NVIDIA GeForce GTX 980 Ti 6 GB

[W1zzard]

Today, NVIDIA releases the GeForce GTX 980 Ti based on the same GPU as the Titan X, with same clocks but fewer shaders. This card is designed as preemptive strike against AMD's upcoming Fiji cards. In our testing, the $650 GTX 980 Ti can almost match the performance of the $1k Titan X.

Show full review

 

[HumanSmoke]

Looks like a solid base to work from for vendors. An 8% deficit in cores/shader modules seems quickly regained and surpassed with the increased clocking headroom the custom designs field.

 

[Luka KLLP]

Now imagine this with a high quality custom design cooler... Very curious what Asus, MSI, EVGA, Gigabyte etc will come up with

 

[ShurikN]

So marginally slower than Titan X, but a lot cheaper. I expected perf more in between the Titan and the default 980.
What a spit in the face of Titan owners. Yet again. Why am not surprised.

 

[the54thvoid]

Stand by for AMD leaks. Frankly, I want to see some because this card is better than I expected it to be (apart from the cooler). Only 2-3 weeks I guess for proper info.

 

[Luka KLLP]

Stand by for AMD leaks. Frankly, I want to see some because this card is better than I expected it to be (apart from the cooler). Only 2-3 weeks I guess for proper info.

Same here. I definitely underestimated this cards performance

 

[the54thvoid]

As for the overclocking Anandtech got 250Mhz too, leading to a max boost of 1477Mhz.

Results speak for themselves...

 

[Fluffmeister]

Damn, Fury better be a hell of a lot cheaper or a hell of a lot faster.

 

[...PACMAN...]

I actually like the UK pricing and the performance looks awesome. Depending on Fury, this may very well be the card I choose for my new setup in the summer

 

[m6tzg6r]

"Slight coil noise". Naughty naughty Nvidia, nobody wants any coil whine, nobody!

 

[HumanSmoke]

Stand by for AMD leaks. Frankly, I want to see some because this card is better than I expected it to be (apart from the cooler). Only 2-3 weeks I guess for proper info.

If AMD have any smarts, they'll release at least some sanctioned benchmarks or allow previews (if the card is comparable with/better than the 980 Ti) within a few days. Uptake of $650 graphics cards isn't fast, and many will be waiting for the EVGA FTW/Classified, Asus DCu3, Galax HOF, Gigabyte G1 Gaming, MSI Lightning, and host of other SKUs as well as EK's and AC's waterblocks. But peoples patience might not last that long once these cards are outed if AMD doesn't offer some solid performance numbers by then.

"Slight coil noise". Naughty naughty Nvidia, nobody wants any coil whine, nobody!

Coil whine is pervasive across both vendors and numerous AIB designs. Eradicating it is something approaching the search for the Holy Grail.

 

[Asbee]

Lol. Look at the chip. 1436A1 2014. 36th week.

 

[xkm1948]

This price is much better than I thought. Well, now I must hold on to my money for a few more weeks until AMD's fury is out, then it will be purchasing time~

 

[BiggieShady]

Did it loose two modules both from the same cluster like this:

or one module from each cluster like this:

or it can be either way?

 

[qubit]

Great card, but I'm sticking this generation out. It niggles me that I can't get the top GPU without spending £900 on a Titan X and even then I can't have a decent, quiet cooler like an MSI Gaming on it.

Also, I'm currently running two MSI GTX 780 Ti in SLI which appear to give roughly the same performance and are really quiet to boot. Yes, there are SLI issues sometimes, but one can't have everything. I don't seem to be missing out on any GPU features that matter by using Kepler, either. Gonna stick this generation out until Pascal and will upgrade to the top GTX model then. I'm especially intrigued by the advancements in SLI coming with Pascal, which appear to fix most or all of the problems associated with the current SLI architecture.

@W1zzard Great review as usual. Just one small point: in the positive bullet points of the conclusion you state "New software features (MFAA and DSR)". This is a tad misleading, because DSR is also available on Kepler and Fermi, which could mislead some less clued-up readers into thinking that they have to buy this generation to get that feature. I think a clarification that it works on older generations is important. Perhaps rewording it something like this would help: "New software features (MFAA and DSR - latter on Kepler and Fermi too)".

 

[radrok]

Kinda confirms what many of us have been saying since the first accurate leaks.

This thing is the usual Nvidia's slap in the face to Titan owners, history repeats itself.

 

[WarHell]

Is there any memory problem since it has some things disabled?

 

[radrok]

It niggles me that I can't get the top GPU without spending £900 on a Titan X and even then I can't have a decent, quiet cooler like an MSI Gaming on it.

But it would be just about having the full GPU, right? I mean two modules do not make a noticeable difference anyway.

If it's like that I can understand

Is there any memory problem since it has some things disabled?

96 ROPs points to single memory partition without strange 970ish allocations, I'm 99% sure it doesn't have memory "tricks"

 

[xenocide]

Is there any memory problem since it has some things disabled?

It says at least twice in the review that there are not.

 

[Darller]

Kinda confirms what many of us have been saying since the first accurate leaks.

This thing is the usual Nvidia's slap in the face to Titan owners, history repeats itself.

Why do you say that? I'm pretty sure TITAN owners knew what they were getting into... I sure did, and I don't feel slighted in the least. I've been using these babies for months, and enjoying every minute of it. The release of these cards doesn't change that one iota.

 

[2big2fail]

Is there any memory problem since it has some things disabled?

No, it doesn't. The reason this is the case is because the 980ti doesn't make use of all its memory address space. If nvidia put 12Gb on this card, then yes, it would have the same problem, i.e. 10.5 GB out of 12GB.

 

[qubit]

Coil whine is pervasive across both vendors and numerous AIB designs. Eradicating it is something approaching the search for the Holy Grail.

I really don't like coil whine and as you say, it happens everywhere. Certainly all of my cards have it to some degree or another.

Anyway, in my opinion, it's only as noticeable as it is, because the cards are built down to a price. Using an "overpowered" power regulation circuitry with higher capacity (likely physically larger) and higher quality coils would eliminate it, or make it so quiet that you'd have to put your ear to it to hear it. It would achieve this by effectively running that circuitry at something like half its rated power perhaps, along with those improved quality components.

I remember forum members from the Fermi era mentioning that there were such cards with almost no coil whine. I think they may have been EVGA custom builds and significantly more expensive than the standard cards.

 

[Darller]

I really don't like coil whine and as you say, it happens everywhere. Certainly all of my cards have it to some degree or another.

Anyway, in my opinion, it's only as noticeable as it is, because the cards are built down to a price. Using an "overpowered" power regulation circuitry with higher capacity (likely physically larger) and higher quality coils would eliminate it, or make it so quiet that you'd have to put your ear to it to hear it. It would achieve this by effectively running that circuitry at something like half its rated power perhaps, along with those improved quality components.

I remember forum members from the Fermi era mentioning that there were such cards with almost no coil whine. I think they may have been EVGA custom builds and significantly more expensive than the standard cards.

I only notice Coil Whine when I'm pushing 500+ FPS in old benchmarks... I've literally never heard it otherwise. It's curious how some people are so sensitive to it, and others like myself are not. I have Tinnitus too, and high pitched noises are usually death for my ears, but in this case I'm thankfully tone deaf.

Aside from my anecdote, isn't there some way to use clear nail polish or something to eliminate coil whine? One of my friends is sensitive to it and I recall him using that solution a long time ago. It still doesn't sound all that safe to me though...

 

[jchambers2586]

Time for 970SLI more performance.

 

[HumanSmoke]

No, it doesn't. The reason this is the case is because the 980ti doesn't make use of all its memory address space. If nvidia put 12Gb on this card, then yes, it would have the same problem, i.e. 10.5 GB out of 12GB.

Well, no it wouldn't.
IF the cut GM200-310 had disabled ROPs and their attendant 32kB L2 cache per ROP while maintaining the attendant memory controller it would have the issue. The GM200-310 in the 980 Ti retains the full 96 ROPs and 3MB of L2, so it doesn't.
I'd swear that the number of people who are unaware of the actual facts actually grows as the correct information becomes more widely disseminated - it's like stepping into BizarroPowerUp by mistake, or a guerrilla marketers training ground.

This thing is the usual Nvidia's slap in the face to Titan owners, history repeats itself.

How many - in your estimation - prospective Titan X buyers would have been unaware that Nvidia would launch a reduced cost 980 Ti in short order - and very likely in the same time frame as AMD's launch - at a much reduced cost?
I don't know of many people who would spend $1K on an individual hardware component that aren't in some way attuned to - at least, the consumer side of the industry.
For those that are willing to spunk $1K on a card without investigating the options, industry forecasts, a general consensus of forum posters, and historical fact?....Maybe you're familiar with the saying " There ain't no cure for stupid"