GTX580's were reviewed....Little to no difference at all with it on or off except the amount of power the card was chugging down and how hot it ran.

[Roadhog]

And did you also notice in the article how varying the powertune levels did very little to nothing, in terms of impacting the performance levels in actual games, so why make the cards run that hard for little to no gains?

I run -20% for most games...

It's the fan profiles that don't speed up to 95c or so.

[shadow001]

The point is why create a scenario that you keep insisting on me doing a run without OCP, when it doesn't increase performance or even improve the overclocking potential of the cards at all.....Frankly given the clocks i'm pushing, i'd rather have it on as i get closer and closer to the 1 Ghz mark overclock as some extra protection, as it is 1500$ of my own money on the line here...

[moshpit]

You'd also have to turn off OCP on your cards and run furmark overnight at 850 Mhz then...It's the only way to satisfy roadhog...:P

I did. Worked fine. What's the point? That's what you're SUPPOSED to do when talking overclocks!!! You find the MOST widely accepted stability testing procedures for that particular item, in this case GPUs, and jump in with both feet airborne style! HOOAHH!!! It's puts the hairs on yer chest, son!

Edit: May I be frank about your position? If you cannot run without OCP enabled, your pushing too hard, simple as that. AMD and Nvidia ONLY implemented OCP because of overclockers who would go too far and make big public stinks about the card's failure and not their own to gauge proper limits and observe total stability rules. Ahhm, hint, hint...

Edit 2: EVGA Precision software logged a max temp on my GTX 580 of 87C when I woke up to check on my all night run of furmark. It fluctuated every once in awhile between 86 and 87 but was an average of almost 87C flat. Totally within limits, never once causing problems. Still running without any artifacts that morning as well. THAT is a successful overclock, anything less is BS. I can confidently say nothing will ever beat my card up so hard ever again, so it is 100% stable.

[Roadhog]

I have an idea. Post some rig pics.

[moshpit]

I have an idea. Post some rig pics.

Me? Okay They're not as quite as monsterous as shadow or demowho's multi-GPU monsters, but I'm quite proud of it none the less:


Before adding Crucial Ballistix Tracers and 60mm adapted to 50mm fan


After additions.

I'm addicted to blue LEDs, sorry. I know I'm cheap

Next addition planned : Corsair H100 to replace the H70!!!

[shadow001]

I did. Worked fine. What's the point? That's what you're SUPPOSED to do when talking overclocks!!! You find the MOST widely accepted stability testing procedures for that particular item, in this case GPUs, and jump in with both feet airborne style! HOOAHH!!! It's puts the hairs on yer chest, son!

Edit: May I be frank about your position? If you cannot run without OCP enabled, your pushing too hard, simple as that. AMD and Nvidia ONLY implemented OCP because of overclockers who would go too far and make big public stinks about the card's failure and not their own to gauge proper limits and observe total stability rules. Ahhm, hint, hint...

Edit 2: EVGA Precision software logged a max temp on my GTX 580 of 87C when I woke up to check on my all night run of furmark. It fluctuated every once in awhile between 86 and 87 but was an average of almost 87C flat. Totally within limits, never once causing problems. Still running without any artifacts that morning as well. THAT is a successful overclock, anything less is BS.

And that's cool, but do you honestly believe that with the clocks as they are right now(925/2100), and doing 65* under load that by disabling OCT it's going to somehow get anywhere near the temperatures your setup got....I don't think so....The variations shown with or without OCT or having powertune at 20% add maybe another 7~8*C and that's on the stock cooler, not water cooling wich can handle more to begin with.


The real test with this setup will come once i flash the Bios and gradually start edging towards that 1 Ghz mark and using 1.2 volts...Then if you insist, i'll disable OCT and run furmark at those clocks for sustained periods and only back off once the load tempertures start aproaching what we see with the stock cooler, at the stock clocks and stock voltages.


Let's see if the water cooling can give me a 230Mhz overclock over stock...

[moshpit]

And that's cool, but do you honestly believe that with the clocks as they are right now(925/2100), and doing 65* under load that by disabling OCT it's going to somehow get anywhere near the temperatures your setup got....I don't think so....The variations shown with or without OCT or having powertune at 20% add maybe another 7~8*C and that's on the stock cooler, not water cooling wich can handle more to begin with.


The real test with this setup will come once i flash the Bios and gradually start edging towards that 1 Ghz mark and using 1.2 volts...Then if you insist, i'll disable OCT and run furmark at those clocks for sustained periods and only back off once the load tempertures start aproaching what we see with the stock cooler, at the stock clocks and stock voltages.


Let's see if the water cooling can give me a 230Mhz overclock over stock...

See, Roadhog? Not getting the point at all. In fact just went off in the opposite direction. He'll learn though, maybe, someday

[moshpit]

Let me make the point as CLEAR as it can be made. NEVER assume stability. Do you not think you can do it, shadow? THAT would be the only reason not to. Keep in mind, you're not in card killing ranges yet, problems will manifest visibly LONG before you threaten the card's life, IF you follow the REAL testing procedures that your skipping on past.

[Roadhog]

See, Roadhog? Not getting the point at all. In fact just went off in the opposite direction. He'll learn though, maybe, someday

lol.. I lol'd really hard at comparing air to water cooling. Once you go over what the radiator can handle wattage wise on water temps just get higher, and higher, and higher. Air you just get hot and stay there.

[moshpit]

lol.. I lol'd really hard at comparing air to water cooling. Once you go over what the radiator can handle wattage wise on water temps just get higher, and higher, and higher. Air you just get hot and stay there.

Regardless of cooling, he's missing the point totally that he could still have his current clocks most likely after proper testing, but would also have a better idea of how much true headroom he has left.

[shadow001]

See, Roadhog? Not getting the point at all. In fact just went off in the opposite direction. He'll learn though, maybe, someday

I'm not going off any opposite direction....


I told you i had the cards running at stock and running at stock voltages and got to 45*C under load.....With a 150Mhz increase on the GPU, a 100Mhz increase on the memory and a 0.1 volt increase on the GPU, the load temperature only increased by 20*C, to 65*C.


There's room to breathe before i start panicking...

[shadow001]

And in all this posting, i still managed to do this:





Created a new bios and flashed all 3 cards with it successfully, so the max limit for voltage on the GPU's using Nvidia inspector went from 1.138 millivolts in the original bios to 1.213 millivolts using my modified bios...


Now we'll see what's what....

[aviphysics]

I think the point they are trying to make is this. Because water has a fair bit of thermal capacity, it takes a fairly long time for your system to heat up to full temperature. If you don't do extensive testing, i.e running a stress test overnight, you aren't going to find out what your real max temperatures are.

[shadow001]

I think the point they are trying to make is this. Because water has a fair bit of thermal capacity, it takes a fairly long time for your system to heat up to full temperature. If you don't do extensive testing, i.e running a stress test overnight, you aren't going to find out what your real max temperatures are.

To a point you're right, but then again how far do you push that standard?...Ambient temperatures makes a huge difference and someone living in more temperate climates, all else being equal, is going to get higher overclocks out of their hardware or at the very least, lower temperatures for a given clock, compared to a user living much further soulth, using the exact same hardware and where ambient temperatures can easily exceed 100*F in the summer, so the cooling requirements are way different in this last example.


There's no pre determined setup that fits all requirements and it really is case by case, and ambient temperature is just one variable, there's many more than just that....If i wanted to be sure that even if i'm living in the middle of the sahara desert, my PC won't overheat no matter how overclocked it is with OCP disabled and running furmark for an entire week if need be..:P, i wouldn't have picked this case, but something like the 2 examples i linked a few posts ago and shove 6 radiators in it with 25~30 fans in push~pull on all radiators and high powered pumps and call it a day....Maybe even use a water chiller whhile i'm at it.


There's no need for charts or experts in any forums of any kind when you go seriously overkill like that, but having something that's basically twice as large as the case i'm already using, wich small it isn't, is just nuts....It would take up way too much room for me, and with all those fans fitted, pretty noisy too....I'm aiming for good enough cooling and silent.

[clerick]

I decided to cat proof my pc since he is so curious.

[aviphysics]

...

In terms of how far you go, you need only run until your temperatures really stabilize. Probably at least several hours to be sure.

If they get too high you need only to turn your overclocks down.

[Jokerzwild]

I decided to cat proof my pc since he is so curious.

Haha! That's perfect, and looks pretty cool too .

[shadow001]

In terms of how far you go, you need only run until your temperatures really stabilize. Probably at least several hours to be sure.

If they get too high you need only to turn your overclocks down.

I did over 2 hours of the heaven demo without interruption at all, and the 65*C temperature was the peak and held that temperature was constant for almost the entire duration with very little variation, and we can't say that it isn't a demanding DX11 techdemo/benchmark program...


The very first pass on the program is when the temperatures were still slowly and very gradually rising, then it stabilized and stayed that way for the next 20+ passes.

[BababooeyHTJ]

To Hap who just asked me yesterday or today in a different thread - "Why do you bother with all that long OCCT/Prime95 testing when all you do is game?"

This is why. So I'm not fending off these kinds of questions.

Edit: When people ask me "But are you REALLY running stable at that speed?" my answer can invariably be "I did every accepted stability test AND went further with combinations of those tests run together, HELL yeah it's bloody well stable!!!"

So when a game crashes to desktop you can be 100% sure that it's not your system and not complaining in the PC gaming section about a game being unstable when it is really your system.

I had one overclock on a P45 that passed small fft but the MCH was mildly unstable. It ran fine for a while and eventually I started seeing artifacts in a couple of games. Even going back to stock speeds didn't fix it due to what I assume would be data corruption. I had to fine tune my oc and reinstall the os.

I did over 2 hours of the heaven demo without interruption at all, and the 65*C temperature was the peak and held that temperature was constant for almost the entire duration with very little variation, and we can't say that it isn't a demanding DX11 techdemo/benchmark program...


The very first pass on the program is when the temperatures were still slowly and very gradually rising, then it stabilized and stayed that way for the next 20+ passes.

I think that it's safe to assume that is about what you will see in actual gaming. On a side note have you tried GPUtool's artifact scanner? It is my favorite utility to stress test a video card. In fullscreen it is just a bit more stressful than actual games.

I'm with you I never run furmark for more than a few minutes. I saw what that did with the vrms on my GTX280 on a uni-sink and there is no way for me to monitor them via software on my current card.

[moshpit]

So when a game crashes to desktop you can be 100% sure that it's not your system and not complaining in the PC gaming section about a game being unstable when it is really your system.

Hallaluya, brother!

[shadow001]

I think that it's safe to assume that is about what you will see in actual gaming. On a side note have you tried GPUtool's artifact scanner? It is my favorite utility to stress test a video card. In fullscreen it is just a bit more stressful than actual games.

I'm with you I never run furmark for more than a few minutes. I saw what that did with the vrms on my GTX280 on a uni-sink and there is no way for me to monitor them via software on my current card.

I also used MSI's kombustor stress program, wich looks like furmark but also tests more features rather than just a spinning furry doughnut, and got the same temperature readings i did when running the heaven tech demo, so i got a long way to go until temperatures are even close to what they would be using the cards stock cooler.


Basically, if the overclock fails and the program crashes to the desktop, it won't be bacause the cards are overheating, but simply because they cards just can't handle the speeds period.

[Roadhog]

Basically, if the overclock fails and the program crashes to the desktop, it won't be bacause the cards are overheating, but simply because they cards just can't handle the speeds period.

You know the higher you overclock, the cooler the card needs to be to stay stable. Why do you think people use liquid nitrogen. lol

[shadow001]

You know the higher you overclock, the cooler the card needs to be to stay stable. Why do you think people use liquid nitrogen. lol

Liquid nitrogen brings it's own set of problems, just like any other sub 0*C cooling method....Water condensation build up from the moisture in the air itself and freezing on any exposed metal, hence why you see all that ice forming around the pots for the CPU and GPU, but motherboards don't particularly like it for obvious reasons, and those who attempt it have to use tons of insulating material precisely for that problem alone....


Add the fact that you always have to top up the pots every 10~15 sec or so as the liquid nitrogen evaporates very quickly as it's absorbing the heat produced by the CPU and GPU(and even the ambient air for that matter)...


Add the special storage containers that liquid nitrogen requires, wich are usually rented and that a lot of LN2 will be used in the process of creating that new world record....


The most i'd ever try beyond straight water cooling would be using a water chiller made of an old A/C unit, but there the evaporator is modded to fit into a container that's part of the water loop, and the pressure switches left alone so the compressor works just like it always designed to, so that the temperature of the evaporator drops to just above 0*C(doing the same for the water in the cooling loop along with it), thus avoiding condensation issues using Ln2 or pelts...Obviously, the condenser has to be mounted outside the house with it's own fans to release the heat produced....With a lot of planning and elbow grease, it could work but that's as far as i'd take it....

[nycdarkness]

Anyone else notice that Shadow is missing a 580?

[Roadhog]

Liquid nitrogen brings it's own set of problems, just like any other sub 0*C cooling method....Water condensation build up from the moisture in the air itself and freezing on any exposed metal, hence why you see all that ice forming around the pots for the CPU and GPU, but motherboards don't particularly like it for obvious reasons, and those who attempt it have to use tons of insulating material precisely for that problem alone....


Add the fact that you always have to top up the pots every 10~15 sec or so as the liquid nitrogen evaporates very quickly as it's absorbing the heat produced by the CPU and GPU(and even the ambient air for that matter)...


Add the special storage containers that liquid nitrogen requires, wich are usually rented and that a lot of LN2 will be used in the process of creating that new world record....


The most i'd ever try beyond straight water cooling would be using a water chiller made of an old A/C unit, but there the evaporator is modded to fit into a container that's part of the water loop, and the pressure switches left alone so the compressor works just like it always designed to, so that the temperature of the evaporator drops to just above 0*C(doing the same for the water in the cooling loop along with it), thus avoiding condensation issues using Ln2 or pelts...Obviously, the condenser has to be mounted outside the house with it's own fans to release the heat produced....With a lot of planning and elbow grease, it could work but that's as far as i'd take it....

lol. you really can't understand simple statements can you...

Here, let me break it down.

You might be able to hit 950mhz @ 75c load

You might be able to hit 1ghz @ 50c load

Pretty simple to understand for the average person.

[shadow001]

Anyone else notice that Shadow is missing a 580?

:P

lol. you really can't understand simple statements can you...

Here, let me break it down.

You might be able to hit 950mhz @ 75c load

You might be able to hit 1ghz @ 50c load

Pretty simple to understand for the average person.

I said what keeps the card from eventually clocking higher isn't because it's overheating, and what you said has nothing to do with the card overheating in the first place, but how to get higher clocks period, in this case by going the sub 0 cooling route by means of liquid nitrogen..... Under those extreme mesures, the GPU is always operating well under 0*C even under load...