[Techspot] Nvidia vs Nvidia 5 generations of cards tested

[96Firebird]

Even they are at stock the 980 is a piffling ~15% faster than the 780 Ti.

Expect that to increase, since according to some Nvidia is neglecting Kepler!

:awe:

 

[Rvenger]

I know Maxwell wasn't the revolution on the performance side but it makes a nice SLI setup with a smaller power supply. I didn't want to run 2x 290x OC on a 5820k OC setup. The 970s are quite nice to have in this regard and they perform exceptional for the power they output.

 

[SirPauly]

My biggest gripe is the naming convention. In the past NV would never label a mid-range GeForce Ti 4200 as 4600, 5600Ultra as 5800Ultra, 6600GT as 6800GT, GTX460 1GB as GTX480 despite those next gen mid-range cards outperforming the previous gen flagships.

Imho,

Naming conventions are dynamic and ever changing based on market place and competition.

The 7900 series was ridiculously small and efficient and yet when released nVidia charged a premium for them.

The 9800 series were small and efficient and yet released were named new generation flag ship sku's but yet charged mid-range prices.

It really isn't about naming conventions with GTX 6XX and 9XX series but a strategy change based on the lessons of Fermi, imho! The bigger dies were getting so complex to bring to market in a timely manner first and nVidia suffered through this with Fermi --- third party reviews were not too kind, late to market, efficiency suffered, mobile suffered, actually lost discrete over-all share, so they changed and dramatically focused on efficiency, rightly or wrongly, some agree or disagree.

Naming conventions here are dynamic and ever changing based on market place and competition much more than anything else, but is always admirable to see posters looking out for the consumer or gamers in general.

 

[Spjut]

It had been fun to see the old GTX 280 as well, it had probably been ok in 1366x768.
The 8800GT is still minimum for alot of games.

 

[Gryz]

Any idea why all forms of AA were turned off in all benchmarks ?

 

[lavaheadache]

Any idea why all forms of AA were turned off in all benchmarks ?

so we weren't seeing single digit numbers often for the 480 and 580 I presume.

 

[Red Hawk]

Neato, I hope they do another article like this for AMD. 5870 to 6970 to 7970 to 290X.

 

[RussianSensation]

]The bigger dies were getting so complex to bring to market in a timely manner first[/B] and nVidia suffered through this with Fermi --- third party reviews were not too kind, late to market, efficiency suffered, mobile suffered, actually lost discrete over-all share, so they changed and dramatically focused on efficiency, rightly or wrongly, some agree or disagree.

Naming conventions here are dynamic and ever changing based on market place and competition much more than anything else, but is always admirable to see posters looking out for the consumer or gamers in general.

Ya, that's why the videocard market is among the markets lacking much common sense in naming compared to most other products in the world. For example, Mercedes and BMW don't magically relabel their new C-Class and 3 series as E-Class and 5 series even if the newer models keep growing in size, features and performance to outclass 1-2 generations old E-Class and 5 series. NV should name cards how they stack in their generation. In that case it would be $550 960Ti and then once the bigdie GM200 comes out, label that as a 980/980Ti. Instead, they'll magically come up with another generation such as 1080 (or w/e) when we are still talking about the exact same generation! They already pulled this non-sense with 480-> 580 (really this is a 485!) and 680-> 780/780Ti (those were really 660Ti and 680/680Ti if you will). Not trying to attack NV solely because as I said AMD did the same non-sense with 2900XT -> 3870 and 5870-> 6970.

In the past ATI's and NV's naming actually made more sense! Since you've owned NV for 15+ years, can you recall a generation where NV's next generation mid-range card was slower than its last generation high-end card? I am having a really hard time finding one (using 560Ti vs. 480 isn't valid since they are the same generation). Pretty much everyone here knows that 680 and 980 are 660Ti/960Ti. It doesn't take away from their perf/watt and performance, feature characteristics but their naming convention is simply a money grab to justify charging $500-600 before the real flagship of each of those generations drops. Whether it's AMD's fault or not isn't really relevant. The point is NV started naming their next gen mid-range cards as flagships primarily to justify $500+ pricing while they work on getting the real flagship out. This also allowed them to raise flagship prices from historical $500-550 all the way up to $700. It makes sense for them since the low-end dGPU market for AIBs is dying and the sub-$100 market is pretty much dead. Many of us talked about AMD/NV raising prices over time to compensate for losing out on these lower dGPU market segments.

Expect that to increase, since according to some Nvidia is neglecting Kepler!

:awe:

That's not necessarily a good thing for early adopters of Titan/780/780Ti. Right now a 780Ti is just 2% faster than an after-market 290/reference 290X at 1440P. Considering that an after-market 290 was $400 1 year ago when 780Ti sold for $700, it's not surprising some people are unhappy with recent performance of Kepler cards. You can make an argument that those who paid $650-1000 for 780/780Ti/Titan are price inelastic / high income consumers so they knew the cost of being an early adopter but it doesn't change the facts that Kepler performance in recent titles is extremely poor given their initial launch prices and supposed future-proofing that those higher prices entailed.

 

[escrow4]

Expect that to increase, since according to some Nvidia is neglecting Kepler!

:awe:

The 980 is stripped down everywhere barring ROPs - less CUDA cores, less TMUs, less bandwidth - see a pattern here? I will not pay full price for a lite pansy poky GPU. The 980 Ti should be 512 bit full fat with over 3000 cores at least to justify the inevitable price.

 

[rgallant]

Woah, from Fermi flagship to Kepler flagship is almost double the performance. Wish we were still getting node shrinks.

almost matches the clock speed drifferential 607 vs 1100 with boost dosn't it.