TR Forums

Whatever, if they release the cards, it might be a good time to upgrade the aging GTX260.

I've long been a fan of Nvidia products, as they always release something interesting from development and API viewpoint, they have very good SDK's, new features to play with on every generation, and usually, a very good performance.

Even dustbusters weren't that bad. It was the first card I could write unlimited length shaders for my projects, and not worry about looping too many times, therefore running out of instructions. And when next gen came out, all of that code just felt natural + executed a lot faster.

I love how Nvidia keep innovating things, and their products have been rock solid since TNT. They are more than just FPS.

flip-mode wrote:

Game_boy wrote:

The 7970 is 365mm^2. The plan can't have been for a 500mm^2+ Nvidia GPU to just match it.

I'm guessing you're not familiar with Nvidia's approach to GPUs for the last few years, which has been to make the most powerful GPUs possible and push die size to the limits. That has generally resulted in 500+ mm die sizes for Nvidia's top of the line GPU for a few generations now. So, no, Nvidia hasn't tired to have a top end GPU that matches up with AMD's top end GPU, rather Nvidia's top GPU has typically been a whole tier above AMD's top GPU.

I've been following this intensely and I'm aware of that. I said the plan wasn't that, Nvidia would never intend to go in and be behind. That was the reality when both cards were out the last few gens, yes.

Airmantharp, supply arguments are going to be irrelevant if 28nm remains shut down for months. There won't be shortages - there just won't be ANY. AMD is on record saying 28nm yields were good, Nvidia is on record saying they were not.

Krogoth wrote:

I love this thread.

Speculation on silicon that hasn't even reached public hands.

Can people just wait for the final product to come out?

Anyway, my take on the matter is that the days of rapid jumps in performance, features and so forth are over.

Silicon has been running out of steam and engineering difficulties involved in making more and more complex designs are catching up.

*puts fingers in ears* LALALALALALA

Krogoth wrote:

I love this thread.

Speculation on silicon that hasn't even reached public hands.

Can people just wait for the final product to come out?

Anyway, my take on the matter is that the days of rapid jumps in performance, features and so forth are over.

Silicon has been running out of steam and engineering difficulties involved in making more and more complex designs are catching up.

Thats the point, this whole deal is becoming more of a fanboy fight.

Looks like the 4850 and ther "vintage" cards are going to stay in the Steam top 10 for a while longer yet.

Surely though it's dangerously close to price fixing when the only two competitors in a market sector maintain the previous generation's pricing when their costs are clearly going down? Isn't this exactly what the DRAM suppliers were nailed for a few years ago?

On the flip-side, maybe there's just a severe supply constraint at the moment, and things will get better once TSMC sort out their problems. I just hope it doesn't take too long.

MrJP wrote:

Looks like the 4850 and ther "vintage" cards are going to stay in the Steam top 10 for a while longer yet.

Wow, Steam has installed application list now? Now it's officially a spyware.

Still, the midrange is what is the most popular. Generation after generation.

Madman wrote:

MrJP wrote:

Looks like the 4850 and ther "vintage" cards are going to stay in the Steam top 10 for a while longer yet.

Wow, Steam has installed application list now? Now it's officially a spyware.

Still, the midrange is what is the most popular. Generation after generation.

Hey, my 5770s are now in 1st place! It still surprises me how small those percentages are. The 5770 is the most popular but it's owned by less than 5% of the user base ... and everything just gets narrower after that. No wonder consoles are the baseline target for pretty much every game out there these days.

And yeah, that installed application list has been there for a few years now I think. It's never sat well with me either, but technically the survey is voluntary.

Krogoth wrote:

I love this thread. Speculation on silicon that hasn't even reached public hands.

Thread title: "Geforce GTX 680 pre-game party". You might have missed it.

Krogoth wrote:

I love this thread. Speculation on silicon that hasn't even reached public hands. Can people just wait for the final product to come out?

No, for the hundredth time, no. But next time I start a rumor thread I'm sure you'll show up with the same comment. Haters gonna hate when others speculate.

Anyway, my take on the matter is that the days of rapid jumps in performance, features and so forth are over. Silicon has been running out of steam and engineering difficulties involved in making more and more complex designs are catching up.

And after criticizing you join in, but only to copy/paste the same thing you've been saying for the last three years

Madman wrote:

Whatever, if they release the cards, it might be a good time to upgrade the aging GTX260. I've long been a fan of Nvidia products, as they always release something interesting from development and API viewpoint, they have very good SDK's, new features to play with on every generation, and usually, a very good performance. Even dustbusters weren't that bad. It was the first card I could write unlimited length shaders for my projects, and not worry about looping too many times, therefore running out of instructions. And when next gen came out, all of that code just felt natural + executed a lot faster. I love how Nvidia keep innovating things, and their products have been rock solid since TNT. They are more than just FPS.

I can respect that. I think GCN has brought AMD roughly up to par with Nvidia on compute, and potentially even ahead in some ways, but I'm not a programmer so I best keep my mouth shut on those topics.

.

Mobile Kepler ... yawn.

I only glanced the thread but i didn't notice anyone talking bout this slide:
http://www.fudzilla.com/home/item/26316 ... x-680-lead

Ofc, be warned, this are supposed to be best case scenario and was posted by nvidia. Wouldn't be suprised if the card will actually lose to the 7970 in some of those games, depending on the resolution.

I wouldn't be surprised if they compare new 4x SMTHNAA vs 8xFSAA to get those numbers.

Still, if I'm buying a GPU, it will be green.

Krogoth wrote:

I love this thread.

*claps half heartedly, clearly unimpressed*

clone wrote:

flip-mode wrote:

AMD's pricing was the first domino, the first shot fired, whatever. It set the tone and the rumor is that Nvidia seems only pleased to match. The fact is that Nvidia has had buckets of problems producing hardware for the past couple of generations. Nvidia is essentially waving the white flag here and what is probably happening is that instead of launching the 500 mm sq monster they had intended to launch, they're shelving it for now and simply hiking up the price of what would have been their midrange GPU. I don't care which fanboy set I happen to ruffle.

thinking AMD's initial pricing was the first shot is a mistake it did nothing but undercut Nvidia at the time and I don't believe Nvidia is waving any flags, my guess is Nvidia wants pricing high for all of the reasons they pushed pricing up for in the past. back when Nvidia first took a stab at motherboard chipsets they tried to sell them for $90 a piece and it was only after every motherboard maker gave them the middle finger that they reconsidered. p.s. speculating anger in response to a weak post with no meat to it doesn't make it so, sarcasm is rarely anything more than what it was in this case..... laughter, no anger your post was a joke, no offense intended but it was just plain empty.

No worries; we're not required to reach any consensus.

Arclight wrote:

I only glanced the thread but i didn't notice anyone talking bout this slide:
http://www.fudzilla.com/home/item/26316 ... x-680-lead

Ofc, be warned, this are supposed to be best case scenario and was posted by nvidia. Wouldn't be suprised if the card will actually lose to the 7970 in some of those games, depending on the resolution.

Added to first post, thanks. Pretty impressive performance gains over the 7970, if true.

If I recall correctly, AMD was also rumored to be working on a more powerful single-GPU solution "7990". If this is true and GK110 isn't to be released yet we may still see some rivalry on the high-end later this year which could lead to more aggressive pricing.

Arclight wrote:

I only glanced the thread but i didn't notice anyone talking bout this slide:
http://www.fudzilla.com/home/item/26316 ... x-680-lead

Ofc, be warned, this are supposed to be best case scenario and was posted by nvidia. Wouldn't be suprised if the card will actually lose to the 7970 in some of those games, depending on the resolution.

Nice to finally see some no's. Pity the graph y-axis doesn't start at zero... so... it's actually hardly any faster. Yah for crap graphs being made to make things look better than they actually are. (And I have an nvidia gpu)

DeadOfKnight wrote:

If I recall correctly, AMD was also rumored to be working on a more powerful single-GPU solution "7990". If this is true and GK110 isn't to be released yet we may still see some rivalry on the high-end later this year which could lead to more aggressive pricing.

HD 7990 should be dual chip...
http://www.fudzilla.com/home/item/26322 ... g-in-april

yogibbear wrote:

Nice to finally see some no's. Pity the graph y-axis doesn't start at zero... so... it's actually hardly any faster.

The graph factored for the HD 7970 as the 1.0 baseline. So GTX 680 looks to be between 5% and 45% faster, and about 15% faster on average. I'd say that's fairly decisive. Heck, AMD would probably slap an MSRP of $650 on that.

Madman wrote:

Wow, Steam has installed application list now? Now it's officially a spyware.

You have to opt in to the survey, so if it's spyware, it's at least exceptionally polite.

flip-mode wrote:

yogibbear wrote:

Nice to finally see some no's. Pity the graph y-axis doesn't start at zero... so... it's actually hardly any faster.

The graph factored for the HD 7970 as the 1.0 baseline. So GTX 680 looks to be between 5% and 45% faster, and about 15% faster on average. I'd say that's fairly decisive. Heck, AMD would probably slap an MSRP of $650 on that.

That's an Nvidia presentation. When has a lead in such a presentation ever been as stated (AMD or Nvidia)?

Game_boy wrote:

flip-mode wrote:

The graph factored for the HD 7970 as the 1.0 baseline. So GTX 680 looks to be between 5% and 45% faster, and about 15% faster on average. I'd say that's fairly decisive. Heck, AMD would probably slap an MSRP of $650 on that.

That's an Nvidia presentation. When has a lead in such a presentation ever been as stated (AMD or Nvidia)?

Yeah, even if the graphical settings were the same for both cards tested, they were probably using AMD's inital "broken/unfinished" release drivers (remember, AMD rushed out the 7xxx series GPUs) compared to their own drivers optimized for the few games they tested. That can make a substantial performance difference in some cases. That slide is mostly marketing hype IMO

On pricing, I think Nvidia is feeling very fortunate that with the 680 they can make up for lost revenue from the delay.

Game_boy wrote:

That's an Nvidia presentation. When has a lead in such a presentation ever been as stated (AMD or Nvidia)?

We work with what we have at the moment until the real reviews publish. What we have is a slide from Nvidia. Nvidia will probably show best case scenarios. That probably means turning on AA if it benefits them in game X, but turning it off if it benefits them in game Y, and so on. It does not seem unrealistic to expect that the GTX 680 is about 15% faster than 7970 on average on the basis of that slide, while at the same time bearing in mind that expectation and reality may diverge to greater or lesser degree when the real benchmarks are published.

I guess I would be more surprised if GTX 680 ended up slower than 7970 on average after the rumors of Nvidia renaming it from 670 ti and the rumors of their surprise at the lower than expected performance of the 7970.

flip-mode wrote:

I guess I would be more surprised if GTX 680 ended up slower than 7970 on average after the rumors of Nvidia renaming it from 670 ti and the rumors of their surprise at the lower than expected performance of the 7970.

Agreed that it won't end up being slower than 7970.

Another thing to consider: the 7xxx chips have shown 20% overclocks. Assuming the real lead is in the 10% range and a realistic overclock gets 10% more fps, AMD could counter this lead by encouraging a lot of factory overclock models.

flip-mode wrote:

Game_boy wrote:

That's an Nvidia presentation. When has a lead in such a presentation ever been as stated (AMD or Nvidia)?

We work with what we have at the moment until the real reviews publish. What we have is a slide from Nvidia. Nvidia will probably show best case scenarios. That probably means turning on AA if it benefits them in game X, but turning it off if it benefits them in game Y, and so on. It does not seem unrealistic to expect that the GTX 680 is about 15% faster than 7970 on average on the basis of that slide, while at the same time bearing in mind that expectation and reality may diverge to greater or lesser degree when the real benchmarks are published.

I guess I would be more surprised if GTX 680 ended up slower than 7970 on average after the rumors of Nvidia renaming it from 670 ti and the rumors of their surprise at the lower than expected performance of the 7970.

If they shuffle settings based on convenience, I don't think averaging the results for an overall lead makes sense ...

.

clone wrote:

it's all rumor.

That's what this thread was created for

in the end I believe Nvidia is lying in an attempt to drag out buying choices because they have nothing and will have nothing to bring to the table for another 3 months.... it's all smoke and mirrors to me and I'll be waiting for the hard launches and real reviews anyway.

So, Nvidia isn't your BFF. It's doesn't make any sense for Nvidia to lie about these things, honestly. Picking best cases is one thing but saying the card is faster than it really is would be counterproductive. Why would they say it's faster than the 7970 and then launch something that struggles to keep up with a 7950? They would not do that because there is absolutely no reason to. AMDs cards are all launched and now Nvidia gets to make a surgical strike, so if they say GTX 680 is faster it is probably so, even if by how much may vary.

I'm always excited by new tech. I can't wait for the review.

I did a theoretical performance calculation of the GTX 680, interesting findings. Note that these are theoretical based on th 640m performance vs the 560m.

Findings:
Kepler is on average 45% as efficient per shader/clock as Fermi, though it has a lot more of them (3x shaders, 90% clock for 680 vs 580).
Vs GTX 580, on average:
GTX 680 is 20.5% faster
HD 7970 is 19.9% faster
This is within statistical error. Nvidia claims -3 to 42% faster, with 10-15% being average. I believe my cherry-picking thoughts are true (games, res, AA/no AA), to show only strengths. Similar to how Nvidia claimed to have the fastest dual card (GTX 590 vs HD 6990), when in reality HD 6990 is clear winner due to architectural efficiency and power limitations (based on every review site), Nvidia just picked a handful of situations where it was barely faster.

I am underwhelmed @ Kepler, I was thinking they would beat the HD 7970 by a greater margin. This is the GTX 680, which isn't the top end, but I don't expect the 690 to be that much faster (expecting more of 570 vs 580 performance). I think the 7970 will OC better too, for those interested. I guess we will see in a couple weeks.

My speadsheet:
http://www.2shared.com/document/r6xxbBmW/Nvidia.html

Sources
http://www.anandtech.com/show/5672/acer ... er-verge/4
http://www.guru3d.com/article/amd-radeo ... 0-review/1