• 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.

Anything on the horizon?

T

Trackr

[H]ard|Gawd
Joined
Feb 10, 2011
Messages
1,786
So, nVidia is set to release 20nm Pascal in 2016.

Until then, nothing? Meaning, no releases in 2015 at all?
 
they'll probably release 8gb variants of their existing "high end" cards, since that's super easy.
 
If I had to guess, I'd say we'd see a Titan (gm200) refresh in the next couple of months. Well also the 960 that should be out soon.
 
I wonder if they will make a card to compete with the AMD R9 295x2 soon? I like nVidia, but man even in SLI the nVidia's can not touch this card on high res monitors...Kinda sad. What are they waiting for?
 
Trackr said:
No, I said 20nm Pascal. 20nm Maxwell.. who knows?



So.. X390 would be a better buy than GTX 980?
Click to expand...

No idea. It was a leaked benchmark, no idea on price, not even sure if the 3dmark score is legit (although they did leak the 970/980 benchmarks and were spot on).

We have no idea if its the 380x or 390x, no idea if its on 28nm or 20nm.

So I mean yes it "could be" a better buy. But if the 3dmark is an indication on performance, it looks to be around 40% faster then a 980 GTX?
 
DASHlT said:
No idea. It was a leaked benchmark, no idea on price, not even sure if the 3dmark score is legit (although they did leak the 970/980 benchmarks and were spot on).

We have no idea if its the 380x or 390x, no idea if its on 28nm or 20nm.

So I mean yes it "could be" a better buy. But if the 3dmark is an indication on performance, it looks to be around 40% faster then a 980 GTX?
Click to expand...

Wow, really? It seemed to be <40% more powerful than a 290x and there are threads of people comparing 290x Crossfire with a single GTX 980..

My main reason for asking is because I've been eyeing the GTX 970 for a while now.. long enough to start wondering if I should just wait since my GTX 660 SLi is till holding its own.
 
Trackr said:
Wow, really? It seemed to be <40% more powerful than a 290x and there are threads of people comparing 290x Crossfire with a single GTX 980..

My main reason for asking is because I've been eyeing the GTX 970 for a while now.. long enough to start wondering if I should just wait since my GTX 660 SLi is till holding its own.
Click to expand...

If 2 980's GTX in SLI is on PAR with a 295x2 at 4k resolution. How can anyone make the claim that 2 290x's in crossfire is as fast as 980 1 GTX? That is called the Nvidia Shill forum marketing.

90% of the time, anyone will feed people a bullshit agenda and pick and choose which graph to show you they are right. But if you take the time to read reviews from multiple websites you can get a sense of what a card is really like.

Thats why when [H] did the 4k review, everyone was shocked that 2 980's in SLI was as fast as 295x2 at 4k.
 
dragonageinquisition said:
In average framerates (aka measured framerate) yes, but in terms of average frame latency aka "observed framerate" its not really close. 295x2 is leagues above 980 SLI. This is mostly due to the fact that the XDMA connections on the 295X2 has like 8 times the bandwidth of Nvidia's SLI bridge.
Click to expand...

The miniscule differences in frametimes don't actually affect the visible framerate as they aren't enough of a deviation (i.e. statistically insignificant and not outside of normal variance between frames anyway). The 295X2 is about on-par with GTX 970 SLI, until you start taking oc'ing into the mix and the 970's surpass it due to having very large headroom on clocks. Please don't spread misinformation that I'd be very surprised if you didn't already know is false.

Trackr said:
Wow, really? It seemed to be <40% more powerful than a 290x and there are threads of people comparing 290x Crossfire with a single GTX 980..

My main reason for asking is because I've been eyeing the GTX 970 for a while now.. long enough to start wondering if I should just wait since my GTX 660 SLi is till holding its own.
Click to expand...

Once oc'd, and the Maxwell 2.0 cards are known to do so extremely well, the GTX 970/980 both end up very, very fast... however, they definitely aren't keeping up with 290X crossfire at all :p. As far as when to upgrade? Once you actually need to. There's little reason to go buy something just because it's new and tempting, if you are happy with the speed you have now.

To answer your original question in the OP, GM200-based card(s) are coming around soon (they've reportedly been ready for a couple of months but weren't released yet, being held back to steal the thunder from AMD's upcoming launches of the 380/390/whatever they call it by releasing shortly after, since nvidia's already currently holding the lead). Rumors pin it as basically what the gtx 780 was to the gtx 680, essentially a 50% boost on most of the key specs... but we'll see what actually launches. 20nm isn't showing up on any video cards now, and probably never will since 16nm is slated for use by amd/nvidia starting in early 2016 onward and the current 20nm process is an SOC.
 
Last edited:
Damn, it's amazing how it still takes 3 years to get to the next manufacturing process even with steady competition.

I miss the old days of the X1900XT. Intel vs. AMD and nVidia vs. ATi.
 
I thought they were skipping 20nm and going 16nm FinFET with Pascal?
 
Armenius said:
I thought they were skipping 20nm and going 16nm FinFET with Pascal?
Click to expand...

The rumor is now 16nm FinFET is going to be delayed longer.....lets hope thats not true. Would LOVE to see a 16nm pascal card!!

Hell even a 20nm Maxwell would be a monster
 
DASHlT said:
The rumor is now 16nm FinFET is going to be delayed longer.....lets hope thats not true. Would LOVE to see a 16nm pascal card!!

Hell even a 20nm Maxwell would be a monster
Click to expand...

Haven't heard that anywhere. The only recent article at all was an ancient bit of news that was already widely known being reposted. That "news" is 16nm being available at the tail end of 2015 or sometime in the first months of 2016, which would come in right alongside Pascal (current rumors have and previously said, that it will be based on 16nm). In short, nothing new, just rehashing of existing info.
 
:(

Was really hoping that the 2Q/3Q '15 refreshes were going to be 20nm or 16nm. I wish TSCM/GloFo/nVidia/AMD weren't being so secretive about the real state of 20nm and smaller processes. I understand that they don't want to divulge info so that people continue to buy current products to keep the coffers filled, but this is getting ridiculous, imo.

Alas, I'm not moving off of Kepler unit a die shrink happens; be it Maxwell 3.0, Pascal, Pirate Islands/PI refresh, or whatever the hell AMD calls the successor to PI.
 
DejaWiz said:
Alas, I'm not moving off of Kepler unit a die shrink happens; be it Maxwell 3.0, Pascal, Pirate Islands/PI refresh, or whatever the hell AMD calls the successor to PI.
Click to expand...

Why wait until a die shrink happens? Not calling you out - just honestly curious.
 
HFBallad said:
Out of curiosity, when was the last die shrink?
Click to expand...

40nm-28nm.
GPUs first used 28nm at the beginning of 2012, Tahiti vs GM104.
The highend versions came in 2013, Hawaii vs GK110.
And it looks like we are getting one more round out of 28nm, Fiji/Bermuda vs GM200, before they move to FinFet.
 
hdgamer said:
The new node shrink hype is the new more vram hype. It just doesn't end. LOL
Click to expand...
It's worse. At least with VRAM, there was a somewhat tangible benefit to the end user. When the card delivers a proportional upgrade in performance, what node it was made on is almost as abstract as the numbering scheme they decided to go with, and it's not like people argue about that.

Wait, oh shit.
 
noobthegreat said:
It's worse. At least with VRAM, there was a somewhat tangible benefit to the end user. When the card delivers a proportional upgrade in performance, what node it was made on is almost as abstract as the numbering scheme they decided to go with, and it's not like people argue about that.

Wait, oh shit.
Click to expand...

What? You are suggesting there is more benefit to gobs of VRAM than the production process? Please, do some research.
 
Trackr said:
Until then, nothing? Meaning, no releases in 2015 at all?
Click to expand...

I just got back from 2017 and I can confirm there are no new video cards that will be released in the next 12 months. In future times, PC gamers will refer to 2015 as "the year our wallets relaxed."

Also, 2016 will be all low-end parts, but the GeForce 1030 will be faster than a 980 and sell for $75, so no one will really mind.

Oh, and you guys are going to be eating your hats when you see Via's new processors.
 
Don't be smartasses. :p

I'm waiting for the smaller process cards because it will undoubtedly be the key to getting a single GPU that at least matches the performance of my 780 SLI while drawing less lower than a 980, maybe even a 970. Which translates to a LOT less heat than even 1 of my 780s.

An increase in VRAM is inevitable with almost every new gen, so I didn't think it even needed to be stated. But since it's a topic of focus, my next GPU will need at least 8GB to make me a buyer.
 
dragonageinquisition said:
In average framerates (aka measured framerate) yes, but in terms of average frame latency aka "observed framerate" its not really close. 295x2 is leagues above 980 SLI. This is mostly due to the fact that the XDMA connections on the 295X2 has like 8 times the bandwidth of Nvidia's SLI bridge.
Click to expand...

Exactly how much, fast, or high is a "league" in this context??
 
LordEC911 said:
What? You are suggesting there is more benefit to gobs of VRAM than the production process? Please, do some research.
Click to expand...
No, you do some less research.

I'm saying what I say more as a reference to the general trend that we here seem to argue about more and more minute things just because we're now more privy to this kind of information and somehow more savvy than the people who designed them.

There was a time when consumers didn't have access to any such details beyond pricing and general performance data. I won't go as far as to say we were "happier" then, but we certainly have gone strides ahead in our presumptuousness.

And as for the VRAM vs. production process:

Don't look at my point so academically. Of course production process is important. So is who the company CEO is, for that matter. They're important details that affect how that GPU will be made... at some point down the line. My point was that at least VRAM is something you can monitor and is directly observable to the end user (not that VRAM will help if GPU power can't match it) If, however, the GPU looks like a duck and moves like a duck, but you crack open that IHS and you see inscribed 28nm instead of 20nm? And you think "Aha! Foul play!"

Hence, "abstract." We're getting further and further from what these cards actually do in our hands and more into the realm of whether it's GMO- and cruelty-free. Or, in the case of numbering schemes, quite actually nothing but the psychological with no basis in practicality.

When these cards actually go on sale, however, I'm betting desire overcomes reason and most of us still buy the "inferior" cards. :rolleyes:
 
LordEC911 said:
Because that is typically when we get the biggest performance jumps over the last generation of cards.
Click to expand...

I understand that. I guess I find it strange that people use something arbitrary like the die size to base their next move on. What if the 390X, 980 Ti, or Titan 2 just blows everything away with an insane price/performance ratio? You're still going to stand pat? Maybe I'm too much of an upgrade nut. *shrug*

DejaWiz said:
Don't be smartasses. :p

I'm waiting for the smaller process cards because it will undoubtedly be the key to getting a single GPU that at least matches the performance of my 780 SLI while drawing less lower than a 980, maybe even a 970. Which translates to a LOT less heat than even 1 of my 780s.

An increase in VRAM is inevitable with almost every new gen, so I didn't think it even needed to be stated. But since it's a topic of focus, my next GPU will need at least 8GB to make me a buyer.
Click to expand...

That makes sense. I am still a bit confused how it must be < 28nm or bust. I get it - you're on 28nm now...but if you need the juice and NVIDIA or AMD bring it - why not? The die shrink bringing xyz% better performance/thermals/etc. is all unknown. It can be safely assumed - but it is not necessarily a guarantee. There could be other unforeseen problems, as well (which is probably why we have not seen 20nm or lower yet).
 
sk3tch said:
I understand that. I guess I find it strange that people use something arbitrary like the die size to base their next move on. What if the 390X, 980 Ti, or Titan 2 just blows everything away with an insane price/performance ratio? You're still going to stand pat? Maybe I'm too much of an upgrade nut. *shrug*
Click to expand...
Hear, hear.

The reality is that die shrinks are not as frequent and not bringing the same exponential jumps in performance as they used to be. Meanwhile, we've had Nvidia recently come out with an impressive increase in efficiency (relatively speaking) on the very same node by simply shifting their architecture. The industry will probably continue to adapt like this in some way.

I mean, I'd love it if the companies stretched their dollar in order to bring us what we're accustomed to, but if the clincher comes down to numbers used mostly only internally? For me, it's a secondary concern (though I will probably also be the first to jump on the flamewagon against Nvidia/AMD when they release a turd and it's because of the node).
 
Low power consumption and low heat output are important criteria for me along with price : performance. The Titan2, 390X, whatever the hell else could come out and blow everything else away. But if it's going to be priced out of my comfort range ($500 tops if the performance is stellar) or spin my electrical meter like a fan, then hell no.
 
noobthegreat said:
No, you do some less research.
*snip*
Click to expand...
While I get your point, there are some of us that are actually involved in the industry...


noobthegreat said:
Hear, hear.

The reality is that die shrinks are not as frequent and not bringing the same exponential jumps in performance as they used to be. Meanwhile, we've had Nvidia recently come out with an impressive increase in efficiency (relatively speaking) on the very same node by simply shifting their architecture. The industry will probably continue to adapt like this in some way.
Click to expand...

Yes, die shrinks aren't coming as fast as they used to be.
Those "exponential" jumps in performance were due to the fact that past GPUs kept jumping up in TDP. We are now at a power wall, ~300w, and low-hanging fruit like power gating smaller and smaller areas of the GPU only get you so far, i.e. Maxwell.

You can't change marchitectures every year or even every 4 years, so yes, they will still be relying on newer nodes to get some extra juice.
 
You must log in or register to reply here.
Back
Top