I've heard that 680 GTX's overclock really well when it comes to memory. I can understand the GPU not clocking up much more than it is due to it being a FTW+, but the memory is still the stock 3004MHz. When I go up even +100MHz, it will crash any benchmark I throw at it.
 
Is this normal with the 4GB cards? Is voltage my limitation? Is my card just a POS?

I was having the same issue last night with 2 of those cards in 3D mark 11. I don't have an explanation for it.

i can get +200 memory but anyhigher crashes

Think of it like system memory. When you add more sometimes you have to loosen timings. Same concept with 4GB models.

I recommend spending the extra $30 to go to the Classified 680. I had 4x FTW+ 4GB and I sold them because they were hot and they did not OC well. I now have 4x Classified 680 and they OC well at both core and memory - and they're 4GB cards.

I got nearly 18000 3d mark 11 scores out of mine just messing around a bit, so I am very happy with the performance.

So I shouldn't get the FTW+ 4GB then?  I am planning on getting 3 for my rig.  Trading with someone else my 2GBs Reference + cash for the FTW+.  Does it perform better than stock 680 as it is?

DO NOT get 3x or 4x FTW+ cards. Backplates are useless in 3x or 4x configuration, anyway - due to the tight space and restriction of airflow. After removing my backplates (FTW+ 4GB) I gained about 3-5C cooler temps in 4-way and 3-way.
 
I have owned 4x reference 680s, 4x FTW+ 680s, and now 4x Classified 680s. If you're going to go 3x or 4x 680 - either get reference or Classifieds for temps.
 
Classifieds have an improved cooler. The vapor chamber cooling in the FTW+ 4GB is a joke. Plus, the Classifieds will get a supported watercooler. FTW+ 4GB will likely not.

Yes, the FTW+ 4GB are great cards - if you have one or if you have SLI (2x). But after my experience with them (I ran them in both 3-way and 4-way due to RMA) I am on a crusade to get people away from them in "sandwich" mode on air. They do not perform.
 
This is not because they need to be watercooled. That is a myth. On-air 680 performance in sandwich mode is actually quite good (reference: my blog post about my 4-way vanilla 680s - http://1pcent.com/?p=43). It's because the FTW+ 4GB is factory OC'd and runs hot. The "upgraded" cooler does not seem to perform well enough for "sandwich" (3-way or 4-way) mode.

whyscotty

I run TRI  SLI  no probs, power110   core +65   mem+465 hottest  card rarley goes over 70.

 
Word. But your temps would be even better with Classifieds. All for only $30/card more. :)

idk about anyone else but my FTW+ 4G idles at 35C with 40% fan .Not hot at all to me :)

USFORCES

Newegg sells the FTW now for $609, it's like $50 more for the classified and ebot another $80. I've heard some of the classified's don't clock as good as the FTW do on 1.175 and I've heard the other way around too, it's luck I guess.
Anyhow I gained about 10FPS from the FTW OC to 1254MHz in SLI so overclocking does help..

 
Well, the EVBot isn't required to get better thermal performance. It's an optional thing. So, $50 for a higher factory OC and a better cooler - plus the option to get an officially supported waterblock later AND get an EVBot for further tweaking. Not bad, I say.

I just received my 680 Classified 4GB.
I am struggling to get the EVGA PrecisionX to work.
I can crank up the Power Target to 120%, change the voltage to 1.175, increase memory speed by 500MHz but no matter what I do with the GPU overclock I don't see much of an improvement.
Running the EVGA OC Scanner with Furry-Tessy-Donut gets me the same 106fps no matter if memory is running at 3000 DDR versus 3500 DDR.
The GPU does whatever it wants and I don't see much of a change no matter if I leave it at default or increase by 180MHz.
It stays at 1071 MHz as per PrecisionX at +182MHz while at default it goes up to 1228 MHz.
Why does the GPU run slower if you increase the GPU Clock Offset?
Am I hitting a maximum power draw target and no matter what setting I choose the GPU does what it wants?

I just noticed that the GPU speeds up to 1280MHz (with 120% Power, +58 MHz GPU and +502 MHz memory) when I have 100% CPU usage (BOINC). Looks like BOINC is also using the GPU for calculations.

Well my GTX 680 FTW+ clocks at +131 power +70 Gpu +200 memory
 
Anything higher and its unstable in Heaven, but at this setting completely stable in heaven. But sometimes I get the nvidia driver had failed error when playing games and the card resets to default clock not sure what causing that.

sgunes

I just received my 680 Classified 4GB.
I am struggling to get the EVGA PrecisionX to work.
I can crank up the Power Target to 120%, change the voltage to 1.175, increase memory speed by 500MHz but no matter what I do with the GPU overclock I don't see much of an improvement.
Running the EVGA OC Scanner with Furry-Tessy-Donut gets me the same 106fps no matter if memory is running at 3000 DDR versus 3500 DDR.
The GPU does whatever it wants and I don't see much of a change no matter if I leave it at default or increase by 180MHz.
It stays at 1071 MHz as per PrecisionX at +182MHz while at default it goes up to 1228 MHz.
Why does the GPU run slower if you increase the GPU Clock Offset?
Am I hitting a maximum power draw target and no matter what setting I choose the GPU does what it wants?

I just noticed that the GPU speeds up to 1280MHz (with 120% Power, +58 MHz GPU and +502 MHz memory) when I have 100% CPU usage (BOINC). Looks like BOINC is also using the GPU for calculations.

 
TEMPS might be your wall. Once you hit 70C, your card downclocks. You can adjust your fan curve to make it at 80% when the card reaches 68C. I did and it got me some results. Also look at HardForums Official GTX 680 overclocking Thread!. These guys are pro's.