NVIDIA GeForce GTX 780 Ti 3 GB

[harry90]

I'm sure with their 512-bit bus they will do fine at higher resolutions.

Personally just think that talk about 4k is still very premature.

In making their decision, people should stick to what matters now, and realize $300 is a lot money and they are not using 4k.

I find people bringing up non-important points for the sake of arguing pointless.

...

The problem is that 780TI suffers at resolutions higher than 1080p. Its 6-7% faster in 1080p but is equal at 1600p or higher resolutions. 3GB just isn't enough with full AA, 4X or 8X. Moreover, R290x still lowers it clock even at 55%fan and Ubermode. So 780TI's victory is pretty narrow 3-4% average. well c an updated review of both cards with non-reference coolers soon.

 

[HTC]

Its a good question, but, the switching is so fast, I do not think it would make a difference. Not to mention, you lose the peaks too by starting out lower.

Turn the fan up.

If the test reveals inefficiency, then there's legitimate reason to avoid any lower quality cooled cards, regardless of which camp.

 

[EarthDog]

If

(Big "if" LOL!)

 

[radrok]

3GB just isn't enough with full AA, 4X or 8X.

To be fair I don't know if I can confirm this statement, after W1zzard asked me to bring him proof about more than 3GB usage on my Titans I couldn't find any game going over 2.8GB on 1600p even with 16X AA.

The only game that would go over 3GB was Skyrim, not vanilla.

Talking about 1600p.

 

[arterius2]

If the test reveals inefficiency, then there's legitimate reason to avoid any lower quality cooled cards, regardless of which camp.

my question is, AMD didn't know or cared about this during their months(or years) of internal testing?

 

[EarthDog]

my question is, AMD didn't know or cared about this during their months(or years) of internal testing?

Great point, which is what I was thinking in formulating my opinion... If the performance difference was more than negligible, they likely would have done something about it.

To be fair I don't know if I can confirm this statement, after W1zzard asked me to bring him proof about more than 3GB usage on my Titans I couldn't find any game going over 2.8GB on 1600p even with 16X AA.

The only game that would go over 3GB was Skyrim, not vanilla.

Talking about 1600p.

Its not the vram that is the problem there it seems... it is the 384bit bus and AA that may be the biggest difference.

I know this is apples and oranges but here is some GREAT testing (IMO) across several games at 1080p and 5760x1080 as well as 3D for vram use...

http://www.overclockers.com/forums/showthread.php?t=718118

IN BF3, I hit 2.3/4GB used in BF4 I hit 2.4/5GB used (1440p default Ultra settings).

 

[arterius2]

Great point, which is what I was thinking in formulating my opinion... If the performance difference was more than negligible, they likely would have done something about it.

things don't happens by chance these days, they have their reasons. my take on this?

It was probably never AMD's intention for 290/290x to be clocked this high to begin with, it seems that somewhere along their development path, they made a last-ditch effort to pretty much push the power envelope to its limit (possibly this happened when they learned news of Nvidia's plans)

the same sort last-ditch effort they made to 290 to "suddenly" lift the fan speed limit to 47% at last moment.

in summary, it looks like AMD was forced to ship their cards clocked this high at last minute (this also explains the inadequate cooler - it was never meant to cool a card of this power envelope) to remain competitive to whatever they thought NVidia had planned, and it seems they were right on the money for that. literally.

 

[PopcornMachine]

Slap a waterblock on it!

Ordered 290, but waterblocks not available just yet.

Going to when I can.

 

[HTC]

my question is, AMD didn't know or cared about this during their months(or years) of internal testing?

Dunno.

If it's proven to be inefficient and this result is then "shoved in their faces", don't you think they'll be forced to up the reference coolers quality?

But don't think i'm only talking about AMD here: i'm also talking about nVidia and should it's throttle efficiency prove bad, i'll "direct my guns" @ them as well.


But this talk is premature: the efficiency may actually be very good. Don't know until someone actually tests this *subtle hint for W1zzard* ...

 

[the54thvoid]

I think the >3GB Vram usage argument is mostly irrelevant, especially with findings from the link given by Earthdog.

People have jumped on it as a validation as to why the 4GB is better than the 3GB of 780 and now 780Ti without even using any evidence. The irony being a single 290X or 780Ti is barely, if at all powerful enough to game at 4K res. Nvidia even state that - single GPU's are not capable yet of rendering for 4K. The proof is in the reviews.

When was the last 'what monitor size do you use' poll? How many people that argue about 4K even have 1440p? Very few. Even fewer have 1600p.

Gaming on BF4 with ultra and max gfx settings I do not see above 2.8GB usage. Most maps are actually settled at 2.5-2.6GB. In fact, i've gamed at 1440p for a good while now and have never seen anything higher for single monitor gaming (and I've got a pretty redundant 6GB VRam on my card).





EDIT:

On a total side note - it's an appalling lack of professionalism that so many reviews I've read have used PCB pics from a 780 or Titan. Well done W1zzard for not taking a half assed lazy approach. W1zz's shots show us there is a different power circuitry in place.

 

[EarthDog]

If you check out the steam stats, I believe that less than 1% of users have a resolution higher than 1920x1200 or multimonitor setups.

 

[the54thvoid]

If you check out the steam stats, I believe that less than 1% of users have a resolution higher than 1920x1200 or multimonitor setups.

I guarantee someone will wade in with a "steam surveys are useless post".

 

[swagnuggets123]

[page=Introduction & Specifications]

Why no bf4 benchmark wiz?

 

[HTC]

If you check out the steam stats, I believe that less than 1% of users have a resolution higher than 1920x1200 or multimonitor setups.

IMO, i think that's the wrong way to think about it.

One should check the percentage of steam users have either camp mid high to high end cards and then, out of those, how many have a resolution higher than 1920x1200 or multimonitor setups.

 

[EarthDog]

I guarantee someone will wade in with a "steam surveys are useless post".

Clearly it is not a completely accurate representation of ALL users, however, its big and broad enough to at least get an idea of what the landscape is like...

 

[manofthem]

Why no bf4 benchmark wiz?

He already said it's going to be added to next round of benchmark reviews, as well as a few others; I think it was the new AC and CoD


Here:

W1z, when will you be implementing BF4 to the benchmark lineup?

next rebench (December). also adding cod ghosts and assassin's creed IV (if it turns out to be usable for benching, ie no fps lock). and kicking out Skyrim

 

[arterius2]

Clearly it is not a completely accurate representation of ALL users, however, its big and broad enough to at least get an idea of what the landscape is like...

last time I checked Intel HD graphics was still the most popular graphics card used by steam users.

 

[EarthDog]

He already said it's going to be added to next round of benchmark reviews, as well as a few others; I think it was the new AC and CoD

yeah, gotta give reviewers time to, for this type of game with no built in benchmark, to find a good location (may I suggest Tashgar on the SP campaign?), and then retest... oh how I hate retesting......

 

[Gadgety]

"While 3 GB doesn't seem like a lot of memory, I am convinced it is enough for all games, in all resolutions, and in the forseeable future. My guess is that we'll also see board partners release additional SKUs with 6 GB, to cater to buyers who think they absolutely need more memory (even though they don't)."

Yes 3GB might be enough for all games, but the "buyers who think they absolutely need more memory" (and really do), are for example those who work with rendering and want to maximise both CUDA cores as well as memory in order to generate faster renderings in larger scenes. I'm sure there's room for 4, 5 and 6GB versions, or even larger, in the marketplace. Value wise this is what it looks like currently:


The Quadro K6000 is $5000 for 2880 CUDA cores: 0.576 CUDA cores/dollar
The TeslaK20 is $3500 for 2496 CUDAs: 0.7 CC/dollar.
When the Titan launched it was a bargain at $1000 for 2688 CUDAS: 2.69 CC/dollar
When the memory limited GTX780 launched it was 3.55 CC/dollar


The GTX780Ti provides 2880 CUDAs for $699: 4.12 CC/dollar. Apart from FP64, the thing the Titan still has over the 780ti is the memory. Bringing out larger memory version 780ti's, and they will have a market among those who really do need them.

 

[W1zzard]

Its not the vram that is the problem there it seems... it is the 384bit bus and AA that may be the biggest difference.

you do realize that gtx 780 ti has more memory bandwidth than r9 290x, right? it's not only how many lanes you have on your highway but also how fast the cars go

oh how I hate retesting..

look at how many cards I have to retest :/ gonna drop a few older ones but should still be 25-30

 

[arterius2]

you do realize that gtx 780 ti has more memory bandwidth than r9 290x, right? it's not only how many lanes you have on your highway but also how fast the cars go

hrm interesting, I used to always thought bandwidth as how many lanes and clock speed for how fast the cars ago, and memory size? I guess you can think of them as how much weight can this bridge hold before it collapse

 

[W1zzard]

hrm interesting, I used to always thought bandwidth as with how many lanes and clock speed for how fast the cars ago, and memory size? I guess you can think of them as how much weight can this bridge hold before it collapse

bandwidth = bus width * clock rate (+ more math for gddr5, and 8 bit=1 byte)

should become obvious if you think about it for a minute: bandwidth = how many cars per hour can go through the toll station on the highway

for memory size you could think of how many cars fit on the parking lot, connect that to your memory bus highway, and now ask how people can get home when work ends and wont be late for dinner (= wife aggro)

 

[theonedub]

Are the heatsink fins behind the plexi window black or did Nvidia tint the window a little bit? The heatsinks on the Ti look jet black whereas on my 780 they look bright silver (sorry if this was mentioned in the review somewhere, I just took a quick glance).

 

[Ja.KooLit]

To be fair I don't know if I can confirm this statement, after W1zzard asked me to bring him proof about more than 3GB usage on my Titans I couldn't find any game going over 2.8GB on 1600p even with 16X AA.

The only game that would go over 3GB was Skyrim, not vanilla.

Talking about 1600p.

max payne 3. turn on all eye candy.... requires alot even my crossfired 7950 cant....

talking about the hbao and ssao... dont know the difference and function

this is at 1600p

 

[johnnyfiive]

This is getting stupid (in a good way).

Here is your call AMD, where is the R290X GHz Edition that comes with a superior cooling solution?! You're losing again! (Already..)

I bet we see a R290X GHz within the next 3 months, and it will cost $629.99... lol