![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
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.
Info on Maxwell?
- Thread starter Emmanon
- Start date
I believe it's still Q3 so will probably be a bit before we get solid info but the 750/750ti are out soon and those will be "Maxwell". I guess it sorta compares to the 7790 launch that was the first of AMD's current GCN 2.0 cards to launch. But aside from grasping at those 750 ti straws and then filling in the gaps with reckless speculation not much info until NV decides to give us something. 
phinix
Gawd
- Joined
- Dec 13, 2005
- Messages
- 914
Quick question: is there a point buying 780 in a month or two or better wait for new Maxwell series in Q3?
Right now I have 670...
Right now I have 670...
nightanole
[H]ard|Gawd
- Joined
- Feb 16, 2003
- Messages
- 2,032
You are assuming that they come out with affordable cards? They are going to come with the 750, and then some $650+ card. Then in six months they will start with all the value cards. Why would they release a $450 card on the first new gpu since 2010, first?
Generally speaking, AMD and nvidia do not share information on new next gen GPUs until 1 month to a couple of weeks prior. Reason being is that giving the information out too early will cannibalize sales - We got the first information on Tahiti in Dec 2011 (released in Jan 2012), the first info on Kepler in Feb 2012 with release in March 2012. I'd expect based on this trend that nvidia will basically stay mum until release is imminent, this is how things have always been done from what i've seen. Same deal with the 290. Roughly one month prior is when AMD shared their information.
Now the interesting part of Maxwell is that they're starting with the mobile and low end parts first, but I imagine big Maxwell will be on 20nm. The 750ti is coming soon according to rumors, but big Maxwell (I assume) is still a ways away. So I wouldn't expect information on "big" 20nm Maxwell to surface anytime soon, although I could be wrong.
Now the interesting part of Maxwell is that they're starting with the mobile and low end parts first, but I imagine big Maxwell will be on 20nm. The 750ti is coming soon according to rumors, but big Maxwell (I assume) is still a ways away. So I wouldn't expect information on "big" 20nm Maxwell to surface anytime soon, although I could be wrong.
New API would be interesting, hopefully they do away with all the HDMI stuff for gamer cards and stick with DP.. keep the hdmi for htpc system cards, but its hard when most consumers can only understand hdmi and never even heard of DP..
Wiseguy2001
2[H]4U
- Joined
- Nov 28, 2001
- Messages
- 3,470
I don't think Maxwell will really get interesting until it has a die shrink, they've done pretty much all they can squeezing out performance on 28nm.
I'd rather have HDMI over DVI, DVI was only a stopgap from VGA. Personally I'm using all three interfaces in this PC and I don't get the hatred for HDMI. HDMI 2.0 is going to be VERY useful! (UHD @60fps)
- Joined
- Aug 5, 2013
- Messages
- 13,451
That is adorable.
If you are watching and listening close enough, we usually have a pretty good idea on what is coming ~6months before hand.
good point, I forgot, I dont even use my DVI much anymore.. VGA and DVI should be rid of lol.. it had its time, but like Flash, it should be forced out for the newer technology..
It really is and makes it hard to take seriously. I wish we were back in those old days where the biggest bad ass GPU's came in at those sizes though. Don't think we'll ever see those times again.
Explains why initial benchmarks showed it sucking pretty hardcore. That memory bandwidth is very eye brow raising.
Six months? Well I don't really agree with that, but non-credible rumors do start to hit the streets around that time frame. Usually two to two and a half months or so beforehand is when more credible rumors hit the streets, mainly from websites that have dudes close to the supply chain in China (since that is where all cards are produced). Websites like chiphell, expreview, etc, will have rumors 2-2.5 months beforehand with i'd say 30-40% of those rumors being reasonably close. I remember the Kepler rumors began hitting hard around late January of '12, with the Feb rumors being pretty close to spot on. Same for Tahiti - two to two and a half months is when the rumors became somewhat reality based.
Six months out? I don't really agree with that, only an insider (IMO) would have info that early. And those insiders usually don't dare peep a word since they're under strict NDA.....I read the rumors around here and OCN and most of the more reliable rumors are in the 2-3 month prior to launch time frame, and those rumors are usually from non NDA guys from China. Case in point, most of the Tahiti rumors from the 2 month prior to launch timeframe were good, same for Kepler. I don't recall any reliable rumors from 6 months out for either, though... Just IMO. As far as "big" Maxwell, I wouldn't expect it anytime imminent since I assume NV wants to put the big parts on 20nm.
Sodapopjones
[H]ard|Gawd
- Joined
- Aug 22, 2012
- Messages
- 1,932
How about no, I like cheap Korean 1440p monitors with low input lag, thanks!
I'm sure they engineered these cards with that bus for a reason.....
Any AMD cards coming out most certainly will be 20nm (outside the R5/7/9 200 Series), but most likely Nvidia will be the first to bring them to market unless AMD does some minor refresh on those hot R9 290 suckers just to test the waters.
Depending on when Nvidia fully drops Maxwell, I would expect a late Q4 '14 refresh or early Q1 '15 at the latest. Nvidia with the Spring and AMD with the Fall are the usual time frames they like. Of course 20nm is kinda screwing that up and right now it's 50/50 whether they will meet those quarters.
Again, we had a good idea of what Tahiti was back in July 2011. I had even more descriptive "speculation" on some other forums, unfortunately one died and the other has a crappy search function.
We knew what GK110 was long before it was released. The same with GF100.
69x0 we had a good idea of the lineup but vliw4/vliw5 threw everyone for a loop.
RV740 was known 5 months before release.
My G90/G91 "speculation" turned into GT200, which I believe was from about 12months before release.
You can only hide so much once an ASIC is taped out and entering mass production. Usually the best defense is a good offense, ie spreading misinformation/disinformation, as seen with RV770.
AMD is much more likely to get 20nm wafers before Q3... Who knows if they will get enough to actually launch anything though.
limitedaccess
Supreme [H]ardness
- Joined
- May 10, 2010
- Messages
- 7,592
This is the very low end for discrete GPUs.
The highest end card GK107 was used in was the GTX 650, which is what GM107 is really replacing. GTX 650 Ti Boost and GTX 660 were using the larger GK106, while GTX 650Ti was a very cut down version of this core (with 128 bit bus).
At the moment it looks like the following comparison -
GK107: 118mm^2 die with 1.3b transistors. Fully enabled it was 384/32/16/128 (cores/tu/rop/bus).
GM107: 156mm^2 die with ? transistors. Assuming fully enabled (750ti) would be 960/80/16/128.
Speculatively it looks like it could be similar to Fermi->Kepler where the cores are not equal. So big Maxwell could very well have those 6k core rumors but don't expect it to actually have that type of performance increase over GK110.
Also looks like DX11.2 and SM 5.0 support now.
Uhm, all that really matters is the performance and the cards they're replacing? The 750 is faster than the prior 650ti, and the 750ti is even faster still, according to WCCFtech. And the replacements are in the 100-125$ price bracket (according to rumors) so as long as the performance is faster than the prior parts they're replacing, and still being similar in price (100-125$ per wccftech), that's all that really matters. Also, these parts are rumored to be more efficient than even Kepler (which IMO is pretty impressive if true). In any case, I don't think we'll see the ambitious designs with large core counts until big Maxwell on 20nm. So with that being the case it's rather worthless to talk about bus sizes and core counts at this point.....
Just my opinion. For a 100$ card, I think only the performance / efficiency matters - even though a 100$ card is not something I would consider except for a HTPC. Totally not getting the talk about bus sizes. I'd do a double take if a GTX 880 had a 192 bit bus, but a 128 bit bus for a 100$ GTX 750 card? It supposedly outperforms the 650ti so.... Meh. Whatever.
Just my opinion. For a 100$ card, I think only the performance / efficiency matters - even though a 100$ card is not something I would consider except for a HTPC. Totally not getting the talk about bus sizes. I'd do a double take if a GTX 880 had a 192 bit bus, but a 128 bit bus for a 100$ GTX 750 card? It supposedly outperforms the 650ti so.... Meh. Whatever.
Last edited:
limitedaccess
Supreme [H]ardness
- Joined
- May 10, 2010
- Messages
- 7,592
Well bus sizes will increase with the higher end cards. The point was a 128 bit isn't really surprising at the price points, product lines and the actual chip used for these cards, the AMDs use 128 bit as well. Once you go even lower you have the cards using DDR3 and not GDDDR5 memory.
Let's face it if you're buying these cards high resolutions and AA (where memory bandwidth matters the most) is probably far (if at all) on your mind for gaming.
If you go by the current rumor/speculation (mainly for fun at this point, no telling the accuracy) -
GM100 - 512 bit, higher than GK110.
GM 104 - 320 bit, higher than GK104. Although kind of annoying/disappointing (well for my plans anyways) if true this mean's VRAM will be 2.5gb (or 3gb with asymm. memory for the last 500mb).
GM 106 - 196 bit, same as GK106
Was kind of hoping for another GTX 680(GK104) but it seems like it's more likely you would have to upgrade from big to big (GK110->GM100) as opposed to before where you could from the GTX 570/580 to the 670/680 (GF110->GK104).
Let's face it if you're buying these cards high resolutions and AA (where memory bandwidth matters the most) is probably far (if at all) on your mind for gaming.
If you go by the current rumor/speculation (mainly for fun at this point, no telling the accuracy) -
GM100 - 512 bit, higher than GK110.
GM 104 - 320 bit, higher than GK104. Although kind of annoying/disappointing (well for my plans anyways) if true this mean's VRAM will be 2.5gb (or 3gb with asymm. memory for the last 500mb).
GM 106 - 196 bit, same as GK106
Was kind of hoping for another GTX 680(GK104) but it seems like it's more likely you would have to upgrade from big to big (GK110->GM100) as opposed to before where you could from the GTX 570/580 to the 670/680 (GF110->GK104).
Semantics
2[H]4U
- Joined
- May 18, 2010
- Messages
- 2,811
660 requires 6 pin 750 ti doesn't. efficiency means more performance per the watt, watt and the heat that comes with it limits gpu's at the high end, so their top end is probably what will change.
They're 100-125$ cards....it's faster and more efficient than the cards they're replacing. That being the 650 and 650ti, while the 750 and 750ti are far more efficient. That's all that matters, nobody in that market (which isn't me) really cares about anything other than that. At least, people buying 100$ cards (HTPC users) as far as I can tell don't obsess about specs like that, although I could be wrong.
Now when you're talking big Maxwell on 20nm, that's when the context becomes clearer. If the mobile maxwell and low end parts are this efficient, and the uarch scales as Kepler did, then that has good implications for however big Maxwell will turn out. The increased efficiency allows nvidia to scale the core count as they did with Kepler, at least that's what i'm thinking - Kepler scaled from low end to extreme high end very well. Hopefully, Maxwell is the same - although since 20nm is only available for SoC right now, and volume production is only about to start (per TSMC's PR anyway), it may be a bit before we see big Maxwell.
Last edited:
Unknown-One
[H]F Junkie
- Joined
- Mar 5, 2005
- Messages
- 8,909
Why would you want HDMI instead of DVI, exactly?
Dual-link DVI supports more bandwidth and higher resolutions than HDMI. Any DVI port can also be adapted to an HDMI port (with audio and HDCP support)
I see no reason for including HDMI ports on video cards at all. All it does is remove the need for a $1 adapter if you happen to be using an HDMI cable to connect your monitor.
You do realize that the standards body for DVI disbanded over a decade ago? I like DVI as well, but it isn't going to be around long term - I give it about 5 years or so before it disappears even for legacy reasons.
Main thing is, HDMI 2.0 will be able to do 4k on a single cable, while DVI-D tops out at 2560*1600. That's the entire problem with DVI right now, it is not 4k capable - theoretically, 2 DVI cables *could* do it, but that is a theory. I have not seen a single screen using double DVI implementation, while nearly every 4k screen is using HDMI 1.4a with 2 cables or displayport with 1 cable. HDMI 2.0 can do 4k with 1 cable.. Heck, 4k wasn't even thought of when the last iteration of DVI was created, and it's a moot point now since the standards body is disbanded. Both HDMI 2.0 and displayport are able to do 4k on a single cable.
edit: nevermind. I suppose you're talking in the context of Maxwell, i'm 100% sure Maxwell cards will have DVI. DVI will disappear at *some* point, but that will be many years from now. I'm getting ahead of myself again - DVI will be around for some time to come simply for legacy reasons. Also, if we're talking HDMI 1.4a (which is honestly a joke) i'd use DVI-D over HDMI 1.4a any day. I can't even get 2560x1600 working on most HDMI 1.4 implementations.
Main thing is, HDMI 2.0 will be able to do 4k on a single cable, while DVI-D tops out at 2560*1600. That's the entire problem with DVI right now, it is not 4k capable - theoretically, 2 DVI cables *could* do it, but that is a theory. I have not seen a single screen using double DVI implementation, while nearly every 4k screen is using HDMI 1.4a with 2 cables or displayport with 1 cable. HDMI 2.0 can do 4k with 1 cable.. Heck, 4k wasn't even thought of when the last iteration of DVI was created, and it's a moot point now since the standards body is disbanded. Both HDMI 2.0 and displayport are able to do 4k on a single cable.
edit: nevermind. I suppose you're talking in the context of Maxwell, i'm 100% sure Maxwell cards will have DVI. DVI will disappear at *some* point, but that will be many years from now. I'm getting ahead of myself again - DVI will be around for some time to come simply for legacy reasons. Also, if we're talking HDMI 1.4a (which is honestly a joke) i'd use DVI-D over HDMI 1.4a any day. I can't even get 2560x1600 working on most HDMI 1.4 implementations.
Going by shots of gpu-z the 750 doubles the pixel and texture of the 650 but memory bandwidth stays pretty much the same.
It could end up starving for bandwidth.
Guess we won't know until reviews come out showing how well it does in games.
For most moderate gamers though the 860 should be the one to look for.
It could end up starving for bandwidth.
Guess we won't know until reviews come out showing how well it does in games.
For most moderate gamers though the 860 should be the one to look for.
I doubt the legacy outputs will will even last that long. AMD and Intel are both going to drop VGA next year, and AMD is also planning to DVI it from chipsets at the same time. I couldn't find anything now but IIRC reading that nVidia was planning to retire legacy outputs at about the same time too. As long as I can get a 4k(5k?) monitor for $1500ish to replace my 3090 before then I'm not overly concerned; for the mass market even if you're keeping your old monitor passive single link DVI-HDMI adapters will work fine for 1080/1200p. It's only legacy 1440/1600 monitors that will be expensive to continue using because they need DL-DVI for bandwidth.
http://www.techworld.com.au/article/412885/vga_dvi_display_interfaces_bow_five_years/
First off, when I say legacy, I was referring to long term stuff like 10 years from now. DVI will not be around that long simply because it does not have the capability to support 4k resolution on a single cable, while HDMI 2.0 and Displayport do. Current HDMI 1.4a is worse than DVI, but HDMI 2.0 will support 4k / 60 IIRC.
Again, main point being: DVI will eventually be the first to go simply because it isn't capable of 4k on a single cable. This will be quite a few years from now, but legacy is a big issue, ESPECIALLY for business. Now for people like us who are on the cutting edge, legacy doesn't matter. But for corporations and enterprise, legacy is a huge deal. This is the entire reason that Lenovo does so well, as all of their thinkpad laptops have every connector imaginable including many with VGA - VGA is still used in quite a lot of projectors.
I wouldn't dismiss legacy so quick. Like I said, for *us* it doesn't matter, but for enterprise it is an absolutely huge deal. Therefore DVI will not just disappear for the next 5 years. It will still be around, trust me. Five years is the absolute minimum time frame I give it, but even for power users - there are a lot of power users still using DVI; a lot of monitors still only have DVI connectors.
Again, main point being: DVI will eventually be the first to go simply because it isn't capable of 4k on a single cable. This will be quite a few years from now, but legacy is a big issue, ESPECIALLY for business. Now for people like us who are on the cutting edge, legacy doesn't matter. But for corporations and enterprise, legacy is a huge deal. This is the entire reason that Lenovo does so well, as all of their thinkpad laptops have every connector imaginable including many with VGA - VGA is still used in quite a lot of projectors.
I wouldn't dismiss legacy so quick. Like I said, for *us* it doesn't matter, but for enterprise it is an absolutely huge deal. Therefore DVI will not just disappear for the next 5 years. It will still be around, trust me. Five years is the absolute minimum time frame I give it, but even for power users - there are a lot of power users still using DVI; a lot of monitors still only have DVI connectors.
Drown Babies
Gawd
- Joined
- May 29, 2005
- Messages
- 700
Really? That to me is the most promising aspect of these new, lower-end cards. Sub 75W GPUs means we might see some low-profile versions of these cards. Something that might finally replace the 7750 in this area, if they can make it single slot.
Nobody knows "exactly" other than in theory. ARM cores could ultimately release some of the burden that CPU's have on GPU performance and simultaneously cut latency by having a dedicated CPU on the GPU handle all the dirty calls locally (including managing system memory) without having to do the little dance with the CPU back to the GPU wasting precious resources.
The real test will be whether we see a noticeable impact immediately, or whether we have to wait for developers to take advantage of the technology like Mantle. Mantle, hUMA, and Project Denver are confusing things and filling us with doubts as to what will be relevant and what wont. At this point we just have to sit back and let things play out. At least we will see improvements in other things so whether these technologies work or don't is irrelevant.
Nvidia doesn't have their own fabs, and are basically at the mercy of TSMC. So they can't really do yearly iterations as Intel does, and TSMC is pretty far behind intel in terms of process - TSMC does not have FinFETs as of yet, and intel is about to hit the ground running with 14nm and has had FinFETS for a very long time now. FinFETs alone is a huge deal, especially for hitting super high frequencies with low TDP mobile parts. Anyway, nvidia presumably could release big Maxwell on 28nm, where gains are still possible. Yet I personally find that doubtful - I think they'll just release mobile and low end on 28nm, with big Maxwell on 20nm. 20nm will allow a far higher transistor budget, which is desirable for the high end.
Last edited: