They are not going to waste time making a pascal on 28nm and waste resources and drive them away from upcoming big product. That is silly!!
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Nvidia Previewed Unreleased Graphics Card At "Secret" Meeting
- Thread starter Megalith
- Start date
They are not going to waste time making a pascal on 28nm and waste resources and drive them away from upcoming big product. That is silly!!
TaintedSquirrel
[H]F Junkie
- Joined
- Aug 5, 2013
- Messages
- 12,691
The idiot Linus already said it's a dual GPU card, he's "leaked" a bunch of stuff in the past... Unless he's trolling. Same guy who waved the 980 Ti around on camera 1 week before launch.
What about a GTX990 AND a refresh of the current Maxwell line-up? Stock +150-200Mhz GPU clocks and 2x the RAM. For only $20 more than the current price of the current model.
Can easily be done given the yields and quality of the GPU's after a year in production. Would give AMD a serious headache as Fury is limited to 4GB, and they need an explanation why the Hawaii-models offer more RAM.
(BTW, the latter is also part of the reason why "full" Tonga is limited to a 256-bit memory bus, and thus 4GB of RAM. Else, it would be 3GB or 6GB. Apart from the extra PCB cost. Now it is just a drop-in replacement on boards for the R9 285.)
Can easily be done given the yields and quality of the GPU's after a year in production. Would give AMD a serious headache as Fury is limited to 4GB, and they need an explanation why the Hawaii-models offer more RAM.
(BTW, the latter is also part of the reason why "full" Tonga is limited to a 256-bit memory bus, and thus 4GB of RAM. Else, it would be 3GB or 6GB. Apart from the extra PCB cost. Now it is just a drop-in replacement on boards for the R9 285.)
I'm betting they're gonna come out swinging. I'm betting the press kit looks something like a GTX 990 AND 990ti. 4096 and 6144 core monstrosities at $999 and $1299 respectively.
Then for good measure toss in a couple benchmarks showing Pascal smoking them both.
Next slide shows Elvis in an NVIDIA shirt playing Rainbow 6 Siege with Michael Jackson and Jesus, who's reading a paper dated 9/20/15.
And finally, a picture of Jen-Hsun Huang teabagging Lisa Su while waving a green flag and chugging a beer as the last presentation slide.
Then for good measure toss in a couple benchmarks showing Pascal smoking them both.
Next slide shows Elvis in an NVIDIA shirt playing Rainbow 6 Siege with Michael Jackson and Jesus, who's reading a paper dated 9/20/15.
And finally, a picture of Jen-Hsun Huang teabagging Lisa Su while waving a green flag and chugging a beer as the last presentation slide.
TaintedSquirrel
[H]F Junkie
- Joined
- Aug 5, 2013
- Messages
- 12,691
What explanation is needed? The 970 and 980 are more than a year old already.
AMD already needs an explanation to potential buyers why their top-model has less RAM than the mid-tier models. So they need to tell people HBM blabla.
There is no need to release a GTX 990 when hardly anyone is buying the GTX 980 Ti right now at the outrageous price it is at. If they do release a 990 then i can see this card costing 1399.00-1499.99
Pascal is coming sooner than anyone thinks, and will make the 980ti obsolete
"August 21st, 2015 at 4:56 pm - AuthorAnton Shilov
Samsung Electronics indicated that it plans to start volume production of high-bandwidth memory (HBM) next year at the Intel Developer Forum this week. At the trade-show, the company revealed its current vision and expectations concerning HBM. Samsung foresees that eventually high-performance applications (such as GPUs) could feature up to six HBM devices to enable unprecedented capacities and bandwidth.
As it appears, Samsung plans to skip the first-generation HBM memory and will only manufacture products compliant with the second-generation HBM specification, which offers higher densities and clock-rates. Such approach will let Samsung to address broader market segments with its HBM offerings. Right now HBM can only be used on consumer graphics cards and certain highly-custom products, but HBM2 will enable professional and high-performance computing GPU-based solutions, which require a lot of memory onboard.
Samsung is currently working on multiple HBM packages featuring two (2Hi stack), four (4Hi) and eight (8Hi) 8Gb memory devices on a base logic die with 1024-bit interface, according to a slide that the company demonstrated at the IDF (which was published by ComputerBase.de). Maximum data-rates of Samsung’s HBM products will be 2Gb/s, which will support up to 256GB/s of bandwidth per chip".
Pascal is coming sooner than anyone thinks, and will make the 980ti obsolete
"August 21st, 2015 at 4:56 pm - AuthorAnton Shilov
Samsung Electronics indicated that it plans to start volume production of high-bandwidth memory (HBM) next year at the Intel Developer Forum this week. At the trade-show, the company revealed its current vision and expectations concerning HBM. Samsung foresees that eventually high-performance applications (such as GPUs) could feature up to six HBM devices to enable unprecedented capacities and bandwidth.
As it appears, Samsung plans to skip the first-generation HBM memory and will only manufacture products compliant with the second-generation HBM specification, which offers higher densities and clock-rates. Such approach will let Samsung to address broader market segments with its HBM offerings. Right now HBM can only be used on consumer graphics cards and certain highly-custom products, but HBM2 will enable professional and high-performance computing GPU-based solutions, which require a lot of memory onboard.
Samsung is currently working on multiple HBM packages featuring two (2Hi stack), four (4Hi) and eight (8Hi) 8Gb memory devices on a base logic die with 1024-bit interface, according to a slide that the company demonstrated at the IDF (which was published by ComputerBase.de). Maximum data-rates of Samsung’s HBM products will be 2Gb/s, which will support up to 256GB/s of bandwidth per chip".
"Hardly anyone" LOL.
Uh Huh. Clearly you haven't been paying attention then. They only got the first test Pascal from the fab last week. That's with no memory or interposer or any of the other stuff. if anything Q12016 looks extremely optimistic.
another hint from a news feed
3DCenter reports that Nvidia has sent the first graphics cards based on the GP100 graphics processor to its subsidiary in India, where it has a lot of hardware and software developers. No actual details about the chip or graphics cards on its base are known, but it is about time for the graphics giant to start testing its GP100.
Nvidia taped out the GP100 in June, 2015. Production cycle of TSMC’s 16nm FinFET process technology is about 90 days, therefore Nvidia got its GP100 from TSMC very recently. Right now the company is testing the chip and its drivers internally.
Pascal benches PLEASE!
3DCenter reports that Nvidia has sent the first graphics cards based on the GP100 graphics processor to its subsidiary in India, where it has a lot of hardware and software developers. No actual details about the chip or graphics cards on its base are known, but it is about time for the graphics giant to start testing its GP100.
Nvidia taped out the GP100 in June, 2015. Production cycle of TSMC’s 16nm FinFET process technology is about 90 days, therefore Nvidia got its GP100 from TSMC very recently. Right now the company is testing the chip and its drivers internally.
Pascal benches PLEASE!
Hornet
Supreme [H]ardness
- Joined
- Oct 4, 2005
- Messages
- 6,624
If they are only testing Pascal chip now, I can't see it being ready this year.
This "secret meeting" GPU is probably something they'll release in time for the holiday season this year. So I think it's unlikely to be Pascal.
Pascal will be released NLT than Feb-March 2016 ,
not 2015.. I think we will see 960ti before 2016 and definitely not a dual 990
I will be very surprised if they release a dual 990 and what will it cost? Probably 1400.00
what a rip off
no one will buy it
Ashbringer
Supreme [H]ardness
- Joined
- Jan 25, 2010
- Messages
- 5,522
Let me explain to you how graphs work. See the one bellow my post? Also to be fair there's a graph from Tech Report which shows the 980 Ti beating the crap out of Fury X, but the Asus Strix GTX 980 Ti that they used for testing is Aftermarket factory overclocked that's 9% faster than reference clock.
So it's safe to say that the 980 Ti and Fury X are trading blows, but not like it matters cause who the hell buys a $650 graphics card anyway? Crazy people, that's who. And the Asus Strix GTX 980 is $870 at Amazon which means those buying it are crazier. Also just look at how AMD's cards are dominating the low and mid range.
Same reason why some 750 Ti's are sold with 4GB of VRAM, cause there's a lot of idiots out there that think memory is that important. Truth is no game today needs more than 4GB of VRAM. Most games are fine with 2GB, hell most are fine with 1GB. The only time you go beyond 4GB of VRAM is 4k+FXAA in Shadow of Mordor and 4k with AA is pointless. A 970 at that point should perform Seppuku.
The 390's with 8GB is AMD's way of saying they don't like the price trend of the 290's so they threw a useless 4GB more memory to keep the price high. Personally I would buy a 290 based on reference design and flash the BIOS to a 290X and crank up the overclocks. Better deal than a 390. If I want to I can flash a 390X bios on a 290, but it doesn't make a difference. A 290 can be had for $250, and that's not even looking on Ebay.
As an Amazon Associate, HardForum may earn from qualifying purchases.
Nenu
[H]ardened
- Joined
- Apr 28, 2007
- Messages
- 20,315
I guess you missed this poll of cards owned.
http://hardforum.com/showthread.php?t=1872058
The 980ti is the card owned by most by a LONG way.
NVidia must have held a lot of competitions with 980tis as prizes if nobody is buying them!
illram
[H]ard|Gawd
- Joined
- Sep 19, 2011
- Messages
- 1,473
My bet is it's an ITX 980ti or something.
Yakk
Supreme [H]ardness
- Joined
- Nov 5, 2010
- Messages
- 5,810
Getting a dual GPU card right before a process node shrink traditionally has been a terrible idea, let alone a process node shrink AND moving from ddr5 to HBM at the same time!
Only reason I can see nvidia releasing a dual GPU card is either the next green card is getting pushed back and they need something to tie them over, or they need it for VR kits.
Either way it looks like a money waster IMO.
Only reason I can see nvidia releasing a dual GPU card is either the next green card is getting pushed back and they need something to tie them over, or they need it for VR kits.
Either way it looks like a money waster IMO.
Except traditionally their dual-GPU's have sold well. They don't make a move without market research (which is why for example you won't see them waste time on a $650 SFF card that no one will buy).
The main reason NVIDIA releases dual-GPU cards is because they can. They know AMD is prepping a dual-GPU Fiji (Fury X2), so just like the 980 Ti was perfectly timed to cut the Fury launch off at the knees, the GTX990 will do the same to Fury X2. Jen-Hsun got spies everywhere, n*gga. <-- You see how I just went urban at the end there. That's urban.
Wonder if you'll say the same when Fury X2 launches?
Ashbringer
Supreme [H]ardness
- Joined
- Jan 25, 2010
- Messages
- 5,522
You know that poll is stupid right? You might as well ask the people on the Ford forum to see what car they buy. Really Ford Mustang? Must be an amazing car!
According to the Steam Hardware Survey about 0.99% of Steam users have a GTX 980. Neither the 980 Ti or the Fury X shows up on the Survey. The GTX 970 is nearly 4% and probably the most popular graphics card for Steam. The GTX 780 is #2 with 3%.
Not saying people don't buy them but certainly not enough to show a blip on the Steam Survey. At least not yet anyway. I almost gaurentee you that new Nvidia product is either a 990 with two GM200's in it or another Tegra product nobody cares about. God knows we need to hear about another useless Tegra product.
cinnamonandgravy
Limp Gawd
- Joined
- Jan 4, 2011
- Messages
- 268
its probably best to just stop saying words now.
cinnamonandgravy
Limp Gawd
- Joined
- Jan 4, 2011
- Messages
- 268
Yakk
Supreme [H]ardness
- Joined
- Nov 5, 2010
- Messages
- 5,810
Yeah, it's a "me-too" product I guess.
Wonder if you'll say the same when Fury X2 launches?
??? I'll say the same, the x2 Fury having HBM already helps though. The card might not age as quickly, but this is STILL all before a node shrink (finally) which will probably give it a much smaller life cycle taking into account the prohibitive (for some) buy-in price.
Maybe he only plays minesweeper. I'm at 2560x1440 and regularly see 2.5+ GB of usage
misterbobby
2[H]4U
- Joined
- Mar 18, 2014
- Messages
- 3,814
I certainly go way over 4 GB and even approach 5 GB in some games at 1440 and well over 5 GB approaching 6 GB at 4k. Of course using that much vram does not mean I actually need it so it is hard to tell sometimes how much is enough.
Its also stupid to try to infer something about discrete GPU buying trends based on steam stats skewed by all the iGPU's. Disingenuous at best. People aren't buying an Intel GPU, they're ending up with one.
TaintedSquirrel
[H]F Junkie
- Joined
- Aug 5, 2013
- Messages
- 12,691
Semantics
2[H]4U
- Joined
- May 18, 2010
- Messages
- 2,811
Except many 980ti are not stock and the one that shows the Fury X quick is AMD supplied benchmark, y'all remember the last Fury X benchmarks provided by AMD.
xXxTiTaNxXx
AGREED!
Either way it looks like a money waster IMO.[/QUOTE]
AGREED!
If they do this time right before Pascal is released they are really dumb
Wonder if you'll say the same when Fury X2 launches?
Is AMD borrowing more money because i thought they were going BROKE
I would have to disagree 125 million users take part of this poll and these users are gamer's.
http://techfrag.com/2015/09/22/nvidia-pascal-arrives-india-internal-testing/
The first batch of samples were sent to India back in August which contained two circuits. The most recent batch sent to India arrived a week ago and featured 7 units of Pascal.
The first batch of Pascal cards might only be 16gb ram for the flagship and then the next batch late 2016-2017 is going to be 32gb.
http://techfrag.com/2015/09/22/nvidia-pascal-arrives-india-internal-testing/
Rumor has it that Nvidia Pascal will feature 17 billion transistors and HMB2 memory under the hood. Reports say that HBM2 memory may go up to 32GB on professional graphics cards. Meanwhile gaming centric cards will feature up to 16GB of HBM2.
Users may notice up to 60-90% performance boost compared to Maxwell.
Ashbringer
Supreme [H]ardness
- Joined
- Jan 25, 2010
- Messages
- 5,522
That's right, people like you exist. Remember to look at the bar graphs kids. These are fairly modern games. Also 1080P is what the majority of people use when it comes to gaming. Don't complain to me about AA off cause if you have a 1GB GPU chances are you can't run the game with it.
And now for games that use about 2GB. Again fairly new games that don't break the 2GB barrier.
The elitism here is just awe inspiring. By any chance maybe people are using iGPU's cause laptops? Even still how does that invalidate anything?
Keep in mind your poll shows more 980 Ti's than 980's but on Steam the 980's show up.
TaintedSquirrel
[H]F Junkie
- Joined
- Aug 5, 2013
- Messages
- 12,691
Those games were 'modern' in 2013. They were originally designed for the PS3 and Xbox360... Except for TW3 which uses some advanced texture streaming to keep usage down.
For 1080p the sweet spot seems to be around 2.5~3GB usage. GTA V and AK spike up to 4. Fury X was tested thoroughly @ 4K and I still don't recall seeing any complaints about textures so it's safe to assume 4GB will be fine for quite a while.
GTA V
AC Unity
Dying Light
Watch Dogs
Arkham Knight
For 1080p the sweet spot seems to be around 2.5~3GB usage. GTA V and AK spike up to 4. Fury X was tested thoroughly @ 4K and I still don't recall seeing any complaints about textures so it's safe to assume 4GB will be fine for quite a while.
GTA V
AC Unity
Dying Light
Watch Dogs
Arkham Knight
misterbobby
2[H]4U
- Joined
- Mar 18, 2014
- Messages
- 3,814
