GTX 780 Ti reviews

[parvadomus]

What really surprised me is that GTX 780 Ti in SLI draws so much less power than R9 290X in Crossfire in Anandtech's Crysis 3 gaming test, while maintaining higher performance too!?!.

Multi-GPU Power Consumption in Crysis 3 (http://images.anandtech.com/graphs/graph7492/59709.png)
GTX 780 Ti SLI: 556 watts
R9 290X "Quiet" Crossfire: 643 watts
R9 290X "Uber"[Loud] Crossfire: 727 watts

Multi-GPU Performance in Crysis 3 (http://images.anandtech.com/graphs/graph7492/59674.png)
GTX 780 Ti SLI: 100.6 fps
R9 290X "Quiet" Crossfire: 80.1 fps
R9 290X "Uber"[Loud] Crossfire: 90.8 fps

This means that, with respect to Crysis 3 multi-GPU gaming perf-per-watt, GTX 780 Ti SLI has 45% higher perf-per-watt compared R9 290X Crossfire in "Uber"[Loud] Mode, and 45% higher perf-per-watt compared to R9 290X Crossfire in "Quiet" mode!

Now let's look at single GPU results.

Single-GPU Power Consumption in Crysis 3 (http://images.anandtech.com/graphs/graph7492/59709.png)
GTX 780 Ti: 372 watts
R9 290X "Quiet": 375 watts
R9 290X "Uber"[Loud]: 405 watts

Single-GPU Performance in Crysis 3 (http://images.anandtech.com/graphs/graph7492/59674.png)
GTX 780 Ti: 62.4 fps
R9 290X "Quiet": 51.9 fps
R9 290X "Uber"[Loud]: 53.8 fps

This means that, with respect to Crysis 3 single GPU gaming perf-per-watt, GTX 780 Ti has 26% higher perf-per-watt compared R9 290X in "Uber"[Loud] Mode, and 21% higher perf-per-watt compared to R9 290X in "Quiet" mode!

For those who want an NVIDIA card, the GTX 780/780 OC models are still a better value than GTX 780 Ti, but in general the perf-per-watt is really good on these high end Kepler models (and one can only imagine how much better performance would be if temp/noise/power limits were raised, especially in SLI systems).

Its barely ~10% more efficient than 290X (running @ high temps :'().
Link
Aftermarket 290X may be on-par efficiency-wise.

 

[24601]

That is incorrect. All GTX 780/780Ti/Titan reference models have much more thermal/noise/power headroom available than R9 290/290X reference models. As an example, Titan owners have dramatically increased performance by raising thermal/noise/power limits. As for perf-per-dollar, no one in their right mind is going to buy $700-$1000 who are on a tight budget. One may as well stick with a $150-$200 GPU to maximize perft-per-dollar.

If you didn't buy a $200 7950 then you didn't give a care to performance per dollar in the first place.

You wanted performance.

Watching everyone bashing me and then most of those same people saying pretty definitively that they would purchase a 780Ti is proof enough that you want performance.

 

[AnandThenMan]

I wonder how long Nvidia has been sitting on this SKU.

 

[wand3r3r]

Well 780 Ti is only faster, cooler, less noisy, better OC, draws less power and comes with bigger bag of goodies than competition.

If only Nvidia could make it cheaper too!

The price is the key drawback, voltage neutering is huge too to me but I won't apply that as critical for others.

Pros:
Slightly faster (7% at 1600p)
Quiet (until aftermarket comes, then that advantage is gone)

Cons:
27% higher price for a piddly 7% performance, wth NV price gouging again!
Voltage neutered :thumbsdown::thumbsdown:



(yeah you can void the warranty and do something about it, they'd love for you to buy a new card when you accidentally wreck it flashing it etc.)

 

[ams23]

I have to say that AT benchmark shows quite a big difference in Crysis 3.

I've seen other benchmarks where the 780Ti also has a big lead in Crysis 3 like hardwarecanucks but I've also seen much less difference like in [H].

The point is that perf-per-watt on this game is so dramatically in NVIDIA's favor (at least in Anandtech's review), especially in multi-GPU configuration. The only reliable way to measure gaming perf-per-watt is to measure both performance and power consumption with the same game and scene tested, and nothing you linked to provides that data.

 

[Gloomy]

I wonder how long Nvidia has been sitting on this SKU.

Same here.

It's time for some intrepid Titan owners to flash their cards with a Ti bios.

Are you man enough to risk bricking your thousand dollaridoo video card?

 

[GaiaHunter]

The point is that perf-per-watt on this game is so dramatically in NVIDIA's favor (at least in Anandtech's review), especially in multi-GPU configuration. The only reliable way to measure gaming perf-per-watt is to measure both performance and power consumption with the same game and scene tested, and nothing you linked to provides that data.

As if the AT review states it is the power consumption at the exactly same scene and settings.

And of course AT review shows 50W difference, same as what guru3d shows.

 

[kawi6rr]

Great numbers but painfully overpriced! Those scores didn’t impress me enough to dish out 100’s of more dollars for a few frame rates. I was contemplating purchasing this card but can’t justify the cost I’ll be purchasing the 290X thanks for making up my mind for me lol.

 

[ams23]

Its barely ~10% more efficient than 290X (running @ high temps :'().
Link

Incorrect. TPU's perf-per-watt methodology is flawed IMHO because they are averaging performance across all games, and then dividing specifically by power consumption in Crysis 2, which is not even a game that they test framerate on (and this affects all of their reviews, and is something I realized just recently). The only proper way to test perf-per-watt is to test game-specific performance and then to measure power consumption on that same game sequence.

 

[ams23]

As if the AT review states it is the power consumption at the exactly same scene and settings.

Ask Ryan Smith yourself, but I am pretty sure he has stated in a past review that the measured power consumption in Crysis 3 is using the same test sequence and 25x16 settings that he uses in the performance benchmark. So it is a valid perf-per-watt measurement when the two results are divided by each other.

 

[Unoid]

so 680 SLI, not even mentioning I run mine both at 1200Mhz is still a good bit better than the 780ti.

Guess I'll wait for 20nm for a single card upgrade

 

[Gikaseixas]

Amazing card for overclocking. Fastest GPU title is back to Nvidia and that, believe it or not is a big deal for both manufacturers.

The price is a big turn off. It's 8% faster, should cost $600. At that price AMD would have find it hard to sell a 290X.

 

[ams23]

Great numbers but painfully overpriced! Those scores didn’t impress me enough to dish out 100’s of more dollars for a few frame rates. I was contemplating purchasing this card but can’t justify the cost I’ll be purchasing the 290X thanks for making up my mind for me lol.

Well for those who want an AMD card and can put up with the high noise/heat from the reference models, the R9 290X has been made completely irrelevant by the R9 290. And for those who want an NVIDIA card and want more perf-per-dollar, go with GTX 780 SC models that are close to $500.

 

[wand3r3r]

That is incorrect. All GTX 780/780Ti/Titan reference models have much more thermal/noise/power headroom available than R9 290/290X reference models. As an example, Titan owners have dramatically increased performance by raising thermal/noise/power limits. As for perf-per-dollar, no one in their right mind is going to spend $700-$1000 on a graphics card when they are on a tight budget. One may as well stick with a $150-$200 GPU to maximize perf-per-dollar.

You contradict yourself.

Praising the power consumption (50w 780ti vs. 290x) difference in every post for the past few weeks doesn't change the fact that 75% of voters here said it's not even a major criteria, which on top of that suddenly you claim nobody buying $700+ cards are on a budget.

Why do you think they would care about 50w but not hundreds of $? Explain that one please. So which one is it? Either way clear bias is evident.

Bolded for emphasis to avoid deflecting.

 

[blackened23]

Roughly 12% better than Titan across various reviews. Impressive. Not worth 700$ for sure, but then again, it's a halo part and not unexpected and therefore forgivable.

The real interesting thing will be seeing aftermarket designs...particularly interested in the Asus DC IIT 780ti. That thing is going to be sick.

 

[JBT]

I'll admit I had a GTX680, sold it and picked up a used 7850 for $100 as it suited my needs. The last couple of weeks have gotten me pretty excited just to play with a new toy, but the sticker shock is at insane levels.

Sorry NV. At this point I'd probably just go with a 290 and OC, and get 90% of the performance for WAY less money. In all likeliness I'll just get nothing and continue on with my 7850.

 

[moonbogg]

OK but that advantage melts to less than 8% at 1600p, and only 15% over R9 290.

Also, R9 290s were built to run at 95C by design. You know the transistors can probably handle 110C as there is no way AMD left just 2-3C headroom.

Let's see, get a $700 card that is 15% faster when CF doesn't work or buy 2 R9 290s for $100 more and get 50-70% more performance when it works?

Also, for a $150 premium, the performance advantage over R9 290X is not enough.

I am not seeing at all how this card is worth $300 more over R9 290 for only a 15% performance increase. AT should have ripped this card apart for how overpriced it is.

[H]'s conclusion is once again spot on:

"It is clear from this breakdown that the AMD Radeon R9 290X is the better value in terms of performance by a long shot. It is $150 cheaper than the GeForce GTX 780 Ti, but delivers the same gameplay experience and very near the same framerate performance. A performance difference that makes little or no real difference. The $150 price difference is a much larger difference than the performance difference."

Agree with all of this. R9 290 crossfire is where its at. Also, this 780ti has how much Vram? 3gb? Really? With so much insane GPU power in an SLI configuration, you can't see yourself running completely out of Vram in a year with these cards? BF4 uses over 2gb at TEN EIGHTY PEE! The cards are great, for today (insane price aside). But in a year these will, yet again, be totally Vram gimped as is now standard procedure for Nvidia. I expect to have two R9 290's in my rig in the near future, unless I find the patience to wait for 20nm.

 

[kawi6rr]

I wonder how long Nvidia has been sitting on this SKU.

Probably since they first built the card but they didn’t need it because they had Titan for $1000 and people were paying for it. Now that AMD brought something to the table it forced Nvidia to step up again and at a lower price than Titan lol. I’ve always purchased both brands cards it just depended on what brand was doing what at the time of purchase but Nvidia has been price gauging its customers for too long. I give this round to AMD!

 

[lavaheadache]

umm, why is amazon trying to charge me tax on this ?

 

[4ghz]

I love how some people determined to take one side totally disregard the 3 free games for the 780ti. Regardless of what you think of the games they are worth a minimum of $70 right now. Even subtract $20 for the 10 minutes of "work" to put up a listing and emailing codes to buyers and that's an extra $50 off. So the card ends up being $100 more for about 10% more performance. Pretty much the way its been for a while between these two companies. If the 780ti isn't "worth it" here's a real shocker, Nvidia top card has NEVER been worth it to you. But please continue blathering on with cherry picked data about how its different this time.

 

[blackened23]

You contradict yourself.

Praising the power consumption (50w 780ti vs. 290x) difference in every post for the past few weeks doesn't change the fact that 75% of voters here said it's not even a major criteria, which on top of that suddenly you claim nobody buying $700+ cards are on a budget.

Why do you think they would care about 50w but not hundreds of $? Explain that one please. So which one is it? Either way clear bias is evident.

I don't care about power use whatsoever but I care about indirect effects (noise). Now, clearly AMD went cheap in this respect and I've stated this a million times. Bad design decision. Now, you are correct that in single card configurations, 780ti and 290X consume about the same power. Therefore, it stands to reason that a better reference blower on the 290X would make it similarly as good in terms of noise, but AMD saw fit to NOT do that.

Nvidia is going for user experience with their Titan shroud and it shows - while I could care less about power, I care when the resulting cooling solution translates that power into noise. All the more reason why AMD should follow NV in terms of optimizing the user experience. Instead, AMD is going for price/performance which is cool - the 290 is the best value high end GPU that we've had in years. Performance is amazing. User experience? Eh. I dunno. I like NV's approach to user experience even though it costs a bit more - not sure how everyone else feels, but since i'm not into custom water loops anymore i'll happily pay a few more bucks for overall user experience. I can't stress enough that I was super excited about the 290X prior to launch but I really wish AMD would focus on optimizing more than just frames - user experience matters. A lot. Aftermarket designs can and will change the game for the 290X but they're not here now, while excellent acoustic solutions are here for the 780/780ti now. That makes a purchase decision somewhat ambiguous - some people won't care, some people will water cool, by a lot of folks want the better out of box aspect. /dead horse

 

[Face2Face]

umm, why is amazon trying to charge me tax on this ?

http://www.bostonglobe.com/business...-tax-friday/acyUMexp4ChUk2I6JtdJyI/story.html

This is why I left Mass.. I saw this coming over 20yrs ago...

 

[96Firebird]

umm, why is amazon trying to charge me tax on this ?

http://www.bostonglobe.com/business...-tax-friday/acyUMexp4ChUk2I6JtdJyI/story.html



Edit - Don't even know why I bother anymore, getting slow in my old (mid-20s) age...

 

[kawi6rr]

Well for those who want an AMD card and can put up with the high noise/heat from the reference models, the R9 290X has been made completely irrelevant by the R9 290. And for those who want an NVIDIA card and want more perf-per-dollar, go with GTX 780 SC models that are close to $500.

I’ll wait for a non-reference card and save the money. When I was looking at the charts the 290X wasn’t much hotter than the 780ti not enough for the price.

 

[ams23]

Praising the power consumption (50w 780ti vs. 290x) difference in every post for the past few weeks doesn't change the fact that 75% of voters here said it's not even a major criteria

What kind of an inane comment is this? Power/noise/heat/etc. (and indirectly perf-per-watt) are important to reviewers and consumers alike, period.

which on top of that suddenly you claim nobody buying $700+ cards are on a budget.

Try reading what I wrote more carefully. I said that no one in their right mind would spend $700-$1000 on a tight budget. Get it?

Why do you think they would care about 50w but not hundreds of $? Explain that one please.

Because people who can spend close to $1k on graphics card all in one shot are usually not penny pinchers?

Seriously, some people just don't seem to understand the market dynamics of a high end halo card.