nVidia: "We expected more from the 7970"

[KCfromNC]

That's all ya can do is compare with what is offered and another example of the importance of execution. AMD is placed in a nice position and can price their small die in monolithic price-points! Good for AMD and for enthusiasts and early adopters that desire and demand the very best. One can make the case, that AMD was too committed to the sweet spot strategy with the 5870 and 5850 and left too much revenue, profits and margins on the table.

Can't argue this - AMD seems to be more nimble and has caught nV off-guard in the past 3 or 4 generations. Now we just have to wait another 3-6 months for nV to respond, assuming history holds.

My biggest concern is that instead of rumors about performance, price, schedule, whatever from nV about their next gen we're seeing lots of FUD trying to tear down what's obviously a good launch from AMD (not from you, just in general). That makes me worried that they don't have much positive to share about the card, which means even less competition when the hardware does arrive. But I could be reading too much into their relative silence.

 

[kidsafe]

If the 480 was a bad card because it was only x amount faster than the cheaper 5870 than the 7970 is bad for the same reason. If the 480 wasn't a bad card and you paid for the fact that it was the fastest card than the 7970 isn't a bad card either.

No, the GTX 480 was a bad card because it was all of the following:

Six months late
$500
So hot, it needed a giant exposed metal heatsink in the reference design (this got burning hot to the touch)
by extension louder
not shipped with Nvidia Surround driver support
only marginally faster than a much less expensive and cooler running 5870

--

Now the GTX 580 fixed many of those issues, but you specifically mention the 480. The HD 7970 is a bigger upgrade over the GTX 580 than the GTX 480 was over the HD 5870. The HD 5970 being available for $500 also presented an interesting alternative.

The HD 7970 also has none of the disadvantages of the GTX 480 other than retail price.

As I already mentioned, timing plays a bigger role than you might believe. If I needed a video card 6 months ago I would have bought a GTX 580.

 

[blackened23]

On the contrary, I think we will all be using GPU's with a few small and efficient cpu cores a few years from now . Both AMD and Nvidia are pushing that way.

Dude, that would really suck. Intel will not make high performance graphics like AMD/nvidia, so we're stuck with intel for high performance CPU's.

What about the GPU? Intel has no reason to push for high performance GPU's. If iGPU's take over and put amd/nvidia out of business, we're screwed. We're already seeing that sandy bridge is
stealing SUBSTANTIAL revenue from both companies! It really does suck for enthusiasts like us - amd and nvidia will not be able to sell mid range or low end discrete cards. That may well put
both out of business, or force them to seek profits in other ways.

Again, it sucks.

 

[bryanW1995]

A nvidia employee who was speaking on their behalf at a trade show-- Sounds like nvidia did say something. At an event like this if you're wearing an nvidia t-shirt, you are speaking for your company.

While I like (and own) NV's products I don't like their tactics. Smearing the competition and making exaggerated claims is pretty unclassy, they did this same sort of thing when 5870 was released - it was nonsense then and its nonsense now.

Would you prefer that they go the AMD route and say "we're going to let the other team win, but we'll be a pretty good #2"? AMD cleaned house a few months ago, and I have zero doubt that a major reason why they did it is that they don't want to play 2nd fiddle any longer. One of the things that I like about Read is his competitiveness/desire to be #1, I'm certainly not going to criticize JHH and his company for having that same trait. Besides, do you really think that even if NV does just a simple die shrink with zero architectural improvements (highly unlikely) they won't still be able to smoke a stock 7970? AMD might come out with a higher clocked premium card, maybe a 7975, when/if kepler finally gets here, but what if they don't? I wouldn't be surprised at all to see 7970 get smoked by a gtx 670 right now. The million dollar question is, will that happen in 6 weeks or 6 months?

 

[BallaTheFeared]

No, the GTX 480 was a bad card because it was all of the following:

Six months late
$500
So hot, it needed a giant exposed metal heatsink in the reference design (this got burning hot to the touch)
by extension louder
not shipped with Nvidia Surround driver support
only marginally faster than a much less expensive and cooler running 5870

--

Now the GTX 580 fixed many of those issues, but you specifically mention the 480. The HD 7970 is a bigger upgrade over the GTX 580 than the GTX 480 was over the HD 5870. The HD 5970 being available for $500 also presented an interesting alternative.

The HD 7970 also has none of the disadvantages of the GTX 480 other than retail price.

As I already mentioned, timing plays a bigger role than you might believe. If I needed a video card 6 months ago I would have bought a GTX 580.

I think the 480 issues, personally are grossly overstated, as is the performance increase of the 7970 over the 580.

Nvidia stated they expected more...

Here's another bench, 470s in SLI @ 900 core, nothing crazy...

Overclocked 470s are 59% faster than a stock 7970, and 40% faster than a 7970 clocked at 1125/1575... In a test where the 7970 is 12% faster than a 580. Granted this is 1080 vs 1200, but the difference isn't enough to account for the disparity in performance.

Hence why Nvidia "expected more". Which is the point of this thread, Nvidia wasn't trying to beat something that was only 6% faster than the 580 at 1200p on avg.

Whatever else we discuss in this thread is pretty much OT.

 

[kidsafe]

Evolve or die. As long as the market exists for high-end graphics, someone will be there to provide it. Discrete GPU development may slow, Nvidia and AMD may put even more focus on the two extremes: portable computing and HPC

Sure AMD Trinity/Fusion and Nvidia's Tegra initiatives could fail, but that is not anyone's fault but their own.

 

[Arzachel]

Dude, that would really suck. Intel will not make high performance graphics like AMD/nvidia, so we're stuck with intel for high performance CPU's.

What about the GPU? Intel has no reason to push for high performance GPU's. If iGPU's take over and put amd/nvidia out of business, we're screwed. We're already seeing that sandy bridge is
stealing SUBSTANTIAL revenue from both companies! It really does suck for enthusiasts like us - amd and nvidia will not be able to sell mid range or low end discrete cards. That may well put
both out of business, or force them to seek profits in other ways.

Again, it sucks.

You misunderstood me, I meant that CPU's are going to become irrelevant for the consumer. That's why Intel is throwing money at every possible diversification, they know that the consumer CPU market is on a decline, because we've long since passed "good enough". Most things you need more horsepower for can be easily parallelized and thus work better on a GPU. 5-10 years from now instead of a CPU with a few weak GPU cores we will see GPU's doing most grunt work with a few weak and efficient CPU cores for legacy/integer crunching. Think Nvidia's "Project Denver" and the whole Fusion initiative from AMD.

 

[Arkadrel]

I wouldn't be surprised at all to see 7970 get smoked by a gtx 670 right now. The million dollar question is, will that happen in 6 weeks or 6 months?

Give it 4 months, going by rumors.
The question isnt weather the 670 is faster than the 7970, but if it ll be cheaper.
Chances are that it wont be... and you ll have to wait 4months from now to get it.

 

[Grooveriding]

Sure it does. The market will tell all. And you're kinda short sided based on the architecture changes with GCN. You're actually making it sound like AMD has this massive desktop discrete advantage. Market speaks much louder than your vocal view, sorry!

Exactly, thanks for making my point for me. You want to use revenue and profits to define the success of something that only accounts for a portion of those metrics and is also affected its self by more factors than how many units sold.

AMD has a good 60%+ of the DX11 market if you use steam numbers and the best selling DX11 card wtih the 5770, but that is only a small part of the big picture. Hence once again DX11 cards sold /= overall success in revenue and margins.

I do understand how important revenue and margins are to you personally though, as you sure do post a lot about them. I could care less myself, I just game with my hardware. I don't concern myself with what sort of profit was turned on the cards I'm using or what market share they account for...

 

[Arzachel]

Hence why Nvidia "expected more". Which is the point of this thread, Nvidia wasn't trying to beat something that was only 6% faster than the 580 at 1200p on avg.
.

'


STOP THE PRESSES, the GTX570 is equal to the 7970!!!

I'd like to know where you're getting that 6% figure, otherwise it seems to be cherry picking on the highest level.

 

[exar333]

You misunderstood me, I meant that CPU's are going to become irrelevant for the consumer. That's why Intel is throwing money at every possible diversification, they know that the consumer CPU market is on a decline, because we've long since passed "good enough". Most things you need more horsepower for can be easily parallelized and thus work better on a GPU. 5-10 years from now instead of a CPU with a few weak GPU cores we will see GPU's doing most grunt work with a few weak and efficient CPU cores for legacy/integer crunching. Think Nvidia's "Project Denver" and the whole Fusion initiative from AMD.

Definitely irrelevant. No one needs to replace an existing PC, emerging markets do not need them, and businesses definitely do not need more power in their workstations and/or servers. /sarcasm

VERY short-sighted post.

 

[BallaTheFeared]

Sorry I thought it was known, when I say 6% I mean on avg over about 14 DX11 programs at 1200p, 8xAA 16xAF. At 1600p the advantage grows to 20% over those same DX11 titles with 8xAA 16xAF.

Source: http://www.tweakpc.de/hardware/test...ition/benchmarks.php?benchmark=pfd11&lang=eng

Test suite includes some 11 DX11 games, as well as a few DX11 benches.

 

[kidsafe]

Overclocked 470s are 59% faster than a stock 7970, and 40% faster than a 7970 clocked at 1125/1575... In a test where the 7970 is 12% faster than a 580. Granted this is 1080 vs 1200, but the difference isn't enough to account for the disparity in performance.

Hence why Nvidia "expected more". Which is the point of this thread, Nvidia wasn't trying to beat something that was only 6% faster than the 580 at 1200p on avg.

Whatever else we discuss in this thread is pretty much OT.

Are you dense? If I wanted to buy two brand new GTX 470s from Amazon or another reputable source, I'd have to pay around $700. It had better be that faster than a single HD 7970.

And please, for the love of God, stop spouting that 6% nonsense.

Why can't you take the review sites' word for it? Here, I'll give you an easy example: http://www.anandtech.com/bench/Product/508?vs=517

Crysis Warhead @ 1920x1200: HD 7970 50.3fps, GTX 580 41.7fps
Metro 2033 @ 1920x1200: HD 7970 54fps, GTX 580 44.5fps
Total War Shogun 2 @ 1920x1200: HD 7970 110.7fps, GTX 580 80.9fps
Batman Arkham City @ 1920x1200: HD 7970 81fps, GTX 580 71fps
Unigine Heaven @ 1920x1200: HD 7970 63.5fps, GTX 580 49.8fps

Some games are indeed closer, but the average of all benchmarks is far greater than 6% difference.

 

[kidsafe]

Sorry I thought it was known, when I say 6% I mean on avg over about 14 DX11 programs at 1200p, 8xAA 16xAF. At 1600p the advantage grows to 20% over those same DX11 titles with 8xAA 16xAF.

Source: http://www.tweakpc.de/hardware/test...ition/benchmarks.php?benchmark=pfd11&lang=eng

Test suite includes some 11 DX11 games, as well as a few DX11 benches.

Even based on that one site's results your numbers do not add up.

Here, I will do the basic arithmetic for you: (70.8-63.9)/63.9 = .108 or 10.8%

And keep in mind these are with a fresh architecture that needs driver optimization.

 

[BallaTheFeared]

Look at the 7970, 32 ROPs, not very impressive on pixel fillrate, look at the difference in tests, 4xAA vs 8xAA.

Picture getting any clearer?

It's 68.7 vs 61.8 @ 8xAA16xAF which is 11%

 

[kidsafe]

(68.7 - 61.8) / 61.8 = 0.112 or 11.2%

Anyway I'm done.

 

[Madcatatlas]

ballathefeared, i like your posts, they remind me of someone who is no longer posting here


Ill get some popcorn to watch how this topic develops, its interesting to say the least..haha.

Nvidia PR working overtime and getting some help it seems

i just hope they can release a killer card within mars month, so i can buy a cheaper 7970

 

[Vesku]

NVIDIA has to release a $300 GTX 580 equivalent on 28nm or it fails to meet my expectations. You hear that NVIDIA, you have expectations to meet.

Although I don't agree with some of the odd arguments I do think halo cards are a bad buy unless you are invested on the display side (30 inch or 3+ monitors). My summer GPU purchase will be looking in the $250 -350 range.

 

[Arzachel]

Definitely irrelevant. No one needs to replace an existing PC, emerging markets do not need them, and businesses definitely do not need more power in their workstations and/or servers. /sarcasm

VERY short-sighted post.

Did you just literally ignored everything I wrote?

 

[SirPauly]

Can't argue this - AMD seems to be more nimble and has caught nV off-guard in the past 3 or 4 generations. Now we just have to wait another 3-6 months for nV to respond, assuming history holds.

My biggest concern is that instead of rumors about performance, price, schedule, whatever from nV about their next gen we're seeing lots of FUD trying to tear down what's obviously a good launch from AMD (not from you, just in general). That makes me worried that they don't have much positive to share about the card, which means even less competition when the hardware does arrive. But I could be reading too much into their relative silence.

To me, nVidia has been very quiet.

 

[OILFIELDTRASH]

At the end of the day AMD is out there to make money, if it can avoid a pricing war with the competition and still sell there cards at a higher price, I say kudos to them.

except for the fact price fixing is illegal.

 

[Vesku]

except for the fact price fixing is illegal.

Ball is in NVIDIA's court. Seem to be a bit of rebating starting on the GTX 580 but not the kind of price drop you'd expect given the appearance of a 28nm card in it's same GPU tier.

Idontcare at one point cited how NVIDIA and AMD were wrist-slapped over price fixing, hopefully they aren't repeating.

 

[thilanliyan]

Sorry I thought it was known, when I say 6% I mean on avg over about 14 DX11 programs at 1200p, 8xAA 16xAF. At 1600p the advantage grows to 20% over those same DX11 titles with 8xAA 16xAF.

Source: http://www.tweakpc.de/hardware/test...ition/benchmarks.php?benchmark=pfd11&lang=eng

Test suite includes some 11 DX11 games, as well as a few DX11 benches.

Seriously, this again...ONE REVIEW does not mean you get a good picture of a card's performance. In one stroke you are TRYING to make every other review site irrelevant. Reasonable people know you have to look at more than just one review. Doesn't matter how many games they test.

 

[SirPauly]

Exactly, thanks for making my point for me. You want to use revenue and profits to define the success of something that only accounts for a portion of those metrics and is also affected its self by more factors than how many units sold.

AMD has a good 60%+ of the DX11 market if you use steam numbers and the best selling DX11 card wtih the 5770, but that is only a small part of the big picture. Hence once again DX11 cards sold /= overall success in revenue and margins.

I do understand how important revenue and margins are to you personally though, as you sure do post a lot about them. I could care less myself, I just game with my hardware. I don't concern myself with what sort of profit was turned on the cards I'm using or what market share they account for...

You keep using those steam numbers! I'll use Jon Peddie and Mercury Research numbers!

 

[Grooveriding]

You keep using those steam numbers! I'll use Jon Peddie and Mercury Research numbers!

It seems you only like Steam if it is working in favour of your point, rather than against ? :|

http://forums.anandtech.com/showpost.php?p=32809419&postcount=71

Originally Posted by Lonbjerg
On Steam and it's more detail.
All for it.
Steam is (like it or not) a better tool than pure guesses.
The more datapoints the better :thumbsup:

I agree with this. The more data points, the better. Sure doesn't hurt discussions and think it is fun and welcomed.

or here http://forums.anandtech.com/showpost.php?p=32818137&postcount=122


Very puzzling.