OFFICIAL KEPLER "GTX680" Reviews

[lavaheadache]

Is it just me or does anybody else smell a thread lock and vacations coming?

680 is not midrange. When was the last time you saw a next gen midrange card decimate the previous gen highend ?

 

[chimaxi83]

The "680 is midrange" is just the current band wagon mantra for the green team, repeat it enough times and you might believe it too.

 

[SirPauly]

So your view has changed on this value you trumpeted ad nauseaum in this forum now that nvidia has released their card with the worst perf/$ improvement in their flagship ever ?

1 2 3 4 5 6 7 8

Those are the tip of the iceberg, probably 20 more posts like that from you.

vs

?

Really is very bad form for you to have done such a flip flop. No consistency and your opinion on the exact same metric going polar opposite shows some clear bias. :thumbsdown:

Right on target those posts and consistent.

 

[Grooveriding]

Biggest factor is locked voltage, really. I'm not sure why [H] review puts the 680 as OC/OV, you can't volt this thing yourself lol.

That TF3 680 is the same bs, locked voltage. Windforce 680, same thing. Most promising one so far looks to be the Zotac Extreme Edition, default voltage of that is around 1.212V which is higher than the current max on the reference cards.

It's been confirmed via EVGA that custom cards will not have voltage control.

http://www.evga.com/forums/tm.aspx?high=&m=1531236&mpage=1#1531283

Signature cards in just a few weeks now.

Yes, extra power phases and 8+6pin power should allow higher OC, but just how much we are still testing.

Unfortunately the voltage limit/ceiling cannot be raised.

I can't see what use extra power phases are if you can't feed the GPU more volts ? I'm sure there is going to be some advantage though, or what is the point.

My guess on that Zotac card is that they will hardwire in more voltage to the core or put some sort of PWM with voltage control ahead of the PWM chip that is hard-locked and takes care of voltage adjustments for the boost feature.

 

[BallaTheFeared]

When was the last time you saw a 195w mid range card?

The 680 is a doubling of the mid ranged GTX 560ti, this isn't news at all.

http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review

The fact that it keeps needing to be repeated is just sad. It's both sad for AMD that they can't beat Nvidia's mid ranged, and it's sad that Nvidia is charging $500 for a small die mid ranged product...

Which is worse? I dunno, I didn't buy anything.

 

[chimaxi83]

So then, generation after generation, card after card, they all have to have the same power consumption? So then you're expecting this generations high end to be another Fermi watt destroyer? We don't believe in tech and efficiency advancements?

Oh, ROP count and bus width are the same as 560ti so it must be its replacement, not that efficiency, reworked memory controller, and a new architecture lowered the components they had to put on board or anything

 

[SirPauly]

The "680 is midrange" is just the current band wagon mantra for the green team, repeat it enough times and you might believe it too.

I don't like smaller chips in monolithic price points -- both AMD and nVidia.

However, using your context if the GTX 680 is a mid-range part then what does this mean for the HD 7970? Also mid-range? Less than mid-range?

I think they're both performance parts but based on maybe added costs, yields, constraints and both being a predator and aggressor -- priced where they are.

 

[blackened23]

Biggest factor is locked voltage, really. I'm not sure why [H] review puts the 680 as OC/OV, you can't volt this thing yourself lol.

That TF3 680 is the same bs, locked voltage. Windforce 680, same thing. Most promising one so far looks to be the Zotac Extreme Edition, default voltage of that is around 1.212V which is higher than the current max on the reference cards.

So they all have locked voltage? :thumbsdown:

I hope the MSI lightning changes this. It looks like the MSI TF3 680 will not be the lightning, and it appears to have 2x 6 pin connectors. Pointless :thumbsdown: I want the MSI lightning monster GTX 680. Unfortunately the reference 680 only seems to scale around 16% with max offset overclocks, the lack of more power and voltage is surely preventing additional scaling. Whats more dismaying is that some aftermarket cards still have the voltage limitation. WTF?

 

[Grooveriding]

The "680 is midrange" is just the current band wagon mantra for the green team, repeat it enough times and you might believe it too.

Pretty much this. It makes some folks warm and fuzzy inside convincing themselves that nvidia is so monstrous their mid-range chip is fast enough to trump AMD's high end chip.

So long as we are not seeing a strategy shift at nvidia in making more efficient, more easily manufactured and smaller consumer GPUs, rather than big, hot PITAs; I look forward to GK110 when/if it gets manufactured on a large viable scale and brought to the desktop consumer as GTX 780.

 

[chimaxi83]

So they all have locked voltage? :thumbsdown:

Everything shown so far, yes. I was disappointed at the TFIII 680, figured that would have it, but its the same as all the others. Some have upgraded PCB components like power phases/caps, but no manual control.

 

[boxleitnerb]

The 7970 is not a high end chip either
Both are performance chips, more or less.

 

[chimaxi83]

I don't like smaller chips in monolithic price points -- both AMD and nVidia.

However, using your context if the GTX 680 is a mid-range part then what does this mean for the HD 7970? Also mid-range? Less than mid-range?

I think they're both performance parts but based on maybe added costs, yields, constraints and both being a predator and aggressor -- priced where they are.

Is it so hard to grasp that perhaps each company has streamlined their designs, improving efficiency and performance/mm2, that they don't need the massive dies Nvidia has you all accustomed to?

 

[blackened23]

Everything shown so far, yes. I was disappointed at the TFIII 680, figured that would have it, but its the same as all the others. Some have upgraded PCB components like power phases/caps, but no manual control.

This makes water cooling near 100% useless. I don't understand it. Throw the power users a bone, or something.

 

[BallaTheFeared]

Pretty much this. It makes some folks warm and fuzzy inside convincing themselves that nvidia is so monstrous their mid-range chip is fast enough to trump AMD's high end chip.

So long as we are not seeing a strategy shift at nvidia in making more efficient, more easily manufactured and smaller consumer GPUs, rather than big, hot PITAs; I look forward to GK110 when/if it gets manufactured on a large viable scale and brought to the desktop consumer as GTX 780.

For our look at Kepler’s architecture, we’re going to be primarily comparing it to GF114, aka Fermi Lite. As you may recall, with Fermi NVIDIA had two designs: a multipurpose architecture for gaming and graphics (GF100/GF110), and a streamlined architecture built with a stronger emphasis on graphics than compute (GF104, etc) that was best suited for use in consumer graphics. As hinted at by the name alone, GK104 is designed to fill the same consumer graphics role as GF114, and consequently NVIDIA built GK104 off of GF114 rather than GF110.

:awe:

Power users should wait for BigK :whiste:

 

[notty22]

Kyle refers to this as the voltage adjustment they were allowed to make within prescision. Everything was dynamically moving , but this slight bump moved the range up. According to him. Though power pull stayed under +132% avg +120%.
It also seems the discussion thread for the article is locked.

EVGA Precision X has a button on the left side called Adjust Voltage. The default voltage is set to 0.987 and the highest you can set this too is 1.150. Now, keep in mind that GPU Boost is auto-adjusting the voltage already in its default auto-clocking routine. Therefore raising the VGPU here, isn't going to give it a lot of extra voltage, but we'll explain that below in our overclocking with voltage testing.

 

[Grooveriding]

This makes water cooling near 100% useless. I don't understand it. Throw the power users a bone, or something.

It's good for silence and cool factor at least Mine run at high 30s to low 40s under BF3 load :ninja:

 

[blackened23]

I like your case layout :thumbsup: I need to do some spring cleaning in mine. I guess you still benefit since you don't hit the 70C downclocking limit, other users are stating that 70c is the soft "cap" where the cards start downclocking. So you run at 100% more often than not, while someone with reference will be downclocked at 70C.

May get the Maximus V when it comes out, I kinda like those mATX boards.

 

[lavaheadache]

When was the last time you saw a 195w mid range card?

The 680 is a doubling of the mid ranged GTX 560ti, this isn't news at all.

http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review

The fact that it keeps needing to be repeated is just sad. It's both sad for AMD that they can't beat Nvidia's mid ranged, and it's sad that Nvidia is charging $500 for a small die mid ranged product...

Which is worse? I dunno, I didn't buy anything.

I think you need to look up stock 7970 power consumption. Maybe it'll change your tune about it being "highend", since the same makes you think the 680 is mid range

 

[chimaxi83]

This makes water cooling near 100% useless. I don't understand it. Throw the power users a bone, or something.

tell me about it. I just might pick up a custom card with upgraded VRM and go for Vmod, it's not like we haven't had to do this before lol

Uncorking Kepler has both a regular Vmod to unlock up to around 1.5V, which doesn't look too bad, and also shows that big ass solder-on board with extra power phases for the insane speeds/power. They had to use one of those for that 3dmark11 record run.

 

[Quantos]

The 7970 is getting beat by Nvidia mid range.

The 680 is a $500 mid range part...

ಠ_ಠ

This again?

Let's see, here's how Nvidia's current offer is like:

[...] GTX 570, GTX 580, GTX 680, GTX 590.

Yep, sure looks midrange to me. I know what you're going to answer. You say that:

- It's GK104, not 110;
- It's a smaller die;
- It was intended all along to be mid ranged and switched names at the last minute because it was enough to beat 7970.

Do I really need to remind you that all this is absolutely irrelevant until Nvidia has other products on the market?

 

[railven]

So your view has changed on this value you trumpeted ad nauseaum in this forum now that nvidia has released their card with the worst perf/$ improvement in their flagship ever ?

<snip>


1 2 3 4 5 6 7 8

Those are the tip of the iceberg, probably 20 more posts like that from you.

vs

<snip>


?

Really is very bad form for you to have done such a flip flop. No consistency and your opinion on the exact same metric going polar opposite shows some clear bias. :thumbsdown:

Proof is in the pudding as they say. All praise for nVidia for charging GTX x80 prices on the GTX x60 replacement (even Keys called out people who were making that claim before it was official) and only mud on AMD for giving up the "sweet spot."

He ignores what I said, he started the Go premiums chant I ran with. I've yet to see him say it once in the face of Kepler. Consistency? Sure.

 

[SirPauly]

Is it so hard to grasp that perhaps each company has streamlined their designs, improving efficiency and performance/mm2, that they don't need the massive dies Nvidia has you all accustomed to?

I would say they do need bigger dies if one desires more performance for 28nm moving forward; to bring in more revenue for GeForce, Quadro and Tesla.

 

[chimaxi83]

&#3232;_&#3232;

This again?

Let's see, here's how Nvidia's current offer is like:

[...] GTX 570, GTX 580, GTX 680, GTX 590.

Yep, sure looks midrange to me. I know what you're going to answer. You say that:

- It's GK104, not 110;
- It's a smaller die;
- It was intended all along to be mid ranged and switched names at the last minute because it was enough to beat 7970.

Do I really need to remind you that all this is absolutely irrelevant until Nvidia has other products on the market?

Repeat it over and over to yourself, you'll start to believe it too lol.

I would say they do need bigger dies if one desires more performance for 28nm moving forward; to bring in more revenue for GeForce, Quadro and Tesla.

I'm sure their bigger die is on it's way. Nothing wrong with being able to provide performance improvements in a smaller package though.

Proof is in the pudding as they say. All praise for nVidia for charging GTX x80 prices on the GTX x60 replacement (even Keys called out people who were making that claim before it was official) and only mud on AMD for giving up the "sweet spot."

He ignores what I said, he started the Go premiums chant I ran with. I've yet to see him say it once in the face of Kepler. Consistency? Sure.

As predicted, they all bashed AMD for their pricing, but when Nvidia provided the same lackluster performance improvement over their 580, it's suddenly TAKE MY MONEY NAO.

I like your case layout :thumbsup: I need to do some spring cleaning in mine. I guess you still benefit since you don't hit the 70C downclocking limit, other users are stating that 70c is the soft "cap" where the cards start downclocking. So you run at 100% more often than not, while someone with reference will be downclocked at 70C.

May get the Maximus V when it comes out, I kinda like those mATX boards.

Nice setup indeed. Gotta love those load temps :biggrin:

Putting mine under water increased my offset by about +20, so nothing to write home about. Decreased the power usage though, at least it looks like it did. Went from very frequent 132% on the Afterburner graphs, to not passing 106% since it's been under water. If only it had more juice.

 

[BallaTheFeared]

I think you need to look up stock 7970 power consumption. Maybe it'll change your tune about it being "highend", since the same makes you think the 680 is mid range

Power draw has nothing to do with what makes me think 680 is mid range. What I'm saying is if you increase the TDP of Nvidia's mid range design it will quickly exceed AMD's "high end" as seen with the 460 vs 5870.

It's design is based on Nvidia's mid range design. GF114, a compute lacking, gaming oriented, lower performance product.

Scroll up and check out what happens when they give GF104 (GTX 460) considerably more TDP with a 900 core clock. It beats the 5870, and the 470.

I'm having a hard time figuring out why I'm having to show this... It should be common knowledge since any reviewer doing a semi in-depth review of the 680 would have mentioned all this.

 

[blackened23]

As far as i'm aware, nvidia specifically sought to make the kepler their most power efficient GPU ever.