![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
so if the 680 was suppose to be the midrange
- Thread starter Filter
- Start date
sirmonkey1985
[H]ard|DCer of the Month - July 2010
- Joined
- Sep 13, 2008
- Messages
- 22,414
the rumored original plan was the 680 being the 660 and GK110 being the 670/680 but they had a ton of issues with the GK110 so they kept pushing it back. in my opinion i think the GK110 thing was a marketing joke and was never actually suppose to be released til Q1 2013 anyways.
I pretty much realized this when they called it the 680. As far as the naming scheme goes, it's a bunch of crap. They could have called it the 660 and it still would have sold for $500.
480 = high-end for 2010
580 = 2011
680 = 2012
780 = 2013
GK110 will be part of a new generation of cards, which will be evidenced by them calling it th 780 and not something like the 685. It'll be part of a Kepler-refresh lineup of cards. Nvidia talked some when AMD released 7970 yet they could only marginally beat it out a few months later...
Last edited:
FaRKle0079
2[H]4U
- Joined
- Jan 11, 2005
- Messages
- 2,686
I'm pretty sure GK100 was supposed to be the high end this generation, not GK110. GK110 is NV re-working GK100 to get it functional. Saying GK110 was supposed to be the high end now doesn't make sense based on Fermi's naming scheme (GF100 was GTX480, GF110 was GTX580, which was a reworked and tweaked GF100).
that is right, the gk100 was supposed to be the 670/80. things didnt work out too good for that die, so they scrapped it. that is why the gk104 ended up being the 670/80, because they gk100 was broken. nvidia got real lucky that the 7950/70 didnt end up being faster.
kostonn760
Weaksauce
- Joined
- May 5, 2012
- Messages
- 111
I'd like to see Nvidia do something the industry hasn't seen before in a future gen VGA. I'm not talking about all these small increases in VRAM and CUDA Cores. It's 2012 guys, LET'S SEE SOMETHING THAT MAKES US DROOL!!
honestly the things we see on the consumer market is probably only a pretty small % of what they could make and what they are capable of..
they're an intelligent business.. they make next gen cards 25-30% better and that means that people are willing to upgrade and it will be easy for them to make a card every year with a performance increase like that.. if they jump 100% in performance people buy less cards because they last longer and they will be hard pressed to make something significantly better. computer technology moves fast so im not complaining but in reality i think its all just intelligent marketing
What? A few months and they're expected to blow it out of the water??? Since when?
Just because the 2 companies aren't on a synchronized release schedule anymore doesn't mean they aren't working on about the same level of technology for a given year.
600 series is the same generation as the 7000 series.... so no shit they'll be about the same. Don't be so entitled, the 8800GT and 5870 were aberrations of their time.
BababooeyHTJ
Supreme [H]ardness
- Joined
- Jan 21, 2009
- Messages
- 6,951
I never even heard of GK100 months before GTX 680 was released when Kepler rumors first started rolling out. I'm not sure where you're getting that info from. I'm seeing a lot of assumptions here with no real proof to back that up.
do you think nvidia decided not to make a gk100? and decided to make a gk110 first? why would they skip the gk100 and go right to the gk110? did they release the gf110 before the gf100? no, they learned from the gf100 and made a better more redefined chip, the gf110.
the logic that nvidia never made a gk100 is crazy. the way their naming scheme work is that middle digit is the revision.
gf104->gf114, gf100->gf110, gf108->gf118, gf106->gf116
so why do you think there never was a gk100 gpu?
BababooeyHTJ
Supreme [H]ardness
- Joined
- Jan 21, 2009
- Messages
- 6,951
Show me one source that claims that a GK100 was ever even taped out.
So because they used those code names last gen that means that it applies to this gen? Thats not what they used for the generation before that (GT200) or before that (G80).
so you think they never made a gf100? instead they started at 110? use your brain and think...
So because they used those code names last gen that means that it applies to this gen? Thats not what they used for the generation before that (GT200) or before that (G80).
lets see, what has changed from last gen? fermi to kepler. gf to gk. why would they scrap their naming scheme?
also, if they changed their naming scheme the gk110 might not even be a top end chip? just because the gf110 was a high end chip doesnt mean the gk110 is a high end chip (by your logic).
BababooeyHTJ
Supreme [H]ardness
- Joined
- Jan 21, 2009
- Messages
- 6,951
Talk to me when you have some evidence to back up your claims. Until then you're making useless assumptions that you can't backup so like the typical internet troll you have to result to insults.
When was GK100 taped out? Since you seem to know it all.
When was GK100 taped out? Since you seem to know it all.
Last edited:
do you think nvidia is going to come out and say "yeah we fucked up real bad on the gk100, so we had to scrap it."
if nvidia started making the gk110 and the gk104 around the same time, why wouldnt the gk110 be out now? or close to being released? why would the gk110 be 1 year behind the gk104?
http://www.3dcenter.org/news/2011-12-16
http://www.theinquirer.net/inquirer...hed-amds-mid-range-radeon-hd-7870-gpu-compute
"Remember, this card is supposed to be 'the crown winner' for Nvidia, since it couldn't make the bigger GK100 die on time"
Last edited:
GK110 is basically gonna be a beefed up GK104. Rumormills around the internet have it besting the 680 (mostly using extrapolated data based on transistor counts) by a moderate margin in 3d gaming. Where it'll REALLY shine is likely gonna be compute based benchmarks. Which would line up nicely with it's initial release being in the HPC/Workstation market.
On double precision it kills the 600 series hands down. On single precision its not that bad.
The difference between amd and nvidia on double is more then 5 to 1 if i remember it right for similar tier card (ie 7970 compare to 680). Also imho any card over 250 is high end. Mid range should be 100-250, sub 100 is low/entry lvl. While 350-400 and up is just for the 3-5% of publics that are hardcore gamer/ppl that get more expensive stuff just so they can say they have the best.
SBSuperfly
Gawd
- Joined
- Feb 1, 2006
- Messages
- 865
you're ignoring competition
geraltofrivia
Limp Gawd
- Joined
- Jun 12, 2012
- Messages
- 409
It feels like AMD is struggling a bit to keep up with nVidia at the top end... It makes no sense for nVidia to blow away AMD by releasing a super expensive big kepler today when they can get away with 6 months "on the house" for the post-kepler R&D team.
Struggling ... thats funny, how are they struggling?
What I see is Nvidia struggling to release cards where they make money at, the low end to mid range cards.
The GTX670 is a the life in nvidia right now on desktop graphics, they seem to have plenty of those and there is a decent market for them. Other then that AMD is raking it in on no competition for their 77XX and 78XX series of cards.
Seriously.
The info was actually from a leaked slide. I can't find the leaked slide as I'm too tired and don't really care to look it up. You are right, it is speculation on our end, speculation created from something that may have been leaked or may have been photo shopped. Either way the final released slide was very similar, just with GK104 replacing the GK100 spot on the leaked slides.
Using assumption and trying to apply some common sense to the situation, one could assume that the GK100 was supposed to release, but didn't. That doesn't mean that the GK104 (680/670) isn't rightfully where it needs to be, because they are great chips. The GK110 announced a few weeks ago, with the official release date of Q1/2013 carries some speculation that there WAS a GK100 in production but the chip was too big and/or the process was too new, which isn't a very unimaginable thing to happen.
Either way, the GK104 carries less computer power with it similar to how the 560 was designed last gen and the GK110 is apparently not losing any compute power like the 580 of last gen. Again more speculation and the only links would be from rumor mills.
geraltofrivia
Limp Gawd
- Joined
- Jun 12, 2012
- Messages
- 409
I mean at the top-end. I think that nvidia has got an ace up its sleeve in Big Kepler. They could've released the 680 as a $400 card and Big as a $600 card if AMD was more advanced at the top end. Instead they can sort of take it easy for bit and ride some comfy margins. This is coming from an lifelong AMD fanboy, I'm rooting for them but right now I just think nvidia/intel makes a better top-end consumer product.
geraltofrivia
Limp Gawd
- Joined
- Jun 12, 2012
- Messages
- 409
All signs from AMD point towards them focusing on mid/low-range graphics (incl next-gen consoles). It's a smart move but the lack of competition at the top is disheartening. AMD wants nothing more than Nvidia to take it slow with their next release...
the next gen cards GTX 700 and HD 8000 are coming only early next year. Not much is known about HD 8900 cards. though Nvidia has provided more details about GK110. What remains to be seen is what are the real specs of both HD 8970 and GTX 780. also how much does AMD improve efficiency and how much does Nvidia's perf/watt in gaming get affected by the massive chip level investments which Nvidia is making for compute in GK110. I think given that Nvidia has gone for a massive 7 billion transistor chip its going to be a really hot and power hungry chip. so people who were happy with how cool the GTX 680 is running (GK104) should be prepared for a hotter running chip. Hopefully its performance justifies that.
geraltofrivia
Limp Gawd
- Joined
- Jun 12, 2012
- Messages
- 409
The size of the chip makes me think it'll be introduced at a higher price point than the 680. If it's priced at $500, then you know nvidia made some serious buck selling 680s for $500 all this time.
I expect price to be around USD 550 - 600. its definitely going to be more costly than GTX 680's launch price of USD 500.
MaelstromOC
Limp Gawd
- Joined
- Feb 28, 2009
- Messages
- 178
I doubt it'll be higher. $550 max really.
needmorecarnitine
Supreme [H]ardness
- Joined
- Feb 10, 2004
- Messages
- 4,609
I think it will come out as a Tesla part first at a high price. The consumer version would come later and be priced at $500 is my guess.
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
The GTX680 (GK104) is the same size and configuration of GF114 in most of the 560's (and many mobile parts). It's nothing like GF100/110, which would lead us to assume that Nvidia co-developed a Big Kepler alongside it, in this case GK100, which would be similar to the size and configuration of Big Fermi. We know that Nvidia isn't about to concede the GPGPU market to AMD.
If GK100 failed for whatever reason, then we can look forward to GK110 sometime this fall; Nvidia has stated this much already, that it is coming as a Tesla product, where it is most valuable.
The real issue here is that Nvidia's mid-range silicon is beating the hell out of AMD's high-end silicon. Assuming yields are equal, Nvidia is beating the hell out of AMD in profit margin right now, due to the slower than expected performance of Big GCN.
Further, even if GK100 had worked, or GK110 were ready today, there's no reason for Nvidia to release it. They have Big Fermi-based Quadro and Tesla cards that are still faster than Big GCN to sell before that becomes an issue.
They can keep their high margins by selling us their mid-range part at twice the price they had expected to when it was on the drawing board, and we have AMD to blame.
If GK100 failed for whatever reason, then we can look forward to GK110 sometime this fall; Nvidia has stated this much already, that it is coming as a Tesla product, where it is most valuable.
The real issue here is that Nvidia's mid-range silicon is beating the hell out of AMD's high-end silicon. Assuming yields are equal, Nvidia is beating the hell out of AMD in profit margin right now, due to the slower than expected performance of Big GCN.
Further, even if GK100 had worked, or GK110 were ready today, there's no reason for Nvidia to release it. They have Big Fermi-based Quadro and Tesla cards that are still faster than Big GCN to sell before that becomes an issue.
They can keep their high margins by selling us their mid-range part at twice the price they had expected to when it was on the drawing board, and we have AMD to blame.
geraltofrivia
Limp Gawd
- Joined
- Jun 12, 2012
- Messages
- 409
Well said.
You are forgetting a lot of things here. AMD didn't attempt to make a big core, Nvidia barely out pacing them on their midrange product means nothing as that is their best product right now.
AMD didn't attempt a big product and failed on it, Nvidia did. Thats a lot of R&D that Nvidia is missing out on, with more spent on fixing it to get it to work with the GK110, which we don't even know if it will work. Nvidia publicly stating they will release a GK110 product brings hope we do see it.
If Nvidia did have a working product, produced the product but decided not to release it is possibly the worst business strategy ever heard of. You don't spend man hours, R&D money, production money just because you have a product that barely beats the competition. You sell and make up for it. We also don't know the margins on the GTX680 or GTX670.
One thing is very well known, AMD is enjoying free reign on the low end to mid range segment where Nvidia has nothing.
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
That's the pessimistic view- and I'm not saying that I disagree. My perspective stems from reports that Nvidia was actually disappointed in GCNs performance, and given that their mid-range part is faster than AMDs largest part, that assessment holds water.
From a business standpoint, though, Nvidia doesn't need to release Big Kepler, and I made the case above that it might even be to their detriment. Right now, again keeping yields equal, Nvidia can charge more for a smaller piece of silicon. That implies that they have the opportunity to enjoy better margins than AMD at the high end, which is something they're not used to- usually, AMDs smaller top-end silicon competes favorably with Nvidia's larger top-end silicon.
Where AMD has made a business out of making gaming-only focused GPUs that generally forsake real GPGPU capabilities and Nvidia has made a business out of making GPGPU parts that are good at games but are otherwise massive in comparison, the situation has reversed, and Nvidia is reaping the benefit by selling what was intended to be a $200-$300 part for $400-$500.
And they're still not losing out to AMD in the GPGPU arena, as Fermi with its exclusive CUDA support and established market share/installed base remain superior to GCN, for now. They have time to make GK110 more profitable (as previously GT200, GF100, GF110 were scraping the barrel with initial yields on consumer parts) and beat AMD over the head with it later. And hell, they may make GK114 bigger/faster/otherwise better in the process, as they've always done.
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
One would expect them to, but not until demand for the higher-priced variants fades, or until they've harvested enough functional dies that don't make the cut for a GTX670/GTX680m etc.