Maxwell Nvidia Questions

JJ91284

Gawd
2FA
Joined
Mar 28, 2007
Messages
696
Does anyone have any links that state any time tables for the release of Nvidia performance GPU's based on Maxwell architecture. I know the 750 TI is out now, but I'd like to get a better performance card based on 20 nm. I know there's not a huge difference in performance, but I'm still interested in the expected release date.
 
I haven't been able to find any hard evidence of a release timetable anywhere. I've heard as early as May and as late as November but the majority of people seem to be saying sometime in Q3 (July-September). With the new(ish) 780ti and Titan Black, I wouldn't imagine they're in much of a hurry. Not to mention AMD just recently released their R9 series and I'm sure Nvidia has seen a nice bump in sales due to all of them being sold out (even at $100-$200 above MSRP).
 
Nothing announced or hinted at yet, but by early 4th qtr this year is probably a safe bet for the high end Maxwell card release.
 
Nvidia implied that there will be a "2nd Gen" Maxwell.
People are thinking these are the GM20x GPUs and they are on 20nm.

If they are on 20nm late Q3/Q4 would be likely.

Will probably be a GM204 and GM206 first with GM200/210 likely coming a good 6-8months after then first 20nm ASIC.

Edit- Seems I was remembering a fake slide about "2nd Gen" Maxwell being released in 2H '14.
 
Last edited:
Nobody knows for sure at this point and the release of the GTX 750/750 Ti really doesn't tell us much. Depends on whether Nvidia wants to push a gimped Maxwell on 28nm quickly on schedule for a normal Q2 release, or whether they are willing to wait until Q3 for 20nm parts, or just do a quick refresh similar to the 400/500 Series.

TSMC has got themselves back on track for the most part late in the game, so much so that they are at least a full quarter ahead of where they were a year ago. Depending on relations they have with TSMC and how quickly they can get it I would say Q4 is really pushing it. Even with the delays it was pushing it and I never saw that number pushed around anywhere other than speculation.

Late Q2 or Early Q3 would be my guess that we'll see the 800 Series in some form. Nvidia could be delaying things further if they want because of Mining causing pricing and shortage issues for AMD. We should be seeing the R9 290X pushing down the GTX 780 Ti, yet that is a long way from happening if ever at this point.
 
Wow... kind of hard to wait 9 months atm, but also kind of hard to sink $5-600 into a card too.
 
Nvidia stated that "2nd Gen" Maxwell will be out 2H '14.
People are thinking these are the GM20x GPUs and they are on 20nm.

If they are on 20nm late Q3/Q4 would be likely.

Will probably be a GM204 and GM206 first with GM200/210 likely coming a good 6-8months after then first 20nm ASIC.

I'm not sure why people seem to think that the big die GPU's will be delayed, Kepler was the exception in this case not rule. Every other card Nvidia has put out in the last 10 years has always had the big die flavour ready to go at launch.
 
Hmm, I was hoping I was missing some important news regarding new Maxwell release dates. I'm building a nice media htpc and wanted to stick with a new gen NVidia gpu. I already have an r9 290 in my main desktop and want to try something new and powerful from NVidia. Figured I could use my htpc as an occasional gaming machine. So it looks like I either wait, or go for a newer 750ti in Maxwell flavor 28 nm or go for a 780 TI on Kepler architecture.
 
I'm not sure why people seem to think that the big die GPU's will be delayed, Kepler was the exception in this case not rule. Every other card Nvidia has put out in the last 10 years has always had the big die flavour ready to go at launch.

GF100 ring any bells?
The only two comparisons that can be made are Fermi and Kepler, everything before that wasn't released right as the new node was available.
 
GF100 ring any bells?
The only two comparisons that can be made are Fermi and Kepler, everything before that wasn't released right as the new node was available.

I'm not sure I follow your point. GF100 was the first GPU out in the wild for Fermi (March 2010), and that was the big die making it GK110's contemporary.

I'm not talking about node availability, I'm simply referring to people assuming that GM200 will see a significant delay over GM204. Seems silly to assume it will follow the same pattern as Kepler.
 
I'm not sure I follow your point. GF100 was the first GPU out in the wild for Fermi (March 2010), and that was the big die making it GK110's contemporary.

I'm not talking about node availability, I'm simply referring to people assuming that GM200 will see a significant delay over GM204. Seems silly to assume it will follow the same pattern as Kepler.

But there was other 40nm ASICs released long before GF100 and by all accounts 40nm was a struggle for Nvidia. GF100 was a good 6months late.
Nvidia also doesn't have TSMC to cover their butts anymore, which is why they needed GK110 and GK100 was cancelled.

To think Nvidia is going to jump into a new node and release a +500mm2 ASIC first is naive.
 
Last edited:
I'm not sure I follow your point. GF100 was the first GPU out in the wild for Fermi (March 2010), and that was the big die making it GK110's contemporary.

I'm not talking about node availability, I'm simply referring to people assuming that GM200 will see a significant delay over GM204. Seems silly to assume it will follow the same pattern as Kepler.

Possibly because NVidia seems to be focusing on mobiles and has talked about designing from the bottom up rather than the top down. Which makes total sense given where the markets for computing devices and profits are moving. That seems like a reason to assume what has been said is true no?
 
Isn't nVidia, starting with Maxwell, going mobile-first, then building up to the big die? Sort of as Intel has already done?
 
Nvidia said in an investor call to expect high-end Maxwell GPU parts by the second half of 2014. That's all that's truly official so far. However, Forbes' tech editor spoke with nVidia in this article which states:

...Nvidia’s high-end Maxwell cards drop later this year (possibly by late March)...

So, we could see something very soon, potentially within the next couple of months or less. Will we? That's anyone's guess right now.
 
While the more popular news is focused on AMD's consumer gaming customers being affected by it's GPGPU (basically mining) customers (at least in certain regions) it gets left out that the same thing has happened on the Nvidia side. Nvidia's professional GPU business segment is growing and huge (much larger and lucrative for them than mining).

Also the other factor is pricing especially if competition is absent. I think with recent history we can assume that both companies (AMD also did this just recently with the 7xxx series release, so I don't know why there is an impression Nvidia only does this) are willing to raise the price bar with new releases. Nvidia could release a GM200 GPU that does perform at 2x the GTX 780 but costs $1000 (or more) maybe as a GTX Titan successor, probably not quite what people were expecting along the lines of a significant upgrade at the same price point.

Isn't nVidia, starting with Maxwell, going mobile-first, then building up to the big die? Sort of as Intel has already done?

The design philosophy (allegedly since we can't really verify how they actually approached this) was to design Maxwell from small to big to be able to better cater to the efficiency requirements of the mobile market.

Actual product release is another story though. While to desktop users (particularly on a site like this) GM107 is considered a "small" die relative to Nvidia's actual line it is in the middle. Smaller would be actual mobile chips Nvidia sells as part of Tegra or a GM108.
 
But there was other 40nm ASICs released long before GF100 and by all accounts 40nm was a struggle for Nvidia. GF100 was a good 6months late.
Nvidia also doesn't have TSMC to cover their butts anymore, which is why they needed GK110 and GK100 was cancelled.

To think Nvidia is going to jump into a new node and release a +500mm2 ASIC first is naive.

Other 40nm ASIC's sure but not Fermi based, that's what I'm getting at here. I wouldn't call it naive to err on the side of how things have gone previously. If they don't roll out GM200 at launch than so be it, but I think it would be a little to presumptuous to assume that it's going to go that way when there is little historical precedent for it.

At least in this case Nvidia has taken the opportunity create a earlier flavour of the architecture on 28nm which should assist when shrinking it down (much like the GTX280 - GTX285 transition).
 
I hope big Maxwell is released this year. Getting bored with my Titans.

Such a pain in the ass situation to be in...

If a Maxwell 20nm 860ti/870 priced around $350-400 doesn't come out by late summer, then I will be a bit surprised. With nVidia seemingly perpetuating their 700 and Titan series with spin-off releases, then maybe the reality is that 20nm Maxwell isn't quite as near as many of us are hoping.
 
I was considering changing my strategy of 1x x70 range card every 4 years to 1x x60 one every two. Then I just did a quick calculation:

Year.........x60.........x70.........x70 (bought at launch price)

1..............250.........350.........400
2
3..............250
4.............................350
5..............250.......................400
6
7..............250.........350
8
9..............250.......................400
10...........................350
11............250
12
13............250.........350.........400

Spent......1750.......1750.......1600

So... if I buy x70 range cards and replace them every 3 years instead of 4, by year 13 I will have spent exactly the same money whether I buy a x60 or a x70 card in different cadences. I'm now wondering then, considering x60 cards have more limitations than x70 specially when you think triple monitor setups like mine... maybe it makes more sense to do a x70 card every three years?

Made this little visual guide for comparison...
 
Last edited:
You're assuming that the pricing structure you listed is going to hold going into the future.

Also pricing and relative performance of each model is not consistent with each generation.

You should also factor in resale value and how much that may recoup.

The other thing to consider is other improvements aside from raw performance such as additional features.

Also industry trends may change which no one can accurately predict. At the moment it looks like companies might be moving towards longer development cycles (at least for major changes).
 
I am VERY anxious as well for any info I can find and of course getting my hands on the new cards. I have 2 MSI GTX 750 Ti GAMING cards and holy crap am I impressed about the performance for that $160-$170 card that requires no external power. I even have it mining and at full load its at 42C... So needless to say I am a believer in Maxwell as I currently know it.
 
I am VERY anxious as well for any info I can find and of course getting my hands on the new cards. I have 2 MSI GTX 750 Ti GAMING cards and holy crap am I impressed about the performance for that $160-$170 card that requires no external power. I even have it mining and at full load its at 42C... So needless to say I am a believer in Maxwell as I currently know it.

Same here. I'm holding on to my 670 and bypassing the 770 completely for Maxwell. I can't wait until I can buy one!
 
I was going to wait for Maxwell and then I just saw this 770 for $309 on Amazon... just couldn't ignore that price and went for it. I guess I'll wait for Volta for my next upgrade!
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Same here. I'm holding on to my 670 and bypassing the 770 completely for Maxwell. I can't wait until I can buy one!

+1

2 680's on 2560x1600. Looking to hopefully pull it all together end of the year with X99 + 20nm. :)

I don't mind the AMD offerings, but living in the heat of 100+ degree days in summer means trying to pick lower TDP parts. :(
 
I don't think Nvidia is in any hurry to release the high end Maxwell gpus. The 780 Ti still sits atop the 290x if only barely so. Then there's the Titan Black if you want to go nuts. Mantle fizzled and didn't live up to the hype with it having no meaningful impact on the high end. Nvidia is going to milk what they already have out for all it's worth before undercutting themselves.
 
I don't think Nvidia is in any hurry to release the high end Maxwell gpus. The 780 Ti still sits atop the 290x if only barely so. Then there's the Titan Black if you want to go nuts. Mantle fizzled and didn't live up to the hype with it having no meaningful impact on the high end. Nvidia is going to milk what they already have out for all it's worth before undercutting themselves.


Except Nvidia's lead wouldn't mean jack if AMD's GPU's were priced at the point they should have been from Day 1 this entire time. It would have either caused Nvidia to stay on the quick release cycle with Maxwell, or at the very least lower that damn GTX 780 Ti down to earth.

Nobody is really winning at the moment other than retailers.
 
Except Nvidia's lead wouldn't mean jack if AMD's GPU's were priced at the point they should have been from Day 1 this entire time. It would have either caused Nvidia to stay on the quick release cycle with Maxwell, or at the very least lower that damn GTX 780 Ti down to earth.

Nobody is really winning at the moment other than retailers.

Yep, the 290 shortages and price gouging are just icing on Nvidia's cake. I picked up 2 new reference 780 Ti's for $620 each when they were on sale right after Christmas. At that price I couldn't say no and I think many on the fence probably wouldn't either given the large markups negating any ati advantage there.
 
Back
Top