NVIDIA GeForce GTX 680 Kepler 2 GB

[RigRebel]

yea man exactly, thats what was confusing me, and i tried spotting the differences but still cant put my hand on what is going on, but owell in general the 680 has an edge in gaming on single gpu, but its pretty close in my opinion
as for scaling yea i saw something like that, and it said that the inconsistent clocks on 4 different cards or so make the scaling very difficult to achieve(remember usualy sli and xfire setups clock according to the lowest card in the bunch, in this case each card clocks differently) so if thats the case im assuming future drivers might disable dynamic clocking(thats the easy solution) or find some clever way to sort it out and i know nvidia's drivers team is pretty talented and can eventually do it, but i could be wrong as i know very little about this issue other than what i read in the article I saw

They did disable dynamic (turbo) clocking I believe in that one.. wish I had the review. I have the link in my work email just can't get to it tonight. I'll re-post on monday with the review link.

 

[N3M3515]

Lol I love these so called "fair" assessments ... they will over clock the 7970 to make it fair with the GTX 680 yet fail to mention the 1GB advantage of the 7970 cards lol. If they wanted to make it completely "fair" just to compare clock for clock, they'd find a way to nerf the extra 1GB of ram the 7970 has lol ... I know that probably can't be done but still not exactly "fair".

I don't know what exactly are you talking about, i just read people that would like to see how the two cards compare when overclocked.
I don't think they overclock the 7970 to "make it fair", they compare it to see which one performs better over the stock clock.

 

[MxPhenom 216]

I don't know what exactly are you talking about, i just read people that would like to see how the two cards compare when overclocked.
I don't think they overclock the 7970 to "make it fair", they compare it to see which one performs better over the stock clock.

yeah they cant just overclock one card. they would have to overclock the other to the same percentage overclock for it to be fair no matter the clock number, since each card acts differently when clocked scaling wise.

 

[RigRebel]

I don't know what exactly are you talking about, i just read people that would like to see how the two cards compare when overclocked.
I don't think they overclock the 7970 to "make it fair", they compare it to see which one performs better over the stock clock.

Yes they did.. They say in the first sentence. Direct quote > "I saw it argued in the comments that the only fair way to compare a GeForce GTX 680 to a Radeon HD 7970 is with both cards overclocked. Because the 680 employs GPU Boost, it purportedly already demonstrates performance close to its limit, whereas Tahiti-based GPUs are known to offer quite a bit of headroom. "

Again notice the word "fair" ... this is the second type review I've read and the first one said same thing but went on to further say something like "the only fair way to tell is to compair clock to clock" .. again they say this but they forget that the 7970 has 1GB more ram so what's "fair" about that ... hypocritical IMO. I'll post the other review on monday it's in my work mail.

Who cares about clock for clock if the end result is getting as much overclocking as you can in the first place. Isn't that the whold point of Overclocking?... to see just how high you can raise the ceiling and how much you can push out of it... The GTX 680 can raise the ceiling to @1340Mhz...and the AMD is somewhere around @1141Mhz I think. Comparing "clock to clock" in a competition where total overclocking is the goal is just another way of AMD side not admiting defeat and wanting to pick it apart and test anything in their favor, IMO. AGain I use the analogy.. so what if your honda is more efficent per cyclinder... a mustang GT will still smoke ya. lol

 

[sergionography]

They did disable dynamic (turbo) clocking I believe in that one.. wish I had the review. I have the link in my work email just can't get to it tonight. I'll re-post on monday with the review link.

are you sure, I think at this point that's not possible, I think wizz even mentions it when over locking and sais even when u overclock u pretty much raise the boast ceiling rather than staying on that frequency, tho I still don't know much about gpu boast,to me it sounds decieving in theory, when I first Thor about it I figured it would give out inconsistent fps, giving you crapload fps on easy to render and scenes that are light on the gpu, but then when the gpu is stressed you get to a limitation, so overall increasing average fps but the experience pretty much remains the same, tho NVIDIA kept the boast minimal I'm assuming to avoid this issue, however if that boast was lik 25% increase in clock just think how inconsistent fps will be, idk if I'm totally looking at this the wrong way, but I'm assuming this is why Keller doesn't seem to scale as much as Tahiti in overclocking

 

[N3M3515]

Yes they did.. They say in the first sentence. Direct quote > "I saw it argued in the comments that the only fair way to compare a GeForce GTX 680 to a Radeon HD 7970 is with both cards overclocked. Because the 680 employs GPU Boost, it purportedly already demonstrates performance close to its limit, whereas Tahiti-based GPUs are known to offer quite a bit of headroom. "

Again notice the word "fair" ... this is the second type review I've read and the first one said same thing but went on to further say something like "the only fair way to tell is to compair clock to clock" .. again they say this but they forget that the 7970 has 1GB more ram so what's "fair" about that ... hypocritical IMO. I'll post the other review on monday it's in my work mail.

Who cares about clock for clock if the end result is getting as much overclocking as you can in the first place. Isn't that the whold point of Overclocking?... to see just how high you can raise the ceiling and how much you can push out of it... The GTX 680 can raise the ceiling to @1340Mhz...and the AMD is somewhere around @1141Mhz I think. Comparing "clock to clock" in a competition where total overclocking is the goal is just another way of AMD side not admiting defeat and wanting to pick it apart and test anything in their favor, IMO. AGain I use the analogy.. so what if your honda is more efficent per cyclinder... a mustang GT will still smoke ya. lol

And yet another AMD vs NVIDIA thing...........

 

[perryra1968]

AMD 5970...$299.00...Black Friday.....Still is the best deal I ever had.....ever! It'll be a while before I upgrade. BF3 at 19X12 is only 11 FPS difference between 5970 and 680. Thanks Newegg!

 

[N3M3515]

yeah they cant just overclock one card. they would have to overclock the other to the same percentage overclock for it to be fair no matter the clock number, since each card acts differently when clocked scaling wise.

They overclocked each card to the max they could i think.......so it is not about clocking the to to the same %, and hey i don't care about fairness, i care about seing how much each one gains from o/c and from the looks of it 7970 scales better imho.

 

[sergionography]

Yes they did.. They say in the first sentence. Direct quote > "I saw it argued in the comments that the only fair way to compare a GeForce GTX 680 to a Radeon HD 7970 is with both cards overclocked. Because the 680 employs GPU Boost, it purportedly already demonstrates performance close to its limit, whereas Tahiti-based GPUs are known to offer quite a bit of headroom. "

Again notice the word "fair" ... this is the second type review I've read and the first one said same thing but went on to further say something like "the only fair way to tell is to compair clock to clock" .. again they say this but they forget that the 7970 has 1GB more ram so what's "fair" about that ... hypocritical IMO. I'll post the other review on monday it's in my work mail.

Who cares about clock for clock if the end result is getting as much overclocking as you can in the first place. Isn't that the whold point of Overclocking?... to see just how high you can raise the ceiling and how much you can push out of it... The GTX 680 can raise the ceiling to @1340Mhz...and the AMD is somewhere around @1141Mhz I think. Comparing "clock to clock" in a competition where total overclocking is the goal is just another way of AMD side not admiting defeat and wanting to pick it apart and test anything in their favor, IMO. AGain I use the analogy.. so what if your honda is more efficent per cyclinder... a mustang GT will still smoke ya. lol

i dont think they mean clock for clock, i think what they mean is how much performance can you squeeze out of both cards for the end user who is interested in overclocking.
im not sure on the top overclocks records but i know some people got real high overclocks with the hd7970.
of about 1180 or something, so while that is still lower than kepler, it is like 20% from stock, 1340 for kepler is also about 20% off from stock(since kepler already clocks to 1100 when needed) yet it doesnt scale well(because overclocking kepler is complex and it pretty much makes the max boast 1340 rather than making it native)
but i get what you mean tho, there is no such thing as "fair"
nvidia did a clever move to get as much power out of the envelope as possible, and they did a darn good job, not to mention how handy that will be for oems who are on a restricted thermal/power envelope for laptops and what not.
having oems and power restrictions in mind its interesting tho to see how both teams had their approach to efficiency and battery life, amd pushing for max idle time with zero power, and nvidia pushing for different power states and turbo and what not(and it makes sense since nvidia for the most part always gets paired with an intel cpu that has a integrated gpu for low power. meaning nvidia's optimus does the job for them)

EDIT: i just came across these 2 benchmarks of both cards maxed out with their clocks
my conclusion was somewhat correct, hd7970 gains more momentum when overclocked

http://www.overclockersclub.com/reviews/nvidia_gtx680/6.htm

http://www.overclockersclub.com/reviews/powercolor_lcs_hd7970/7.htm

i specifically posted the links on the battlefield because nvidia seemed to have an edge in that game, but overclocked they match eachother, even tho nvidia is clocked at 1300 something while amd is in the 1200 range.


http://www.overclockersclub.com/reviews/nvidia_gtx680/5.htm

http://www.overclockersclub.com/reviews/powercolor_lcs_hd7970/6.htm

same thing going on with arkam city

tho i would take this with a grain of salt because im not sure how legit this website is, but what do you think?

 

[xtremesv]

poeple : "if your 7970 wont drop to $400 - 450, i will give my wallet to nvidia!"

I don't know why people are asking for $100-150 price reduction when GTX 680 is just 5% faster than 7970 on average, that'd mean a price tag of $475 ($75 reduction) for AMD offering to be competitive.

Am I the only one who is not really impressed by this new card

 

[W1zzard]

i dont think they mean clock for clock, i think what they mean is how much performance can you squeeze out of both cards for the end user who is interested in overclocking.
im not sure on the top overclocks records but i know some people got real high overclocks with the hd7970.
of about 1180 or something, so while that is still lower than kepler, it is like 20% from stock, 1340 for kepler is also about 20% off from stock(since kepler already clocks to 1100 when needed) yet it doesnt scale well(because overclocking kepler is complex and it pretty much makes the max boast 1340 rather than making it native)
but i get what you mean tho, there is no such thing as "fair"
nvidia did a clever move to get as much power out of the envelope as possible, and they did a darn good job, not to mention how handy that will be for oems who are on a restricted thermal/power envelope for laptops and what not.
having oems and power restrictions in mind its interesting tho to see how both teams had their approach to efficiency and battery life, amd pushing for max idle time with zero power, and nvidia pushing for different power states and turbo and what not(and it makes sense since nvidia for the most part always gets paired with an intel cpu that has a integrated gpu for low power. meaning nvidia's optimus does the job for them)

EDIT: i just came across these 2 benchmarks of both cards maxed out with their clocks
my conclusion was somewhat correct, hd7970 gains more momentum when overclocked

http://www.overclockersclub.com/reviews/nvidia_gtx680/6.htm

http://www.overclockersclub.com/reviews/powercolor_lcs_hd7970/7.htm

i specifically posted the links on the battlefield because nvidia seemed to have an edge in that game, but overclocked they match eachother, even tho nvidia is clocked at 1300 something while amd is in the 1200 range.


http://www.overclockersclub.com/reviews/nvidia_gtx680/5.htm

http://www.overclockersclub.com/reviews/powercolor_lcs_hd7970/6.htm

same thing going on with arkam city

tho i would take this with a grain of salt because im not sure how legit this website is, but what do you think?

did nobody see that graph ?

both cards oc'd to max

 

[sergionography]

did nobody see that graph ?

http://tpucdn.com/reviews/NVIDIA/GeForce_GTX_680/images/max_oc_vs_7970.gif
both cards oc'd to max

yea i saw these, but ive seen the cards getting pushed more than 16% for an hd7970, that makes it around 1050 or something for the core right? ive seen posts of people getting higher numbers and thats when i got curious. in the benchmarks ive posted they were being overvolted i believe and were pushed further
but you are correct most users will not overvolt and wont achieve the extreme overclocks i was looking at, so pretty much ur conclusion applies

what do you think about the sli scaling mentioned earlier? will you be doing a review about that by any chance? im really curious about that

 

[W1zzard]

that makes it around 1050 or something for the core right?

1080 MHz GPU. Higher was not stable in all tests

We already posted a SLI review, I won't invest more time in this. Got more cards coming this week that need my attention

 

[N3M3515]

i dont think they mean clock for clock, i think what they mean is how much performance can you squeeze out of both cards for the end user who is interested in overclocking.
im not sure on the top overclocks records but i know some people got real high overclocks with the hd7970.
of about 1180 or something, so while that is still lower than kepler, it is like 20% from stock, 1340 for kepler is also about 20% off from stock(since kepler already clocks to 1100 when needed) yet it doesnt scale well(because overclocking kepler is complex and it pretty much makes the max boast 1340 rather than making it native)
but i get what you mean tho, there is no such thing as "fair"
nvidia did a clever move to get as much power out of the envelope as possible, and they did a darn good job, not to mention how handy that will be for oems who are on a restricted thermal/power envelope for laptops and what not.
having oems and power restrictions in mind its interesting tho to see how both teams had their approach to efficiency and battery life, amd pushing for max idle time with zero power, and nvidia pushing for different power states and turbo and what not(and it makes sense since nvidia for the most part always gets paired with an intel cpu that has a integrated gpu for low power. meaning nvidia's optimus does the job for them)

EDIT: i just came across these 2 benchmarks of both cards maxed out with their clocks
my conclusion was somewhat correct, hd7970 gains more momentum when overclocked

http://www.overclockersclub.com/reviews/nvidia_gtx680/6.htm

http://www.overclockersclub.com/reviews/powercolor_lcs_hd7970/7.htm

i specifically posted the links on the battlefield because nvidia seemed to have an edge in that game, but overclocked they match eachother, even tho nvidia is clocked at 1300 something while amd is in the 1200 range.


http://www.overclockersclub.com/reviews/nvidia_gtx680/5.htm

http://www.overclockersclub.com/reviews/powercolor_lcs_hd7970/6.htm

same thing going on with arkam city

tho i would take this with a grain of salt because im not sure how legit this website is, but what do you think?

Yeah, the same pattern can be seen here: http://vr-zone.com/articles/asus-gtx-680-2gb-overclocking-review-win-some-lose-some/15322-4.html

Scalability of the 7970 seems better.

 

[sergionography]

Yeah, the same pattern can be seen here: http://vr-zone.com/articles/asus-gtx-680-2gb-overclocking-review-win-some-lose-some/15322-4.html

Scalability of the 7970 seems better.

that and it seems the early benchmarks hd7970 did real bad in battlefield, like 30% slower than kepler, but here its only like 10fps away at 78 vs 89
but yea it seems the battle in this round is closer than ever, now lets see the prices go down!

and im curious on how the 670 and 660 will stack up against the 7800 series, considering the 7800 series being way more efficient than tahiti.

 

[Moatsim]

think of it in quad SLI !!!!!!!!!!!!!!

 

[Disruptor4]

Can someone explain wtf is with the 5970 being up there challenging the 680 with the performance and actually beating it sometimes? Not to mention the 7970 doesn't seem worthy enough over the 5970. In fact, it looks like nothing has been really worth it over the 5970 according to those graphs... What have I even been waiting for???

 

[xenocide]

and im curious on how the 670 and 660 will stack up against the 7800 series, considering the 7800 series being way more efficient than tahiti.

The GTX680 proves Kepler is was more efficient than Tahiti as well. I expect the 670 to be around the 7950 (with maybe an elusive Ti model edging it out), 660 to be around the 7870 (just like last gen), and 650 to be below or around the 7850.

Can someone explain wtf is with the 5970 being up there challenging the 680 with the performance and actually beating it sometimes? Not to mention the 7970 doesn't seem worthy enough over the 5970. In fact, it looks like nothing has been really worth it over the 5970 according to those graphs... What have I even been waiting for???

The HD5970 is a very good dual-gpu card. There's a reason it cost a lot at launch. You'll notice more often than not the HD6990 and GTX590 beat the GTX680 as well, for the same reason. Dual-GPU cards hold their performance over a much longer period of time than true single card solutions. It's kind of the benefit of the investment.

 

[N3M3515]

The GTX680 proves Kepler is was more efficient than Tahiti as well. I expect the 670 to be around the 7950 (with maybe an elusive Ti model edging it out), 660 to be around the 7870 (just like last gen), and 650 to be below or around the 7850.

I just wish they return to the JUST prices, way overpriced except for the GTX680....
670 with 7950 perf and $350 pricetag anyone? , one can dream...

 

[Melvis]

I just wish they return to the JUST prices, way overpriced except for the GTX680....
670 with 7950 perf and $350 pricetag anyone? , one can dream...

Ebay in 8 months time

 

[Disruptor4]

The HD5970 is a very good dual-gpu card. There's a reason it cost a lot at launch. You'll notice more often than not the HD6990 and GTX590 beat the GTX680 as well, for the same reason. Dual-GPU cards hold their performance over a much longer period of time than true single card solutions. It's kind of the benefit of the investment.

Oh wow I forgot that it was dual gpu! I'm so used to the dual gpu being a x90 or x990 that I forgot they did that! Cheers! Makes so much more sense now!

 

[xenocide]

I just wish they return to the JUST prices, way overpriced except for the GTX680....
670 with 7950 perf and $350 pricetag anyone? , one can dream...

I would honestly not be surprised if we saw;

GTX680: $499 (Slightly better than HD7970)
GTX670: $399 (Slightly better than HD7950)
GTX660: $299 (Slightly better than HD7870)
GTX650: $199 (On par for HD7850)

Would make it really hard for AMD to compete at those price points. Keeping their equivalent cards slightly better and lower priced than AMD alternatives--by $50-80 at every price point--would seriously hurt AMD sales. Nvidia is at a huge advantage since GK104 was designed to be a mid-range GPU, and would have been priced accordingly.

 

[ISI300]

They are almost 750 USD here :shadedshu

The Asus HD 7970 is more than 1500$ over here... hem ... iran. in the only online shop that I could find it.

 

[N3M3515]

I would honestly not be surprised if we saw;

GTX680: $499 (Slightly better than HD7970)
GTX670: $399 (Slightly better than HD7950)
GTX660: $299 (Slightly better than HD7870)
GTX650: $199 (On par for HD7850)

Would make it really hard for AMD to compete at those price points. Keeping their equivalent cards slightly better and lower priced than AMD alternatives--by $50-80 at every price point--would seriously hurt AMD sales. Nvidia is at a huge advantage since GK104 was designed to be a mid-range GPU, and would have been priced accordingly.

Yes, in the case of a price war, NV has a huge advantage.

 

[sergionography]

I would honestly not be surprised if we saw;

GTX680: $499 (Slightly better than HD7970)
GTX670: $399 (Slightly better than HD7950)
GTX660: $299 (Slightly better than HD7870)
GTX650: $199 (On par for HD7850)

Would make it really hard for AMD to compete at those price points. Keeping their equivalent cards slightly better and lower priced than AMD alternatives--by $50-80 at every price point--would seriously hurt AMD sales. Nvidia is at a huge advantage since GK104 was designed to be a mid-range GPU, and would have been priced accordingly.

yea im aware this will most likely happen, but my concern was, will nvidia beat amd in efficiency against the 7800 series. and will it have a smaller die size yet perform a bit better(just like the case with the gtx680)
cuz if gk104 is as efficient as kepler can be, then the smaller chips might not see a much better performance/per watt the way we saw with pitcairn vs tahiti
because nvidia being second after amd in the release date gives them all the flexibility to place cards a bit faster than their amd counterparts and place them in the market were they please in terms of price, however it wont be a total win situation for nvidia if they end up creating a bigger and more expensive chip to manufacture compared to its direct rival.
because gk104 does that, and even if hd7970 and gtx680 had the same performance(thats assuming the worst cases considering all the second thots people might have and sli scaling and what not) nvidia is still a winner because gk104 is cheaper to manufacture than tahiti, therefor margin is better for nvidia