OFFICIAL KEPLER "GTX680" Reviews

[-Slacker-]

2 Reviews already and GTX680 has won in both so far. Let's wait for 20 more then.

It's pretty funny being on a hardware enthusiast site where people don't welcome new features such as Turbo boost out of the box for all users and a $50 lower price, that might force price cuts making HD7900 even more attractive for everyone and make Turbo boost a standard feature for new generation of videocards much like it is today for CPUs.:thumbsup:

So wanting to see more benchmarks form somewhere other than some obscure chinese website that shows both competing cards do less than 75fps in MW3 at 1080p, and a very fishy tomshardware leaked review that has the 7970 perform 10fps slower than in their launch review equals "people don't welcome new features such as Turbo boost out of the box for all users and a $50 lower price"??

 

[3DVagabond]

Logically, you cannot prove that HD7970 is a better card at $550 unless you bring some outlier benchmark or game or some specific case that only applies in a particular situation.

The fact that GTX680 is actually better in pretty much everything and costs less and has a free Turbo Boost? What's not to like? What exactly is there to discuss?

It seems you are dead set on proving that HD7970 is better. Not sure what you are trying to accomplish here. You've enjoyed your 1340mhz HD7970 and GTX680 isn't going to make that level of performance obsolete. For most consumers looking to buy a $500 GPU starting tomorrow, HD7970 at $550 is irrelevant.

Nothing's free, mate. You are paying for everything they give you.

 

[Sind]

Man folks gonna get this thread locked, yet another sigh.

I'm liking what I see so far but I'll wait for H, HWC and anand before I form my own opinion. Just hoping we get some details on the rest of the lineup.

 

[Grooveriding]

I don't really like the sound of the 'turbo-boost' thing. It's sounds to me like something I would prefer to just turn off and leave the card running as fast as possible at all times with whatever overclock I chose to put on it.

I also wonder what sort of negative affects it could have on your gaming with it constantly changing up clock speeds. Just sounds like a feature that could easily introduce bugs and issues.

 

[MrK6]

I'm hard up for an upgrade, so I've lowered my expectations from 40% to 30% @ 2560x1600 4xAA over a 580 to get me to buy a pair. So long as there is a respectable amount of OC headroom 15%+ I am buying. Heck, I think these cards are going to sell fast and I want EVGAs so I can keep my warranty and put waterblocks on them. I plan to buy two as soon as they show up on newegg and cancel the order if they don't meet my needs.

If you do upgrade, I'd love to see some benches. If they overclock to the ~1150MHz+ range on stock volts, that'd be inline with what the 79xx cards do on 28nm. I'd hope for them to hit at least 1300Mhz with volts, better yet 1400MHz on water.

 

[RussianSensation]

So wanting to see more benchmarks form somewhere other than some obscure chinese website that shows both competing cards do less than 75fps in MW3 at 1080p, and a very fishy tomshardware leaked review that has the 7970 perform 10fps slower than in their launch review equals "people don't welcome new features such as Turbo boost out of the box for all users and a $50 lower price"??

The 2nd review posted here are charts from PC Perspective I believe. What Chinese website? That's a pretty reputable site.

Nothing's free, mate. You are paying for everything they give you.

You know what I mean though. For example, SB gives up to 4 bins Turbo with 1 core and 1 full bin with 4 full core active. Is that cheating? No because most programs aren't like OCCT, Prime95 or LinX that load SB to full 95W TDP. All that's being done here is when the card is not using 100% of its TDP potential, the TurboBoost is activated.

Before NV would cap clocks more conservatively since they'd target the most demanding GPU usage possible. Now they are actually taking into account more real world usage, which is why the Turbo is possible. Also, those who don't want this feature will likely be able to turn it off.

 

[-Slacker-]

to groov:

Sounds like a straight up performance improvement to me ... ups the clocks when you need them, keeps them they way you set them when you don't...

 

[blastingcap]

I don't really like the sound of the 'turbo-boost' thing. It's sounds to me like something I would prefer to just turn off and leave the card running as fast as possible at all times with whatever overclock I chose to put on it.

I also wonder what sort of negative affects it could have on your gaming with it constantly changing up clock speeds. Just sounds like a feature that could easily introduce bugs and issues.

True but Intel Turbo Boost and AMD Turbo Core seem to work fine.

 

[MrK6]

I don't really like the sound of the 'turbo-boost' thing. It's sounds to me like something I would prefer to just turn off and leave the card running as fast as possible at all times with whatever overclock I chose to put on it.

I also wonder what sort of negative affects it could have on your gaming with it constantly changing up clock speeds. Just sounds like a feature that could easily introduce bugs and issues.

Much like Intel's Turbo Boost, it seems to be a feature that gives more performance at stock. But for people like us who don't run things anywhere near stock, I'm not sure what the benefit is (or if it actually harms, like Turbo Boost does).

to groov:

Sounds like a straight up performance improvement to me ... ups the clocks when you need them, keeps them they way you set them when you don't...

Is it though? It seems like it only ups the clock when the card is running under TDP. If the card is being pushed to the wall, I imagine it's close to TDP, so I'm not sure what kind of benefit you'll see. I've postulated before that if a scene loads a GPU lopsidedly, it could theoretically use the "saved" power from those portions not being stressed and apply to those portions that are to increase performance. This is just a guess though. Hopefully AT does an awesome write up of the technology in the GPU, like they usually do.

 

[3DVagabond]

For the 2nd time, I am not implying [H] is biased.

They usually have a review of 4-5 sh1ty games, but they do it honestly as far as I can tell.

If you weren't implying bias...

dirty Nvidia mother^&^&$#@ complicating things, and making life less simple for Kyle [H] (AMD partnership website?)...

Just so they can get more performance(!)

How low can they go?!

...then why did you post this?

 

[tviceman]

GPU boost can't possibly be considered cheating if it works like CPU turbos do, but if it isn't helping minimum frame rates below 60fps then it's a worthless feature. That said, based on the charts posted the gtx680 has noticeably better averages and doesn't dip as low or as often as the hd7970. GK104 is more than very likely to end up superior in every way, INCLUDING price.

People set on buying an AMD card should be happy - prices should be dropping soon because, despite the complaints by many (including myself) that the hd7000 series was overpriced from the start, now their prices look even worse in comparison to Nvidia's first 28nm card.

And to think, so many people argued there was no way Nvidia would price a better performing part for less...

 

[3DVagabond]

You know what I mean though. For example, SB gives up to 4 bins Turbo with 1 core and 1 full bin with 4 full core active. Is that cheating? No because most programs aren't like OCCT, Prime95 or LinX that load SB to full 95W TDP. All that's being done here is when the card is not using 100% of its TDP potential, the TurboBoost is activated.

Before NV would cap clocks more conservatively since they'd target the most demanding GPU usage possible. Now they are actually taking into account more real world usage, which is why the Turbo is possible. Also, those who don't want this feature will likely be able to turn it off.

I don't have any problems with "turbo boost", assuming it works as you say. Although I don't think it's going to work like it does with CPU's. More likely when games/apps are poorly optimized and the card is cruising along at <100% usage it will up clocks. I also don't understand why some people seem to have issues with it. If it improves performance, great.

If it merely allows them to claim some artificial TDP, then it's just marketing BS. I'm assuming it's like you say it is. Wouldn't be too surprised though if it wasn't.

 

[RussianSensation]

If you look at those charts from PC Perspective, GTX680 has a pretty large lead over HD7970.

It's 20% faster in Unigine with Extreme Tessellation, has a 20-25 fps lead in Dirt 3, and the performance difference in SKYRIM is huge. Basically, it would take an 1100mhz+ HD7970 to just match this card out of the box. Most air cooled 7970's top out at 1200mhz and 1250mhz requires 1.3V. Some air cooled models can't even hit 1200mhz. The power consumption of HD7970 at 1.3V and 1250mhz just takes off. Normally I could care less about power consumption but here is a card that gives HD7970 overclocked level of performance out of the box and consumes less power and costs $50 less. That's not even discussing more overclocking headroom on the 680.

I think the TurboBoost is only 1058mhz. So if GTX680 can net another 200mhz overlcock, it's prob. going to be a solid win. Looking at 3DMark 11 scores, theoretical performance of HD7970 @ 1100mhz or so just is needed to match a stock 680. Their gaming benches are also reflecting that.

 

[boxleitnerb]

Skyrim is without MSAA, though. Maybe a suggestion of the reviewer guide.

 

[Grooveriding]

Skyrim is without MSAA, though. Maybe a suggestion of the reviewer guide.

I think so. The leak from Tom's was the same, Skyrim only used FXAA. This is why I want to see all the reviews myself, particularly from the sites that don't follow the review guide and/or offer overall summaries at different resolutions/AA levels.

 

[MrK6]

Skyrim is without MSAA, though. Maybe a suggestion of the reviewer guide.

If they're using the complete Ultra preset, AA should be at 8x MSAA. Hopefully the article itself clears that up.

 

[Diceman2037]

I'm guessing the more reputable and meticulous reviewers will run games with GPU Boost disabled on the GTX 680 when comparing vs. 7970, stock vs. stock, but will turn GPU Boost on and overclock when comparing them OC vs. OC.

Also, minimum fps matters. If GPU Boost is a cheap trick to boost averages without also raising minimums, it's useless. (Check out that huge spike in Skyrim that subsequently craters. It helps the average framerate but doesn't seem to help minimums much. Then again, who knows whether GPU Boost was enabled or not in these possibly-real charts?)

Soon enough we'll find out how much OC headroom is in the GTX 680.

if they want their review worth anything they'll leave the cards settings as it was sent to them.

you don't see reviewers disabling Intel turbo because it puts amd at a disadvantage. It is part of the stock design and interfering with it nulls the review.

 

[kidsafe]

if they want their review worth anything they'll leave the cards settings as it was sent to them.

you don't see reviewers disabling Intel turbo because it puts amd at a disadvantage. It is part of the stock design and interfering with it nulls the review.

What I personally want to see is the the boost turned off and both cards OC'd to their fullest potential...but that's me.

 

[RussianSensation]

Size comparison to HD6990:

 

[blastingcap]

What I personally want to see is the the boost turned off and both cards OC'd to their fullest potential...but that's me.

As an enthusiast, I agree. I don't buy cards to run them at stock, boost or no boost.

 

[Diceman2037]

what i personally want to see

i want a card that beats the oppositions without requiring screwing around with clocks, voltages and drivers.

 

[hawtdawg]

i want a card that beats the oppositions without requiring screwing around with clocks, voltages and drivers.

because overclocking a GPU is incredibly difficult...


edit: how am i quoting a post that is below me?

 

[RussianSensation]

As an enthusiast, I agree. I don't buy cards to run them at stock, boost or no boost.

There should be 2 sections in the review:

1) Stock vs. Stock out of the box performance. They can even throw the highest factory pre-overclocked HD7970 they want, which means almost a $600 card.

2) OCed vs. OCed performance.

However, I am pretty sure the TurboBoost is only 1058mhz. Looking at those charts, and comparing GTX680 to HD7950 reveals extreme overpricing of HD7950 already. HD7950 for $450? That's absurd because it will need a 40% overclock just to match an out of the box 680. It's like buying an FX8120 and overclocking it 40% to match a stock 2600K with Turbo Boost enabled.

 

[jivephish]

i want a card that beats the oppositions without requiring screwing around with clocks, voltages and drivers.

I want a card that also makes a decent tuna sandwich.

 

[Silverforce11]

For the 2nd time, I am not implying [H] is biased.

They usually have a review of 4-5 sh1ty games, but they do it honestly as far as I can tell.

You mean 4-5 of the best current games from different genres.. if you aren't implying anything why do you feel the need to add your second line denigrating [H]? Is that really necessary?