[H] HD 7970 Dual-X Review

[Elfear]

We haven't gotten more stats from people in this thread, but so far, we only have a variance of a whopping 13Mhz from my card and the one other person who replied's base max boost. That's slightly over 1%. I think you're trying to make a problem where none exists.

There have been two or three cards compared in this thread. That is not an adequate sample size. As has already been discussed, [H]'s card hit 1200Mhz. Either they were sent a ringer by Nividia for their review or there are cards out there that boost a significant amount. With such a variable range of stock performance, it makes benchmarking very difficult and gives no true baseline of what the end-user can expect from the stock card. 1097Mhz to 1200Mhz is a 9% difference. That's more than a little variance.

 

[RussianSensation]

There's already plenty of links to show OC vs OC reviews and benches (max OC 7970 beat a +135% gtx680), you can dig them out of the OLD review thread yourself. This is beating a dead horse.

"The implementation of hardware monitoring and power/performance balancing doesn't negatively impact framerates though and at 1920x1080 the GTX 680 almost always exceeds the performance of a reference clocked 7970, in fact it tends to match or exceed the 7970 running at its maximum rate for air cooling." ~ Hardware heaven

On average GTX680 OC and HD7970 OCed trade blows. However, if it starts coming down to modern vs. older games, GTX680 wins in modern games while HD7970 wins in older games (Metro 2033, Crysis 1/Warhead).

In BF3, Witcher 2, Dragon Age 2, Batman AC, Old Republic, SKYRIM, Dirt 3, GTX680 sweeps the 7970 without any problems up to 1920x1200. At 2560x1600 both cards are too slow to provide playability in the most demanding games without a 2nd card. So it's a moot point. Considering HD7970 costs more in US, it's a non-starter.

Once aftermarket 680s launch, any overclocking advantage 7970 has right now will completely disappear. Here is a Zotac that hit 1400mhz on Air.

We already know it takes at least an 10500-1070mhz HD7970 just to match a stock 680 from MSI Lightning Reviews.

Also, AMD continues to focus heavily on texture optimization (which is an acceptable practice) but not when it reduces image quality and unfairly increases performance on HD7000 series of cards. It was criticized immensely for this when HD6870 was launched with cheating drivers and produced inferior texture image quality to HD5850/5870 with a 5-6% increase in performance as consequence and now this is happening again with HD7970:

"Considering the data we’ve seen up until this point, we have to come to the disturbing conclusion that AMD's Radeon HD 7000-series cards currently enjoy more aggressive benchmark results at their default driver settings, resulting in reduced texture quality compared to the Radeon HD 6000s and GeForce GTX 500s. Using the highest Catalyst A.I. setting appears to be the remedy, though it costs additional speed." - Toms Hardware

It looks like AMD released a driver that fixes this issue. However, shouldn't the image quality be almost a non-factor in 2012? Why does a professional review website have to go out of their way and investigate such a thing? Similar image quality should be almost a standard.

It's not whether the issue was fixed or not but because it has now happened with HD6870 and HD7870, it means we can't automatically assume that AMD won't do this again. That's disappointing.

you then do not understand what a baseline is

for it to be a baseline it needs to be utterly repeatable. on different cards, in the same game with the same settings. this prevents that. teh boost #'s in the [H] review were all over the place

All benchmarks have a variation of 1-3% as a result of margin of error. The fact that NV's GPUs Boost out of the box is how the consumer will get those cards. Therefore disable GPU Boost is counter to how 100% of GTX680 owners will use their cards.

On the other hand manual overclocking on an HD7970 or GTX680 is not something each user will perform. The fair comparison would entail:

1) Out of the box GTX680 vs. out of the box HD7970 with no manual overclocking at all

OR

2) Manual overclocking on GTX680 vs. manual overclocking on HD7970.

The baseline performance for any product is what performance the user will get after he takes that component out of the box and puts it in his/her system. Any speed above that speed which results from manual adjustments is overclocking.

Just wait until AMD implements the same feature and then all these arguments will be put to the backside. Dynamic overclocking is the future because it allows for a better balance between performance and power consumption.

There have been two or three cards compared in this thread. That is not an adequate sample size. As has already been discussed, [H]'s card hit 1200Mhz. Either they were sent a ringer by Nividia for their review or there are cards out there that boost a significant amount. With such a variable range of stock performance, it makes benchmarking very difficult and gives no true baseline of what the end-user can expect from the stock card. 1097Mhz to 1200Mhz is a 9% difference. That's more than a little variance.

If HD7970 with aftermarket coolers was $499, then we can start getting into these types of discussions. As it stands, it takes cream of the crop HD7970 that can reach 1250-1280mhz on air to beat a factory default 680. Since pretty much most 680s can hit 1200mhz+ with a GPU offset, that would bring 680 to parity with an OCed 7970, while retaining a cheaper price and lower power consumption and more features.

And then there is the case of after market non-reference 680s that will hit 1300mhz+. HD7970 is still a good card but until there is a price cut, it comes down to what games/programs you run specifically since an overclocked-on-air 7970 still cannot convincingly beat a manually overclocked 680.

Also, I think the debate is more for people running 2560x1600 screens. 1920x1080 or 1920x1200, it's not that close. GTX680 has a healthy 15%+ lead in more recent games at those resolutions.

I think it's best to buy based on the games/programs you run. Like if you play Anno 2070, Bulletstorm, Metro 2033, Crysis 1/Warhead, get the AMD card. If you play BF3, Crysis 2, Batman AC, Dirt 3, SKYRIM, get the 680. For current HD7970 users, there is no point in wasting $ to side-grade.

The choice is rather very simple. When AMD and NV had cards which were similarly priced (i.e., X850XT PE vs. 6800U), this is how we used to recommend them.

 

[Ferzerp]

I'm still suspect of the blurb in that first HardOCP report. I'd like to see them revisit that because they're the only ones who said anything about that, and it totally conflicts with the info they posted in their OC review (as in, their non-voltage changed, just offset raised OC clock in BF3 doesn't even reach 1200Mhz).

 

[Keysplayr]

I agree. This idea that you are getting something sturdier with no evidence, is a strange angle to debate. A faster card, that does it pulling less power, that only needs 2 6 pin connectors, should be viewed as a positive.
I'm pretty sure the gtx 480 with it's header heat pipes and ability to pull 300-400 watts was never thought impressive.
Or are we talking about AMD's card being longer ? I swear in poll's, length isn't important !
I'm impressed with the cards design /gpu design that allows gddr5 memory to be clocked stock at 1500mhz, highest ever.
I swear , I used to read that was important. Even without relevance to overall performance. But of course now, Nvidia is running the memory faster, so it's not.
And the memory o/c's further

It's like feeling victory that you have a higher red line/rpm shift point in a slower car.
It's a bad car analogy, but there it is.
I've already noticed the z77 Intel boards look barer in the voltage area's, and that's because they can be. They pull less power. Period.

I agree. Of course the arguments are strange. They "have" to be strange because that is all that remains. Strange arguments that are unfounded and fabricated in the hopes word travels virally through the web. A rumor becomes fact if you repeat it enough times in enough places, right?

7970s are great cards, there isn't any doubt about this. Anyone who has one is sure to be happy with it (I would imagine, I don't own one). The 680, however, is greater still. And this thread referencing a review that pits a severely o/c'd 7970 against a stock 680 is quite misleading. Not on the OP's part of course, who did change the thread title.

 

[Zanovar]

Just ordered this card,hope its a good clocker.

 

[Lonbjerg]

Doesn't the 680 overclock itself with GPU boost?

No, if you think the GPU boost is overclocking, you need to read up.

By that definitions speedstep is also O.C.

 

[Lepton87]

7970s are great cards, there isn't any doubt about this. Anyone who has one is sure to be happy with it (I would imagine, I don't own one). .

How is 7970 relevant when talking about GTX680 blocking the ability to control voltage? 7970 has nothing to do with that discussion. It only concerns REFERENCE GTX680. We don't yet know if custom 680s will allow voltage control. The inability to control voltage is a huge minus. PERIOD. It's perplexing to see so many people claiming otherwise. Hard modes were a thing of the past for normal overclocking I can't fathom why you so enthusiastically embrace regressing back to that "stone age"

I agree. Of course the arguments are strange. They "have" to be strange because that is all that remains. Strange arguments that are unfounded and fabricated in the hopes word travels virally through the web. A rumor becomes fact if you repeat it enough times in enough places, right?

Unlike you many people don't care who makes what.

 

[Ferzerp]

How is 7970 relevant when talking about GTX680 blocking the ability to control voltage? 7970 has nothing to do with that discussion. It only concerns REFERENCE GTX680. We don't yet know if custom 680s will allow voltage control. The inability to control voltage is a huge minus. PERIOD. It's perplexing to see so many people claiming otherwise.


Unlike you many people don't care who makes what.

Except they don't block voltage control. They just don't allow for going over 1.175V, but you can most certainly change the voltage.

 

[RussianSensation]

How is 7970 relevant when talking about GTX680 blocking the ability to control voltage? 7970 has nothing to do with that discussion. It only concerns REFERENCE GTX680. We don't yet know if custom 680s will allow voltage control. The inability to control voltage is a huge minus.

Ya, so let's wait until they launch. Beating a stock 680 is easy. HardOCP didn't even bother manually overclocking the 680 in that review. That's basically pitting a maximum overclocked 7970 on air against a stock 680. Let's see what happens when a max overclock is applied to a good after market 680.

Also, even without voltage control, a manually overclocked 7970 cannot convincingly beat a manually overclocked reference 680 (even without voltage control). So it's a moot point really, especially since 680s cost less.

No, if you think the GPU boost is overclocking, you need to read up.

By that definitions speedstep is also O.C.

Never thought I'd see the day where cream of the crop card from one brand would square off against a reference card of another brand to prove its superiority. Did you hear that GTX580 MSI Lightning with a maximum overclock destroyed a stock 6970?

In other news a review had a 2600K with an $80 aftermarket cooler @ 5.0ghz pitted against a stock FX8150 where Intel's setup wiped it off the face of the earth.

Interesting how dynamic factory overclocking based on load would be considered unfair and should be "turned off" even though it's a feature out of the box but top 98th percentile manual overclock on air for an $80 more expensive card that's largely a luck of the draw is considered almost guaranteed.

Next up, HD7970 on watercooling @ 1400mhz against an air cooled GTX685. More will be revealed in Q4.

 

[Lepton87]

Except they don't block voltage control. They just don't allow for going over 1.175V, but you can most certainly change the voltage.

It is the same because GTX680 hits 1.175V at stock.

http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review/4

but you are right voltage control isn't completely blocked, increasing the voltage is.

 

[BallaTheFeared]

I like how heated things get even though we're discussing the worst generational performance jump in memory.

 

[Keysplayr]

How is 7970 relevant when talking about GTX680 blocking the ability to control voltage? 7970 has nothing to do with that discussion. It only concerns REFERENCE GTX680. We don't yet know if custom 680s will allow voltage control. The inability to control voltage is a huge minus. PERIOD. It's perplexing to see so many people claiming otherwise.

Hehe. Still speaking in absolutes I see. Look, you are purposely trying to swing the conversation from poor PCB and components to voltage control. I don't blame you because that crashed and burned so now you must move on and find something else to complain about the 680. I get it.

You feel lack of direct voltage control is a huge problem. I respect your opinion and anyone else who feels that is the case. Voltage control is passive on the 680. Raise base clock and GPU offset, and you'll see higher voltage being supplies to the 680. IMHO opinion, you are really blowing it out of proportion as needed. The 680 seems to be overclocking just fine as is. As for non reference GTX680's, who knows. It may be exactly the same setup just with a lot more headroom and greater offsets and base clock adjustments available.

I'm telling you Lepton, don't back yourself into a corner as your comments can be referenced later on when AMD offers the same feature as Nvidia has done with the 680. It'll be RadBoost or some such name.
In many ways, no access to voltage is a good way to protect the product as well, costing companies less in warranties by over zealous overclockers. Pros and cons to everything.

 

[Ferzerp]

It is the same because GTX680 hits 1.175V at stock.

http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review/4

but you are right voltage control isn't completely blocked, increasing the voltage is.

When you change it though, you do set the new floor, so if the auto voltage adjustments aren't providing enough voltage at a given clock, you can provide more voltage. Sadly, you end up raising power usage in scenarios where you don't need the extra voltage though.

It is extremely limited in range, but you can affect it. Saying it is totally locked is misleading. Saying it has a very low limit is accurate.

What would be better would be to have access to the voltage curve so that manual adjustments could be made for given clocks/temps/power, but that would be more complex than most people could handle.

 

[chimaxi83]

Except they don't block voltage control. They just don't allow for going over 1.175V, but you can most certainly change the voltage.

This is incorrect. Whether I use EVGA Voltage Tuner to manually set max of 1.175V or minimum of .825V, the card will change voltage between 1.1V at idle and 1.175V at load on it's own. Changing voltage, even within max spec, is out of our hands.

 

[Elfear]

Ya, so let's wait until they launch.

Also, even without voltage control, a manual overclocked 7970 cannot convincingly beat a manually overclocked 680 (even without voltage control). So it's a moot point really, especially since 680s cost less.



Never thought I'd see the day where cream of the crop card from one brand would square off against a reference card of another brand to prove its superiority.

In other news a review had a 2600K with an $80 aftermarket cooler @ 5.0ghz pitted against a stock FX8150 where Intel's setup wiped it off the face of the earth.

Interesting how dynamic factory overclocking based on load would be considered unfair and should be "turned off" even though it's a feature out of the box but top 98th percentile manual overclock on air for an $80 more expensive card that's largely a luck of the draw is considered almost guaranteed.

Next up, HD7970 on watercooling @ 1400mhz against an air cooled GTX685. More will be revealed in Q4.

RS, I don't think you're getting it. No one is arguing that it was a fair comparison between an overclocked 7970 vs a stock 680. Go back and read the first page. The discussion changed to how to bench a 7970 against a 680 when GPU Boost is so variable. Some people were saying overclocking is YMMV and stock is stock. Well in this case it isn't. Now overclocking AND stock are YMMV. Makes these comparisons less meaningful IMO.

 

[BallaTheFeared]

Are you saying when you downclocks at idle and lowers voltage you've lost voltage control?

 

[Lepton87]

I'm telling you Lepton, don't back yourself into a corner as your comments can be referenced later on when AMD offers the same feature as Nvidia has done with the 680. It'll be RadBoost or some such name.
In many ways, no access to voltage is a good way to protect the product as well, costing companies less in warranties by over zealous overclockers. Pros and cons to everything.

I never criticized turbo boost it's a great feature for those folks who don't manually overclock their cards. It's usefulness for overclockers is questionable but it's not the problem. The problem is lack of direct control. And why would we consumers care for return rates? For us setting the hard ceiling for voltage doesn't have any advantages only disadvantages. I said GTX680 PCB is poor in relative terms it's not held up to the standards I expect from such an expensive card. I said that it is adequate for normal use.

 

[Keysplayr]

I never criticized turbo boost it's a great feature for those folks who don't manually overclock their cards. It's usefulness for overclockers is questionable but it's not the problem. The problem is lack of direct control. And why would we consumers care for return rates? For us setting the hard ceiling for voltage doesn't have any advantages only disadvantages. I said GTX680 PCB is poor in relative terms it's not held up to the standards I expect from such an expensive card. I said that it is adequate for normal use.

Are you able to overclock a 680? As a wise member pointed out, and you conceded to, direct voltage control isn't there, but passive control with a ceiling is. So, it isn't blocked, just more automated.
And, you should be concerned with return rates. It means more money for R&D, less overhead for a company when returns are lowered. Translates into better products in the long run. Are you a stockholder? I'm sure there are some on these boards who are, and would be happy with less of a return rate when dividends come to visit. It's a big world out there with a big picture Lept.

 

[Keysplayr]

RS, I don't think you're getting it. No one is arguing that it was a fair comparison between an overclocked 7970 vs a stock 680. Go back and read the first page. The discussion changed to how to bench a 7970 against a 680 when GPU Boost is so variable. Some people were saying overclocking is YMMV and stock is stock. Well in this case it isn't. Now overclocking AND stock are YMMV. Makes these comparisons less meaningful IMO.

Misleading. Anyone who buys a 680 is guaranteed a base clock of 1006MHz for reference model. There isn't any guarantee of how much boost there is for any given application though. You are guaranteed a minimum and that is all that's warranted. Anything else is a bonus.
By what I am reading, and by these standards, the GTX680 can never really be benched according to some here..

 

[RussianSensation]

Just ordered this card,hope its a good clocker.

:thumbsup: Congrats. Let us know your overclocked results. Looks like the card now has a $30 Off coupon. :awe:

 

[IlllI]

well i read 1 page of this topic. sorry i did. wish i could get my 5 minutes of life back. all those complainers and bashers. funny how they bash a topic b/c they are saying 'you cant compare an overclocked card to a stock one' which is ironic because i thought the 680 automatically overclocked it self too

 

[RussianSensation]

RS, I don't think you're getting it. No one is arguing that it was a fair comparison between an overclocked 7970 vs a stock 680. Go back and read the first page. The discussion changed to how to bench a 7970 against a 680 when GPU Boost is so variable. Some people were saying overclocking is YMMV and stock is stock. Well in this case it isn't. Now overclocking AND stock are YMMV. Makes these comparisons less meaningful IMO.

I get your point of view. The 680 is YMMV above 1058mhz. But all of them should boost to at least 1058mhz. I think it's reasonable to expect some 680 to get to 1084-1110 (or w/e the proper 13mhz increments are). Some might not. I don't think it's fair to say that reference 680s will go to 1200mhz frequently. That seems like an outlier in HardOCP's review.

Look at Computerbase. Their reference card hovered between 1033 and 1097 mhz.

If you actually look at the benchmarks, for the games you personally play, it will become clear that in games where HD7970 is faster, 1097 isn't enough to catch up to the 7970. While in games where GTX680 is tangibly faster, it would require a 20% overclock to the 7970 to just match the 680 in those instances. So even despite dynamic OCing, the conclusion doesn't really change. In games where HD7970 is faster (Metro 2033, Crysis 1/warhead, Anno 2070), an overclocked 680 won't ever beat an overclocked 7970. In games where 680 is faster, an overclocked 7970 won't beat an overclocked 680 either.

Because GTX680s performance varies more, we can just take more reviews into consideration to reduce the margin of error.

In your case, this shouldn't even be an issue since you have a huge overclock on your 7970 that 680 won't match on reference coolers. It'll be a long time before anything much faster than an HD7970 @ 1375mhz comes to the market. You might even be able to skip HD8790. Think about it, if AMD increases SPs to 2304 and raises clocks to 1200mhz, that would only match your card right now.

GTX680 isn't meant to be an upgrade for HD7970 users. People who purchased 7970 3 months ago shouldn't have any regrets. I think if Sapphire HD7970 OC drops to $499, it'll be a formidable contender. Prices for these cards vary widely worldwide though.

For example, here in Toronto we have Zotac 680 with 3 Assassin's Creed games for $539.99 vs. Gigabyte 7970 for $649.99.

However, in Europe and other parts of the world, 680s might cost more. I think it comes down to the pricing in your country, the games you play and how high of an overclock you think you can achieve.

 

[chimaxi83]

Are you saying when you downclocks at idle and lowers voltage you've lost voltage control?

Voltage control just isn't there, period. Not on reference cards. Here, some Afterburner screens for comparison. Three 3DMark11 runs, first one is undervolted, second is completely stock, third is overvolted and overclocked (the minimum voltage on this one was 1.087V as opposed to 1.1V for some reason). I used Afterburner, undervolt is -100, overvolt is +100. The voltage results are the same as in Precision, though Precision voltage adjustment lets you select an actual voltage. Still doesn't do anything.

undervolt with stock speed

stock voltage with stock speed

Afterburner overvolt and +155 offset overclock

 

[Ferzerp]

Voltage control just isn't there, period. Not on reference cards. Here, some Afterburner screens for comparison. Three 3DMark11 runs, first one is undervolted, second is completely stock, third is overvolted and overclocked (the minimum voltage on this one was 1.087V as opposed to 1.1V for some reason). I used Afterburner, undervolt is -100, overvolt is +100. The voltage results are the same as in Precision, though Precision voltage adjustment lets you select an actual voltage. Still doesn't do anything.

undervolt with stock speed


stock voltage with stock speed


Afterburner overvolt and +155 offset overclock

The app that's packaged with Precision is a different app with a slider from the minimum up to 1.2x volts, but it only works up to 1.175. When I set it with that app, it doesn't go below what I set.

 

[Ferzerp]

This is incorrect. Whether I use EVGA Voltage Tuner to manually set max of 1.175V or minimum of .825V, the card will change voltage between 1.1V at idle and 1.175V at load on it's own. Changing voltage, even within max spec, is out of our hands.

I wonder if it only works properly with the EVGA bios? It appears to do exactly what I claim it does with my evga card.