What do you expect from the GTX680?

[blackened23]

Actually rumor has it GK110 is coming in a few months with a 50% increase in performance over GK104

You might want to hold on to those 7970 until the high end chip comes.

Why bother, i'm pretty sure I can sell 680s for 90% of their value on ebay if and when another card is coming out. Nvidia doesn't have a strong track record of releasing stuff on time, so i'll just get the 680s (as long as high resolution results are in line with expectations)

 

[lavaheadache]

Actually rumor has it GK110 is coming in a few months with a 50% increase in performance over GK104

You might want to hold on to those 7970 until the high end chip comes.

So are we in a flat out denial of anything higher coming from AMD?

nVidia is allowed endless amounts of rumors upcoming and unreleased cards but AMD is only allowed a card they released well over 2 months ago that was clearly put out the door very conservatively?

interesting

 

[Grooveriding]

How so? I think people should temper their expectations, the days of 100% increases from gen to gen are probably gone...I don't think its disappointing at all. If the results hold for the 680 i'm sold. I just need to see more surround and super high resolution results.

Using the 1600P Tom's results it's 20% faster @ 1600P. Now take a look at the historical increases seen there - http://translate.google.ca/translat....de+gtx+480&hl=en&biw=1082&bih=571&prmd=imvnshttp://www.computerbase.de/artikel/...force-gtx-480/21/#abschnitt_performancerating

1080P 8XAA = 84%
1600P 4XAA = 64%
1600P 8XAA = 133%!!

100% increases are long gone, that was 8800GTX. But there is temper and there is jump in a vat of ice water temper. 20% ? Really ? I think it's an awesome card for 1080P, but it just is coming up short on perf. increase overall, bus, VRAM etc.

I do still want to see the rest of the benches though, I generally don't look at Tom's anyways. Anand, [h], hwc and c.de I would like to see.

 

[Lepton87]

I think this process shrink is the worst ever. Thanks to underwhelming performance of 7970 and predatory tactics of nvidia we have to pay high-end prices for midrange designs. It's akin to nvidia releasing GF104 as high-end. If they screwed up badly with GF100, they could have released GF104 clocked as high as it could handle and it would have competed with 5870 just fine. Fortunately they left that headroom, but not this time.

 

[BallaTheFeared]

Like Dragon Age 2? Took NV a few months. I can understand if its no name indy game, but bioware, for real?

And on...

ps. CF and SLI will have issues in games, period. Relying on up to date profiles makes it an unavoidable problem.

DA2 wasn't very good, and AMD's AO is what caused the nvidia driver issues I believe. The rest of those games, while you tried real hard, aren't very meaningful.

But I think we're going OT now, let's stick to the point of with all of 2011's releases at your disposal the only semi meaningful game you could list was DA2. Which didn't see negative scaling like AMD cards which lack driver level AO support in Skyrim, but just not optimal scaling.

I don't think I need to speak anymore on that, since you've already spoke well enough for me.

 

[Silverforce11]

This doesn't look good for gk106, with the rumored specs of ~half a gk104.

Not going to compete well vs 7870s, hopefully it will be priced really low.. at least have some price competition in the mid-range. Near $400 for "mid-range" is nuts.

 

[Silverforce11]

The rest of those games, while you tried real hard, aren't very meaningful.

Crysis 2 is meaningful? It's a shit game which nobody even plays online anymore, collecting dust and only brought out for benchmarks, and to gloat at tessellation performance (which NV no longer has a lead). How bout games like anno, shogun etc.

Alan Wake? Hi? Even SWTOR had issues.

If those aren't meaningful, then pray tell which recent "meaningful" game did AMD not provide CF support for in a timely manner?

If it really comes down to it, the only meaningful game of late, that is, people spent many hrs on it and still spending many hrs on it, would be BF3. Every other recent PC game is meh.

ps. Skyrim on ultra runs fine on AMD hardware, your attempt bring up this NV "AO" feature is fail. My mods make skyrim look much better thanks.

 

[BallaTheFeared]

Crysis 2 is meaningful? It's a shit game which nobody even plays online anymore, collecting dust and only brought out for benchmarks, and to gloat at tessellation performance (which NV no longer has a lead)..

You sound bitter, is everything ok? I'd like to dicuss hardware in a civil matter, however if you're going to take everything personally and lash out I don't see how that will be possible.

 

[Silverforce11]

No really, I want to see proof of this so called superior NV driver support. I keep seeing fanboys post it on forums as if its a mantra, like Apple worshippers claiming the iphone hardware is the best!

Edit: Perhaps you should do a comparison between recent titles for CF/SLI scaling, have a look around for facts for once.

Edit 2: I presented a few recent titles out of many where SLI is furbar, if all you got is disregard for non "meaningful" games.. that's the problem isn't it. There's a LOT of games for the PC, many of which do NOT have proper SLI support at all. You still gloat about NV's superior driver support?

ps. "You sound bitter" - So you can hear my text? Glad to know the voice of reason and logic is present.

 

[chimaxi83]

DA2 wasn't very good, and AMD's AO is what caused the nvidia driver issues I believe

Source? Proof?

So many biased statements, so few objective facts.

 

[BallaTheFeared]

Source? Proof?

So many biased statements, so few objective facts.

Oh I'm sorry, that probably came out as if it was AMD fault.

Let me clear that up, it was Nvidia's problem and they had a hard time fixing it.

After they fixed the performance they had shadow flickering issues, it was a pretty big mess for awhile.

Lucky the AMD like phenomena isn't that common with Nvidia...

We have to comment on our CrossFire issues in this article and of course point out that these are NOT XFX's fault. Instead, we seem to have found a fault with the AMD drivers and its QA program. The problem of not seeing any scaling at 2560x1600 in Battlefield 3 (and ONLY 2560x1600) while seeing no scaling in Batman: Arkham City, crashes at 2560x1600 in Skyrim and the large dips in frame rate in Deus Ex: Human Revolution are pretty dramatic and damning for the brand new architecture.

That's what, every big AAA title from the last 6 months?

 

[blackened23]

Welp i'm throwing the towl in on the fanboy wars, i'm done. Gonna buy 2x toys tonight (unless some radically different info comes out)

I need confirmation of when NDA lifts and product availability occurs! ANYONE?! Someone tell me.

 

[Lepton87]

Welp i'm throwing the towl in on the fanboy wars, i'm done. Gonna buy 2x toys tonight (unless some radically different info comes out)

I need confirmation of when NDA lifts and product availability occurs! ANYONE?! Someone tell me.

I would at least see 2560 benches. OC GF vs OC tahiti. it might be faster at stock but oc worse or be faster only at low res.

 

[blackened23]

I would at least see 2560 benches. OC GF vs OC tahiti. it might be faster at stock but oc worse or be faster only at low res.

Yeah i'd like to see super high resolution / oc results too, thats the only thing that could sway me. But I don't think things will change drastically.

 

[Lepton87]

Yeah i'd like to see super high resolution / oc results too, thats the only thing that could sway me. But I don't think things will change drastically.

256bit seems rather crippling for 2560. If you're basing expectations on those tomshardware benchmarks you might be disappointed. In those benchmarks 7950 is slower than 580. We all know 7950 is a faster card.

The only meaningful thing we can infer from those benchmarks is that GTX680 is 31% faster than 580. Exactly as much as most sites claim 7970 to be faster than 580. And that advantage would probably shrink with resolution, question is by how much.

Will you still switch cards if it turns out that average OC 680 is 5% faster than your overclocked 7970? If it's less than 10% seems like futile exercise to go into hassle of selling cards, but you probably like switching hardware in itself not for the benefit it brings, right?

 

[thilanliyan]

So are we in a flat out denial of anything higher coming from AMD?

nVidia is allowed endless amounts of rumors upcoming and unreleased cards but AMD is only allowed a card they released well over 2 months ago that was clearly put out the door very conservatively?

interesting

Yes, nV are allowed to charge premiums, have more unannounced cards (for which an announcement of an announcement will be made when they feel like it), JHH is allowed to come to your house for dinner and borrow your car anytime he pleases, and only nV's driver team is allowed on your Christmas list.

EDIT: JHH is automatically on your Christmas list FYI.

 

[notty22]

256bit seems rather crippling for 2560. If you're basing expectations on those tomshardware benchmarks you might be disappointed. In those benchmarks 7950 is slower than 580. We all know 7950 is a faster card.

The only meaningful thing we can infer from those benchmarks is that GTX680 is 31% faster than 580. Exactly as much as most sites claim 7970 to be faster than 580. And that advantage would probably shrink with resolution, question is by how much.

I disagree.
http://www.hardwarecanucks.com/foru...eon-hd-7950-review-tahiti-pro-arrives-24.html
They are pretty equal. Even 10% is marginal. Situational at least.

 

[blackened23]

I disagree.
http://www.hardwarecanucks.com/foru...eon-hd-7950-review-tahiti-pro-arrives-24.html

D: seems to bolster his argument

 

[blackened23]

256bit seems rather crippling for 2560. If you're basing expectations on those tomshardware benchmarks you might be disappointed. In those benchmarks 7950 is slower than 580. We all know 7950 is a faster card.

The only meaningful thing we can infer from those benchmarks is that GTX680 is 31% faster than 580. Exactly as much as most sites claim 7970 to be faster than 580. And that advantage would probably shrink with resolution, question is by how much.

Will you still switch cards if it turns out that average OC 680 is 5% faster than your overclocked 7970? If it's less than 10% seems like futile exercise to go into hassle of selling cards, but you probably like switching hardware in itself not for the benefit it brings, right?

Yeah, unless reviews indicate that the GK104 is a dog at 2560 and/or doesn't overclock well i'm going to get 2 of them. I just don't see that happening really, i've heard rumors that the GK104 oc's fairly well.

We'll see though :thumbsup: I'm a sucker for new and shiny toys.

 

[Lepton87]

I disagree.

They are pretty equal. Even 10% is marginal. Situational at least.

But according to tomshardware it's 8% slower when in fact it is 2-5% faster. It's totally insignificant lead but going from 8% slower to 4% faster starts to matter. It's 13% disparity with other reviews, if we add 13% to 7970 score, the lead becomes trivial, 4% to be exact. Those cards can easily switch places at 2560. I guess those cards are so similar that it will be up to reviewers to say which card is faster by game selection and settings like AA. Not unlike GTX590 and 6990 that are within 5% of each other, even though GTX590 is generally faster you can easily show that 6990 is faster by game selection.

 

[notty22]

D: seems to bolster his argument

You think AVG improvement of 4% and 0% with 4xaa@1080p bolsters the point of one card being faster than another.
Game selection or just variables in testing could account for that and more.
So your comment is lost on me. Maybe I should have super exaggerated my point for you.

 

[MisterMac]

20% at minimum, or all of us get fukked pricewise.

If the 550 USD MSRP is correct, and gtx 680 doesn't push the pformance envelope enough (even if quieter/lesspowerhungry) a shift in prices won't occur.

And i'll be unhappy -_-

Wether i go for a 7900 or a gtx 680 -_-

 

[RussianSensation]

Using the 1600P Tom's results it's 20% faster @ 1600P. Now take a look at the historical increases seen there - http://translate.google.ca/translat....de+gtx+480&hl=en&biw=1082&bih=571&prmd=imvnshttp://www.computerbase.de/artikel/...force-gtx-480/21/#abschnitt_performancerating

1080P 8XAA = 84%
1600P 4XAA = 64%
1600P 8XAA = 133%!!

100% increases are long gone, that was 8800GTX. But there is temper and there is jump in a vat of ice water temper. 20% ? Really ?

Ya, that's why I like to revisit reviews 6-12 months later to see re-assess how the newer generation of cards performs in more modern games. If we just went off launch TPU reviews of GTX480, the performance increase over GTX285 would have been very underwhelming. Put a GTX285 ~ GTX460 1GB in a modern game and GTX480 will easily lead by 50%+.

However, due to consolitis, it looks like Metro 2033, Crysis 2, BF3, Witcher 2, Dragon Age 2 are as demanding as things are going to get in the near future, at least until GTA5 is out this year (Metro 2034 got delayed into 2013 I believe).

So unlike initial reviews of HD5870/GTX480 that didn't allow those cards to truly shine vs. their predecessors, GTX680 is put in the best possible light right now since the games where we need the most performance is where GTX680 is nowhere near 50% faster than GTX580. So if it can't beat GTX580 by 50% now in the most demanding games, I doubt it will come up with some magical boost later.

Having said that, I still think GTX680 is an upper mid-range Kepler card: 256-bit bus @ 192GB/sec, 2GB VRAM, not a 400mm^2 die, neutered DP performance. In that regard its performance is in line against GTX580; its naming and price are not. :biggrin:

I think you just have to wait until GK110 or get a high overclock on 2x GTX680 (1300mhz+) before you can ditch 3x GTX480s @ 580 speeds and still get a nice 40-50% performance increase.

 

[Lepton87]

You think AVG improvement of 4% and 0% with 4xaa@1080p bolsters the point of one card being faster than another.
Game selection or just variables in testing could account for that and more.
So your comment is lost on me. Maybe I should have super exaggerated my point for you.

Reviewers agree that 7950 is the faster card. Anand:

"While the 7970 sails past the GTX 580—and AMD has priced it based on that—the 7950 and the GTX 580 are trading blows on a game-by-game basis, similar to what we saw last year in comparing the GTX 500 series and the Radeon HD 6900 series. But when the 7950 wins it wins big, while the same cannot be said of the GTX 580; the only real weakness for the 7950 right now is Battlefield 3, and while that’s an important game it’s but one of several."

Given our experiences, we don't think the Radeon HD 7950 is a worthy upgrade from a GeForce GTX 580 in performance or gameplay experience. If you have a GeForce GTX 580 buying a Radeon HD 7950 won't significantly impact your gaming experience in a meaningful way. There will be some games where the Radeon HD 7950 is a percentage faster than the GTX 580, but that percentage could be made up by overclocking or having an overclocked GeForce GTX 580.

See those charts.

http://www.computerbase.de/artikel/...70-und-hd-7850/5/#abschnitt_leistung_mit_aaaf


it's marginally faster at 1080p but it is clearly superior at 2560. going from 8% slower to 14% faster makes a big difference. After looking at those charts it becomes clear why only 1080p charts leaked. For me 1080p is irrelevant as it is for most high-end cards owners.

Don't get me wrong 7950's lead is totally insignificant and thus purely academic but tests consistently paint 7950 as marginally faster than 580.

 

[JAG87]

256bit seems rather crippling for 2560.

People really have to stop with these ridiculous statements. There is PLENTY of bandwidth and VRAM on the 680 to handle 2560 resolutions. Any discrepancy you saw in the THG charts is either THG being THG, or drivers.

Proof of concept: 7870 beats 580 and 6970 despite having less bandwidth.