[Rumor] GTX 980 Ti Release Date Set For February 2015 Rumored As Specs Leaked Online!

[patrickjp93]

are you stupid? I said its approximate numbers... 

 

who said you would have a 6GB model? also, im pretty sure it will have a 512b bus, but i was playing devils advocate that there is an option to do it... as i said, it would not be elegant, but possible

Nah, 768-bit. Gotta do something to combat the 390X.

[LukaP]

Nah, 768-bit. Gotta do something to combat the 390X.

 

Im not even sure how serious you are at this point...

 

You do know Maxwell is really efficient with bandwith due to the enormous L2, and the compression algorythms... This is just trolling again. BEGONE, TROLL!

[Lord Baal]

actually there was a gtx 285 

Huh I run Sli BFG 260's at that time and didn't pay alot of attention to the rest of the series.  So there we go my comment could have been more accurate then I'd even realized lol

(BFG At the time were the only ones that would still cover overclocked cards under warranty here in Australia)

[BiG StroOnZ]

are you stupid? I said its approximate numbers... 

 

who said you would have a 6GB model? also, im pretty sure it will have a 512b bus, but i was playing devils advocate that there is an option to do it... as i said, it would not be elegant, but possible

 

If you read what I posted before you quoted me. The article in question claims the card will come in a 6GB and 8GB models. That's exactly what the article says.

 

I wouldn't call someone stupid when you are saying things like, "I  said before that i wasnt calculating it." You didn't say once you were using approximate numbers, let's just look at your posts:

 

LukaP, on 15 Nov 2014 - 05:18 AM, said:

Not entirely true. Yes this is correct if you are using 4/8Gb memory modules. But if they used higher capacity ones they could get that amount over 384bit bus. Remember, bus only affects the number of "channels" that memory is attached to. Not the size of that memory, and with 384 that is 3 channels of 128 bits, with 64bits for upstream and 64 for Downstream simultaneously. If you double the amount of chips, you lose simultaneous up/down transmission.

 

 

 

LukaP, on 15 Nov 2014 - 2:01 PM, said:

no, youre not getting it. you can use chips that are (cba to calculate it) lets say 768MB in size, and use 6 of them to get 8GB 

 

 

LukaP, on 15 Nov 2014 - 2:07 PM, said:

how? you take 6 chips that are 1.33GB in size (or 12 that are 0.665GB) and put them on the card. granted its not an elegant way, but its a way...

 

 

LukaP, on 15 Nov 2014 - 2:18 PM, said:

I said before that i wasnt calculating it. and now it does add up almost perfectly. its around that size (+- 1%)

 

 

LukaP, on 15 Nov 2014 - 2:35 PM, said:

are you stupid? I said its approximate numbers... 

 

who said you would have a 6GB model? also, im pretty sure it will have a 512b bus, but i was playing devils advocate that there is an option to do it... as i said, it would not be elegant, but possible

 

 

Not once did you claim these were only approximate numbers you were using (until after you made all your posts and claimed I was "stupid"). Moving along, the article doesn't claim it will be on a 512-bit bus. It claims it will be on a 384-bit bus, with 6GB and 8GB models. Which is the point I'm trying to make. If you are going to respond to someone, and start a counter argument to their point. Read thoroughly about what the person is trying to say before you respond to it:

 

 

 

As you can see, from the articles. Having a 384-bit bus and 8GB (8192MB) of memory makes no sense. What makes no sense even more is there being 8GB and 6GB models. Which is all I'm trying to say.

 

Hope that clears everything up.

 

http://www.game-debate.com/hardware/?gid=2490&graphics=GeForce%20GTX%20980%20Ti%208GB

 

http://www.kdramastars.com/articles/52474/20141113/gtx-980-ti-release-date.htm

[LukaP]

(cba to calculate it) was my exact line, in post #2.

 

and i admit, the 6GB variant would be stupid, if they were to release 8GB, and vice versa. why i think it will be 8GB is because titan was 6.

 

snip

[Lord Baal]

 

 

 

At this point I'd say the article is pure BS and we're debating over what's 100% pure fiction/speculation the simple fact they're calling it 980ti and clearly stating it differs from 980 in a lot of unlikely ways pretty much suggests that.  

[patrickjp93]

 

 

Im not even sure how serious you are at this point...

 

You do know Maxwell is really efficient with bandwith due to the enormous L2, and the compression algorythms... This is just trolling again. BEGONE, TROLL!

WOW! I'm not even allowed to make a joke... And it's not that bandwidth efficient. If it were higher resolutions wouldn't be seeing performance hits against the R9 280/x and 290/x. And we can't just blame release drivers anymore.

[LukaP]

At this point I'd say the article is purely made up and we're debating over what's 100% pure fiction/speculation the simple fact they're calling it 980ti and clearly stating it differs from 980 in a lot of unlikely ways pretty much suggest it's not based on fact at all to me. 

well its 2 really suspicious sites (worse than WCCF)... i was just playing devils advocate... i believe the GM210/200 will be a 512b bus, 8GB card

[Lord Baal]

well its 2 really suspicious sites (worse than WCCF)... i was just playing devils advocate... i believe the GM210/200 will be a 512b bus, 8GB card

If I were speculate(much like whom ever wrote the article) I'd have to agree with you.

[patrickjp93]

well its 2 really suspicious sites (worse than WCCF)... i was just playing devils advocate... i believe the GM210/200 will be a 512b bus, 8GB card

 

Alternatively, Nvidia could go with just 384 to leverage its "efficiency in bandwidth" (which is debatable given Maxwell's shortcomings in 1440p and 4K) and maintain cost competition against AMD.

[LukaP]

Alternatively, Nvidia could go with just 384 to leverage its "efficiency in bandwidth" (which is debatable given Maxwell's shortcomings in 1440p and 4K) and maintain cost competition against AMD.

and still offer only 6GB of ram? nope.

[patrickjp93]

and still offer only 6GB of ram? nope.

6GB is a lot more than most people need even for 4K. The only time this isn't true is when you have a crap ton of uncompressed textures or you more than modded skyrim.

[Lord Baal]

Alternatively, Nvidia could go with just 384 to leverage its "efficiency in bandwidth" (which is debatable given Maxwell's shortcomings in 1440p and 4K) and maintain cost competition against AMD.

Until AMD release their 300 series we have no idea of pricing a 390x may well cost a good deal more then you're expecting (I'm not stating that you're wrong just that we don't actually know)

If we're getting 8gb 980s I doubt a higher end card I'll call it Frank(because its about as likely as 980ti)will use 6 I personally have hit limitations with the 4gb on my 770 so I'm hoping for 8gb because while you're correct most people today wont need more then 6 most people also don't buy flagship cards.

[patrickjp93]

Until AMD release their 300 series we have no idea of pricing a 390x may well cost a good deal more then you're expecting (I'm not stating that you're wrong just that we don't actually know)

I'm fully expecting the AMD 300 series to cost more to pay for the HBM research, and if benchmarks come out in serious favor to AMD, prices will spike even higher just as they did in the data coin mining era.

[LukaP]

6GB is a lot more than most people need even for 4K. The only time this isn't true is when you have a crap ton of uncompressed textures or you more than modded skyrim.

i agree its way too much (im a gamedev) but NV have to stay competitive. and their lastgen top card had a 6GB buffer (and even a 780 had it optional) so from a marketing perspective, its stupid to not go for 8GB

[eshbop]

This is how it'll play out...

 

 

GM200 is 28nm. ~40% more shaders than the 980. ~500mm² die. Increased effective memory bandwidth by tweaking existing technolgy

390x is 20nm. ~40% more shaders than the 290x + GCN 1.2 improvements. ~?400mm². Increased effective memory bandwidth with new memory tech. Lower power consumption & heat output = maybe better OCing headroom than GCN 1.1?

 

 

Sooooo

 

980Ti - ~20% performance over 980

290x - ~30% performance over 290x

 

The 390x will ship for the cost of a 980 at the time. The 980Ti will ship for $700.

 

 

 

Just as about every other GPU launch seems to go

[BiG StroOnZ]

(cba to calculate it) was my exact line, in post #2.

 

and i admit, the 6GB variant would be stupid, if they were to release 8GB, and vice versa. why i think it will be 8GB is because titan was 6.

 

Yes very stupid, and nonsensical.

 

The only reason why the Titan came with 6GB of memory was because it was a cut-down version of the Quadro K6000 which had 12GB of memory. In order for the next Titan card to have 8GB of memory, it would mean that the next Quadro Card would have to have 16GB of memory. Of course this goes back to what bus-bit interface the GM200 or 210 is on. If it's on 384-bit or 512-bit will dictate how much memory they have available. But let's just say for arguments sake that the next Quadro Card will be a on 512-bit bus interface and will have 16GB of memory. If they make another Titan out of that, it means that the Titan will have 8GB of memory. Because it would make no sense to have a cheaper Titan card than a Quadro card both with 16GB of memory. Because then the Titan card would cut into the Quadro sales. Then this is where the difficult part comes in and your head starts to spin. If they make a "980 Ti" card out of this GM200/210 512-bit 16GB monster. How much memory will the 980 Ti have? The only thing I can think of is the Titan is a fully unlocked GM200/210 card with 8GB of memory and the 980 Ti is a cut down version with less Cuda Cores but also with 8GB of memory (unless they pull a fast one like with the 780 Ti and only put on 4GB of memory [where the 780 Ti only had 3GB of memory]). But they would have to remove a decent amount of Cuda Cores to make the performance differences seem relatively big enough to warrant even having a new Titan card in the first place. Since both of them now have 8GB of VRAM (unless of course again, they lower the VRAM amount on the Ti card and put 4GB of memory). Where before the 780 Ti only had 3GB of memory, while the Titan Black had 6GB of memory (and the Quadro K6000 had 12GB of memory). The only reason why someone would get a Titan Black over a 780 Ti for gaming would be because of the memory constraints on the 780 Ti with only 3GB of VRAM. Regardless of what people say the actual purpose of the Titan Black was (for compute and double precision is what they say) it was still marketed as a gaming card in many instances. Which means they will have to take a different approach for the next Titan card and Ti card. Unless of course they stick to their original pattern and the Quadro Card has 16GB of memory, the Titan Card has 8GB of memory, and the 980 Ti has 4GB of memory.

[LukaP]

snip

well the GM210 (lets call it that) is rumoured to have 22-24 SMMs. i think it will go: Quadro M20: 24SMMs (16GB), Titan 2: 22-23 SMMs (8GB), GTX 1080 20-21 SMMs (4GB, maybe a version with 8GB), 1070 18-19 SMMs (only 4GB)

[-TesseracT-]

Yes!! Bring on the GPU's! I really hope that performance gain is true.

[eLucid]

Still waiting for nvidia to release a next gen dual gpu. It's been far too long...