Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [TT] NVIDIA should launch its next-gen Pascal GPUs with HBM2 in 2H 2016
New Posts  All Forums:Forum Nav:

[TT] NVIDIA should launch its next-gen Pascal GPUs with HBM2 in 2H 2016 - Page 55

post #541 of 724
Quote:
Originally Posted by CrazyElf View Post

The money matters a lot more than you would like. Money means Nvidia has more to spend on R&D, which means that they can make relative gains or use their money in a way that locks AMD out. That's the real problem with the money. It means you can spend more on research to build a better product. The average user is not well educated and having a good halo product means sales. To use another example, on the CPU front, what is limiting AMD is not engineering talent, but money. With enough money, they'd be able to build a much better CPU. I'm worried the GPU front could end up like this as well.

What is inexcusable in my opinion is that with their money, they do not use it to pay a few software developers to give a longer support period for their products. That's just planned obsolescence taken to its logical extreme. You get 1 year worth of optimization and that is it. That's not a good deal for the end consumer, but it does translate into sales. We consumers lose - big time. What happens when the gain a monopoly, raise prices on even their middle-end GPUs, all while keeping a year of support?

Unfortunately, their unethical business practices seem to have paid off. Part of it is due heavily to AMD's missteps, but a lot of it is because right now what they are doing has paid off. They've gained market and mind share. Their Titan-series of products, despite being overpriced and everybody knowing that they are not a good value, have sold quite well. They've gained marketshare and control the high margin Compute market. Look at this thread. There is an army of people who will defend Nividia in a bizarre Stockholm syndrome like way. That sort of thing translates into sales, which means more money, which means the potential for more R&D.

At this point, I'd support anti-trust legislation to break both Intel and Nvidia into 2 companies. I think it is necessary to ignite real competition.
This.

Unless Nvidia is willing to dilute the Titan brand name (although there is a precedence for this - Intel did so with the Pentium series and AMD did so with the FX series), they won't launch if first. The first die will be a mid-ranged part. Only after the big part launches will we see a "true Titan".

I have seen a lot of companies with "money" that has been squandered on fruitless endeavors , while they neglect the core business and its customers.

My thesis is that Nvidia's gig is up, as customers in this segment are usually about "what have you done for me lately". So, screwing them over with limited optimization support isn't gonna help make them any more stickier, particularly if AMD comes out with a compelling value proposition for this market, no matter how many hired guns come out to advocate on Nvidia's behalf.
Simplicity
(11 items)
 
Apotheosis
(10 items)
 
 
CPUMotherboardGraphicsRAM
4770k Asus Z87 Pro TBD Corsair Vengeance (2x8GB) DDR3 1600 RAM 
OSMonitorKeyboardPower
Windows 7 Pro Dell U2713HM Alienware TactX gaming Seasonic 850W Gold  
CaseMouse
Cooler Master HAF XB Alienware TactX premium mouse 
  hide details  
Reply
Simplicity
(11 items)
 
Apotheosis
(10 items)
 
 
CPUMotherboardGraphicsRAM
4770k Asus Z87 Pro TBD Corsair Vengeance (2x8GB) DDR3 1600 RAM 
OSMonitorKeyboardPower
Windows 7 Pro Dell U2713HM Alienware TactX gaming Seasonic 850W Gold  
CaseMouse
Cooler Master HAF XB Alienware TactX premium mouse 
  hide details  
Reply
post #542 of 724
As I've said countless times, there is no precedence whatsoever for Nvidia to release a big-Pascal Titan first and you'd have to be naive to really believe they would do so. As Magnek said, yields are going to be crap for a long time on this new node and the profit motive for releasing a mid-range GPU as the new flagship long before releasing a true flagship GPU has been well established since the GTX 680 launched way back in 2012. They will release GP204 first as the 1080 (or whatever they call it) and we will see a GP100 Titan launch probably early next year if I had to guess. The guys who have been feverishly getting their wallets ready for a surprise April GP100 launch are simply allowing their hopes to overrun common sense and history....
post #543 of 724
I agree with Eric but will add one exception: IF (that's a big IF) AMD is hellbent on clawing back the performance crown, and comes flying out the gate with big Polaris, nVidia may have no choice but to cobble together a GP100 part unless they want to lose in every performance metric at every single resolution (well save for Gameworks games I guess tongue.gif).
post #544 of 724
Quote:
Originally Posted by Glottis View Post

yeah.. i don't see DX12 exclusivity becoming a thing for a few years at least. a few MS games will do it this year, but that's a promo thing for Win10. Other publishers don't have anything to gain from DX12 exclusivity.

I see DX12 and Vulkan being incorporated into games when DX11 dies off, that way you have a Win7/8 compatible option for those of us who aren't going to Win10. Pascals DX12 performance matters to a greater degree than Maxwell but I doubt it'd make or break the product, especially if its DX11 performance is greater than Polaris. That said if it's only typically a few % above Polaris in DX11 and still much worse in DX12 I can see people getting Polaris just for that as people tend to keep cards for longer and longer these days, all my friends who upgraded this generation have full intentions to be using their cards for 4 years or so at least.
My system
(22 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i5 3770k @ 4.6Ghz ASRock Z77 Pro3 Powercolor Radeon HD7950 3GB @ 1150/1370 4x4GB G.Skill Ares 2200Mhz CL11 
Hard DriveHard DriveHard DriveOptical Drive
Samsung 840 250GB Western Digital Black 1TB WD1002FAEX Samsung Spinpoint EcoGreen 2TB Pioneer DVR-220LBKS 
CoolingCoolingCoolingOS
Noctua NH-D14 Scythe Gentle Typhoon 1850rpm Corsair AF140 Quiet Edition Arch Linux x86-64, amdgpu 
OSMonitorMonitorKeyboard
Windows 10 Bloatfree Edition BenQ G2220HD BenQ G2020HD Ducky Shine III Year of the Snake, Cherry Blue 
PowerCaseMouseMouse Pad
Silverstone Strider Plus 600w Lian Li Lancool PC-K60 SteelSeries Sensei Professional Artisan Hien Mid Japan Black Large 
AudioOther
ASUS Xonar DX NZXT Sentry Mesh 30w Fan Controller 
  hide details  
Reply
My system
(22 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i5 3770k @ 4.6Ghz ASRock Z77 Pro3 Powercolor Radeon HD7950 3GB @ 1150/1370 4x4GB G.Skill Ares 2200Mhz CL11 
Hard DriveHard DriveHard DriveOptical Drive
Samsung 840 250GB Western Digital Black 1TB WD1002FAEX Samsung Spinpoint EcoGreen 2TB Pioneer DVR-220LBKS 
CoolingCoolingCoolingOS
Noctua NH-D14 Scythe Gentle Typhoon 1850rpm Corsair AF140 Quiet Edition Arch Linux x86-64, amdgpu 
OSMonitorMonitorKeyboard
Windows 10 Bloatfree Edition BenQ G2220HD BenQ G2020HD Ducky Shine III Year of the Snake, Cherry Blue 
PowerCaseMouseMouse Pad
Silverstone Strider Plus 600w Lian Li Lancool PC-K60 SteelSeries Sensei Professional Artisan Hien Mid Japan Black Large 
AudioOther
ASUS Xonar DX NZXT Sentry Mesh 30w Fan Controller 
  hide details  
Reply
post #545 of 724
Vulkan might get more publicity, at the very least it'll provide an even smoother cross-platform compatibility.
that is to say, SteamOS, XB and PS platform will likely support Vulkan, which makes porting games much easier.

i doubt DX12 would ever go into Win7/8, its similar to WinXP never getting DX10/11.
Edited by epic1337 - 2/24/16 at 10:46pm
post #546 of 724
Thread Starter 
Quote:
Originally Posted by magnek View Post

I agree with Eric but will add one exception: IF (that's a big IF) AMD is hellbent on clawing back the performance crown, and comes flying out the gate with big Polaris, nVidia may have no choice but to cobble together a GP100 part unless they want to lose in every performance metric at every single resolution (well save for Gameworks games I guess tongue.gif).

I hope AMD will get the performance crown back.
MyCleanPC
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 4770k ASUS MAXIMUS VI EVGA master blaster Corsair Vengence  
Hard DriveOptical DriveOSMonitor
OCZ SSD raid0 samsung Win 7 Samsung 
PowerCase
Enermax rev 1050 Stacker 832 
  hide details  
Reply
MyCleanPC
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 4770k ASUS MAXIMUS VI EVGA master blaster Corsair Vengence  
Hard DriveOptical DriveOSMonitor
OCZ SSD raid0 samsung Win 7 Samsung 
PowerCase
Enermax rev 1050 Stacker 832 
  hide details  
Reply
post #547 of 724
nvm
Andromeda
(12 items)
 
  
CPUMotherboardGraphicsRAM
7700K Asus Z270G Strix GTX 1080 Strix Triden Z RGB 3600 CL16 
Hard DriveCoolingOSMonitor
M.2 Samsung Evo 960 Kraken X62 Win 10 Acer XB270HU 144hz IPS GSYNC 
KeyboardPowerCaseMouse
Razer Black Widow Tournament Corsair 860i NZXT S340 Elite Corsair Sabre Optical 
  hide details  
Reply
Andromeda
(12 items)
 
  
CPUMotherboardGraphicsRAM
7700K Asus Z270G Strix GTX 1080 Strix Triden Z RGB 3600 CL16 
Hard DriveCoolingOSMonitor
M.2 Samsung Evo 960 Kraken X62 Win 10 Acer XB270HU 144hz IPS GSYNC 
KeyboardPowerCaseMouse
Razer Black Widow Tournament Corsair 860i NZXT S340 Elite Corsair Sabre Optical 
  hide details  
Reply
post #548 of 724
Quote:
Originally Posted by Majin SSJ Eric View Post

As I've said countless times, there is no precedence whatsoever for Nvidia to release a big-Pascal Titan first and you'd have to be naive to really believe they would do so. As Magnek said, yields are going to be crap for a long time on this new node and the profit motive for releasing a mid-range GPU as the new flagship long before releasing a true flagship GPU has been well established since the GTX 680 launched way back in 2012. They will release GP204 first as the 1080 (or whatever they call it) and we will see a GP100 Titan launch probably early next year if I had to guess. The guys who have been feverishly getting their wallets ready for a surprise April GP100 launch are simply allowing their hopes to overrun common sense and history....

GTX 480 launched first. With some issues. I get the profit motive on the mid-range flagship then big chip but maybe I should blame AMD for that. GTX 580 was first. 8800 GTX ( I remember that day) but I don't think either was a new node.
Edited by dubldwn - 2/25/16 at 3:28pm
Twilight
(14 items)
 
   
CPUMotherboardGraphicsRAM
Intel 5820k ASUS X99-A EVGA GTX 1080 G.SKILL Value 16GB 2400 
Hard DriveCoolingOSMonitor
Samsung sm951 512GB Custom liquid Windows 10 ACER XB270HU 
KeyboardPowerCaseMouse
Corsair RGB Strafe Silent EVGA 850 G2 Caselabs S8 Razer Deathadder 
  hide details  
Reply
Twilight
(14 items)
 
   
CPUMotherboardGraphicsRAM
Intel 5820k ASUS X99-A EVGA GTX 1080 G.SKILL Value 16GB 2400 
Hard DriveCoolingOSMonitor
Samsung sm951 512GB Custom liquid Windows 10 ACER XB270HU 
KeyboardPowerCaseMouse
Corsair RGB Strafe Silent EVGA 850 G2 Caselabs S8 Razer Deathadder 
  hide details  
Reply
post #549 of 724
Quote:
Originally Posted by Brutuz View Post

I see DX12 and Vulkan being incorporated into games when DX11 dies off, that way you have a Win7/8 compatible option for those of us who aren't going to Win10. Pascals DX12 performance matters to a greater degree than Maxwell but I doubt it'd make or break the product, especially if its DX11 performance is greater than Polaris. That said if it's only typically a few % above Polaris in DX11 and still much worse in DX12 I can see people getting Polaris just for that as people tend to keep cards for longer and longer these days, all my friends who upgraded this generation have full intentions to be using their cards for 4 years or so at least.
4 years? I'm sure that's what their intentions are now, maybe not so much next year. My 780tis are a little over 2 years old. Could you imagine keeping Keplers through 2014, through 2015 and Maxwell, through 2016 and pascal, through 2017 and big pascal/little volta, waiting to 2018 and big volta? Sure, my Volvo has 300k miles, but new cars don't give me twice the driving performance on my way to work.
If they let people drive 160 mph that Volvo would be on craigslist.

Sounds like making a stop for just one or two rolleyes.gif
post #550 of 724
Quote:
Originally Posted by provost View Post

I have seen a lot of companies with "money" that has been squandered on fruitless endeavors , while they neglect the core business and its customers.

My thesis is that Nvidia's gig is up, as customers in this segment are usually about "what have you done for me lately". So, screwing them over with limited optimization support isn't gonna help make them any more stickier, particularly if AMD comes out with a compelling value proposition for this market, no matter how many hired guns come out to advocate on Nvidia's behalf.

I agree with everything else you've said, but that will only happen if they start losing fans in large numbers to make a material difference. Look around. There is a very loyal fan-base right now that has thoroughly drunk the proverbial Kool Aid, so to speak. So long as they have that base, they have no incentive to do that, unless AMD comes up with something good. You have a large base insisting right now that Nvidia's business practices are perfectly ok and that it is acceptable right now to have such short support cycles. Until they wake up in large numbers, that isn't changing.

In regards, to the compelling product, I'm hoping that will be the case with Polaris. See my response below to SSJ Eric: AMD too desperately needs to pull a very compelling product in both Zen and Polaris alike. If those fail, they are in really, really big trouble.

I'd also argue that AMD needs to pull a more compelling user experience together. Faster driver updates, a better UI for things like Crimson. Sure they have a solution with better CF scaling than SLI, and better performance over time, but they need to focus on more aggressively getting that up. It isn't easy making drivers that make the most of GCN right now, but it would arguably pay for itself.

Bottom line is that neither company is perfect. AMD though is certainly in regards to the consumer, the "lesser evil" in terms of business ethics though.

Quote:
Originally Posted by Majin SSJ Eric View Post

As I've said countless times, there is no precedence whatsoever for Nvidia to release a big-Pascal Titan first and you'd have to be naive to really believe they would do so. As Magnek said, yields are going to be crap for a long time on this new node and the profit motive for releasing a mid-range GPU as the new flagship long before releasing a true flagship GPU has been well established since the GTX 680 launched way back in 2012. They will release GP204 first as the 1080 (or whatever they call it) and we will see a GP100 Titan launch probably early next year if I had to guess. The guys who have been feverishly getting their wallets ready for a surprise April GP100 launch are simply allowing their hopes to overrun common sense and history....

It depends.

If what Mahigan is saying is true, even a 350mm^2 variant of the Polaris should be a decent improvement of >20% perhaps over the Fury X and perhaps the 980Ti (the 2 are pretty close, especially at 4k). More importantly, with HBM2; there may be more than 4GB of VRAM with HBM2, which is an important limitation to overcome. That's not a negligible improvement, even if the bigger dies add another 30% or so on top of that.

Another important consideration is that on the Green side, Nvidia's chips, if the Ashes of Singularity benchmarks are representative of DX12 performance. IN that regard, a good DX12 GPU might be a good investment. The analogy might be the HD 7970. It was the first high end 28nm GPU. Sure it was later eclipsed by the 290X and the Fury X, but if you bought one in 2012, it wasn't a bad deal at all considering the life you got out of it. If you had held your wallet out against the GTX 580 and the 6970, then bought the 7970 on release, you got a pretty good value all things considered. It still is going strong by virtue of GCN today.

But you are absolutely right that a 600mm^2 chip is improbable. TSMC has struggled mightily with yields. A couple of months ago, TSMC couldn''t even get 100mm^2 chips with acceptable yields. Asking for a 600mm^2 chip isn't happening. If Global Foundries is building the AMD chip, then one other weak point is their lack of experience with giant dies. We won't see the big dies until 2017, perhaps as late as 2018. Let's just say that these giant dies are a marvel of engineering and it takes a lot of effort behind the scenes to get them out. Plus a lot could go wrong that could lead to other delays. It's hard to say for sure. Even the best engineers won't know.
Quote:
Originally Posted by dubldwn View Post

GTX 480 launched first. With some issues. I get the profit motive on the mid-range flagship then big chip but maybe I should blame AMD for that. GTX 580 was first. 8800 GTX ( I remember that day) but I don't think either was a new node.




The GTX 480 Fermi launched in March of 2010:
http://www.anandtech.com/show/2977/nvidia-s-geforce-gtx-480-and-gtx-470-6-months-late-was-it-worth-the-wait-

Historically, AMD has been the first to launch on big dies (Nvidia actually conceded that after a previous disaster). I believe that the Fermi disaster was why Nvidia quietly adopted AMD's small die strategy.

GTX 680 > Titan
750 Ti > GTX 980 > Titan X


Cypress was more power efficient partly due to the fact that AMD had through the smaller dies understood the limitations of the 55 nm process that it was working with, while Nvidia jumped right in with a giant die. The story is a lot more complex, but it a part of the role.

To Nivida's credit, they learned for Kepler and were able to build on those previous lessons, exceeding AMD's expectations (which is why we say the 7970 GHz and the aggressively clocked 290X / Fury X).
Edited by CrazyElf - 2/25/16 at 4:50pm
Trooper Typhoon
(20 items)
 
  
CPUMotherboardGraphicsGraphics
5960X X99A Godlike MSI 1080 Ti Lightning MSI 1080 Ti Lightning 
RAMHard DriveHard DriveHard Drive
G.Skill Trident Z 32 Gb Samsung 850 Pro Samsung SM843T 960 GB Western Digital Caviar Black 2Tb 
Hard DriveOptical DriveCoolingCooling
Samsung SV843 960 GB LG WH14NS40 Cryorig R1 Ultimate 9x Gentle Typhoon 1850 rpm on case 
OSMonitorKeyboardPower
Windows 7 Pro x64 LG 27UD68 Ducky Legend with Vortex PBT Doubleshot Backlit... EVGA 1300W G2 
CaseMouseAudioOther
Cooler Master Storm Trooper Logitech G502 Proteus Asus Xonar Essence STX Lamptron Fanatic Fan Controller  
  hide details  
Reply
Trooper Typhoon
(20 items)
 
  
CPUMotherboardGraphicsGraphics
5960X X99A Godlike MSI 1080 Ti Lightning MSI 1080 Ti Lightning 
RAMHard DriveHard DriveHard Drive
G.Skill Trident Z 32 Gb Samsung 850 Pro Samsung SM843T 960 GB Western Digital Caviar Black 2Tb 
Hard DriveOptical DriveCoolingCooling
Samsung SV843 960 GB LG WH14NS40 Cryorig R1 Ultimate 9x Gentle Typhoon 1850 rpm on case 
OSMonitorKeyboardPower
Windows 7 Pro x64 LG 27UD68 Ducky Legend with Vortex PBT Doubleshot Backlit... EVGA 1300W G2 
CaseMouseAudioOther
Cooler Master Storm Trooper Logitech G502 Proteus Asus Xonar Essence STX Lamptron Fanatic Fan Controller  
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Rumors and Unconfirmed Articles
Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [TT] NVIDIA should launch its next-gen Pascal GPUs with HBM2 in 2H 2016