nVidia: "We expected more from the 7970"

[Lepton87]

I just want 3-monitor support so I can keep all of my displays hooked up at the same time... Have the main screen, the work screen, and then the TV hooked up for movies/Netflix/etc. So when I'm doing CAD work, or Excel work, I can keep that screen up all the time while using the other monitor for whatever. I am probably a pretty unique case, but it would be a nice thing to have...

And Vesku, I just disable my TV using Win + P when its not in use, keeps the temps cooler.

Can't you connect one of your displays to iGPU? That way you could already have all of your displays connected with an nV card.

 

[96Firebird]

Yeah having it all chained to one card would make it easier to selectively disable screens, too, to save power and wear and tear on monitors that I am paying zero attention to. Right now because I stupidly sold off my DP adapter, I can only run 2 monitors on my GPU and the third off my mainboard's IGP, making for more headaches when switching monitor modes.

I really want to try running my TV off my i5's GPU, but haven't gotten around to looking in to how to do it. I know I need to use Virtu, which I have installed, but haven't gotten any farther than that. That would allow my to get 3 displays, or just use 1 monitor and my TV without having to use Win + P and the GPU would clock down. Might look into this a little more...

But this is getting way off topic, should probably just create a new thread...

Edit - Beat by Lepton, but yeah I'm going to look into it...

 

[Arzachel]

So first to market = win? Unfortuantely the marketshare-gain retention by AMD with the first DX11 cards was poor.

The 5770 begs to differ.

Charlie repeatedly mentioned that Kepler would have features that he couldn't talk about... that it would hold a big surprise that Nvidia would exploit a la Crysis 2's tessellation. Perhaps the added features and debugging them, are what are holding up Nvidia. A card being delayed doesn't necessarily mean it's the GPU hardware that is at fault (indeed, Charlie even says that it looks like relatively smooth sailing on that front, though I have my doubts). It could be a software driver or debug delay.

It's far to late to be working on features, specs get locked prior to any prototypes.

 

[Attic]

I'd prefer to hear nVidia talking about their own tech. To me this would signify a card that is worthy of mention in the works at their labs.

But this is nVidia, what do you expect from their reps? We've seen this kind of behavior before from the green team when they are beat and without a competing product on the shelves.

 

[blastingcap]

It's far to late to be working on features, specs get locked prior to any prototypes.

I don't think you read what I wrote very carefully.

 

[Arzachel]

I don't think you read what I wrote very carefully.

If the delay would be on the software side, Nvidia would already be stockpiling cards and we would see some actual leaks.

 

[mak360]

It’s amazing how people are saying a mid-end card will beat out a high end without any info whatsoever. Lol
It’s even more amazing how people are saying overclocking doesn’t count. Lol


What’s really amazing is how AMD kept the new cards from showing its true potential by capping speed & Drivers, so nVidia can’t pin-point the performance. There was allot of noise a week back to push AMD in releasing drivers (Hardocp & others). Expect magic drivers to come out once nVidia pops its little head out.


Ps todays & Last week’s drivers are not for the 7 series also. (Drivers for 7 series is on the DVD)

 

[Elfear]

I had much higher hopes for this oft-delayed node. 6XXX was barely an improvement over 5XXX, which was released on what date? I'll leave your arbitrary percentages out of the conversation.

If you read my whole post instead of just picking one thing that offends your self-worth due to your brand-association, you will see that there was plenty of reason to expect more from the 28nm cards so far. And I am not laying that blame soley on the card makers.

Arbitrary? I was a little high in my estimate according to one website, but it still shows a 36-40% disparity between the 6970 and 7970 (at 1080p and 1600p respectively). Or approximately 56-60% faster than the 5870. If you want to compare the 7970 to it's 2.5-yr old ancestor, I'm still trying to see how 56-60% faster on average equates to Bulldozer which gets beat in many instance by the 3-yr old Phenom II?

Would I have liked more performance out of 28nm? Absolutely. Something like the 8800GTX would have been great but that hasn't happened for awhile and it doesn't mean Tahiti compares in any way to the epic failure that is Bulldozer.

 

[Obsoleet]

Bring that vaporware Nvidia.

 

[blastingcap]

If the delay would be on the software side, Nvidia would already be stockpiling cards and we would see some actual leaks.

Although such leakiness is not a given, good point. But my point was that it could be anything, not necessarily software debugging. Maybe it's a thermal solution problem or shortage of a key component that AMD doesn't use, or who knows what. I guess we'll have to wait and see. Frankly I hope Nvidia doesn't take too much longer, as I'm itching to buy a card but want there to be a price war first.

 

[DarkKnightDude]

A company says they aren't impressed with their competitor's product.

More at 11!

 

[3DVagabond]

Nvidia have been known to shoot for 2x the speed in flagship cards after a die shrink and architecture redesign in the past, so it wouldn't be a very big stretch to see 2x the 580 hitting later in the year, especially if they maintain large GPU sizes of late.

I'm not going to quote every post where numbers are just made up. There are a lot of them. I'm quoting this one simply because it was only a couple of posts prior where it was clearly pointed out that, "No, nVidia doesn't come anywhere near to an automatic 2x performance." Keep in mind that with the 285 to 480 they were also pushed hard up against the 300W pcie wall. Actually putting a couple of cracks in it. They later fixed it a bit with software, but from Fermi they don't really have anywhere left to grow in TDP.

 

[MrK6]

I'm not going to quote every post where numbers are just made up. There are a lot of them. I'm quoting this one simply because it was only a couple of posts prior where it was clearly pointed out that, "No, nVidia doesn't come anywhere near to an automatic 2x performance." Keep in mind that with the 285 to 480 they were also pushed hard up against the 300W pcie wall. Actually putting a couple of cracks in it. They later fixed it a bit with software, but from Fermi they don't really have anywhere left to grow in TDP.

Glad to see someone else is holding people accountable on this forum :thumbsup:.

Even after the GTX 580 was released almost two years after the GTX 285, they still didn't even break the 70% faster mark. It also consumed about 35% more power, so it's actually even less of an improvement. Progress is slowing down folks until either company finds the next big architectural breakthrough.

 

[3DVagabond]

Let's be fair here, 28nm may be having teething problems, and AMD also had to sacrifice more die space for HPC-related stuff (double precision float, etc.).

If HD 7970 had been a simple die shrink from HD 6970 I suspect that HD 7970 would be significantly faster.

I think AMD was running into problems feeding the VLIW arch. That's why they went from VLIW5 to VLIW4, and now to GCN. I don't think doubling the shaders would have given them anything close to a linear scaling of performance and used a crapload more power.

 

[AtenRa]

GTX280 @ 65nm
GTX285 @ 55nm <--- Half node of 65nm
GTX480 @ 40nm <---- Full node over 65nm ~2.3x transistor density over 65nm not 55nm

HD4870 @ 55nm
HD5870 @ 40nm
HD7870 @ 28nm <--- Full Node over 40nm ~2x transistor density over 40nm

HD7870 has a nice performance increase over HD5870 and HD6970 and larger in DX-11 Tessellation. They kept almost the same die size as before at a reasonably TDP.

Kepler could have the same or better % increase in DX-11 Tessellation games that HD7970 has by keeping the same die size of GF110 and same TDP.

Im sure NVIDIA was expecting AMD to raise the die size and performance of Tahiti in relation to 40nm Cayman. AMD chose to stick with the same strategy (smaller die size) as before, but i believe this time it will not have the same effect.

 

[Keysplayr]

A company says they aren't impressed with their competitor's product.

More at 11!

If they were impressed, it's more likely they would have said nothing at all.

IMHO

 

[wahdangun]

If they were impressed, it's more likely they would have said nothing at all.

IMHO

so with this statement, its mean they was disappointed ?


hahah it like HD 5870 all over again, so do we expect woodscrew graphic card for nvdia ?

 

[wahdangun]

btw did all of you remember this :

and few months latter we get this :

its even more power hungry than HD 5970. ???

 

[MrK6]

btw did all of you remember this :

and few months latter we get this :

its even more power hungry than HD 5970. ???

Yep, the GTX 480 was a joke. NVIDIA hasn't put out a decent high end card since the GTX 295. Hopefully Kepler turns that around.

 

[cliftonite]

I had much higher hopes for this oft-delayed node. 6XXX was barely an improvement over 5XXX, which was released on what date? I'll leave your arbitrary percentages out of the conversation.

If you read my whole post instead of just picking one thing that offends your self-worth due to your brand-association, you will see that there was plenty of reason to expect more from the 28nm cards so far. And I am not laying that blame soley on the card makers.

What about his post made you think it offended his self-worth. What does that even have to do with the discussion?

 

[wahdangun]

Yep, the GTX 480 was a joke. NVIDIA hasn't put out a decent high end card since the GTX 295. Hopefully Kepler turns that around.

yeah it must have snapped some nvdia fans i mean waiting for six month for just 10% more performance ? and much higher power draw.

and here quote from wizzard himself :

In idle the power consumption of the card is simply unacceptable. AMD's HD 5000 Series is basically 3x as efficient here compared to NVIDIA's latest product. When you look at the rendering performance of GTX 480 it is roughly similar to GTX 295. In most of our tests, we see the GTX 480 close to the GTX 295 in regards to power consumption.

 

[AtenRa]

Yep, the GTX 480 was a joke. NVIDIA hasn't put out a decent high end card since the GTX 295. Hopefully Kepler turns that around.

Sorry to say but GTX480 was far from a joke,

Even today it can play all DX-11 games at 1080p much faster than HD5870 and in games like BF3, Crysis 2, metro 2033 etc, at 1080p with high/ultra settings it can produce more than 30fps when HD5870 can't.

Im not saying it was the best card of all times, but it was more future proof (DX-11) than HD5870 and it is very fast even for today's standards

 

[blackened23]

Sorry to say but GTX480 was far from a joke,

Even today it can play all DX-11 games at 1080p much faster than HD5870 and in games like BF3, Crysis 2, metro 2033 etc, at 1080p with high/ultra settings it can produce more than 30fps when HD5870 can't.

Im not saying it was the best card of all times, but it was more future proof (DX-11) than HD5870 and it is very fast even for today's standards

I agree with him. The performance was there but EVERYTHING else was a joke -- it was too expensive, too hot, it was basically a power guzzling dust buster piece of junk. Obviously NV went on to perfect it with the 500 series, and the GTX 580 was and still is one of the best GPU's ever. But IMO the 400 series were a disappointment, and many reviewers felt the same way at release.

The main takeaway from this is how nvidia boasted that GTX 480 would be 2-3 times faster than 5870 and what happened? 6 months later and 10-15% faster. Guess that was their viral marketing then as well If we've learned anything from history, it usually takes nvidia 2 tries to perfect a new architecture.

 

[SirPauly]

The GTX 580 was only 15% faster than the GTX 480, and the GTX 480 was only 15% faster than the Radeon HD 5870. Your point? The Radeon HD 7970 is only 20% faster than the GTX 580, but it's 40% faster than the HD 6970. That grows with overclocking, and by the time NVIDIA comes out with something AMD will have the HD 8970 ready to launch.

Accept the fact that the GTX 580, 480, 5870 are on the same node and both using their respective 40nm arch and the 7970 is on a substantial and significant node change and new arch.

Here is the other side of the coin:

AMD softened their commitment to the sweet spot.

AMD is charging around 50 percent more MSRP than their past chip

This may be the first time in history that a GPU's MSRP percentage increase offered more than its over-all performance increase.

This may be the first time in history that a new node and arch didn't actually redefine price performance.

And, there is such strong defense for this; as if the sweet spot never existed. It's like I've entered the twilight zone after over a decade of posting.

 

[Lonyo]

Accept the fact that the GTX 580, 480, 5870 are on the same node and both using their respective 40nm arch and the 7970 is on a substantial and significant node change and new arch.

Here is the other side of the coin:

AMD softened their commitment to the sweet spot.

AMD is charging around 50 percent more MSRP than their past chip

This may be the first time in history that a GPU's MSRP percentage increase offered more than its over-all performance increase.

This may be the first time in history that a new node and arch didn't actually redefine price performance.

And, there is such strong defense for this; as if the sweet spot never existed. It's like I've entered the twilight zone after over a decade of posting.

1. HD5870 was basically 33% more than the 4870 launch price ($400 effective street price vs $300 launch price), and more than that over the 4870 price when the 5870 came out, although I don't have a hard number for the 4870 end of life price.
2. GTX 8800 Ultra?
3. GTX 8800 Ultra?