NVIDIA GeForce GTX 980 Ti

True, but to maintain 65 C @ 1539MHz on air quite a "feather in the hat"!

 

Don't let the meager resolution to fool you. DSR is just too tempting to resist the upgrade to 980Ti.

 

Oh, I like it and can use it in some titles. I enjoy my upgrades and it's something of a ritual. I'll just say that the 980 Ti is a great card but it doesn't fit into my upgrade cycle, or my price range. But it's still an important card to me as it signifies the level of performance that has to be beat by the next generation of top tier cards.

 

yes the cooler on that Kingpin card is pretty good, should be about 5c better than the regular Classified cooler, the impressive thing about that video/card though is that he is only giving the card 1.16 volts, so it's basically undervolted and still reaching those speeds.

 

Big Pascal should be adding back reasonable FP64 capabilities so I'm wondering how much FP32 (FP16?) performance that will soak up.

Since GM200 basically dropped FP64 it's probably going to hold up fairly well in games for quite some time.

 

Windows 10 VGA graphics card performance benchmarks (quick test) ...

http://www.guru3d.com/news-story/windows-10-vga-graphics-card-performance-benchmarks.html

 

I personally noticed a performance increase in my laptop's Radeon 5650M when comparing to the late builds of Win10 Preview, but I'm also sure this was specifically linked to the far newer graphics drivers that come with W10 that were otherwise previously "unavailable" for a card of this age. I think the last AMD driver update for the 5000-series cards was back in May of 2014 or something.

 

http://www.guru3d.com/news-story/corsair-bulldog-back-in-black.html

 

Gigabyte GeForce GTX 980 Ti Extreme Gaming WaterForce Review

http://www.guru3d.com/articles_page...80_ti_extreme_gaming_waterforce_review,1.html

 

So that's what happens when you put a watercooler on a 980Ti. And it still uses less power than the FuryX

 

Other than the obvious reduction in noise, I"m not as impressed as I hoped to be. The real challenge is having enough voltage to take advantage of the excellent cooling. With a max of +87mv, the attained clocks are only ~50MHz (max) from what my Gigabyte 980Ti G1 Windforce can do.

Still, Maxwell is a honkin' beast.

 

Now if only it were as small as the Fury X...

Regards,
SB

 

Actually why is it so huge? This goes for GPUs in general. Most are 2x bigger than they really need to be.

I hope AMD continues the trend of making small reference designs. And I wish NVIDIA would follow suit. But I guess with even mITX systems capable of handling 13" cards, it isn't a huge deal in the grand scheme. Just a pet peeve.

 

I was under the impression that Fury / FuryX were able to be much smaller because they don't need the individual GDDR5 memory traces. It's not just the traces by themselves either; the traces to each chip need to be equidistant to ensure you have proper clocking phase alignment between all the individual chips. This means that the length for ALL chips ends up being the "maximum length" of whichever chip had to be furthest away.

 

Maxwell isn't like Kepler with regards to voltages. Maxwell's frequency scales poorly with voltages unless extreme cooling is used. You don't really get benefit with water, not much anyway. In fact using too much voltage can easily hurt your OC and cause instability with Maxwell, that's why and because it often just makes the card hit the power limit sooner annoys me when every reviewer just always apply the max +87mv as a default with their OCs (The card doesn't necessary use it all though). They should see how far it goes with stock voltage and then try incrementally adding voltage to see if they get benefit from it, not just put the max allowed instantly.

 

As an anecdote, I arrived at the same conclusion that you just explained. My best overclock on my 980Ti was with zero additional voltage. If I upped the volts, I ended up performing less because of the power limit -- exactly as you described. I'm super impressed with how my well my factory-overclocked 980Ti was able to go even further on the stock voltage, so I'm really not complaining.

I guess the statement here is the water cooling is really only useful for keeping it silent. The clocks just don't have much more room to grow...

 

Why does NV implement a hard power limit on their cards?

 

It began with Nvidia's Green Light program.
Tends to cut down the warranty claims for board vendors, and Nvidia would prefer people griping about the nanny state than having to shell out for warranty replacements (and carrying the RMA inventory that is associated with it.)
It's pretty easy to see how certain SKUs get tainted when OC is pushed hard, fails, and the model becomes an instant internet meme. Gigabyte's GTX 780Ti GHz Edition and Asus's GTX 670 DCII TOP are prime examples of how a single product can have a wider ranged detrimental effect on a brand.

 

Makes sense. Thanks!

 

I fully expect the boards with Pascal chip/s that use HBM2 will likely be similarly small.

Regards,
SB