• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

980 Ti @ $599?

will Nvidia actually launch the 980 Ti at a significantly lower price point then the rumored 390XT's price of $849??...Nvidia will be taking a page out of AMD's- if you can't beat em in performance then beat em in price- motto
 
Nvidia would never undercut AMD on price/perf since they don't have to.
So take whatever AMD does and add 20-30% price = 980 Ti.
 
Nvidia would never undercut AMD on price/perf since they don't have to.
So take whatever AMD does and add 20-30% price = 980 Ti.

The tech forum view of AMD and Nvidia's pricing structure in relation to each other seems to be something that is extremely inconsistent with reality and history.

When Nvidia was the the one initiating a new product series with the GTX 7xx and GTX 9xx cards they moved costs downwards compared to both themsleves and AMD. GTX 6xx cards also did the same when they were releasing behind AMD. Nvidia also cut prices in relation to AMDs R9 2xx (Nvidia's repositioning of the 770/780 actually led AMD to respond with last minute change at the driver to up the power limit , and therefore performance, of the R9 290. This was talked about by reviewers and done when they had already began testing the cards for review). Nvidia by themselves in fact do the value in their products signfiicantly generation to generation, and they do respond to the competition.

Now let us look at AMD. When AMD released 7xxx look at 7850/7870 launch prices in relation the 6950/6970. Or when AMD released the R9 285.

This is the problem with AMD's pricing in actual practice. Enthuasists get fixated on the very high end so they look at things like the Titan X however this price bracket is irrelevant to the majority of actual consumers.

Look at what the GTX 970 improved for people buy at its pricepoint and what the R9 285 "improved" at its price point.
 
Nvidia would never undercut AMD on price/perf since they don't have to.
So take whatever AMD does and add 20-30% price = 980 Ti.

GTX 680 $500
7970 $550

Nvidia will do whats necessary to sell cards.

The 900 series will see price drops due to the Hawaii refresh. The 380X will probably match the GTX 980.
The Ti will not be as expensive as you think because the 390 will be near the same performance for much cheaper.
 
And people are thinking it will have the horribls segmented memory like the 970 GTX

5.25+0.75 of slow ass speed :/

What a horrible move
 
And people are thinking it will have the horribls segmented memory like the 970 GTX

5.25+0.75 of slow ass speed :/

What a horrible move

The segments have to be in 0.5gb (512mb) chunks.

Configuration is not yet known.
 
The segments have to be in 0.5gb (512mb) chunks.

Configuration is not yet known.

Well 1 can pray. Otherwise it will be all over the media like the 970 GTX was.

I don't think Nvidia would make the same mistake twice.
 
And people are thinking it will have the horribls segmented memory like the 970 GTX

5.25+0.75 of slow ass speed :/

What a horrible move

That is because people don't understand why the 970 is the way it is and ignorantly apply the same to the 980 ti.

NV royally screwed up with hiding the issues with the 970 and will be haunted by it for at least a few gens.
 
That is because people don't understand why the 970 is the way it is and ignorantly apply the same to the 980 ti.

NV royally screwed up with hiding the issues with the 970 and will be haunted by it for at least a few gens.

Well why wouldn't you apply it the same say since it has disables 1 SMM?, but yet no info if the ROPS are affected.

Nvidia themselves stated this is how Maxwell works (when i came to the .5GB memory being slow).

Not saying it's true, but people who brought out the leaks are also wondering if its the same issue.

We will find out soon enough.
 
Well why wouldn't you apply it the same say since it has disables 1 SMM?, but yet no info if the ROPS are affected.

Exactly as you said. We don't know yet what the 980 TI has.

But considering the Ti has half the ram of the Titan, as opposed to the 970 matching the 980, the cut down should not affect the speed of the Ti's ram.

The 970 appears to be a unique case and I highly doubt NV will allow another mistake like it.
 
But considering the Ti has half the ram of the Titan, as opposed to the 970 matching the 980, the cut down should not affect the speed of the Ti's ram.

This comment right here shows you have no clue what the issue is with the 970 GTX and the memory.

http://techreport.com/news/27721/nvidia-admits-explains-geforce-gtx-970-memory-allocation-issue

"The GeForce GTX 970 is equipped with 4GB of dedicated graphics memory. However the 970 has a different configuration of SMs than the 980, and fewer crossbar resources to the memory system. To optimally manage memory traffic in this configuration, we segment graphics memory into a 3.5GB section and a 0.5GB section."

The 980ti is a cutdown version of the Titan X.
 
This comment right here shows you have no clue what the issue is with the 970 GTX and the memory.

"The GeForce GTX 970 is equipped with 4GB of dedicated graphics memory. However the 970 has a different configuration of SMs than the 980, and fewer crossbar resources to the memory system. To optimally manage memory traffic in this configuration, we segment graphics memory into a 3.5GB section and a 0.5GB section."

The 980ti is a cutdown version of the Titan X.

Before you start throwing stones maybe you should re-read my comment. I know exactly that issues found in the 970.

Maybe you either read my comment wrong or don't understand the 970 issue yourself. If you did you would understand the 970's issues most certainly won't apply to the 980ti due to the memory being halved. Less memory, less memory traffic for SM's & crossbar.

The 980Ti would need to be cut down so much to suffer the same issues as the 970 that NV would render it completely worthless. Only other way is if the Ti had the same amount of vram as the X while still being cut-down.
 
I think there is still some confusion about what happened.

Perhaps this diagram (modified in MS paint from Anandtech's GTX 970 article - http://www.anandtech.com/show/8935/...cting-the-specs-exploring-memory-allocation/2) will help -

Each block is identical to GM204 chips, GM200 just has 50% more of them.

Assuming a GTX 980ti with 2 cut SMMs and 8 ROPs cut (half a ROP partition) -

jtvuj7.jpg


So the GTX 980ti in this configuration would have 5.5GB+0.5GB memory and the equivalent of 352 bit access to 5.5GB (308GB/s) and 32 bit access to 0.5GB (28GB/s).

With no cut ROPs it would be 6GB with 384 bit access to 6GB (336GB/s).

Now in terms of what the real world impact would be you obviously lose performance in the memory subsystem simply due to the lower actual bandwidth and also the loss in L2 cache just like GTX 970 vs GTX 980. However because of the larger amounts of VRAM in this case quantity issues may be less likely. As while 3GB-4GB is currently straddling the borderline scenario in terms of actual VRAM usage (not utilization but actual meaningful usage) in many cases 5GB+ is still rather limited (if any practical scenarios) and also based upon current information you will likely have a limited amount of practical alternatives which carry more VRAM. Future usage is also another consideration but again the lack of alternatives crop up unlike with the GTX 970 you had full 4GB R9 290/290x as an alternative for example in that product segment.

Also again I have to state the configuration itself is not technically flawed as cutting down a chip design is a valid method to provide product segmentation. We as the public are not privy to detailed information but you have to assume Nvidia chose to configure the GTX 970 that way as they felt it was the best balance of cost:desirability (of the end product). Would I prefer a GTX 970 with full ROPs? Sure but I don't have cost figures for that nor how that would affect the end pricing. Would I have maybe preferred an alternative full ROP configuration but using lower clocked 6gbps GDDR5 chips instead? On paper performance wise it'd be comparable but without the memory segmentation but again I don't know what the actual cost difference would be.

Not disclosing the configuration was an issue however as it was potentially misleading at least to some consumers.
 
Even 6 gb is cutting it close especially in high resolutions with sli configs. Any lower and you might as well go for titan x or amd's 8gb versions if they have one. 8 gb is a safe amount of ram 5.5 gb or less probably won't cut it for certain games. Witcher 3 was surprisingly one of those games with very low vram usage.
 
It's not really a mistake if they do it for budget purposes.

It seems the AMD fans were the ones most affected by it. As they were screaming the loudest. Everyone else looked at the benchmarks and bought one instead of a 290x.

Just as most people will buy a 980ti over a 390x
 
It seems the AMD fans were the ones most affected by it. As they were screaming the loudest. Everyone else looked at the benchmarks and bought one instead of a 290x.

Just as most people will buy a 980ti over a 390x
GoldenTiger returned his dual 970's over it. Enough said.
Still a great card @ 3.5 GB. Could have been 3 GB at its price point and nobody would have complained.
 
Last edited:
Folks, this is the 980Ti nomenclature as in better than 980. It's not called Titan Lite for a reason. If the 980 doesn't have memory issue, why would the improved version have it? The fact that the performance is extremely close and even slightly superior to the Titan in 4K means they did something with the memory controller as 4K moves tons of data. That's the only logical conclusion based on the data presented so far.

Also, 6GB is a weird ass number. Normally it should be 8GB. So based on that fact, 2GB is already disabled.
 
It has 6 GB due to 384-bit bus.
It's a cut-down Titan X the same way the 970 is a cut-down 980.

You can't blame people for being skeptical. :)
 
Also, 6GB is a weird ass number. Normally it should be 8GB. So based on that fact, 2GB is already disabled.

Normally you say?? Where have you been hiding, 6GB has been already released several times in the past by Nvidia, Titan, Titan Black, and 6GB GTX 780. How on earth did you come with the conclusion 6GB isn't normal but 8GB is, when Nvidia has never released an 8GB nor ever hinted at one.
 
Normally you say?? Where have you been hiding, 6GB has been already released several times in the past by Nvidia, Titan, Titan Black, and 6GB GTX 780. How on earth did you come with the conclusion 6GB isn't normal but 8GB is, when Nvidia has never released an 8GB nor ever hinted at one.

My bad. Computers usually go by power of 2, so 6GB is unusual but I guess it's not the case with video cards. Never owned a Titan before so I guess I'm out of the loop. However, the 970 was the only card that had the weird memory configuration. AMD and Nvidia have always had cards with defective units disabled with full RAM. the 970 was an anomaly not a rule so my point stands, the 980TI should have full memory.

Funny that people are attacking the potential for 5.5GB while ignoring the 4GB of 390x. If the AMD geniuses can figure out memory compression, don't you think Nvidia can do the same? Last time I checked, compression is always slower than no compression. lossless compression is also not very efficient. The fact that the 980TI can overclock like the Titan X means that the 390x needs to be able to overlook that well also to even be considered a competitor. That may be why they need the water block because 4096 cores is going to run hot.
 
My bad. Computers usually go by power of 2, so 6GB is unusual but I guess it's not the case with video cards. Never owned a Titan before so I guess I'm out of the loop. However, the 970 was the only card that had the weird memory configuration. AMD and Nvidia have always had cards with defective units disabled with full RAM. the 970 was an anomaly not a rule so my point stands, the 980TI should have full memory.

Funny that people are attacking the potential for 5.5GB while ignoring the 4GB of 390x. If the AMD geniuses can figure out memory compression, don't you think Nvidia can do the same? Last time I checked, compression is always slower than no compression. lossless compression is also not very efficient. The fact that the 980TI can overclock like the Titan X means that the 390x needs to be able to overlook that well also to even be considered a competitor. That may be why they need the water block because 4096 cores is going to run hot.

So you have no idea what you are talking about and expect people to listen to you?
6GB on a 384bit bus is a power of 2...
GTX970 was NOT the only Nvidia GPU to have a weird memory config.
 
So you have no idea what you are talking about and expect people to listen to you?
6GB on a 384bit bus is a power of 2...
GTX970 was NOT the only Nvidia GPU to have a weird memory config.

You sound pretty defensive and butthhurt. I don' expect anyone to listen to me as I'm just another voice on the Internet with an opinion,

Name another video card with the 3.5+5GB memory configuration. Also 2 to the what power is 6?
 
Funny that people are attacking the potential for 5.5GB while ignoring the 4GB of 390x.

The difference is that AMD openly and honestly disclosed that the 390x will have 4GB of VRAM. People were hoping otherwise.

Nvidia not quite as honest about it. Then they gave conflicting messages about the workable VRAM after they were called out on it.
 
Last edited:
That is because people don't understand why the 970 is the way it is and ignorantly apply the same to the 980 ti.

NV royally screwed up with hiding the issues with the 970 and will be haunted by it for at least a few gens.

Hah! You must be joking, I guarantee they've already forgotten about it.
 
Just like Pearl Harbor and 9/11, people will eventually forget.
3 years from now, who will even care anymore? As long as it's not an ongoing issue. The lying, I mean.
 
You sound pretty defensive and butthhurt. I don' expect anyone to listen to me as I'm just another voice on the Internet with an opinion,

Name another video card with the 3.5+5GB memory configuration. Also 2 to the what power is 6?

The GTX 550ti, 660 and 660ti were sold in configurations where there was not full speed access to the entire VRAM.

Again, I have to restate this, the key difference is that in those cases they were properly labeled from the start. The issue isn't how the GTX 970 was configured but that it was not disclosed.
 
Folks, this is the 980Ti nomenclature as in better than 980. It's not called Titan Lite for a reason. If the 980 doesn't have memory issue, why would the improved version have it? The fact that the performance is extremely close and even slightly superior to the Titan in 4K means they did something with the memory controller as 4K moves tons of data. That's the only logical conclusion based on the data presented so far.

Also, 6GB is a weird ass number. Normally it should be 8GB. So based on that fact, 2GB is already disabled.

I hope you are joking?
 
benchmark does not always equate to real world performance. having the extra vram got rid of stutters i had when barely on the edge.
 
Last edited:
Back
Top