Radeon 5970 Overclocking: The VRM Temperature Bottleneck

[Manicdan]

might exist? AnAndTech clearly shows it exist and even ATI acknowledges it as the "power virus"...



it's funny that ATI says their cards shouldn't have to handle stress tests like OCCT/Furmark because it's not "realistically possible". Show me another video card on the market that can't run Furmark or OCCT without dropping clocks or crashing at their stock speed.

by might exist, i was referring to gaming, since i also dont play furmark

[Vinas]

People like the 5970 because it is new and fast. Like others have said, if it becomes a problem people can always buy aftermarket coolers. It's not that hard to put an aftermarket air cooler on these cards just to spank the competition a bit.

Personally I think all dual GPU solutions should have better cooling. When I was in the market for a dual GPU I specifically bought an Asus TOP model so that I had extra cooling capacity. If I didn't plan on overclocking, that would not even be a concern.

I get what your concern is about the 5970 cooling capacity. But I just don't think it's an issue for many of the folks that will buy this card. These buyers either A) will run at stock speed or B) will W/C it and OC the crap out of it. Either way ATi has a winner on their hands just because of its speed.

[Manicdan]

It's great all you ATI boys say you don't play FurMark, but if this card is already giving problems you can only imagine what is going to happen down the road as more and more stressful games are released. It's a formula for heat problems and AnAndTech is using stress tests to bring this problem to the surface now with our current applications. People turning their noses up at the test are overlooking the bigger picture.

I mean think about how stressful Doom 3 is compared to Crysis, you are talking about 10-20C difference in GPU temps between those two games. Eventually there are going to be games that push the cards just as hard as FurMark does. Also since most people buying these cards think they are future proofing, they might like to know about this VRM temperature problem early on (maybe to give time to save up for a water block).

games cannot stress like a synthetic program. ati uses a different style for their shaders where 1 can have the power of up to 5, yet most games will never be simple enough that the logic can be divided to all 5 perfectly. this is why i find it very important to post the FPS of things like furmark. if a 4870 is doing 2x the work of a 280, then guess what, drop the speed in half, and match the competition without stressing the cards at all. thats whats missing in the whole debate, how WELL does the ati card work with these power viruses.

[defect9]

wow, it's an epeen party in here

surprised the internet is still so serious

[Movieman]

I'm going to say it just once. This crap stops now.
Be smart, take the hint.

[postumus]

games cannot stress like a synthetic program. ati uses a different style for their shaders where 1 can have the power of up to 5, yet most games will never be simple enough that the logic can be divided to all 5 perfectly. this is why i find it very important to post the FPS of things like furmark. if a 4870 is doing 2x the work of a 280, then guess what, drop the speed in half, and match the competition without stressing the cards at all. thats whats missing in the whole debate, how WELL does the ati card work with these power viruses.

Interesting and valid point imo.

[Smartidiot89]

FurMark this, OCCT that, fanboy there

How does the card actually work, I meen REALLY work? That is what matters not what some stupid test program does seriously. If the stockcooler can handle stockspeeds in real games then where's the problem?

[STEvil]

Can it handle the card at stock? Seems so for everything except maybe F@H assuming you dont have high room/case ambient temperatures.

[annihilat0r]

It can handle everything at stock but AMD begs you to overclock this card, so they produce a card that is not capable of being really overclocked but advertise it like it is. But of course, AMD can't do anything wrong. So I am talking in vain.

[cowie]

sry your not allowed to say ati and doing something wrong in the same sentence

[vab206]

I would rather have a card that can maintain it's advertised speeds without having to throttle down. I have never heard of people praising a feature that cuts performance drastically. It's not going to be a "few fps", it's going to be almost cut in half if it throttles to low power 3D speeds.

advertised speeds??? the whole point of that article was the problem that arises when trying to overclock the card to 5870 speeds, there shouldnt be any problems at advertised speeds.

[annihilat0r]

Advertised speeds: OC'ing to 5870.

[eXa]

So the really cool thing about the HD5970, is that not only is it the fastets graphics card in the world, but its also an extreme overclocker.

We designed alot of headroom into both the board and some special software features, to allow the end user and enthusiasts to really push this ting to the limits.

From this promotion video:
http://www.youtube.com/watch?v=tUn5qkF3tjY

[tajoh111]

I think we have to take things like furmark into consideration because we never know how stressful future programs and games will be. Also if it is not stable in something brief like OCCT or furmark, who is to say we are not causing longterm damage to the card over a longer period of time.

People saying we will never see loads like furmark simply want progress to be held back for the sake of their argument. Something like this program can easily be fixed if another phase of power was added to the card. When the cards are 600 a piece and such a add on would be insignificant to the overall cost, they should do it for the benefit of the customer, especially if these cards are being marketed to overclock.

Honestly, I think most people would love if a game was programed so well, that it was using 100% of the videocards power. This might be a tad unrealistic because its difficult to programs in such a manner when there are so many hardware configurations there are, unlike consoles.

However, I can see a program(not game), especially if open Cl does take off, which could be as stressful as furmark. If these cards ever want to be taken seriously in the supercomputer world, they better be beefed up because there will be programs that use 100%.

It was not until the 4870 series, which had crashing with furmark, that we started calling it a power virus on AMD accord.

However what is the point of a stress program if it does not show long term stability. A stress program has to prove stable in the short term at above average loads to show a piece of hardware can stay stable in the long term at lower loads. If elevator manufacturers only tested to the maximum rating, most of us would feel alot less safe in a crowded elevators. Or if tires were only tested up there speed rating, we would see alot more tires burning out.

Programs like Prime95 or Intelburn in, test beyond realistic loads, but do we call them power viruses? No we don't.

[LedHed]

very nice post, informative and non-biased

[Manicdan]

^^if they have programs that are able to utilize every shader perfectly, ATI can just drop the clocks in half for that program using drivers, and still stomp the competition.

[LedHed]

^^if they have programs that are able to utilize every shader perfectly, ATI can just drop the clocks in half for that program using drivers, and still stomp the competition.

I don't see why everyone in this thread tries to make it a competition of NVIDIA vs ATI. Why can't we just talk about thermal issues on the power regulator chips?

Also no one has any proof that ATI's current throttle mode doesn't drop the clocks like NVIDIA to 0.8mhz across the board or even low Power 3D speeds would severely hurt performance.

[gumballguy]

I think we have to take things like furmark into consideration because we never know how stressful future programs and games will be. Also if it is not stable in something brief like OCCT or furmark, who is to say we are not causing longterm damage to the card over a longer period of time.

People saying we will never see loads like furmark simply want progress to be held back for the sake of their argument. Something like this program can easily be fixed if another phase of power was added to the card. When the cards are 600 a piece and such a add on would be insignificant to the overall cost, they should do it for the benefit of the customer, especially if these cards are being marketed to overclock.

Honestly, I think most people would love if a game was programed so well, that it was using 100% of the videocards power. This might be a tad unrealistic because its difficult to programs in such a manner when there are so many hardware configurations there are, unlike consoles.

However, I can see a program(not game), especially if open Cl does take off, which could be as stressful as furmark. If these cards ever want to be taken seriously in the supercomputer world, they better be beefed up because there will be programs that use 100%.

It was not until the 4870 series, which had crashing with furmark, that we started calling it a power virus on AMD accord.

However what is the point of a stress program if it does not show long term stability. A stress program has to prove stable in the short term at above average loads to show a piece of hardware can stay stable in the long term at lower loads. If elevator manufacturers only tested to the maximum rating, most of us would feel alot less safe in a crowded elevators. Or if tires were only tested up there speed rating, we would see alot more tires burning out.

Programs like Prime95 or Intelburn in, test beyond realistic loads, but do we call them power viruses? No we don't.

A lot of that is what I said 10 posts ago... but just watch, your post will get overlooked, just like mine was...

At least there's a couple of people who get it

And yeah I'm really just quoting you to shove this post into peoples faces one more time. Hehe

[postumus]

A lot of that is what I said 10 posts ago... but just watch, your post will get overlooked, just like mine was...

At least there's a couple of people who get it

And yeah I'm really just quoting you to shove this post into peoples faces one more time. Hehe

I think if there are cards aimed at high performance computing they may address the issue then, these cards are aimed at gaming, and for gaming it's the best card on the market atm. As for future games, who is to say? Most games are console ports anyway. Also, I thought this was XS, I don't really expect ppl here to keep the card for years and years to come when some hypothetical game may or may not manage to load all the cores to 100%, which I personally don't think will ever happen anyway. It's a minor issue on a small niche card, I think Anand was looking to score some more hits with their amazing discovery.

[safan80]

I don't see why everyone in this thread tries to make it a competition of NVIDIA vs ATI. Why can't we just talk about thermal issues on the power regulator chips?

Also no one has any proof that ATI's current throttle mode doesn't drop the clocks like NVIDIA to 0.8mhz across the board or even low Power 3D speeds would severely hurt performance.

If Anandtech added a GTX295 to the article and tested it then we would be talking about the issue at hand. Flame bait makes great web site traffic

[Manicdan]

I don't see why everyone in this thread tries to make it a competition of NVIDIA vs ATI. Why can't we just talk about thermal issues on the power regulator chips?

Also no one has any proof that ATI's current throttle mode doesn't drop the clocks like NVIDIA to 0.8mhz across the board or even low Power 3D speeds would severely hurt performance.

im not trying to make it an ATI vs Nvidia thing. the point is when these cards are pushed 100W past their TDP, its not wasted energy, its actually being used for processing, which equals performance. so they can drop the clocks and keep power consumption within a safe range.

[Chumbucket843]

^^if they have programs that are able to utilize every shader perfectly, ATI can just drop the clocks in half for that program using drivers, and still stomp the competition.

gpgpu programs that use every shader (or get very close) on ATi hardware dont overheat the vrm's or anything else. that might be something worth noting.

[LedHed]

If Anandtech added a GTX295 to the article and tested it then we would be talking about the issue at hand. Flame bait makes great web site traffic

the 295 has no thermal problems though, in the dual or single PCB version.

Mine idles at 45C and never goes above 75C even after 4hrs of Crysis.

[InCredible]

the 5970 doesnt have thermal issues either at stock..
and AMD mentions that the card is very overclockable for ENTHUSIAST...which it is...enthusiast know that cooling is important and will be monitored/changed to adequately cool this beast of a card.

theres nothing "wrong" with the 5970..yea it'd be nice to have better cooling for the vrm's to withstand a higher overclock but really thats just asking for something extra..the card performs as it is intended too..and if you are an enthusiast and want higher clocks..then be one and put some water on it or wait for an aftermarket air cooler or do some modding to it..
not that u need it since it can play everything at stock clocks but if ur enthusiast u always want more and u know that u need things like better cooling more volts etc etc...
back in the day u couldnt expect 15-25% overclock on cpu's with stock cooling and volts..
l

[gumballguy]

I think if there are cards aimed at high performance computing they may address the issue then

I'm an engineer for a living and I will correct you - everything has a design life. If it isn't designed to meet that life, it doesn't work. Within the design life of this card, it is reasonable to expect that gpgpu will be stressing it out such that the cooler (which works for todays loads) will not work in the future. dnet is the first example, there will be others.

The future life of this card is longer than how long you have it in your hands. First of all, your personal circumstance is completely irrelevant, it doesnt matter how often you personally upgrade. Second of all - when you're not using it are you throwing it in the bin or reselling it ? Just because its not in your hands doesnt mean its design life is over.

Most games are console ports anyway.

Ahem. Thats rationalisation if ever I've heard it.

the 5970 doesnt have thermal issues either at stock..

AGAIN. People, please think about this in the time dimension. It works right now, and I haven't said otherwise. Its dubious in the longer term, and I've given two reasons why.

theres nothing "wrong" with the 5970..yea it'd be nice to have better cooling for the vrm's to withstand a higher overclock but really thats just asking for something extra..the card performs as it is intended too..and if you are an enthusiast and want higher clocks..then be one and put some water on it or wait for an aftermarket air cooler or do some modding to it..
not that u need it since it can play everything at stock clocks but if ur enthusiast u always want more and u know that u need things like better cooling more volts etc etc...
back in the day u couldnt expect 15-25% overclock on cpu's with stock cooling and volts..
l

For an overclock, rather than stock, I don't disagree with you. Some people are arguing about ATI's advertising of it as an overclocker, I'm not. I didn't expect it to overclock without a better cooling solution.





Personally I think AMD didn't revisit their assumption about what loads these cards will be under with gpgpu adoption at stock settings, when they should have. People tend to run gpgpu 24/7... how long is it going to last? My bet is 6-12 months.

PS: I'm trying to buy a 5870 vapor x for the improved cooling. Just cant get one in stock in AU for less than a $250AUD premium. Point being, there's no anti ATI in my camp.