Radeon 5970 Overclocking: The VRM Temperature Bottleneck

[LedHed]

What exactly is your point? What do you think long periods of furmark would have done to my card on the stock cooler? Again, if you can afford a $600+ card you can afford better cooling. So this card has a temp bottleneck? So does yours. It's max clocks (hell lets say stock clocks) are still a better bang for your buck than a GTX295 with lets say your max clocks. Anyone who has seen my post history knows that I'm no ATI fanboy, not by a long shot, but stop comparing your 295 to a 5970. Just stop.

I keep saying over and over this isn't the 295 vs the 5970, it's strictly 5970 thermal issues when OCing on stock cooling. I thought many people would find this interesting and it may give a heads up if they want to OC to buy a better cooler.

The only reasons I've mentioned the 295 is because it's the only other dual GPU card on the market anywhere near the 5970 launch. Pretty logical in my opinion the 2nd best and the best temperature comparisons.

I guess you could say my point is ATI needs better cooling on the VRMs of the card, plain and simple. Nothing NVIDIA vs ATI, just a thermal problem.

[BababooeyHTJ]

I keep saying over and over this isn't the 295 vs the 5970, it's strictly 5970 thermal issues when OCing on stock cooling. I thought many people would find this interesting and it may give a heads up if they want to OC to buy a better cooler.

The only reasons I've mentioned the 295 is because it's the only other dual GPU card on the market anywhere near the 5970 launch. Pretty logical in my opinion the 2nd best and the best temperature comparisons.

I guess you could say my point is ATI needs better cooling on the VRMs of the card, plain and simple. Nothing NVIDIA vs ATI, just a thermal problem.

I think that a better comparison is GTX280. First gen vs first gen card with the same heat issues on the vrms or 4870 with all of the early cards with memory chips degrading also due to heat. Your card is pretty much a refresh of a lineup with similar issues. You really don't think that 5970 wont see a refresh? If I run furmark for a good long burst I'll see similar vrm temps with my uni-sink which performs about as well as the stock cooler for the vrms, maybe a bit better.

Long periods of furmark will kill any high end card over time., yours and mine included. Both of our overclocks are being held back be heat. Water cooling would gain the both of us some frequency across the board just like it would with 5970. The card is still pretty impressive performance wise even for the price. I don't see where there is a problem.

I agree, this isn't ATI vs Nvidia, most high end first gen cards seem to have these types of issues. Even though there is no issue sine all is well at stock clocks.

[InCredible]

I have a 4870X2 and have modified the fan profiles to keep the GPU's and VRM's under 75c with 20c ambients. When ambient temperatures rise in the summer I use a different profile... its a little louder but keeps the card stable (I have had crashing issues at 85c+).

i have one as well and never had any issues from crashing when it hit into the 90c+ ...now i did modify the cooler and took the stock fan out added a 120mm on top of the card to lower temps and get it silent which helped a lot...one thing it messed up was the vrm cooling so i had an issue there..now im on water..but ive had 127c vrm's and 90c core temps and still playing benching ..not the best i know but it did it and never crashed or saw performance decreases while gaming..
now its much better since its watercooled and never seen anything over 80ish vrm unless under furmark and core max at 50c..but this change was just in the past month or so and ive had the card since launch.


EDIT. for the arguements between Bababooey and led....baba u make a good point all Ledhed's overclocking is on refresh cards and not on the first generation(compare 4870 to 4890 for instance)

now maybe nvidia is smarter for not using their first generation of a gpu to make a dual gpu card since i think all their gx2/gtx295 etc were a refresh of the previous gpu's....while ati does their dual gpu on the first generation of the gpu..however it still works at stock like its suppose to and yes it can overclock at 1000mhz..we've seen this multiple times..just might have to change some stuff to make this work 24/7 etc..

ati could of made better cooling and maybe will even come out with a better cooled refresh but right now you get the best card with coolin for stock clocks for the most part...atleast when ur playin furmark.

[LedHed]

EDIT. for the arguements between Bababooey and led....baba u make a good point all Ledhed's overclocking is on refresh cards and not on the first generation(compare 4870 to 4890 for instance)

now maybe nvidia is smarter for not using their first generation of a gpu to make a dual gpu card since i think all their gx2/gtx295 etc were a refresh of the previous gpu's....while ati does their dual gpu on the first generation of the gpu..however it still works at stock like its suppose to and yes it can overclock at 1000mhz..we've seen this multiple times..just might have to change some stuff to make this work 24/7 etc..

ati could of made better cooling and maybe will even come out with a better cooled refresh but right now you get the best card with coolin for stock clocks for the most part...atleast when ur playin furmark.

Where did I list every card I've OCed? Assuming much are we?

Just a heads up I OCed the original ATI Rage 128 Pro and the original ATI 9500 Pro and the original NVIDIA 6800 just to name a few.

[Jamesrt2004]

Where did I list every card I've OCed? Assuming much are we?

Just a heads up I OCed the original ATI Rage 128 Pro and the original ATI 9500 Pro and the original NVIDIA 6800 just to name a few.

he meant on the 4 series

[LedHed]

he meant on the 4 series

ok well I also OCed my 4850 Crossfire setup before buying the 295.

[InCredible]

Where did I list every card I've OCed? Assuming much are we?

Just a heads up I OCed the original ATI Rage 128 Pro and the original ATI 9500 Pro and the original NVIDIA 6800 just to name a few.

wow dont take words so seriously..the cards u mentioned that overclocked great from nvidia were both refreshes...

[LedHed]

...I don't see any harshness in my post...

[InCredible]

it wasnt im just saying that when i respond its going to be to your direct post..and it just happens all the cards you mentioned were refreshes..thats all..never mentioned that it was all you've ever done in your entire life

[Calmatory]

carefull, i havnt seen a figure of if more 5800/5900s are being sold than high end g200s. would be interesting, even if ATI is selling out, are the selling more?

Didn't Dell buy loads of RV870 derivants from AMD?

[gumballguy]

one thing i didnt take into account is 24/7 gpu crunching and ...now like mentioned in my last post above..the 4870x2 hit 90c while gaming..probably while crunching as well...did it die in 6-12 months? im not making a point cause i dont know..just asking out of curiosity...if it did..then id say yea we can assume the 5970 would probably follow suit..however if it didnt...well...same thing itll probably follow suit..but im not sure of this since i dont crunch

Fair enough.

I think the real worry is gpgpu - what can it do?

The experiences with my mates is a lot of dead 4870's. No comments being made on 4850's / 4890's.

Well why not take that a step further and analyze why it is many ppl appear to have thrown their weight behind AMD? What has either company done that seems to have created a pro AMD "bias".

1. Snip
2. Snip
3. Snip
4. Snip
5. Snip
6. Snip

So instead of crying about how ppl appear to have developed a pro AMD bias, why not analyze why it has come about; what NV can do it to change it, assuming they care to.

Whilst I prefer ATI business practices lately, if they do something wrong, then they should be called out on that.. We'll leave it at "if" here - as that's what we've been debating in the thread.

and all this talk about how ATI is innovating and original why NVIDIA isn't is just customer bias. If no one remembers let me remind you, when the new 5870 came out NVIDIA still had the fastest card over a year after it was released (295) not until the 5970 hit the market (in very very small quantities) could AMD claim the performance crown back. It took AMD a full 13 months (even w/ GDDR5/DX10.1/DX11) after the 295 was launched to catch up with it.

Regardless of the number 13, once you've put your cards on the table, theres always going to be one winner. That winner will stay there until the new architecture. You should expect 12 months leads. Yeah yeah there's refreshes, but overall theres little change, especially in the top end. Top end often adjusts when x2's are released, I'm not counting that as its obviously going to be the case.

i dont think engines make a good comparison for this

I dont think car examples make a good comparison, period. Lol.

How do you know the VRMs are underbuilt?
I had some very nice posts in the OCCT thread that went unnoticed and were buried by the common sheeple, so I feel your pain.
...
No longterm study has been done on the longevity of these cards being stressed. Any comment made on that subject is pure baseless speculation with no evidence or facts.

Edit- Also AMD/ATi did take the blame on the Furmark/OCCT VRM problem on the 4 series. They said they didn't see the need to increase the cost and the price for the cards to statisfy an outlier of a situation and that has no effect on the actual operation/performance of the card.

For me, it is personal experience. 4870s seemed to die a lot. And I have made the assumption that the vrm's are the most likely candidate for the blame, largely as they get so hot. We do know operating temps for them should be <120. My understanding is that this is a short term maximum though, and lifetime of electronic parts decreases non-linearly at those kind of temps. Of course theres potential for misdiagnosis.

Do you think ATI did any long term testing? They'll have tried to emulate it to some extent. Nobody will know the truth for at least a year... I personally think a few dollars on a better cooling solution for them is common sense.

From Anand's article

so the answer is even Dnet allows the 5870 be OCed up to 850MHz as long as the fan is at 70%, if you can deal with 100% who knows what you can do. the only other cases with proven throttling are both moot points as far as i think anyone should be concerned.

I think that a better comparison is GTX280. First gen vs first gen card with the same heat issues on the vrms or 4870 with all of the early cards with memory chips degrading also due to heat. Your card is pretty much a refresh of a lineup with similar issues. You really don't think that 5970 wont see a refresh? If I run furmark for a good long burst I'll see similar vrm temps with my uni-sink which performs about as well as the stock cooler for the vrms, maybe a bit better.

Long periods of furmark will kill any high end card over time., yours and mine included. Both of our overclocks are being held back be heat. Water cooling would gain the both of us some frequency across the board just like it would with 5970. The card is still pretty impressive performance wise even for the price. I don't see where there is a problem.

Its the long term with gpgpu that's a bit scary. We dont know where it'll end up. Again, a few dollars for better cooling would have helped, don't you think?

I wonder if having a better stock cooler would make angry gpu partners. But overall I still think ATI needed to reassess expected gpu temps as gpgpu becomes more popular.




Again, the card works at stock at todays loads for a relatively short duration. I'm only questioning duration.

[grimREEFER]

might i add -
buy an xfx, push it hard, when it dies, you get a new one thanks to the lifetime warranty

[Smartidiot89]

might i add -
buy an xfx, push it hard, when it dies, you get a new one thanks to the lifetime warranty

afaik XFX "lifetime warranty" accounts for the whole life the product actually is alive on the market, once it's EOL that lifetime warranty is gone?

[DragoonXX]

you wont likely get the same generation card back, you'll get either an equivalent to its performance, or to its performance bracket

[LordEC911]

For me, it is personal experience. 4870s seemed to die a lot. And I have made the assumption that the vrm's are the most likely candidate for the blame, largely as they get so hot. We do know operating temps for them should be <120. My understanding is that this is a short term maximum though, and lifetime of electronic parts decreases non-linearly at those kind of temps. Of course theres potential for misdiagnosis.

Have they not been using the same/similar components for the last 3-4 years, if not longer? These components are tested and rated to run at certain specifications. AMD/ATi's job is to make sure those specs align with what they plan on using them for.

You seem to think AMD/ATi is doing/using something new...

[gumballguy]

Have they not been using the same/similar components for the last 3-4 years, if not longer? These components are tested and rated to run at certain specifications. AMD/ATi's job is to make sure those specs align with what they plan on using them for.

You seem to think AMD/ATi is doing/using something new...

The parts are the same but they'll be pushed harder than before by gpgpu. So yeah sure we know what the capacity is, but we dont know what the load is. You need both capacity and load to know whether something works or not.

I did query if their view of capacity & load in the last generation was correct as well, thats what all my 4870 vrm babble was about.

But as to your last sentence - no.

[Solus Corvus]

I'm a bit disappointed that AMD made the same mistake that they made with the 4000 series. The furmark issue was one of their biggest pr blunders of that generation and I had figured AMD is a company that can learn from its mistakes. Maybe they will finally learn with the next generation.

But, IMO, the VRM problem isn't as big of an issue as some people are trying to make it out to be. For the vast majority of users this is going to be a total non-issue. It plays most games and runs most GPGPU apps just fine, even overclocked. For those who are concerned about running these particular apps or long term reliability then they have plenty of other options. They can wait for a 4 phase version, get 2 5870s, watercool it, or save their money for something else. Nobody is making you spend a huge wad of cash on something you feel is inadequate.

Also, comparing a 295 to a 5970 is laughable. Try comparing a 295 to a 5870 instead. They have about the same performance but the 5870 costs less, doesn't have multi-gpu issues, dumps less heat, consumes less power, overclocks higher, also has 4 VRM phases, and software voltage control - not to mention supports DX11, etc.

[kadozer]

I'm a bit disappointed that AMD made the same mistake that they made with the 4000 series. The furmark issue was one of their biggest pr blunders of that generation and I had figured AMD is a company that can learn from its mistakes. Maybe they will finally learn with the next generation.

But, IMO, the VRM problem isn't as big of an issue as some people are trying to make it out to be. For the vast majority of users this is going to be a total non-issue. It plays most games and runs most GPGPU apps just fine, even overclocked. For those who are concerned about running these particular apps or long term reliability then they have plenty of other options. They can wait for a 4 phase version, get 2 5870s, watercool it, or save their money for something else. Nobody is making you spend a huge wad of cash on something you feel is inadequate.

Also, comparing a 295 to a 5970 is laughable. Try comparing a 295 to a 5870 instead. They have about the same performance but the 5870 costs less, doesn't have multi-gpu issues, dumps less heat, consumes less power, overclocks higher, also has 4 VRM phases, and software voltage control - not to mention supports DX11, etc.

eyeinfinity +1 to add

[Manicdan]

I'm a bit disappointed that AMD made the same mistake that they made with the 4000 series. The furmark issue was one of their biggest pr blunders of that generation and I had figured AMD is a company that can learn from its mistakes. Maybe they will finally learn with the next generation.

they did fix the issue, the 48xx cards would crash due to overheated VRMs, these ones dont let you overheat them. its not ATIs fault your running a card 100W higher than what they rate its TDP for. (has anyone even shown us what the max power draw is before they throttle?)

the only thing adding better VRMs will do is let you play furmark, and let them loose alot more margin on their cards, which will probably cause even more price jumps people seem to love to hate.

[Solus Corvus]

they did fix the issue, the 48xx cards would crash due to overheated VRMs, these ones dont let you overheat them. its not ATIs fault your running a card 100W higher than what they rate its TDP for. (has anyone even shown us what the max power draw is before they throttle?)

the only thing adding better VRMs will do is let you play furmark, and let them loose alot more margin on their cards, which will probably cause even more price jumps people seem to love to hate.

It throttles instead of crashing, that's a partial fix. But if they just added an extra phase per gpu then it wouldn't throttle either. If I was spending a large amount of cash for this card I would definitely prefer the latter solution to the former - even if it meant ~$20 higher price. I would prefer that because, as an overclocker, it would mean I could take the card to higher clocks before reaching its limits.

I have a 5970 in my system right now (just testing, it's not mine). The performance, even at stock, in games is nothing short of astounding. IMO the people making a big stink about the VRM limitation are just scratching around to find something, anything, wrong with the fastest video card ever made (thus far). As I already stated I think this is basically a non-issue for the vast majority of users. Even mildly overclocked it runs almost all apps just fine with a bump in fan speed. But I still wish they had increased the VRM capacity - these cards have so much potential that it's really a shame to limit them like that.

Also, I imagine that ATI omitted the extra phase for market positioning reasons, not cost. They probably have a 5990 waiting in the wings with high clocks and 4 phases. It'd be the perfect tool to deflate the enthusiasm when fermi launches.

[Manicdan]

good solid post, people should learn to debate like that more often.

i also agree, it would be very nice if they did put more into the VRMs, (and it sounds like a great way for some of their partners to get a jump on the bandwagon if ATI wont)

[Splave]

so play games with the card, bench with the card. It wasnt made to fold/gpgpu.
Thats like saying my microwave is fail because it doesnt do a good job washing my clothes. It wasnt meant to.
Wait for a firegl variant that you cant overclock and there will be no issues. just MO

[Solus Corvus]

good solid post, people should learn to debate like that more often.

i also agree, it would be very nice if they did put more into the VRMs, (and it sounds like a great way for some of their partners to get a jump on the bandwagon if ATI wont)

I as well wish more people understood inductive and deductive logic. It would make my average internet debate much more interesting.

I don't think ATI will let the partners make any changes to the 5970 till a later date. But the 5850 might not be under the same restriction. The 5850 has a place for an extra phase. If a partner made a 4 phase version then you could probably pick up 2 for roughly the price of a 5970. Then you'd just have to decide what matters to you more, the extra shaders and a single card or higher clock potential and VRM capacity in two cards.

so play games with the card, bench with the card. It wasnt made to fold/gpgpu.
Thats like saying my microwave is fail because it doesnt do a good job washing my clothes. It wasnt meant to.
Wait for a firegl variant that you cant overclock and there will be no issues. just MO

Except it was made to do GPGPU, among other things. It says so right on the back of the box.

[initialised]

so play games with the card, bench with the card. It wasnt made to fold/gpgpu.
Thats like saying my microwave is fail because it doesnt do a good job washing my clothes. It wasnt meant to.
Wait for a firegl variant that you cant overclock and there will be no issues. just MO

Not exactly, it's more like saying I'd rather use a conventional oven than a microwave for some things but for speed it wins every time.

Anyway, I was testing two of these today, one on an open bench with a 4GHz CPU I'm not allowed to mention and 24GB of RAM and refused to OC to anything like 850/1200, while one in an under-ventilated Cosmos with an i7 960 @ 3.8GHz had no problem running at 875/1250. Both were on P6T Deluxe V2 powered by Tagan 1.1kW units. So yeah, your mileage may vary with these cards, I'm doubting that it's a thermal limit given the success of the card in the case over the card on the bench. No throttling, just a system restart, probably power related.

[LedHed]

Also, comparing a 295 to a 5970 is laughable. Try comparing a 295 to a 5870 instead. They have about the same performance but the 5870 costs less, doesn't have multi-gpu issues, dumps less heat, consumes less power, overclocks higher, also has 4 VRM phases, and software voltage control - not to mention supports DX11, etc.

tell me another GPU with dual GPU on a single card that was made within a year of the 5970. No body said it was a perfect comparison, but it's the closest one we have right now. You can't compare the 295 with the 5870 because a dual GPU card produces more heat, this is why I compared the 295 to the 5970. There just aren't any other cards out to compare to that are dual GPU besides the 295; I guess you could also compare it to the 4870X2, but we know how hot they got also.