-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Nvidia announces 2014 CES Press Conference and Live Webcast - MAXWELL?
- Thread starter artist
- Start date
Stallion Free
Cock Encumbered
No 20nm in 2014 would be a huge disappointment. It's hard enough waiting for Q4.
Expanded manufacturer support for G-sync before it even launches would make me slightly forgive them.
Expanded manufacturer support for G-sync before it even launches would make me slightly forgive them.
TheThunder
Member
No 20nm cards in 2014 would suck.
I'm sticking to my 670 until the "real" Maxwell arrives, hopefully this year.
I'm sticking to my 670 until the "real" Maxwell arrives, hopefully this year.
Terbinator
Member
Unlikely we'll get anything significant due to the essence of CES. Best highlight we can expect is a raft of high-quality G-sync monitors announced I would wager.
So they'll hold a livestream event twice for G-Sync? Sounds more than that, may not be Maxwell but just G-Sync and Tegra doesnt seem to warrant a livestream event.
If we get a 1440p non-TN G-sync monitor that would actually be a bigger deal for me than Maxwell. But I don't think that will happen so soon.
LiquidMetal14
hide your water-based mammals
Need to start seeing more on the GSYNC front since having only 1 available GSYNC compatible piece is unacceptable. I already have a 27"IPS and want to get the ASUS 27"144hz monitor but unlike its 24" counterpart it's mysteriously not GSYNC compatible. Stuff like this is giving me a headache.
I don't want to mix a 24 and 27 together as I'm going for dual monitor gaming here.
I don't want to mix a 24 and 27 together as I'm going for dual monitor gaming here.
After reading so many articles and posts about how 20nm (the type they use for GPUs, since there different types of processes) won't be ready for H1 2014, or 2014 at all, I'd be surprised to see GM104/204 or even GM100 make it as early as nVidia hoped for. I guess we'll see how things pan out.
From what I know, having coil whine is just a case of bad luck. It can happen to any brand. Even then it can depend on the actual card's manufacturer and whatnot.
From what I know, having coil whine is just a case of bad luck. It can happen to any brand. Even then it can depend on the actual card's manufacturer and whatnot.
demosthenes
Member
Hope to see lots on Gsync.
A Feb launch of a 28nm GPU is kind of strange on first inspection, but it makes complete sense if you've been following the state of 20nm GPU fabrication. It's going to be interesting to see where a GTX860 would sit for price/performance.
I'll be disappointed if the integrated ARM CPU is not ready, as that was a really interesting idea to me.
I'll be disappointed if the integrated ARM CPU is not ready, as that was a really interesting idea to me.
If they do go that route, my guess would be that GM104's size would end up between GK104 and GK110. They can toss out the compute stuff like they did with GK104 so they can achieve bigger gaming performance than GK110 without it being huge and soaking up a lot of power.
Well, the only difference the framerate in games made was the frequency of the noise, if that's what you mean. It went from this in games to the more high pitched sound in menues. You could even hear it when i was watching a movie with 2 monitors connected, although it was quiet. As sound some gpu load was applied the thing started making noise and i'm glad the shop took it back because it would have drove me insane
Wouldn't that cripple performance when a game is using advanced Physx stuff?
Future PhaZe
Member
Bad optimization aside, I doubt there's much coming this year that will make my 680's sweat aside from Witcher 3. G-Sync should make a 40 fps experience feel really solid for that, so I'll deal until 20nm's are out, although I'm kind of just waiting for 780tis to drop in price so I go SLI on those which should destroy everything until probably mid-gen at worst.
McHuj
Member
It shouldn't. A lot of the compute in something like GK110 is targeted for scientific computing that requires 64-bit double floats. You don't need that for a game engine (at least, I hope their only doing it in single precision floats).
Terbinator
Member
See below your post. G-Sync is a massive thing and fits in more with CES than any Maxwell type stuff does.
I see. I asked because I heard that the main cause of TressFX bad performance on latest nVidia's card is the lack of proper compute support.
I know TressFX isn't a Physx effect, but both require lots of computing power, right?
chaosblade
Unconfirmed Member
28nm would be unsurprising but still disappointing. No regrets buying a 760.
Interested in G-Sync though. Hopefully we see some monitors that aren't like $400, but I'm not counting on it.
Interested in G-Sync though. Hopefully we see some monitors that aren't like $400, but I'm not counting on it.
TC McQueen
Member
Looks like I'm sticking with my 560 Ti 448 core until the 20mm stuff comes out or I find a good deal on a used/new card.
Terbinator
Member
It just happens to be that AMD cards are faster in certain compute scenarios than not. This is the reason why they absolutely mince Nvidia at mining (and quite a lot of other stuff, really).
Edit: http://www.extremetech.com/computing/153467-amd-destroys-nvidia-bitcoin-mining
The bottom half of this article explains why Nvidia lack in certain computing scenarios.
From my understanding, TressFX uses DirectCompute, which is DirectX"s implementiation of compute (pretty self-explanatory and all). Somehow it ends up running faster on AMD/GCN cards than on nVidia's for reasons I don't remember. It depends on the type of workload I guess.
I've read that shrinking the process does not necessarily reduce power consumption overall. There is power leakage and power usage, with leakage increasing as the process is shrunk. There are technologies designed to reduce the issue, but obviously the problems are complex. I expect maxwell to include some good design improvements over kepler to even make a 28nm maxwell worthwhile designing and manufacturing.
Great article here on the subject: http://www.extremetech.com/computing/123529-nvidia-deeply-unhappy-with-tsmc-claims-22nm-essentially-worthless
McHuj
Member
Yes, that's true to an extent. Leakage is going up and up with shrinking nodes. It's really problematic in the mobile space, where it can dominate power consumption.
That's why Intel moved to the FinFet transistors, they help a lot with leakage. The 16nm node from TSMC is basically 20nm node with FinFet, so the power should be better.
20nm should provide a power or performance boost. http://www.tsmc.com/english/dedicatedFoundry/technology/20nm.htm
TSMC is advertising a 25% power reduction or a 30% performance increase. Granted that's marketing, but it at least it will provide some gains. For desktop GPU's this should ok, for laptop components it maybe an issue.
My expectation is that we'll probably get boards that run at the same power consumption levels but with better performance (maybe ~1.25 GHz clocks).
Dictator93
Member
No maxwell so far. Oh man.
I couldn't find anything online, so I assume Maxwell wasn't even mentioned at the event? That's massively disappointing. The Tesla K1 is pretty impressive-looking for a mobile chip I guess, but meh. With the GTX 880M being a rebadge, Nvidia still releasing Kepler parts (Tegra K1 and GTX 790), and AMD recently releasing their Volcanic Islands parts I'm upset we might not be seeing any 20 nm parts soon. :/