What would the possible performance be on a 980 Ti

M

Marcdaddy

2[H]4U
Joined
Feb 21, 2003
Messages
3,635
If they are adding more cuda cores and increasing speed what kind of gains could we expect over a 980 GTX ?
 
If performance isn't optimal, add at least 1 racing stripe on the card or some chrome trimming. Everyone knows that's worth at least 20HP.
 
KENNYB said:
If performance isn't optimal, add at least 1 racing stripe on the card or some chrome trimming. Everyone knows that's worth at least 20HP.
Click to expand...

I personally draw a big, red "R" on my card to maximize performance.
 
There are 4 GPC's with the GM204 so it wouldn't be too far off to think Nvidia adds another 2 GPC's (3072 Cores) most likely gimped due to yield issues at such a large die. I'm guessing the GTX 980 Ti would have 2304 to 2560 cores which should easily offer 15% improvement over the 980 upwards of 25-30% solid. Depends on if they're going to do a Maxwell refresh at all let alone 28nm vs 20nm. Wouldn't be surprised if they skip a refresh entirely like some earlier rumors a couple months ago and jump straight to Pascal. They do have some room for those two versions though.
 
new2019 said:
Its going to give your about 10% max boost from gtx 980 .
Click to expand...

10%??? 980 has 5 bill. transistors while Titan 7 billion. Adding 2 billion plus inherently better efficiency of Maxwell makes improvement minimum by 1/3 if not by half, no?
 
wirk said:
10%??? 980 has 5 bill. transistors while Titan 7 billion. Adding 2 billion plus inherently better efficiency of Maxwell makes improvement minimum by 1/3 if not by half, no?
Click to expand...

This is the logic I'd use. Gotta think OCs probably won't be as high due to the higher heat?
 
http://hardforum.com/showthread.php?t=1806694

Speculation from Feb 15th (a few days prior to maxwell 1.0 aka GTX 750 Ti launch) about what we'd see 28nm maxwell and 20nm maxwell in the future.

What was my speculation back then about what would become the GTX 970 & 980?

GoldenTiger on Feb 15th 2014 said:
3 GPC units on 28nm taking around 420-430mm2 with this imaginary chip that would have 15 SMM units. 15 SMM units times 128 per unit would mean 1920 Maxwell cores.

Pretend their scheduler is great and the performance scales well with core count and clocks, and that they kept the idea of triple everything there in this hypothetical, non-existent card that is an illustration only. So we'd have a 384 bit bus with 1920 Maxwell cores, probably 7ghz memory speed of GDDR5 like Kepler does at least, and a TDP that fits inside of 200 watts. Now let's say that you only get about 75% scaling from core count here, which is reasonable even though Kepler scales pretty linearly, but it's a new architecture with Maxwell, so let's make the safe assumption. So a GTX 660 performs 12% better than a GM107 with 640 cores. Triple the core count there with our rough napkin math again with everything else and you would have a card performing around the same as GK110 fully unlocked by that theory-crafting, at least, and it has better potential for higher clocks thanks to the lower power usage.

a card at 28nm with a 1085mhz GPU clock
Click to expand...

The things that turned out wrong? The bus width was determined to not be needed, and they used 2048 shaders instead of 1920 (6.6% difference). :) Oh, and I was a whopping 40mhz off on the 1126 MHz base clock if you want to count that (a whopping 3.7% margin). :p Pretty solid for being 7 months before launch for a random forum guy if I say so myself :).

Sound familiar? :D
 
My thoughts are:

Jan/Feb 2015 -> Titan 2....4000 cores...possible 20nm shrink

Q3/Q4 2015 -> Maxwell refresh Geforce XXX... 16nm shrink

H2 2016 ->16nm Pascal
 
Scenario 1 use the GM204 GPU on a pcb that has 1 8 pn 1 6pin and a higher TDP. GPU voltage can go just over 1.3 v on the GTX 970/980 cards. They'll likely utilize a higher voltage (maybe not that high) and use an amazing cooler to tame a 250 Watt TDP version of the GM204 chip. Clocks will probably be stock around 1.5Ghz with a boost around 1.7Ghz. The card will probably be 6GB of GDDR 5 on a 256bit bus using faster than 7ghz memory.

Scenario 2. a more powerful maxwell GPU (GM214) with more cuda cores, more power just as mentioned above, and more ram. Still 28nm.

In either case at least 30% faster than GTX 980. Just overclocking a 980 those results are being realized hitting a 165 watt TDP wall, imagine a fully 250Watt TDP monser version of this architecture. ;)
 
Love all the early responses, high octane drivers, and the "9000" one, etc. hahah. Pitty people started trying to answer the silly question with serious answers.
 
picket23 said:
Love all the early responses, high octane drivers, and the "9000" one, etc. hahah. Pitty people started trying to answer the silly question with serious answers.
Click to expand...

How Dareith Thou!! (Thunder Rumbles)

Are you mocking the Forum Famous Crystal Ball, it tells me all that is all

:D

2013-real-estate-crystal-ball.jpg
 
Nice pic, very appropriate, hehe =)

On the real though, the GTX-980-TI-Google-Edition is going to blow the 980ti out of the water so save you money. I heard its going to have a Nexus 5 strapped to it so you can just call the card and ask it to run faster when your fps drop.

Game.

Over.

AMD.
 
Lord_Exodia said:
Scenario 1 use the GM204 GPU on a pcb that has 1 8 pn 1 6pin and a higher TDP. GPU voltage can go just over 1.3 v on the GTX 970/980 cards. They'll likely utilize a higher voltage (maybe not that high) and use an amazing cooler to tame a 250 Watt TDP version of the GM204 chip. Clocks will probably be stock around 1.5Ghz with a boost around 1.7Ghz. The card will probably be 6GB of GDDR 5 on a 256bit bus using faster than 7ghz memory.

Scenario 2. a more powerful maxwell GPU (GM214) with more cuda cores, more power just as mentioned above, and more ram. Still 28nm.

In either case at least 30% faster than GTX 980. Just overclocking a 980 those results are being realized hitting a 165 watt TDP wall, imagine a fully 250Watt TDP monser version of this architecture. ;)
Click to expand...
Scenario 3: Stick two GM204's on one PCB and release a new dual-GPU card ;)
 
GoldenTiger said:
http://hardforum.com/showthread.php?t=1806694

Speculation from Feb 15th (a few days prior to maxwell 1.0 aka GTX 750 Ti launch) about what we'd see 28nm maxwell and 20nm maxwell in the future.

What was my speculation back then about what would become the GTX 970 & 980?



The things that turned out wrong? The bus width was determined to not be needed, and they used 2048 shaders instead of 1920 (6.6% difference). :) Oh, and I was a whopping 40mhz off on the 1126 MHz base clock if you want to count that (a whopping 3.7% margin). :p Pretty solid for being 7 months before launch for a random forum guy if I say so myself :).

Sound familiar? :D
Click to expand...

Nothing new coming from you.. I can remember your post before the launch of kepler and damn you guessed lot of things perfectly lot of hits there.. ;)
 
Unknown-One said:
Scenario 3: Stick two GM204's on one PCB and release a new dual-GPU card ;)
Click to expand...

Haven't thought of that one but look deeper into this. With the power efficiencies and the mere 165 watt TDP, why stick 2 and have a dual when you can have 3!!

Mwahahahahaaa!! :D

Hell if they made GTX 590 happen with that hardware who can say a tri GTX 980 card isn't possible.
 
EternityZX9 said:
My thoughts are:

Jan/Feb 2015 -> Titan 2....4000 cores...possible 20nm shrink

Q3/Q4 2015 -> Maxwell refresh Geforce XXX... 16nm shrink

H2 2016 ->16nm Pascal
Click to expand...


I don't see why they'd blow there load on 16nm so soon if they don't have to, both AMD and Nvidia. With the length of time its been taking TSMC to do each node it would be wise to go slow with the node changes as it may be another 4 years going from R&D to 100% production.

I see a gimped Titan (just like last time) using a big ass die on 28nm (lot of room to play with still) followed a few months later by the full 980 Ti on 20nm. The rest of the Maxwell refresh comes on 20nm later this year and perhaps even Pascal starts on 20nm and switches to 16nm in late 2015 or early 2016.

I don't think TSMC will have anything close to ready by the end of 2016 <16nm
 
BroHamBone said:
I use the stickers provided with the card for extra overclocking fan speed cooling......:cool:
+10 to bandwidth speed.
Click to expand...

Trimlock said:
Brb I'm going to pepper my pc with those stickers, didn't realise how awesome they were.
Click to expand...

I've been wasting these stickers on a bookshelf. Of course they won't make the bookshelf faster that is totally stupid. They should have been placed on my pc where they would have made a difference.
 
Liger88 said:
I don't see why they'd blow there load on 16nm so soon if they don't have to, both AMD and Nvidia. With the length of time its been taking TSMC to do each node it would be wise to go slow with the node changes as it may be another 4 years going from R&D to 100% production.

I see a gimped Titan (just like last time) using a big ass die on 28nm (lot of room to play with still) followed a few months later by the full 980 Ti on 20nm. The rest of the Maxwell refresh comes on 20nm later this year and perhaps even Pascal starts on 20nm and switches to 16nm in late 2015 or early 2016.

I don't think TSMC will have anything close to ready by the end of 2016 <16nm
Click to expand...

It's just speculation...but search google for nvidia+20nm or even 16nm and see what comes up. Where there is smoke there is usually a fire.

Example: http://www.tomshardware.com/news/tsmc-apple-nvidia-denver-finfet,27538.html

Again, just speculation, but it's got me thinking it could be a possibility.
 
EternityZX9 said:
It's just speculation...but search google for nvidia+20nm or even 16nm and see what comes up. Where there is smoke there is usually a fire.

Example: http://www.tomshardware.com/news/tsmc-apple-nvidia-denver-finfet,27538.html

Again, just speculation, but it's got me thinking it could be a possibility.
Click to expand...


Oh I don't doubt it or it being within the realm of possibility. It would at least give us a very rapid performance boost in the next 2 years, but we might be right back where we are after that with TSMC. I would hope both sides are considering other fabs because of it. The 20nm debacle really set both of them back from where they would have liked to be.
 
Will 980Ti use 512Bit bus, to provide more room for frame buffer, or will it be same 256 Bit as 980? If nVidia design 980Ti, I presume they will keep 4K gaming into consideration.
 
KENNYB said:
I've been wasting these stickers on a bookshelf. Of course they won't make the bookshelf faster that is totally stupid. They should have been placed on my pc where they would have made a difference.
Click to expand...

I bet that bookshelf could play crysis!
 
Just saw on overclock.net there is some talk about 610mm^2 die still on 28nm coming up as next high end card. That's some big ass die, supposedly 990/Titan 2... Probably Q1 15, rumor mill never stops...
 
Maybe they will announce something on sept 26t after whatever AMD is announcing on the 25th
 
GibsonEX said:
Just saw on overclock.net there is some talk about 610mm^2 die still on 28nm coming up as next high end card. That's some big ass die, supposedly 990/Titan 2... Probably Q1 15, rumor mill never stops...
Click to expand...

Gibbo (one behind most of the 980 leaks) claimed that the Titan II would be 561mm², same size as GK110, so we'll see.

If it's the same size as the Titan it'll be at least 40% faster, maybe even 50% if the clock speed is boosted.
 
dragonageinquisition said:
Gibbo (one behind most of the 980 leaks) claimed that the Titan II would be 561mm², same size as GK110, so we'll see.
If it's the same size as the Titan it'll be at least 40% faster, maybe even 50% if the clock speed is boosted.
Click to expand...

OK, then how about additional boost when shifting to the 20nm mode? Another 50% giving in total 100% over the current Titan? This happening in March ´15 would be nothing short of amazing.
 
dragonageinquisition said:
Gibbo (one behind most of the 980 leaks) claimed that the Titan II would be 561mm², same size as GK110, so we'll see.

If it's the same size as the Titan it'll be at least 40% faster, maybe even 50% if the clock speed is boosted.
Click to expand...

That sounds more reasonable, and Gibbo knows what he's talking about, i'd take his word over OBR or many others. Still, even at that size it would be a very good chip with monster performance. That's a good (rumor) news ;)
 
Lord_Exodia said:
Hey Marcdaddy, after pooping on launch day, what did you end up doing? Buy a card, two, or 3 or wait it out?
Click to expand...

Waiting on my 3 Evga Sc 980s to he here but I'm thinking of doing a full system upgrade since I'm running that 2600k at 4.8ghz with only PCI express 2.0. It's hard to leave the 2600k behind just to upgrade to PCI express when 4.0 might be coming out next fall with the release of Skylake.
 
You must log in or register to reply here.
Back
Top