AMD 7970m vs GTX 680m

^^ I couldn't agree more Kernal. The 680m may have potential, but we just don't know right now. What keeps worrying me is that 900mhz vram that they put on there. We don't know what the ram is really rated at, but there's no way they would put say 1250 rated vram on it just to downclock it to 900mhz. They'd be wasting a lot of money....my point is that even if you can bump the clocks closer to desktop speeds, what's the point when you are running at such low bandwidth

what will make or break the 680m for me is how well it OC's, so i'm crossing my fingers that they at least put good rated vram on the final production 680m's

 

I'm pretty sure it was made to be overclocked, as the 685m will just be an overclocked 680m, 760m will be an overclocked 685m its Nvidia marketing at its best.

 

A reminder that the original 680m's shipped might very well not bother to use slower RAM so they can get it out the door.

Nvidia loses more by delaying right now than they will by shipping exceptional 680m's with downclocked faster RAM.

Note that even if they do, heat and power are still issues.

However, could you imagine almost a full GTX 670 in a laptop? THAT might indeed be worth the price hike.

 

If this is the case, i will go green and never look back lol

but nvidia put in 3.6ghz effective (115.2 GB/s) for a reason on these cards. We will soon see why.

It may be a moot point in m17x Alienware though as all Alienware laptops have a 78c hard throttle. Clocking the 680m that high will no doubt create a heat problem

 

I won`t be buying it ASAP. I`m still waiting to see how hot it runs. Plus I want to see a review of it, not synthetic stuff that have been posted by notebookcheck

 

I´d probably take 680M instead of 7970M, if it just would support 3 external monitors on Clevo P150. Sadly its been tested, that mobile Kepler doesnt support it.
Im not even completely sure, if i can get 3 external screens working with 7970M.
Oh and theres the price difference... mysn didnt tell me exactly how much.

 

yeah there is certainly a catch to this, also notebookcheck is citing their benches from nvidia.com, so it needed to be downclocked a lot to get into mobile land, but still 1344 CUDA cores, are you kidding me? (btw Slick, reinstalling hwinfo 1640 solved all my problems including fans/brightness/performance, thanks to Long! now it is half-perma-smile )

as soon as someone tests 680m (COME ON WIDEZU!!) I think I will buy it just for the kicks

 

You get 3 external monitors on the P150EM through the DisplayPort, but you will need extra hardware.

 

oh that's fabulous! good that 1640 solved your issues. I'll be downloading it... +

and when you get your 680m, i'll buy your 7970, ok? you think $100 is a fair deal ?

 

Unfortunately that's exactly the level of performance it provides :S (around i3 Sandy Bridge). Though if the CPU would be no bottleneck for the 7970m then for 1200 dollars it would be amazing.

 

I'm pretty sure I predicted that the 680m was going to be based on the 670 before everyone else.

Looks like I'm always right!

 

Uhm did you even look at the results? Both 680M and 7970M had the tessellation disabled. I tried to tell you that earlier
Tessellation is kinda what the whole benchmark is about

 

Not exactly. Fewer cores, 1344 vs 1536 (i think) and about 9% better performance from the 680 gtx. You still get more bang for your buck from a 670.

 

but if you go a gigabyte or zotak or asus and tace one of there overclocked ver then it beats all you dont need to tuch the card

 

Cloud, i was referring to this:

GeForce GTX 680M benchmaked on MSI GT70 - Notebookcheck.net News

It was a 680m test not a 7970m test? I've done my own testing with Ungine 3.0 and tess set at normal so i know what the 7970s will do. And you are right, the whole point of Unigine is tess...so what is the point of running a Unigine test with a 680m with tesselation off? Except to pump up the score and make the card look faster than it is

for example at 850/1200, 7970 @ stock clocks:

 

I do wonder if its possible to buy the Asus G75 with 670M and then buy GTX 680M and install it myself. How does that work? Is it difficult to do?

 

Seems like you missed it. Here is the 680M compared against 7970M with tesselation disabled. You don`t have a 680M with tessellation enabled to compare against, that is what I meant.

 

If we are talking about a OC'ed 670 vs a stock 680 you''re right. But with both GPU's OC'ed the 680 will still come out on top. You're also likely to pay more for an OC'ed 670 which closes the price gap between the two (OC'ed 670 & Stock 670).

 

i have 2 gtx 670's in SLI that clock to 1400mhz core. I don't think most 680's will do that

 

and what does the 7970m score there? Sorry, I can't see the pic nor can i understand the German

I'm scoring almost 60fps in Unigine with Tess normal...without tess, it would be faster than the 65fps the 680m got on that MSI laptop

Here is my Unigine score ....7970m @ 950/1400, tess normal

 

7970M got 69.4FPS, 680M got 65.9FPS.
Tessellation is kinda important with Heaven benchmark.

 

if that is the case is not looking good for nvidia once they enable tess the score is going to drop like 8-10fps and yours is with a 3820QM they have a 3920XM

 

As far as I know you can't because ASUS cards have a unique connector so normal mxm cards may fit but not work. In other words, its not upgradeable.

Your choices are Sager, MSI, or Alienware.

 

Congrats, you have a good unit. But that still doesn't change the averages when comparing performance between the 670 and the 680

 

Well that sucks.
I really like the G-series. Its a shame they lack the biggest GPUs. Sneaky bast*rds, bet they use a different connector just to make people buy a brand new notebook each time they want a new GPU

 

i was just saying to give a ide so you dont need to Oc and as well you get the cooler

 

Ah, i gotcha. The 670 is definitely a great card for the price.

 

Hmm although it's best to wait for user benchmarks, if this does indeed turn out to be the case, the 680M is in some trouble.

 

What does the 7970M get with tess disabled?

 

~69FPS, based on that German site. So, if those scores are valid, the 7970m has an edge in both Unigine and Vantage GPU. There's a notebookcheck review in the works, so we should see more benches in a few days.

 

Great, 7970M have a slight edge with 1280×1024 resolution, a tessellation benchmark but without the tessellation. These synthetic tests have been done on a computer fair

Really interesting indeed :wub:

 

you can get a peek on the notebook from msi on sweclockers latest vid from computex

 

You do realise that amd cards keep getting better as resolution increase.

 

I think he's starting to get nervous.

 

You do realize that its the other way around? Extreme is what 680M and 7970M should have been tested at. All serious reviewer sites test the desktop models at that settings now. Also in 3DMark 11.
Oh and did I mention that 6990M CF get 10FPS more in Heaven benchmark than 580M SLI @1080p although 580M SLI is 5% better in games? Thats why I don`t trust these synthetic benchmarks

 

can some one run the 3.0 with tess set to application controlled and not amd optimized. i want to see what some of you guys get.

 

You are forgetting the rather drastic change in bandwidth.

 

Look forward to seeing a post where you state you purchased a notebook with either a 7970M or 680M. Then anyone who has the other brand you bought, can start swarming the forum with their overclocks beating yours, both in games and synthetic benchmarks

 

He's right, the latest cards should be run with extreme mode. Furthermore, amd tends to fall behind in the latest releases, especially multiplayer games like bf3 which have a lot of replay value. Has amd learned how to program a decent driver yet?

Sent from my GT-N7000

 

AMD sat down and said. Know what we should do? Annoy 5150Joker. Let's not learn how to program a decent driver.

 

yeah, but those score not true. i benched both cards and the hd 7970 clearly is beating the 680. both cards are on air while cpu was under liquid nitrogen. and the 680 lost in air testing.

 

7-ish months from now there will be something better anyways. A life in glory is short lived with GPUs and basicly any IT stuff. Not to mention we might have Haswell out that time which will put our Ivy bridge to shame

But I appriciate the hostile remark

And thats why it is important to benchmark in extreme mode because you will see if 680M is bottlenecked by the low memory bandwidth.

 

Can anyone with a gtx 680 or gtx 670 do some test with lower memory clock to see how much of an impact it makes. the score does seem kind of low for an underclocked gtx 670, maybe thats the reason that gtx 670 is so close to the gtx 680, both have the same bandwidth.

yes, in extreme mode it would push more the core and less the memory, we could see some differences.

 

I also would like to see some scores with a lower memory bandwidth to compare the scores and to see if kepler is really that memory efficient that it still performs well.

 

there you go my CPU is 3720QM but overclocked to 3920XM stock well close lol the GPU is stock and tess is set on CCC

 

[note edited in the interest of brevity]


It's been my practice to consider a poster's opinions expressed on this board to be trustworthy until proven otherwise. I have seen nothing to contradict your self-designated "clueless moron" title, therefore I have no grounds upon which to disagree.

Having said that, perhaps the observation is a bit harsh, but your title made for soooo tempting a target..

As for myself, I'm faced with a "good look" every morning while shaving, which unfortunately reaffirms that I'm no longer as young as I used to be.

Now, as for all the benchmarks, leaked or otherwise, that I see out there, I have yet to see a really apples to apples comparo. The machine that put up the 680 bench's is running what I thought was the fastest available Ivy Quad, while the the one with the 7970 was using the slowest one. There are other performance benchmark impacting differences as well. Anyone that has done any serious competitive benching will tell you that you need to plug and play the differing cards in the same unit if you want truly comparable scores on the same software. Furthermore, being a *bang for buck* kind of guy myself, I'd want to see what each contender can do with solid 24/7 capable overclocks.

Throw in heat considerations, drivers etc. and then I'll be in a position to choose one for myself. Others' mileage, of course, may vary. I have to admit, tho, on an initial exam, the nvidia pricing looks a bit inappropriate to me. But then again, that's just me, and for as long as I can remember, its been just *them* too.


Speedy, thanks for posting. Those are some very interesting numbers

seer

 

The score is down 100 points, perhaps the driver 12.5 reduced the score? Your Vantage is very nice!

 

seeratlas is correct, we have yet to see a comparison of equal systems running the exact same config with the only variable being the gpu, we will also need this to be done by a consumer, not "legit" sites receiving benchmarks from the card makers themselves.

I for one will absolutely buy the gtx680m if i see that it does indeed perform 15% or more over the 7970m as i would like ambient occlusion and TxAA, as they are very appealing technologies.

 

setting the tess to use application settings on CCC really forces tess on all the time that's why but id they enable tess on the 680m run it may affect the frames even more

 

Agreed. And I'd likely to see a fair comparison not just in synthetic benchmarks but modern games as well. About the ambient occlusion and TxAA comment, you say ambient occlusion won't work with the 7970M? I thought the only thing I'd be missing out on was PhysX...