Jump to content

GTX 680


Recommended Posts

the reference models are out now. and while its a good card, its nearly identical in performance to the 7970. only slightly better then the 580. it is however, quieter and uses less power then both. bu other then that, it didnt blow either out of the water.

Average power is actually the same under load (3W more for 680 over 12 seconds, 3W more for 7970 in peak). In idle 7970 consumes less, but 680 consumes less under some loads like video playback. Overall, they're neck to neck here.

 

GTX680 has less consistent performance than 7970; it delivers stellar fps under some conditions and lags behind under others. Unfortunately, it works best in low resolutions or under lighter loads, and drops the worst under the heaviest loads.

 

Overall performance results under medium-heavy load (1920x1080) are as this:

680 OC > 7970 OC > 7950 OC > 680 stock > 7970 stock > 580 stock > 7950 stock

The results under heavy load (2560x1440 or 1920x1080 3D) change:

7970 OC > 680 OC > 7950 OC > 680 stock > 7970 stock > 7950 stock > 580 stock

 

It would be a tough choice today. On one hand, 680 is faster on the average, on the other hand, 7970 is faster when you need it the most, i.e. when your fps is at its lowest. The worst-case scenario recorded was in Metro 2033, where 680 OC provided 31 fps average and floored at 18, while 7970 OC had 36 average and floored at 20. This lack of performance in tough scenarios is a direct result of 680's narrow bus width of only 256 bits.

 

If you're into 3D displays and gaming with 3D glasses, 7970 might be a better choice, but check 2560x1440+ or stereo 3D benchmarks for your game specifically. If the most you play is Skyrim, 680 seems to be your card. But then you are only getting 76 fps rather than 70 fps with 7970... big deal when your display caps it at 60 fps. Well, at least you know you have more headroom - but on the other hand HD texture packs will skew things back towards 7970 with its wider memory bus. Really it will boil down to price.

 

 

Both cards are unreasonably loud with stock cooling, but 680 is a little quieter (43 vs 46 dB). However, you definitely do not want either of them in reference design. Both cards suffer from poor power supply design with only a 4-phase VRM on 680 and some issues noted with 7970 VRM. Remembering how many 570's died from overclocking (and RMA rejects these), you might want to avoid overclocking a reference 680.

 

Custom PCB and cooling designs offer better reliability and better cooling with less noise. As always, Asus is going to be the best (sorry folks, I'm not being a fanboy, but triple-slot DirectCU II cards with 12-phase VRM's are just that good), but unreasonably expensive. Look out for less expensive offerings from other makers, basically anything with 2 or 3 fans on it, that may cost only a bit more than reference designs, but work much quieter. Powercolor, Gigabyte, EVGA, what have you.

Link to comment
Share on other sites

I overclocked my 7950 to match the 7970 performance, they say the 7950 can actually out perform a 7970 if overclocked properly :biggrin: :biggrin:

 

Not to mention it stays cool :thumbsup:

 

I highly recommend Gigabyte :thumbsup:

Edited by Thor.
Link to comment
Share on other sites

The GeForce GTX 680 is actually the midrange card that would normally have been the GTX 660(TI or not.) nVidia renamed it to the 680 to emphasize that their midrange card performs at or even better than AMD's current flagship card. nVidia still has an even better GPU yet to be released at this time. The linked article mentions this in the first page:

 

http://www.tomshardware.com/reviews/geforce-gtx-680-review-benchmark,3161.html

Link to comment
Share on other sites

About the number of cores, it is misleading is taken alone, Nvidia uses much more complex cores and even then manages to keep on pair with the sheer amount of shader cores from AMD/ATI due to sheer speed values.

 

The two cards are conceptually different and then performs very differently depending on the circumstances.

 

At the moment I stick yet with Nvidia mostly because previous problems getting ATI to run nicely with Linux but seriously admit both are very good options.

 

 

PS: A bit off topic, just want to brag a little, I managed to push two mid range GTS450 in SLI (on a crossfire only mobo) and I'm amazed with how good it is performing in almost all games compared with the single card :) actually I can shove textures packs on them better than I could, say, in a 470 with 1GB of VRAM.

 

This combo can run the unigine benchmark with high shaders, moderate ambient occlusion and 4x AA and AF in above 30 FPS all the time, the same card in single is just too sluggish to run unigine even in the most basic setup. Still to try benchmark them with more aggressive setup.

Edited by nosisab
Link to comment
Share on other sites

The GeForce GTX 680 is actually the midrange card that would normally have been the GTX 660(TI or not.)

That's an older and now disproved misconception, based on the chip codename being GK104. As most people expected GK100, GK104, GK108, etc, it was presumed that GK104 would be the midrange card.

 

What happened is both companies hoped to launch HD7xxx/GT6xx in mid-late 2011, which would be respectively Tahiti XT and GK100. Designs themselves, then, were complete in early 2010 already. However there were severe troubles with manufacturing these chips, so the launch has been delayed. While production issues were being fixed, NV had the time to develop an updated chip design, which got the next alphanumerical designation in sequence.

 

 

About the number of cores, it is misleading is taken alone, Nvidia uses much more complex cores and even then manages to keep on pair with the sheer amount of shader cores from AMD/ATI due to sheer speed values.

That used to be so with the previous generation, although not exactly. More precisely, ATi counted each scalar processed in its 5D or 4D vector unit as a 'core', while NV counted each 4D vector unit as a 'core' (this had some basis in architecture).

Currently both architectures are very similar, and both count each scalar processed as a 'core'.

They are not really cores anyway, so the distinction is academic only.

Link to comment
Share on other sites

Hmm Tessellation i was getting 60fps completely maxed out with two gtx470's, Although temps sky rocketed from 80 102c in matter of minutes :teehee: :teehee:, one problem with Nvidia.

 

Witch are the running temps, but i was afraid for my other components over heating...

 

I have a Sabertooth 990fx motherboard quad sli-crossfire, next build i think will be Nvidia. Just for a change...

Edited by Thor.
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...