nVidia: "We expected more from the 7970"

[lavaheadache]

Although HD7970 (default 925MHz) produces more avg fps it has high dips when GTX580 frame rate is more constant and perhaps GTX580 can have smoother gameplay because of that.

I imagine there isn't gonna be very many people who conclude the same thing as you by looking at those graphs

 

[Munky]

I don't think that will happen. 3XX mm2 vs 5XXmm2 on the same process, GCN would have to be hugely more efficient than kepler for that to happen. GCN is a step backwards in terms of efficiency per mm2 compared to VLIV4, it's not that surprising, VLIV4 was focused solely on graphics and GCN is a jack of all trades. (compute and graphics)

You mean kinda like the AMD 4800 series was much smaller, cheaper, cooler, and less power hungry than the Nvidia 200 series GPU, while offering 90% of the performance in the high end, and beating them soundly in the midrange? Yeah, that could never happen....

 

[MrK6]

I'll keep my comments to two parts of your book.

1. My entire point was that civ5 is entirely relevant to use as a benchmark. It is a AAA game, it uses a lot of advanced dx11 features, and everything that it uses is relevant to either improving IQ, FPS, or both. You didn't say anything at all to refute that. You earlier stated that nvidia somehow convinced firaxis to implement a bunch of nvidia-specific improvements into civ5 for their own advantage, yet you and I both later linked to an article which refuted that claim. Are you now saying that nvidia did nothing sneaky here, but that their driver team was more on the ball than AMD's at implementing MTR, or do you still somehow maintain that nvidia is at fault for AMD's incompetence on this issue?

If you're going to insist on using hyperbole this discussion isn't going to go anywhere. It's no secret Civ V developers have worked with both NVIDIA and AMD to implement features into their game. The fact that NVIDIA sells the game on its site and bundles it with its cards leads me to believe there's a pretty decent partnership there. Again, if the game is using a feature that only one manufacturer supports, what's the point in testing it? Remember how up in arms NVIDIA fans got when tessellation only worked on 5xxx cards? Where's that faux rage now? If it's fair for one instance, it's fair for all. That's what I'm saying.

because people don't want to be told to go out and buy a new graphics card if you want to play a game properly. i buy games on sale on steam all the time that i wouldn't normally buy. i'd be pissed if i found out my $800 worth of video cards can't run the game maxed out with no slow down compared to even a $300 nvidia card. </sarcasm on prices>

Your $800 worth of graphics cards are still going to blow away this game: http://www.anandtech.com/show/5261/amd-radeon-hd-7970-review/24 . The 7970 is still faster than the GTX 580, MTR or not. The point is in the comparison.

 

[BallaTheFeared]

tl;dr edit

 

[SirPauly]

I am pro-nVidia, at this time, and my views have been always positive about tessellation from the Radeon 8500 with hardware n-patches to the DirectX 11 5870; to nVidia raising the bar with tessellation and with the HD 7970.

Any title that offers tessellation with the help of IHV's are welcomed to me, including Crysis 2, and great to see the HD 7970 handle that title splendidly.

 

[Vesku]

I actually believe NVIDIA will have some good 28nm product WHEN they finally release it. They have a lot of experience from 8800->580 and the Radeon 7970 shows that TSMC 28nm can provide a nice bump in density and clockspeed. Will the yields and price TSMC charges result in a 28nm tweaked GTX 580 @ 900+MHz $300 card? Anything less will be below my expectations. ;p

 

[Quantos]

I imagine there isn't gonna be very many people who conclude the same thing as you by looking at those graphs

What? No, that makes perfect sense. My card gets 0 FPS all the way in that game, it's perfectly smooth!

 

[MrK6]

The 480 isn't a bad card, are we really having this discussion?

Was the i7-920 a bad chip because you could only get 3.2Ghz with the stock cooler? Is the i5-2500k a bad chip because it hits 80C at 4Ghz with the stock cooler?

Was the 6970 a bad card because the 480 was cheaper and provided the same overall performance while offering drastically more features and tess performance?

Because CPU's have any relevance in this discussion. Also, my buddy gave me a 6990 for $10 and a stick of gum, so I'm going to use that as the basis of the 6990 for all future comparisons, ok? Honestly, some of the points made here are ludicrous. The GTX 480 always had a more expensive MSRP, used 40-50% more power, less vRAM, no triple monitor support, and was one of the loudest video cards on the market. The way you consistently omit relevant information, cherry pick benchmarks, and run or hide from defeating arguments is really quite amazing.

The 5870 was priced lower than the 7970 is, while beating the 285 by more than the 7970 beats the 580. You can make any old excuse you want, the fact remains the reasoning for ignoring the benefits of Nvidia over AMD no longer include performance per dollar with competitive release prices on competitive products. The 580 wasn't even mentioned, you're thinking of the 480.

The fact remains that NVIDIA has yet to put out anything on 28nm, so that makes your whole argument ridiculous, doesn't it? I'm glad you think you're a good enough armchair engineer that you can decide what is an appropriate performance jump to 28nm at each point in the processes's development. Really, why AMD or NVIDIA hasn't hired you yet is beyond me. Also, you last point is completely incorrect, since the 7970 currently offers the best bang-for-your-buck among flagship GPU's.

I didn't pull any numbers Sherlock, I used the wrong one. There is a difference and it would be nice if you could figure that out on your own without me having to explain it. However the difference between 6%, which is what I said, and 11%, which is what I meant doesn't change the logic behind anything I said. Point of fact is you were unable to actually discuss what I said, and instead used a common board tactic of picking irrelevant points to discuss.

Oh, so now when you get called out on your blatant lies, you're just going to be insulting and try to deflect from your incompetence. Nice try, not going to fly around here :thumbsup:.

I don't think AMD is moving as many 7970 as you want to believe. I don't really care what Nvidia is doing with their last gen product that was overpriced in it's own right. The state of the gaming industry isn't as such where releasing a few months early is going to make or break anyone. Are you seeing the same benches I am? Everyone's getting playable FPS, only tools and idiots assume their gaming experience changes because they have paid out the wazoo for the latest and greatest.

At least the 5xxx series offered the first DX11 gpu, which was a selling point people are still waiting to matter. 7xxx doesn't bring anything more than a few more fps.

So now that your smear campaign against the 7970 has been soundly defeated in every inane iteration you've come up with over the last week, you're going to switch the argument, yet again, to "oh well we don't need that kind of power anyway." Please, what a joke.

@Lonyo Your opinion doesn't make things fact.

Right, he should just outright lie and cherry-pick information to back it up, amirite?

Would it make sense for me to run around comparing my 950 core 470s to factory clocked 7970s? No, it really wouldn't air makes more sense than comparing water clocks, but even still I use the 1300 core water cooled results.

Except that's what you already did. Remember this thread: http://forums.anandtech.com/showthread.php?t=2221009&page=2 , the one where you insisted your GTX 470's were faster because you only compared them to a lightly overclocked 7970. So now that the shoe's on the other foot that's OK? Once gain, lying, changing the argument, and hypocrisy.

Also 38 mins in a competitive FPS isn't really what I'd call acceptable for play, nor are 21 stock mins and 29 overclocked mins in Deus. Having those types of requent fps drops would detract from the overall game experience more so than lower avg fps. Not that the 580 is producing ideal fps, it is however from what I've experienced producing a smoother gaming experience.

And yet you completely negated the advantages of minimum FPS improvement when the 7970's minimums destroyed those of your GTX 470's. Interesting, again, a complete reversal of opinion once the shoe's on the other foot. I hope everyone reading this thread sees the hypocrisy going on here and why ballathefeared is really posting.

In Metro 2033 the graph shows the actual fps in solids, the bench reads something different and produces stupid numbers like mins lower than you'll see with any fps recorder and max higher than anything you'll see too.

Yet just above, you held minimums on [H]'s graph as gospel. So now only certain application's minimums count, especially if they back up your (inane) points? Interesting.

 

[BallaTheFeared]

Because CPU's have any relevance in this discussion. Also, my buddy gave me a 6990 for $10 and a stick of gum, so I'm going to use that as the basis of the 6990 for all future comparisons, ok? Honestly, some of the points made here are ludicrous. The GTX 480 always had a more expensive MSRP, used 40-50% more power, less vRAM, no triple monitor support, and was one of the loudest video cards on the market. The way you consistently omit relevant information, cherry pick benchmarks, and run or hide from defeating arguments is really quite amazing.

The fact remains that NVIDIA has yet to put out anything on 28nm, so that makes your whole argument ridiculous, doesn't it? I'm glad you think you're a good enough armchair engineer that you can decide what is an appropriate performance jump to 28nm at each point in the processes's development. Really, why AMD or NVIDIA hasn't hired you yet is beyond me. Also, you last point is completely incorrect, since the 7970 currently offers the best bang-for-your-buck among flagship GPU's.

Oh, so now when you get called out on your blatant lies, you're just going to be insulting and try to deflect from your incompetence. Nice try, not going to fly around here :thumbsup:.

So now that your smear campaign against the 7970 has been soundly defeated in every inane iteration you've come up with over the last week, you're going to switch the argument, yet again, to "oh well we don't need that kind of power anyway." Please, what a joke.
Right, he should just outright lie and cherry-pick information to back it up, amirite?
Except that's what you already did. Remember this thread: http://forums.anandtech.com/showthread.php?t=2221009&page=2 , the one where you insisted your GTX 470's were faster because you only compared them to a lightly overclocked 7970. So now that the shoe's on the other foot that's OK? Once gain, lying, changing the argument, and hypocrisy.
And yet you completely negated the advantages of minimum FPS improvement when the 7970's minimums destroyed those of your GTX 470's. Interesting, again, a complete reversal of opinion once the shoe's on the other foot. I hope everyone reading this thread sees the hypocrisy going on here and why ballathefeared is really posting.
Yet just above, you held minimums on [H]'s graph as gospel. So now only certain application's minimums count, especially if they back up your (inane) points? Interesting.

It's relevant because it's not an uncommon occurrence and wasn't an exclusive deal unlike your story about your "boy friend". It's also relevant because nobody expects to get the most out of their gpu with stock cooling, unless of course you had a 5xxx or 6xxx AMD, then it's the metric by which all standards are held. Then start drumming the horns of overclocking since those cards did so poorly for the most part all the while having downplayed the 4xx and 5xx series OC ability.

Less vram, don't care. No triple support for one gpu, don't care it's not typical like 1600p isn't in a group which already represents a fraction of the overall market.

I have better things to do than spell out everything for the unfortunate few who don't know or can't google that information for themselves if they feel they need it.

More like 20%, but then again you're used to making things up and benching with incorrect settings anyways. There is a reason your results are inconsistent with the rest of the internet, and it's not because you were the only one who overclocked the ram (you aren't btw).

No it doesn't make anything ridiculous, if Nvidia can't do better than the 7970 there is no reason for me to upgrade because what's already out now is considerably slower than what I currently have. You last point is oddly misplaced because AMD now has one of the worst price/performance cards on the market. But then you know with AMD you get a gpu, that's it. Graphics Processing Unit. With Nvidia you get PhysX, you get CUDA, programs to actually use CUDA, and other features that AMD simply doesn't have. So it's a GPGPU compute unit vs Graphics card still.

Flys just fine when people are capable of clicking links, sadly you must not be one of those elite people.

I'm not on a smear campaign, though I'm not going to sit here and listen to people talk like it's the second coming while faking their benchmark results.

He should fake bench results, then talk about the 7970 being twice as fast if not more than 6970 CF just because in one bench it was faster than his poorly clocked 6950.

1125 is something some reviewers couldn't even get, and what I said stands true, just like when my 470's beat the card you got back after paying the 15% restocking fee for your replacement and putting that on water which clearly isn't a typical scenario since your first one didn't clock well hence you took it back herp durp. It was still 23% faster than your water cooled replacement 7970 @1300 in Shogun 2. That's assuming you maxed everything and used 4xAA, but I'm only guessing since you have a strange habit of omitting your settings making your results impossible to question without actually having a 7970 to compare.

When has the 7970 ever beat my MIN fps? When have I ever said that didn't matter? You got proof, you were able to link things prior to this, yet failed to do so here.. Why is that?

Are you so affected you can't read now? The benchmark doesn't produce the same min and max as the graph shows, nor does it produce the same min/max as the fraps will show. As far as I know [H] doesn't use the benchmark results for their reviews, and if they even use the benchmark itself, they'd use fraps and not the readout the benchmark reports.


Well played sir.

Everyone, please thank Balla here for ruining this thread. This was an excellent example of how not to constructively post in a technical forum.
-ViRGE

 

[taltamir]

I imagine there isn't gonna be very many people who conclude the same thing as you by looking at those graphs

every serious FPS player knows that only minimum framerate counts. not average.

 

[GaiaHunter]

Just to get this straight, we are talking about Launch prices.

GTX285 was launched at $400
GTX295 was launched at $500
HD5870 was launched at $379 and it was 30% faster than GTX285
GTX480 was launched at $499 and it was 40-50% faster than GTX285 and 5-10% faster than GTX295

HD6990 was launched at $699 and it was ~60-70% faster than HD6970 on avg
GTX590 was launched at $699 and it was ~50% faster than GTX580 on avg

If the GTX580 costed $300 and the GTX590 costed $470 the 7970 at $550 would be silly.

Actually people are saying that a GK104 at $300 with GTX580 performance will be a big success and I agree.

Market conditions can't be ignored.

 

[Lepton87]

More like 30%, but then again you're used to making things up and benching with incorrect settings anyways. There is a reason your results are inconsistent with the rest of the internet, and it's not because you were the only one who overclocked the ram (you aren't btw).

Seriously? Those graphs are comparing total system power consumption! You need to subtract platform power consumption to isolate the cards. In that case the rest of the computer draws about 150W so it's 169W vs 271W so GTX480 draws 60% more power than 5870. Alternately you can just use CF and SLI numbers.

CF power draw - single and SLI-single which gives similar figures.

 

[blackened23]

Although HD7970 (default 925MHz) produces more avg fps it has high dips when GTX580 frame rate is more constant and perhaps GTX580 can have smoother gameplay because of that.

Is this a joke? I hope it is because all 3 lines are proportional to each other Or more succinctly as someone who has used both, no. Comparing single card performance: Playing Batman: AC at 2560x1600 with high tess, DX11, very high everything, physx off, on a gtx 580 oc'ed results in framerates of 30 or lower from the getgo. 7970 oc'ed : 60 fps or higher solid. This is comparing directly to each other in the first act with Catwoman.

 

[BallaTheFeared]

Seriously? Those graphs are comparing total system power consumption! You need to subtract platform power consumption to isolate the cards. In that case the rest of the computer draws about 150W so it's 169W vs 271W so GTX480 draws 60% more power than 5870. Alternately you can just use CF and SLI numbers.

CF power draw - single and SLI-single which gives similar figures.

Sorry

There is a 100 watt difference in power draw between 6970 CF and 480 SLI.

I don't know how well 480 scale with voltage/clocks, but a single 6970 @ 1000 core pretty much increase overall power draw by close to 100 watts for an extra 40Mhz.

I never really looked too far into it, as I don't really care.

100 watt difference is 10 hours of gaming per day costing the 480 sli user an extra 6 cents for that day.

 

[OILFIELDTRASH]

Is this a joke? I hope it is because all 3 lines are proportional to each other Or more succinctly as someone who has used both, no. Comparing single card performance: Playing Batman: AC at 2560x1600 with high tess, DX11, very high everything, physx off, on a gtx 580 oc'ed results in framerates of 30 or lower from the getgo. 7970 oc'ed : 60 fps or higher solid. This is comparing directly to each other in the first act with Catwoman.

Just looking at that chart that was posted it looks like the 580 would provide just as good gaming experience. The dips down into the 20's would be annoying. Especially if I had just paid $550.

 

[sontin]

Actually people are saying that a GK104 at $300 with GTX580 performance will be a big success and I agree.

It will not be a big success. It's only the progress in the business that every 18 months you get the same performance for half the price - Moore's law for perf/$.

But AMD didn't lower the price or will increase the performance. They give you exact the same performance. So from a performance standpoint there is no improvement...