OFFICIAL KEPLER "GTX680" Reviews

[blackened23]

Because I fabricated some results over at Guru3D. Don't waste forum space with your paranoia replies. Nothing was sensational about my post. It's pretty cut and dry that SLI is not scaling well on Kepler yet.

Nothing in my comments said it's a nail in the coffin or final.

Dude, just don't reply to my posts if you can only offer your usual subtle and immature jabs.

Taking passive aggressive jabs when anything isn't completely and utterly 100% positive, news at 11. Just take your 680s to the AMD lab and dissect them for any dirt you can dig, thats obviously what people on another board think. I know I spit my coffee up in laughter from reading it the last time I went there, some paranoid conspiracy theorists. I think they all believe in UFOs too.

 

[krumme]

Was looking at the PCB layout, its components and other things and one things for sure. The card is downright cheaper to produce than the HD7970 period. I mean for starters they got away with a 4 phase VRM design using cheap components, a very simple controller for the VRMs, heatpipe based heatsink that looks like ~20% smaller than the vapor chamber the HD7970 uses, yet the card can sustain high clocks, and draw less power than the HD7970 even when overclocked. That is simply impressive in my view. The margins for this card could be huge!

It has "performance" card screaming all over it rather than being a high-end/enthusiast card. Wish they priced this at $349 or $399.

Unfortunately its not that simple, even though it seems like a major step forward for NV.

On the old fermi platform NV could sell the lower binned chips for the professional market, leaving the high performance for the consumer segment. That was a brilliant strategic move. They dont have that opportunity any more. How does that effect the total cost? - it remains to be seen.

Secondly, the pitcarn 7870 still have the efficiency edge aiming for the most interesting segment at 200usd (the cards prices will get there). Basicly AMD architechture still seems to have the edge perf/mm2. What is NV response for this important segment?

Third. 680 is just pushed to the max, quite opposite to the 7970, giving AMD all the oportunity for

a. Keep the OEM releasing oc models, giving them a good sales force here - they need that more than NV, because their brand is not so strong
The added benefit here is more working dies keeping cost down

b. Simply releasing a faster card at 1200Mhz in small number, reclaiming the halo title for marketing purpose

The point is. AMD have the opportunity to choose.

Then regarding the cost. The cost of the fast ddr5 for NV, must be a real burden now, the cost of VRM wider bus, simple pcb layout, whatever must pale in comparison to that. But i guess its a much more future oriented way of adressing the total cost when ddr5 fast ram lowers in price. NV have a solid platform for selling to the gaming market the next year.

 

[n0x1ous]

http://www.guru3d.com/article/geforce-gtx-680-sli-review/1

Guru3D SLI benchmark to see SLI scaling on 680s.

Making me grimace a bit at buying two. I hope this is just a case of driver issues. Fermi SLI scaling is generally close to 100% with two cards, whereas Kepler is about 50-70%.

Not the best, but drivers are young. OTOH, scaling looked excellent in BF3 at your Rez and I think you said that was one of your main games

 

[Grooveriding]

Not the best, but drivers are young. OTOH, scaling looked excellent in BF3 at your Rez and I think you said that was one of your main games

BF3 does look good, no doubt.

I expect it is just drivers and a new architecture. I was just linking the review and mentioning what I noticed. I think we'd all like to see the sort of scaling we saw on Fermi, and I'm sure with driver improvements we will.

Am I doing it right now ?

 

[railven]

BF3 does look good, no doubt.

I expect it is just drivers and a new architecture. I was just linking the review and mentioning what I noticed. I think we'd all like to see the sort of scaling we saw on Fermi, and I'm sure with driver improvements we will.

Am I doing it right now ?

No, you're not doing it right, because there is no problem. Use the newest WHQL drivers and stop trolling.

Clearly drivers will make a huge difference with this card. So much to balance. Hell, that Tweak Town review on old drivers showed these cards in a bad light (power consumption.)

 

[Keysplayr]

What begins? He only made one factual observation. If we're supposed to wait for the next whql drivers before we can express opinions on this release, then the thread you've made and the 700 comments wouldn't have much of a purpose, since no card tested has a younger forceware than 300.99

Very true. I was more concerned for Groover's seemingly knee-jerk superfast purchase of the 680s. Perhaps he should have waited a bit longer just to ensure his satisfaction with all areas of performance was met. He probably should have waited for SLI benchmarks to come round before buying two. I'm sure you would agree with me on this.
Then again, it is possible that he did consider that Kepler is a new arch with brand new drivers into his purchase. Pretty much justifying his quick trigger if he anticipated that.

 

[Olikan]

interesting review of David kanter

http://www.realworldtech.com/page.cfm?ArticleID=RWT032212172023

it seems that even bigK, won't be better than fermi at HPC

 

[n0x1ous]

Am I doing it right now ?

Doing what? Did I miss something?

 

[-Slacker-]

To Keys: Fair enough

 

[antef]

Can anyone comment on the chances of a 660 Ti being out by May? I want to buy then and will probably be looking in the $250 price range.

 

[Pheran]

Can anyone comment on the chances of a 660 Ti being out by May? I want to buy then and will probably be looking in the $250 price range.

That is highly unlikely. You'll probably just be getting good availability on the 680 by May.

 

[Gryz]

Quick question:

Will TXAA be available in all games ? Can we enable it via the control panel ?

Or will each game have to change code to allow for TXAA ? So only new games might allow us to use TXAA ? Or older games which get a specific patch to support TXAA ?

I've read different things in different reviews. Some reviews mention the control panel. But other reviews say that we have to wait for games to support TXAA. Confused. Btw, the main game where I would like to try TXAA at the moment is Skyrim. I don't expect Bethesda to include new technology in their patches ever. So for Skyrim we would depend on the control panel.

 

[bryanW1995]

Small comparison here, TXAA doesn't necessarily look blurry, but it's also not a full screen shot. I like how it looks a lot better than MSAA though (in that shot).

From Legit Reviews:


Not implemented until later this year, meh. But they claim 16xMSAA quality with the cost of 4xMSAA. Hopefully we get an in depth IQ comparison. Until then, I'll do my own comparison when my card gets here

With apologies to grooveriding, this sounds more like the anti-physx. Physx kills performance for tiny visual gains, TXAA HELPS performance for tiny visual gains. Hopefully, their patch will also cover gtx 480. And those games that will theoretically support it in the future will actually be decent titles. And, um,...ok, it does kinda sound like physx. At least this time it doesn't matter b/c they have the performance crown, anyway, and wreckage isn't here to talk about TXAA in every single thread that doesn't prominently feature positive NV spin.

 

[bryanW1995]

What begins? He only made one factual observation. If we're supposed to wait for the next whql drivers before we can express opinions on this release, then the thread you've made and the 700 comments wouldn't have much of a purpose, since no card tested has a younger forceware than 300.99

(withholding comment until next whql drivers are released)

 

[Creig]

Nvidia's "Dawn" to return as a GTX 680 demo?

http://www.youtube.com/watch?v=SjhM_xc6Yes

 

[bryanW1995]

Quick question:

Will TXAA be available in all games ? Can we enable it via the control panel ?

Or will each game have to change code to allow for TXAA ? So only new games might allow us to use TXAA ? Or older games which get a specific patch to support TXAA ?

I've read different things in different reviews. Some reviews mention the control panel. But other reviews say that we have to wait for games to support TXAA. Confused. Btw, the main game where I would like to try TXAA at the moment is Skyrim. I don't expect Bethesda to include new technology in their patches ever. So for Skyrim we would depend on the control panel.

According to the statements I've read, it requires actual work from the game developer to be usable at all. There are supposed to be one or more games coming out later this year that have it. Basically, it sounds like 16xMSAA for the price of 4xMSAA.

 

[Grooveriding]

With apologies to grooveriding, this sounds more like the anti-physx. Physx kills performance for tiny visual gains, TXAA HELPS performance for tiny visual gains. Hopefully, their patch will also cover gtx 480. And those games that will theoretically support it in the future will actually be decent titles. And, um,...ok, it does kinda sound like physx. At least this time it doesn't matter b/c they have the performance crown, anyway, and wreckage isn't here to talk about TXAA in every single thread that doesn't prominently feature positive NV spin.

You're actually right in that respect, TXAA will not have a performance impact. I don't like post-AA modes so was never swayed one way or the other as I won't use it. At least for people who don't mind trading IQ for better performance it will be a benefit in however many games end up using it.

 

[Vesku]

Consensus on the 680 out of the box having almost no remaining performance to unlock through additional OCing? Personally I'm all for this kind of auto clock and power monitoring as I'm not into watercooling. I have to wonder about whether there is anything to gain from improved cooling other than lower noise.

 

[blackened23]

I got my 2 cards this morning. I am *not* happy with the SLI scaling and overclocking.

Thats all I will say for now until I do some more testing. On the other hand, the cards are audibly silent. Much, much more silent than the 7970s by a TON. I am downright amazed at how cool and quiet these cards are , they are a home run for nvidia in that respect. Props to them for finally nailing efficiency.

 

[Keysplayr]

Consensus on the 680 out of the box having almost no remaining performance to unlock through additional OCing? Personally I'm all for this kind of auto clock and power monitoring as I'm not into watercooling. I have to wonder about whether there is anything to gain from improved cooling other than lower noise.

Probably best to ask an engineer in this field, but I have read and heard
that less heat = less power consumed.

 

[exar333]

Unfortunately its not that simple, even though it seems like a major step forward for NV.

On the old fermi platform NV could sell the lower binned chips for the professional market, leaving the high performance for the consumer segment. That was a brilliant strategic move. They dont have that opportunity any more. How does that effect the total cost? - it remains to be seen.

Secondly, the pitcarn 7870 still have the efficiency edge aiming for the most interesting segment at 200usd (the cards prices will get there). Basicly AMD architechture still seems to have the edge perf/mm2. What is NV response for this important segment?

Third. 680 is just pushed to the max, quite opposite to the 7970, giving AMD all the oportunity for

a. Keep the OEM releasing oc models, giving them a good sales force here - they need that more than NV, because their brand is not so strong
The added benefit here is more working dies keeping cost down

b. Simply releasing a faster card at 1200Mhz in small number, reclaiming the halo title for marketing purpose

The point is. AMD have the opportunity to choose.

Then regarding the cost. The cost of the fast ddr5 for NV, must be a real burden now, the cost of VRM wider bus, simple pcb layout, whatever must pale in comparison to that. But i guess its a much more future oriented way of adressing the total cost when ddr5 fast ram lowers in price. NV have a solid platform for selling to the gaming market the next year.

You are WAY over-simplifying a very complex process. NV likely has better yields + lower costs with a much smaller wafer size. They should be able to continue to re-use for consumer products just fine, as as for mobile offerings.

The 'big die' kepler will be released later this year (presumably) and would continue to leverage the gains you speaking of that they had with Fermi. With the high-end availability issues, I doubt NV will likely lose anything, and will have much more for gain.

I would not be surprised to see AMD follow-suit in the next iteration and cut-out (or disable) much of the computational parts of their consumer GPUs. I think we will see a much more stratified GPU sector that will look much like what we have on the Intel CPU side. 'Enthusiast' line that is essentially re-badged professional parts vs. efficient mainstream parts. I don't think this is a bad thing at all actually, if it keeps the mainstream part costs down.

 

[Janooo]

Probably best to ask an engineer in this field, but I have read and heard
that less heat = less power consumed.

It does not work like that. Less power -> less heat to take care of.
With better cooler you are opening the door for more power, higher performance.

 

[Lonyo]

It does not work like that. Less power -> less heat to take care of.
With better cooler you are opening the door for more power, higher performance.

Check the CPU forum.
IDC did some testing with his CPU, where he applied different levels of cooling to set clock speeds and voltages, and measured power consumption.

As the CPU temp increased, the power used by the CPU also increased.
Load was constant, clockspeed was constant, voltage was constant. CPU temperature was the only variable changed by altering the level of cooling on the CPU (adding and removing fans from the heatsink, putting a box over it to trap heat etc).

As the CPU temp went up, power consumption went up.

 

[Tempered81]

hey janoo i disagree with keys 9 times outa 8, but hes right about temps and power. not saying you are wrong, but i wouldn't immediately discredit his statement.

1.2v gpu at 30C liquid is going to use less power than the same 1.2v gpu with the block off on an air cooler running at 90C.

 

[railven]

Check the CPU forum.
IDC did some testing with his CPU, where he applied different levels of cooling to set clock speeds and voltages, and measured power consumption.

As the CPU temp increased, the power used by the CPU also increased.
Load was constant, clockspeed was constant, voltage was constant. CPU temperature was the only variable changed by altering the level of cooling on the CPU (adding and removing fans from the heatsink, putting a box over it to trap heat etc).

As the CPU temp went up, power consumption went up.

hey janoo i disagree with keys 9 times outa 8, but hes right about temps and power. not saying you are wrong, but i wouldn't immediately discredit his statement.

1.2v gpu at 30C liquid is going to use less power than the same 1.2v gpu with the block off on an air cooler running at 90C.

Yet, if you read what he said - that's basically what he said.

Better cooler == less heat == less power

Better cooler == less heat thus you can use more power BEFORE heat causes the increased power.

What IDC proved that a better cooler would result in less power used and thus opening the door for better OCing.