New GTX 680M review (The truth)

Here is a brand new review, and clearly shows that the Notebookcheck review is a bunch of BS
Take a look at the Battlefield 3 especially. Notebookcheck found out that 7970M scored 12% better. This review found out that GTX 680M is 2% faster!

So here you have it. Just like I said earlier. Notebookcheck review was flawed!
So god know what other games they tested that was also wrong.

GTX 680M wins 4/6, draw 1/6 and 7970M is faster in 1/6

Test: NVIDIA GeForce GTX 680M im Schenker P702




GTX 680M is 32% faster


GTX 680M is 15% faster


GTX 680M is 2% faster (Not 12% slower!)


GTX 680M is 25% slower


GTX 680M is 44% faster


GTX 680M and 7970M is alike.

Power consumption:

 

that is still not a fair comparison because the 7970m is completly jacked by enduro in clevos ATM.

 

The 670 stomping the 675 in World in Conflict sounds like some mistake to me.
Just one question, how are these any more valid than any other site or form member?

My thinking is we should wait a couple few months for driver refinements and do this comparison then.

Let's also talk about the 675 winning 3D Mark. That shouldn't happen.

Edit: Also they don't talk about, unless I just missed it skimming though the first few pages where it should be, what differences there are in the comparison side from gpu. Is it the same cpu, mem, etc? Looks like one was an Asus g73, one was an msi.. I'm on my phone didn't feel like digging in, nice if they had it in a table. What driver versions did they use?

 

Looks like Enduro is the problem for sure, cuz even the GTX 670m beats it in some tests...
Best Apples-to-Apples test would be with Alienware laptop

 

Because now we have 3 reviews giving a totally different result than the Notebookcheck review. Plus we have hackness results which also was different than the NBC review.

So we can conclude safely that the Notebookcheck review was flawed. GTX 680M runs much better than what they found out

BTW: Look at the power consumption of the GPU compared to 675M. Nvidia have reduced it by 63W, yet it perform 50% better. It even draw less wattage than 670M...

Thats a pretty good accomplishment

 

When are you gonna get a GTX 680m so you can validate the performance of the card Cloudfire?

 

Cloud, why is it such a mission for you to prove that the 680m is faster than the 7970m? I think we are all burnt out with this now.

From talking with dell qa testers, the 680m is faster...not much faster, but it is faster. So without reading a million reviews, i can confirm to everyone that the 680m is faster than the 7970m but depends on the game.

I wish we could put all this to bed and just enjoy the damn laptops the way they were meant to be used which is to have fun and game

People have to ultimately decide what card they are comfortable with and get what suits their needs and their budgets

NOW LETS GAME!

 

Here here! I'd rep if I wasn't on my phone on tapatalk.

Finally got the dark brotherhood missions last night on Skyrim, I think the last thing I have left side from those elusive floating berenziah stones. Even on my lowly 675M I was having a blast.

Edit: it's funny how long I went without sleeping in a bed.. that's what finally triggered it. I must have completed 66% if the quests in the game while this quest was waiting for me to sleep...

 

i just think its funny that he is on this mission when he doesnt own either card, giving him no reason to care whatsoever

 

Just a reminder going forward to keep the thread as civil as it currently is. Mods have mounted scopes to their ban hammers over this topic.

Also this review has enduro enabled and gpu boost apparently disabled so both cards aren't going 100%. Need another test.

 

Actually it does.

My mission was to discredit Notebookcheck review since as you can see, it was clearly wrong. I along with other forum members said from the day it came out that it was flawed. I`m not looking to boast it in front of 7970M since its in a league of its own price/performance. So sorry if it came out that way.

Well we don`t know if they disabled it or not.

 

thanks for posting Cloudfire +rep
i will be getting one soon

 

And to you people who is using Enduro as an excuse for Notebookcheck`s bad testing of the 680M:

TheVerge, hackness, hardwareluxxx all got 45FPS for Battlefield 3 with Ultra.
Notebookcheck got 35FPS...

So since they got so much worse result than the other sites in BF3, how can we trust them in the other games they tested?
That is what I`m trying to say with this thread. Nothing more.

 

I really dont see the point of being a fanboy of video cards, starwars vs star treck - fine, friendly arguments on what piece of film is better is nothing bad. Manchester United vs Chelsea - fine, sports fans always have been competitive, its the nature of sports.

But to seriosuly go so far to prove such a small difference in a graphics chip and trashtalk another gpu is just rediculous. People get what they feel fits best their needs, i personally cannot constitute a 350$ difference with such miniscule performance differences, I can afford it, i just dont think its a good value, others perhaps do, good for you! Leave it at that.

But i do rep the individuals that simply do these tests just so we have some sort of hard numbers to know what each card offers to help us educate ourselves. Just please dont be a gpu fanboy, its just sad at that point. Reminds me of console fanboys who go out of their way to defend their purchaise

 

More Reviews should come soon right? Around these days. I find it a bit weird how nvidia seems to be struggling making different kepler cards at the moment. Technically there are only three kepler cores right?

I hope more reviews, more indeep, can come soon. I don't think I have read that much of the HD7970m either, at least from reviewers. Mostly users.

 

GPU-Z screenshot shows the boost clock being the same as the standard clock. I'm no expert on the matter but I think that means its disabled. Its also typical that different review sites have different benchmark scores because how they run their tests varies but I do agree the NBC article wasn't the best.

 

wow the 675m is really power hungry. Anyway 7970m holds its own against 680m so I don't get what cloudfire is on about the truth. I was right that 680m and 7970m consume similar to a 670m and 675m is the most power hungry single gpu.

Test: NVIDIA GeForce GTX 680M im Schenker P702

cloudfire the 7970m easily beats 680m on 3dmark11 when notebookcheck says otherwise. I said before that 680m and 7970m are about equal as you get performance per watt and overall performance. Im shocked by 675m power consumption.

 

I can see why Cloudfire open this thread to clarify the mistakes of NBC. If I remember correctly NBC was one of the very first sites to review a 680m, and apparently their review was flawed. I guess if you were in the market deciding between AMD and Nvidia you would probably chose AMD based on that review, something that I wouldn't. I think Cloudfire's motive is to clarify the performance of the 680m, nothing more, and I appreciate it.

 

One might also consider that Notebookcheck pulls the computer out of the box, loads up their tests programs and run them.

Drivers make a difference

EDIT: Lets see some more current games on there? Stalker: CoP is 3 years old, World in Conflict came out in 2007! What about Max Payne 3, something more relevant to today?

EDIT 2: I just loaded up Stalker: CoP and enabled DX11 settings - I am getting 117 FPS at very first load in area, with stock GPU speeds and crossfire disabled. Drivers, eh?

EDIT 3: Sweet Mary and Joseph, Cloudfire - I just went through that article and looked at max settings for the games/synthetics. How on Earth do you draw your conclusions? You DID read that article, right?

EDIT 4: There will be no convincing you, Cloudfire, so I retire with my now 99.99% functional 7970m Crossfire (sans The Witcher 2, which is playable but dips to 40 FPS when looking around). Bon chance, and good day.

TRUTH IS IN THE EYE OF THE BEHOLDER.

 

Seriously another thread about this?

we already knew that the 680m was faster than the 7970m, not 300 bucks more faster, still faster.

I dont understand the need for this crusade, notebookcheck is notoriously flawed when giving their gpu performance. Why did you think this time they would get it right?

 

Yep I'm happily $300 dollars richer...

 

arestavo,your comments are goingto force me to read the review now when i get back from lunch. im on my phone so can you elaborate on edit 3?

Sent from my SGH-T989D using Tapatalk 2

 

All that shows is that I am happy that I did not spend an extra $300 as it would have been a waste. I generally go nVidia for everything, but there is no justification for an extra $300 in any of the benchmarks. Especially since the 12.7 beta drivers came out. Mine now runs cool and performs great.

 

3Dmark06 - 7970m winner (listed will be winner)
3DMark Vantage - 680M
3Dmark11 - 7970M
Uniengine 3.0 - 7970M
World in Conflict - 680M (came out in '07, and if I had it, I'd test this)
Anno 2070 - 680M (don't have this either)
BF3 - 680M (tested, I get 45 to 55 FPS, but not sure where they tested)
Metro 2033 - 7970M
Stalker: CoP - 680M (I tested this, and got 5 FPS more than the 680M in the start area)
Street Fighter IV - 7970M (why is this even in here?)

So, by that reviewer, it's 5 for 7970M, and 5 for 680M. I can personally refute the Stalker: CoP results (yes I used DX11). So that would be 6 to 4 for me, but we're focusing on this reviewer.

Seeing through the eyes of a looking glass - who will play a game that can get more than 60FPS at less than 1080P and less than max settings if that is what their monitor can card can push? Not me. Max settings all the way, which is why I look at the max settings benchmarks EXCLUSIVELY.

Guess that means that I cherry pick. Because those are the settings I would play at

 

Thanks, Cloud, but we already found out that notebookcheck's review was flawed/biased a few weeks ago, so no point in more reviews, especially from small time, nearly anonymous hardware sites. If it was one of the larger ones, then it might have been more interesting. Still, thanks.

As for this whole price/performance debate and who's the better card is - you went for the 7970? Excellent, you got a great gaming card and got more for less. Why do you find the need to try to rub it in and attempt to spoil the fun at all costs with those aggravating, never-ending ego wars, then? That's the whole problem with this issue.

As for me, I didn't want to go for another AMD card if I can help it. Why? Because that's how I feel about it, and that's it. All I wanted was an nVIDIA card that is comparable to AMD's, and that is more or less what I got. If it was a much lesser card for $300 more lilke what notebookcheck's review suggested, that would be a different story, but that's not the case, so I really don't see the point in continuing those ego measurements.

Anyway, that's all water under the bridge now, and just like what SlickDude said, let's just game and move on already.

 

I'm wiping away tears of sorrow that the 680 might be 5% faster in some games than my beloved 7970.

 

My own comments:
3DMark06: Interestingly, the 680M performs as well as the Enduro-limited 7970m. Optimus problem or raw performance problem?

3DMark Vantage: Advantage to 680M because of PhysX, I'm guessing. Isn't the 680M score around what nVidia quoted w/ PhysX enabled?

3DMark11: What the heck happened to the 680m?!?!

Unigine 3.0: Willing to bet Enduro again limits the 7970m at low settings. At higher settings, the only ones that actually matter, 7970m rocks. Amazing if tess was enabled.

World in Conflict: nVidia has much better support for this game, IMO, even comparing my nVidia- and AMD-powered desktops.

Anno 2070: Seriously, what the heck is up with the 680M results? 42 FPS at 4xAF and 72 FPS at 16xAF? Still a victory for 680, though.

BF3: Pretty clean draw.

Metro 2033: Framerates far above what I see on my GTX 580 in the Frontline benchmark. Must have been custom. I thought Metro 2033 was always better on nVidia's equivalent chips. Interesting.

Stalker: Clean win for nVidia it seems.

SF IV: Yeah, you're right. Seriously wondering why this is here. Pretty much a CPU benchmark, I would assume.

In general, these results seem off if the 680M is as fast as we believe it is. Particularly the 3DMark11 score. 5125? Really?

 

Again, notebookcheck should be considered the absolute worst either GPU will ever do. They have this habit of testing as the "lowest common denominator" user. Such people don't update drivers, they don't optimize their OS, they don't look for the latest firmware or BIOS, they don't do anything but run benchmarks, and they do them at default settings only.

The new benchmarks once again confirm that the 680m is ever-so-slightly-faster stock-vs-stock. (the expected result)

Once again, the real result all of us want is the one where we take the "stock" blinders off and OC both the 680m and 7970m to "reasonable" levels (where stock cooling is adequate and where we can assume most users will get to) and see what we get.

I also want to see the "crazy overclock" results

To be entirely fair Cloud's reaction is pretty much caused by the AMD fanatics and their constant hen-pecking about Nvidia never coming up with anything even remotely close to the 7970m. Nvidia did, and the AMD fanatics have just changed stories.

That being said... the 7970m remains an amazing price/performance.

None of this is a surprise for those of us who aren't fanatics on one side or the other.

 

I hear you there. as I'm still contemplating which and what laptop to buy, i will have to read every thing i can find on this argument.

I'm not exactly a fan boy of either. I just don't want to spend big again on a lemon. I did that last year on a factory Over Clocked GTX 570 desktop card...

 

Ive heard many people on this site tell me "every card is different"

So honestly, im starting to think there may never be a definative awser as to how much each card is faster than the other in each specific game and benchmark. I still think that price is the biggest difference between the two cards...

 

I dont think its proper to call NBC's review as flawed. If that's what they got then that's what they got. Every review is going to be different especially in games benchmarking because there are too many variables. Maybe not unless in some games that have the benchmarking features.

We cannot even guarantee 2 different 680s or 2 different 7970s to give us the same results so how can we make an assumption that one review is better or more correct than the others. Review results are trends, there will be highs and lows. Its not going to be exact.

680 is a great card, its already been confirmed by recent test from nbr user and other sites; and so is 7970. It will always be personal preference. Its just going to be the same as AW vs Clevo, its not gonna end.

 

It's just unfortunate that Notebookcheck reviews are often incorrect. I take a great deal of what they have on their web site with a grain of salt.

 

notebookcheck normally have different models with same gpu. I reckon notebookcheck are the best as they have multiple data and you can get a good idea of the results all in one website and compare them as well.

 

so 7970m < 675m < 670m < 680m... BS. period.

(the WoC bench of this beloved review)

 

ok, got to play with a 680m in a NON gaming environment today at a trade show for media creators in Vancouver.

now since I dont have a 680m yet, my opinions from reading about 90 threads here and elsewhere

1: both 7970m are great gaming cards and seems to be another 580m/6990m too close to care in performance. its personal preference

2: AMD needs to fix enduro drivers and the crossfire drivers.

3: Nvidia still needs to fix a couple aspects of Optimus, the IGP passthrough still messes up occasionally artifacting some applications.

now what I KNOW from a bit of hands on at this time

1: 680m and current drivers = heavy DPC latency in our test system, get ready audio people, its a repeat ( 8000us latency seen 2-3000 common)

2: OpenCL IS as bad as speculated. rendering on Photoshop 6, Premier Pro with 6.01 update showed the 680m behind by 70% and behind by 60% when switched to CUDA support instead of CL

3: solidworks and ensight was a wider disparity. ( 120 - 150% advantage AMD )

4: still no DeepColor support for dealing with 10+ bit color screens properly ( only the Quadros )


so, both are great gaming cards, and the 7970 is also a good pro app card as where the 680 sacrificed alot of its pro level support. as to the driver bit ... Ive had nightmares with both the last few years, that argument is beating a dead horse IMO

 

The main problem with relying on notebookcheck's results is that Notebookcheck's results are not typical for users at this site.

Most people here will get advice to update drivers, update VBIOS, and optimize their OS in order to get the best performance they can. Most of the above isn't rocket science, and we are able to walk even laptop neophytes through it.

This doesn't even count the moderate and usually responsible overclocking usually advocated here. Even those who choose not to overclock can do MUCH better typically than Notebookcheck.

 

Guys, you simply don't get it.

He's not trying to prove that the GTX 680M is simply faster than the 7970M.

He's trying to prove that the GTX 680M is 15% faster than the 7970M.

So what if the GTX 670M is as fast or faster than the GTX 675M and the 7970M in some games despite physically weaker hardware, just as long as the GTX 680M is faster than the 7970M by 15% or more! As long as it shows that, it is a legit review guys.

On a serious note, please stop making a thread for every 680M review that comes up. You already have like 6 of them. Best thing to do would be to consolidate these 680M review threads into one so I wouldn't have to jump around to get a good laugh at shameless fanboyism on display

 

All I know is, future 680M users and current 7970M users should be astonished, amazed and happy they have such incredible performance.

I've seen benchmarks showing 3610M is on par with a desktop 2500K from last year and the 7970M on par nearly with a desktop GTX 580 from last year. All this is just simply astounding. I thought I had an amazing deal when I got my G73JH. But my NP9150EM blows that away. 15" laptop with this much power? Just incredible.

 

at the end, 28nm should be blessed, happy gaming everybody! (whichever card you own / whatever company you root / how thick is your wallet, we have a card for whatever setup which will max / run admirably any game on the market! that's the important part imho)

 

I have an i7-2 600k and GTX 580 in my desktop. I can tell you that at maximum performance my i7-3610qm and 7970m come within 10% of my desktop components at stock in most games/benchmarks. So, saying they are equivalent to an i7-2500k/GTX 570 system is certainly legitimate.

 

Yep! Even with HBAO/SSAO and Tessellation on! And almost all the new games have FXAA implementation, so no worries on AA either.

 

I am just really amazed in the difference in consumption power levels between the 675 and 680 (65W less) while providing about 45% greater performance.

 

what is wrong in NBC I don't get it.. 3dmark11 is about < 2-3% apart from whatever bench we have seen so far.. their WoC bench is flawed no comparison for that, ANNO I don't care.. BF3 is the same with NBC except for they have a weird bench for the other machine with 12% over 680m, Metro 2033 is I am sure wrong in this review as my 7970m doesn't give me that frames (no where near) on that setup, stalker is also clearly wrong in this review, and the other two games are not even famous...

I think NBC is perfectly right in their benches so far, and it will turn out to be that way.. 680m is about 5% faster in 3dmark11 and 7970m is 5% faster in vantage, that's it.

 

The 660/680M are based on Kepler 28nm, the other 6XXM (675M, here's lookin' at you!) series are just rebranded 5XXM series 40nm, which in turn are just rebranded 4XXM series 40nm as I understand it.

 

I think part of the problem is the interpretation of the results and their abnormalities, such as: why is the 680m performing worse in 3DMark11 by a large margin? Why is the 670m nearly outperforming everything in World in Conflict?

The 680m scores wins in Stalker, Anno, and WiC. Correct me if I'm wrong, but everything else is an AMD win (a few) or (most of the remainder) a very close tie (1 FPS does not equal a "win" for either side IMO). This is why you should always perform ~7 runs (my preferred minimum for small-sample situations) or so of each game, then run ANOVAs and post-hoc pairwise tests to confirm differences.

 

If you get a 3720qm and an overclocked 7970m you basically have your desktop on stock clocks

Which is awesome

 

*Smirks while shaking head*

 

If you have time, can you make a plot of 680m/7970m related threads you have closed vs time? lol

 

Credit to the opening poster, it's always healthy to watch the watchers, or in our case to review the reviewers.

The problem is, as others have pointed out, that there are so many variables that it's impossible to say who has the correct benchmarks, or is the more reliable reviewer.

 

Abandon ship!