Test - GeForce-GTX-680M vs Radeon-HD-7970M

Test GeForce GTX 680M vs. Radeon HD 7970M - Notebookcheck.com Tests

HF

 

Intriguing. So hot games like Crysis 2, Starcraft 2, and Battlefield 3 performs way better on 7970m. Only hot games like Max Payne3 and Modern Warfare 3 performs better on 680m gtx. Skyrim has a minuscule edge on 680m gtx.

 

wow that is blow for blow there! such a close race! its really gonna come down to OC potential

 

Skyrim test is invalidated by the setup. Primarily, look at the M17x with enduro off on 7970M to the 680M rather than the Clevo 7970M to avoid the performance loss. Then factor in that the M17x was tested with a slower CPU (3610qm/7970M vs 3720qm/680M). CPU will make a world of difference in Skyrim and will probably give the edge back to the 7970M in any scenario in which the two were even or barely (2-3%) apart.

They did however use older drivers on both systems so its a bit skewed either way. Good test, could've been much better and more definitive. Also this should be continued in this thread - edit : that thread was closed? disregard.

http://translate.google.com/transla...-680M-vs-Radeon-HD-7970M.77072.0.html&act=url

^ english version of the notebookcheck article.

 

BF3 is all I wanted to see.

 

Code:

So your saying that if 7970m is paired with good CPU, SKyrim will be better on 7970m?

Amazing this 7970m card is.

 

Can anyone please explain how come the review sums up that the 680M has less power consumption thus causing less heat, when the graphs clearly show that the 7970M is much cooler (about 3-5 degrees less) than the 680M?

And again, these are the exact same numbers we've seen from Notebookcheck, so no news there, but thanks for the review!

 

Lol fail, didn't realize the 7970M was already leading in Skyrim. It'll have a better lead with a faster CPU.

 

680m furmark = 84°C / BF3 = 69°C
7970m furmark = 95°C / BF3 = 83°C

 

Thanks, I missed that part.

So how come the 7970M case was so much cooler, then?

Edit: SkittlesXD, yeah, this actually saddens me to no end. Skyrim was the main reason I preordered my 680M laptop based on the early Nvidia PR sheet (more with an 'i' than with a 'ee') that claimed Skyrim benefits greatly from this card (they claimed 77 FPS against the 55-60 of AMD's). What a blatant lie. This is almost lawsuit worthy, right there.

 

Hmm according to notebookcheck 680m takes less power but the m17x r4 review proved different. All in all I reckon nvidia and ati top end cards have different strengths as seen on the review.

Looks like on cad 7970m is twice as much better but 680m just about beats the 7970m on raw performance. kepler seems quite good.

I personally think this is the closest identical generation performance per watt cards have come.

I wouldn't mind getting a gt 640m or 650m.

 

for the clevo, what driver they use on 7970m to make it work properly on 3dmark06?

 

Also why MP3 is getting like 36fps when on my system I can max it out while having 85fps?
These cards are enough for such a game, my old 485m was able to output 35fps maxed out, while 58fps/60 using very high and hbao.

 

Clevo 7970M seems to run hotter than the AW version. I wonder why...

But do you see the difference in sheer computing power? Wow, the 7970M simply unreachable in CAD apps and even beats some Quadros and FirePros there!

 

Although the M17x-R4's power consumption numbers seem lower than this Clevo system, I think the 680M's consumption numbers are very impressive, especially considering that it's derived from a GK104. I think Kepler's management of Furmark has a lot to do with it, but it's good to see nVidia tone down consumption to the point where R2 users who want to experiment like myself wouldn't have too much trouble dealing with excessive heat output.

Would be nice if they tested overclocking potential on both of these cards too.

It's most likely due to the differences in die size. The 680M has a bigger die size so although the die itself might operate at a lower temperature than the 7970M die, the 680M die could still generate more heat through the system.

The best example I can use to help clarify this is taking extremes: Which do you think will heat up a room faster? A large oil heater that is 90 degrees celsius or a small curling iron that is 200 degrees celsius?

 

Why the 680m doesn't have a boost clock like other kepler cards? Is that due to the high power consumption?

 

Wow if this is it, I may have to cancel my preorder.... What a shame Nvidia, completely lying on its PR... What you guys that didn’t buy any of those yet (or are on preorder for either) are going to do? Risk the AMD and the supposedly bad driver support or pay premium for the driver and extra feat. since performance it is the same.

 

I can accept this explanation, but that doesn't explain why the comparison reviewer doesn't comment on this behavior. And the die size difference should basically be actualized in just the way these two GPUs operate, no? I mean, it's not like the 680M die is twice as large in size or something. Or is it?

But really, when thinking about it from the reviewer's perspective, why do I care if my GPU is all cool and drinking margaritas in there whil I'm having to cope with all the excessive heat? I've had an old laptop with a 8600GS GPU that ran games of that time perfectly, but it had such a cheap cooling system where the WASD area was so hot that I couldn't keep my left hand on it, and had to play in short bursts or stick to strategy games. And no notebook cooler could help. I remember this experience so badly that I really wouldn't want to experience it again.

And 42 degrees (the 680M's case) seems dangerously close to that.

Edit: nanias, this is the exact same site we had the benchmarks from since last week, so no wonder it stays the same. They just reused the same results from their previous review, so no surprises there. I'm waiting at least until the next big review from a major hardware site (anandtech, guru3d, etc.) before I pass judgement. But yeah, if this remains the same I will probably cancel my preorder. Waiting a few months and paying a premium extra for a card that has similar performance to what is already on the market is not cool. Plus, that whole Optimus stuff is useless to me. This is a giant desktop replacement, I don't buy it for the battery life, nor do I plan on using it for browsing away from a nearby power socket.

 

I am in the same situation, debating if I should contact XoticPC to swap my GPU
from 680M to 7970M for my order... But I always had less issues with Nvidia drivers than AMD, but is it worth the $300 that I would save? Maybe that $300 can go towards a faster CPU... Tough decision.

 

I am really torn on this.. Got my brand new M17x with the 7970m sitting untouched in its boxing still... and I got until thursday to return it to wait for the 680m..

I am not on a budget, but obviously I don't feel like throwing money out of the window either.

I have a feeling the 7970m will OC the most out of the two and if this is the case it will boil down to performance vs features on this round..

I just can't make up my mind

 

Shizm, if you already have the laptop, I say open the box and enjoy.

Even if Nvidia miraculously improve stuff via drivers, the 680M is definitely not leagues and bounds from the 7970M.

 

A reminder, notebookcheck's results are the absolute worst either GPU will ever do.

He doesn't update drivers, he doesn't unload bloatware, he just runs the tests out of the box. Note, this applies to both GPUs.

What we learned is that the 680m and 7970m are roughly equivalent at their baseline with old drivers. What we didn't learn is how good either one can be when using up-to-date drivers and tweaked.

Note please, notebookcheck's evaluation of the 570m was so totally incorrect as to be rediculous. A reasonably savvy user with only moderate overclocking luck would have it performing almost 50% better than notebookcheck listed it.

The 680m is STILL a heavily underclocked Desktop 670 with slow RAM. (compared to the 7970m being based on the 7870, but just slightly underclocked)

If the 680m can OC anywhere near its Desktop cousin's speeds, it will be very impressive. If it is that badly shackled by heat and slow RAM, then maybe the overall speed crown goes to AMD this time. The jury is still out.

Note the 7970m's price and its level of performance at that price is currently unrivaled no matter what the 680m does.

 

Guys, i've been saying +/- 5% between these cards all along and it looks like that's what it is going to be.

With regards to enduro, this is new tech and will be vastly improved as time goes on. I remember when optimus first came out, it had its issues too. AMD aren't going to just let it sit and forget about it because Enduro is something that is in the plans for the future and it will be optimized

As far as I'm concerned, the cards are even. It all boils down to overclocking for me. We have already seen that the 7970m overclocks very well, but we don't have a 680m yet to play with.

My everyday gaming clocks is +12% core/+20% memory (stock voltage). I bench at +18.5% core/ +30% memory (stock voltage). and my temps are in the 60's for both furmark and games.

If you are on the bench regarding whether or not to get the 7970m, I'd say go for it. The card is a beast. From initial reports, the 680m is also a beast. Nvidia has exceeded my expectations as I didn't think they would get a card out at this performance level until Sept.

From a consumer perspective, you have to decide if the 680m is worth the additional $300. If it is going to help you sleep at night, I say get it. Having two in SLI, I can tell you that the desktop gtx 670's are great cards and the 680m is a downclocked version

I agree Kernal! It is hard to argue about the 7970m price/performance

 

Wow.... the nVIDIA loses in BF3. I don't understand at all. It must be dependent on memory bandwidth (it's the only explanation).

Congrats to AMD, they win this round (assuming they fix their drivers).

 

So from the math I've done thats 952/1440?
Looks like 950/1450 is the best OC for stability in Clevo/AW.

 

yes, i usually run 950/1400-1450 for my gaming in the more demanding fps shooters i play.

However, in Diablo3, i just run the card at stock clocks. i'm already getting 100+fps with D3 maxxed so there is no need to go above 850/1200

 

And dude. As long as temps don't go over 70C you 100% sure that 950/1450 will not decrease the life span or damage my 7970m right?

 

the general rule is 90C

 

Wait, if the 680M is 4GB, why does alienware sell write 2gb in the website?

All clevo sells with an option for 4GB 680M's

 

Performance is very close between the two cards, but the lower temps and power usage are pushing me more towards the 680m. Plus I'll have the option of CUDA development with the 680m.

 

7970m runs cooler than 680m Test GeForce GTX 680M vs. Radeon HD 7970M - Notebookcheck.com Tests

7970m can be used to programme too.

 

I don`t believe this test for a second. 7970M 13% faster in Battlefield 3, a game where Nvidia have always beat the AMD cards?

Nah-a

 

The 680m itself runs cooler than 7970m.. not the other way around.

Stress test with BF3 and Furmark:

7970m: 83c - 95c
680m: 70c - 85c

 

"confused 7970m fanboy" haha, i dont really care whether nvidia is faster i just want the cards to keep getting better, although it looks like nvidia is running a lot cooler than the amd and it uses a lot less power, like 20 watts less, which means epic overclocking.

 

Wow! Such a deja-vu...

Can we please stop relying and repeating our posts based on the exact same content (albeit expanded) from notebookcheck.com taken from a week and a half ago and at least wait for a second opinion from others as well?

Shizm, he was referring to the whole case temperature thing. But again, if the 680M is cooler as mentioned, why the case is so much hotter? Not that I mind it that much, because I'm not basing my judgement on a brand, but something is definitely off in this whole review.

 

personally i believe that part of the temperature issue in clevo is based off of the inability to turn off enduro... AW users have confirmed that when they turn off enduro they see temp drops. i know for a fact that optimus doesnt cause this so (assuming i am right) once enduro gets the kinks worked out of it maybe the 7970m will see a slight temp drop in the clevo models. this would bring them closer in temps (crossing fingers)

 

Meant card temperature not case temperatures (Typo in my last post). I know NVidia open sourced their CUDA compiler not sure if there is support for AMD cards yet though.

 

Notebookcheck found out that:

Temperature:

7970M:
Battlefield 3: 83 degree celsius
Furmark: 95 degree celsius

680M:
Battlefield 3: 70 degree celsius
Furmark: 85 degree celsius

Power consumption:

7970M:
3DMark 06: 106-131W
Battlefield 3: 143W
Maximum power consumption: 200W

680M:
3DMark 06: 99-109W
Battlefield 3: 126W
Maximum power consumption: 183W
----------------------------------------------------------------------------------------

If the performance they found out is true, and that 680M consume about 20W less than 7970M, why didn`t Nvidia increase it 20W to beat 7970M? Nvidia absolutely have the thermal headroom enough to do it, considering 680M runs 10+ degree celsius colder than 7970M. Sooo weird

Atleast there must be some crazy overclocking potential here?

 

Test is obviously flawed. Disregard it and wait for a proper one.

 

Not sure about this, just look at the voltage supply circuits of the 680m, seems a bit weak... but we'll see.

 

My 7970s don't push past 74c. That Clevo must suck then, lol.

 

You will come no where near these temps in an Alienware

My 7970m does 65c in furmark and the same or less in games

 

You're very observant

The Alienware have more robust cooling compared to Clevo...as discussed in this thread:

http://forum.notebookreview.com/gaming-software-graphics-cards/665964-7970m-temps-11.html

 

Well, since you get what you pay for, and Alienware usually charge a hefty sum over other barebones, they sure better show something for it besides a glowing alien head

But really, an Alienware seems great, but the mere thought of its cost in the US, let alone here in Europe makes me shudder.

 

Yes, I think so too. Why the hell would they want to release a GPU that runs 20W lower than 7970M when they could have matched the power consumption of the 7970M and beat it? I mean, having the best GPU must be their priority right? Why "stop" at 20W lower and come out worse in a lot of games?

Then there is the whole Battlefield 3 performance that baffles me. Worse than an AMD card when they beat AMD with all the desktop Kepler GPUs? The Verge review got 30% better FPS results in Battlefield with 680M than notebookcheck did.

 

Another unaware? Its not expensive if you know how to negotiate. You may end up to about $100-200 more than barebone laptops or maybe the same price for same base specs. I'll gladly pay that premium for some better features like better cooling, better GPU Ram, better Customer Support (at least for me), and for the glowing alienhead and keyboard

On topic: When can we expect an actual test of 680M from nbr peeps?

 

These are minimum fps, not average fps.

 

+1 to AMD from me, really good job

as for the 680M the only hope is, that I'm thinking that it has tremendous overclocking potential and can (theoretically) win, if cooled good

but f*** this, fail nVIDIA

 

I went with an Alienware back in March because it was about $100 cheaper than the cheapest Clevo reseller with the exact same specs. I ended up saving an extra $200 by ordering from the AAFES website and using my Military Star Card. So overall, I saved about $300 by going with Alienware.

Others have been able to save just as much, even more, by just calling in and negotiating with the sales team.

 

Id consider alienware if they offered a toned down chassis and a matte screen option. I dont wanna walk around with something that looks radioactive