AMD 7970m vs GTX 680m

So anyone with Crysis 1 and a 7970m please check how many fps with Max settings? The reason why I want to use Crysis 1 as a benchmark for games because it is really demanding.

 

Find AMD 7870 benchmarks and minus 4-9 FPS ..lol..

 

That's just not going to happen, lol. The 100W limit stops this. Maybeeeee Maxwell, but NOT Kepler.

 

I would like to hear explanation why GTX660M is 75W and as weak as it is. Below 2300 gpu points in 3dm11 isnt just cutting it. Im sure that Nvidia didnt just put 75W TDP there for nothing. There must be reason behind it.

IF GTX680M will be around 6000-6200P 3DM11, when NV needs to put something in between them as well. Old Fermi isnt going to be around for long and people dont really want to buy old 40nm cards when 28nm is available.
Is it likely that we see something like GTX670ti in between? Theres clearly too large cap from GTX660 to GTX680 like this.

 

i agree,even the gap between 675m and 680m would be way too large,so what are they gonna do,introduce a 685m to counter the 7970m and the 680m as competition for 7950m? naming-wise,nvidia shot into their own foots

Sent from my GT-I9001 using Tapatalk 2

 

Kepler just isn't that good performance per watt until it reaches 150W+.

 

It's probably rated for 75w. In all reality it probably pulls something like 35w at most.

 

Lets see some proof of your claim.

 

From the wall measurements I saw it draws at least 45W during 3dmark.

 

I've done some test with the 7970stock along 2720qm with BF3 MP Ultra (Maxed out)

- STOCK CLOCKS - Metro 64 players - ULTRA Full HD - Max temps: 82°C CPU - 72°C GPU - Avg fps: ~60-65 (read with FRAPS)
Screenshots (taken with FRAPS):
ImageShack® - Online Photo and Video Hosting (just click on Suivant to see them all)

- STOCK CLOCKS - Caspian Border 64 players - ULTRA Full HD - Max temps: 82°C CPU - 72°C GPU - Avg fps: ~30-35 (read with FRAPS)
Screenshots ( taken with FRAPS):
ImageShack® - Online Photo and Video Hosting (just click on Suivant to see them all)

Regarding power consumption:
http://forum.notebookreview.com/msi/659226-7970m-heatsink-modification-9.html#post8501603

The 28nm process definitely helps for reducing considerably heat&power draw.

 

Thanks man! Rep up!

 

Which is more demanding?crysis 1 maxed out or battlefield 3 maxed out?

 

BF3 and Metro are the most demanding shooters i think.

 

People really need to stop using Crysis 1 as a benchmark for modern GPUs...

 

Thanks for the benchies mate, but it's only me that notices not-so-jaw dropping performance there?
60-65fps in metro looks legit, but 30-35 on Caspian?
Is that a bit low? Just for reference I tried a 64man server with my 485m at 700/1800 and I'm getting an average of 42fps?
What's wrong? Its the same AMD driver related issue all over again, or my 2760qm is helping me out there?

Because if that's the performance I may hold into investing in the 7970m..I was expecting a gtx 570 like performance, at least around 45-50fps..like I had on my desktop..thoughts?

 

For reference. I have Phenom II X6 @3900mhz, HD7970 @ 1100/1600 and Caspian 64man fps is around 40-45 as average on Ultra 1920x1200.
Well, if you feel that fps on caspian border is too low than perhaps switch to less punishing AA type.
64man Caspian is insanely challenging to whole PC. Not just GPU.

 

But crysis 1 is very demanding. But if bf3 is more demanding then lets use that.

 

@Kingpingzero

I think you might need to try MP Caspian with full 64 players as this is very very very demanding - I thought that was a bit low too but I was playing very fluently on any other maps (Gran Bazaar, Operation Metro) and Caspian was one of the most demanding map with 64players.

I'm sure every single card will struggle with this map, even desktop card (like Supranium mentioned)

 

I am eagerly waiting for a complete review of 7970M with full comparison to see the difference. But I guess we will have to wait for June to see it

GTX 680 get around 60FPS+ in average with 4XMSAA in Ultra with 64 people in that map. I am guessing that Nvidia have an advantage with this exact game that may reflect on your comparison, Kingpin
http://www.youtube.com/watch?v=FMjqAgxKIoM

Did you undervolt/underclock the 7970M when running the test or was everything set at normal? If so, then bloody hell thats impressive

 

Nope, and actually just re-done the test with the 7970m overclocked @984/1500 (no voltage mod) and the power didn't even increase noticeably (the same I'd honestly say)...Same for the temps, they stay really low and the o/cing doesnt seem to affect them...

That's insane. The 580M was o/c and o/v though so it has to be taken in consideration.

Also Cloud watch carefully the video, this is in 900p

When I play BF3, I play in 900p too with a mix of Ultra/High settings without MSAA => Avg 90-100fps on MP maps with the 7970M o/c.

edit: you're right to mention that the 580M had an edge @ BF3 over the 6990M - don't know bout the 7970M but it's been good so far.

 

Crysis is more demanding than BF3.

nVIDIA wins in BF3

AMD wins in Crysis 1

nVIDIA wins in Crysis 2

??? wins in Crysis 3

 

Nvidia wins in newer games, especially those using DX11.

AMD wins in older games not using DX11.

Crysis 3 is going to be basically the same as 2, DX11 heavy engine and Nvidia should be a clear winner there.

 

I was into a server full 64 ppl when i benched that - but - truth to be told, i was using a custom ultra setting, not the preset. Which means all settings to ultra but no AA or post-aa.
So if im getting from 35 to 41, much like you do in those screenshots, i have to take into account that im not running either 4xMsaa or Post AA. Thing that you did thought.

In other words, performance wise, its a jump around 40-45% ahead of the 485m/580m.
If my calculations are correct, if you run the same map with my same settings, youll get a boost in perfomance around 10fps. Which translates to smooth 50fps?

 

But the 7970m can run those new dx11 games on max settings great, right?

 

Not in all cases though...

AMD wins in Metro, which is more demanding.

 

Metro kind of has an awful engine and it is also not true DX11. It can be adjusted to be 9/10/11. Even with DX11 enabled it isn't using a lot of the DX11 stuff, just tessellation (which AMD often turns off in drivers) and DOF.

Anyway, the rule of thumb holds up in most cases. For the vast majority of modern DX11 games Nvidia is going to win simply based on architecture and drivers even if the cards are the exact same speed.

 

Well, I don't want to take this argument to awful engines, Crysis 2 also has a bad rep with poor optimization with the DX11 patch.

 

YodaGoneMad,
Not true.
Games are always balanced better for one or another. Theres no ground rule or any proof on your claim. Thats simply speculative.

 

To be very fair and right, I think we just need to compare a HD7870 desktop over a 560ti desktop from a review to see what's the real fps difference

You need to turn to max both MS and FXAA, you will see especially MSAA slow down the game tremendously...FXAA is fine though.

 

Wait a minute. Your going to knock the "validity" of Metro as a DX11 title but then use Crysis 2 as some meter for a newer game utilizing DX11???

LOL

Also FTR Crysis 2 Warhead most review sites show AMD to have better performance vs Nvidia.

 

Good news for those who are waiting for GTX 680M. Earlier we saw around P4600 (P4900 overclocked) in 3DMark11 for 680M, seems like they went back to the drawing board after seeing the mighty 7970M and created a new 680M.

To be honest it makes me a little pissed off that they originally planned to release the P4600 and holding back so much. Thank god we have competition

 

Yea, it is a whole can of worms. I guess I just don't consider Metro a modern game, it is many years old (just like Crysis 1, I just don't feel it is relevant for modern performance). I am talking things like Witcher 2, BF 3, etc. True ultra-modern graphics engines that use all the latest tricks. Not console port jobs with some extra effects tacked on. I guess using DX version as a rule was not good, because there are modern engines (like Skyrim) that use lots of the latest tricks and run much better on Nvidia, even though it is not really using the flashy DX11 stuff.

I agree it is speculative, but just what I have seen and taken note of. It seems that the more modern the engine the more of an advantage Nvidia has, if an engine is older than it is more even or AMD could have the advantage.

It will be interesting going forward because there will be more instances to test my theory. For example, AC3 is going to use a brand new (hopefully) modern engine. In the past the AC games generally ran better on AMD, will AC3 suddenly run better on Nvidia? It will be interesting to see.

A great example of why competition is always good and early leaked ES samples are not very credible. If Nvidia uses either the 660 or 670 as the base for the mobile 680 then it will be a pretty great card. The 670 seems like the clear choice, it would crush the 7970m and already has very low power consumption. Plus, as your quote says, with their auto-OCing they can cheat. They can release a 150 TDP card and have it rated at 100 watts, then just let their software boost it when needed. It could be possible to see a 680m that runs surprisingly close to 670 desktop speeds at full OC.

 

Metro 2033 is to Metro 2034

AS

Crysis 2 is to Crysis 3

And IMO, Metro looks far superior to Crysis 2/3.

I own ALL of them too.

 

That's interesting. The GTX 670 Ti desktop card is a gem? That is the card they are basing it from.

 

wow 6500-7000 points in 3DMark11 at stock? thats very impressive indeed! with a nice OC one could even hope to break 8k! but tbh, seeing these numbers it makes me a little scared of what nvidia is gonna charge for such a premium performance....

 

Yeah its pretty sick.

And this GPU is announced at Computex June 5-9 and is ready for shipping in July according to latest talk. With the latest leak of the photo of the 680M MXM card and the "680M in late production" it looks to be true

Zero: Its either 670 or 660 Ti. Looking at the 3DMark score, GTX 680M gets 77% of the 670 3DMark score.

Yoda: Yeah, this baby overclocked, man would love to see that score

 

I wonder if there are still 768 shaders...

 

If the card truly hits 6500, theres no way it is 768 shaders.
Anyway. I cant wait until July. Its way too far away. So no cheap 7970M to me.

 

Nope, also I slightly worry if they use powertune to go all out limiting it, it might not bode well for clocking if it cant be adjusted.

 

@Cloudfire: Source for this new speculation?

 

Yeah, I would like to see that source also. Thanks

 

One problem, the actual performance difference between the 675M and 680M is MASSIVE compared to what the naming scheme would suggest.

 

yep, my thoughts exactly. so maybe nvidia decides to name this flagship 685M with a 680M to fill the gap to the 675M?

 

Source

As for the 768 shaders, I think that is the specs from their previous 680M that scored P4600, the GK106.

This P6500-P7000 680M is a GK104 like we seen from the MXM card leaked few days ago, which means most probably either 670 or 660Ti.

Meaker: Who knows, we might see Nvidia releasing the P4600 GPU at the same time. But this P6500-7000 is coming pretty soon

 

660 - (20%) -> 670 - (20%) -> 675 - (100% Roflcopter) -> 680

 

"chh have broke the news ... in fact gtx680m would have really only 3d11 p5k less than ... the results you know, 7970 had 6k .... now melted down to re-create, broke 6.5k ~ 7k (6.5k I think fly) ... The gk104 the power control is indeed a great progress, and this automatic overclocking can write to tdp lower than the actual point"

lolwat

I think he means the 7970M (from a stock of 6k) has broken 6.5-7K (thanks to overclock) and the 680M would only score less than 5k...

Seems like you have linked us to a Chinese version of this thread.

"Personally, I think, AMD mobile graphics performance has been done far better than NVIDIA.

Whether it is cost-effective, power, temperature, and performance."

Well said.

 

To me it sounds like 680M had P5000, which it had overclocked, 7970 was released and had P6000, which it about had, and then Nvidia "re-created" 680M, and the new one hits P6500-P7000 which is why he talks about GK104 and the automatic overclocking, which varies from P6500-7000.

Google translate is useless. We need someone who knows chinese

 

P7000 not overclocked is faster than a 580 desktop AND a 7950 desktop. I'm really doubting it. I'm willing to believe that as a max overclocked score. Something like 6900 fully overclocked is possible, IMO, maxing the 100W limit.

 

GTX 670 scores P9075, its a 150W GPU. It beats the 7950 severely, which gets P7000. Why is it so hard to believe?

 

hmm okay, that would put the 680m and 7970m on approx. equal ground...