Overclock.net › Forums › Industry News › Hardware News › [MT] CEO Will Announce Big GPU News On May 15: NVIDIA
New Posts  All Forums:Forum Nav:

[MT] CEO Will Announce Big GPU News On May 15: NVIDIA - Page 10

post #91 of 142
Quote:
Originally Posted by prava View Post

Certainly, but forget about it being released in a Geforce product. They want it to be big and probably exclusive for gpgpu, as they are introducing it into a compute conference (GTC).

They've released similar GPUs as GeForce cards before. Actually afaik they've never had a tesla exclusive product and it wouldn't even make sense to have one since they're going to end up with chips that are too leaky etc. for GPGPU use so it makes sense to release those as GeForce.

Also with around 7 billion transistors the big kepler chip would have a similar die size as the previous NV flagship GPUs like GF100, GF110, GT200, GT200b, etc.

I just can't see them releasing a tesla-only product.
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
Reply
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
Reply
post #92 of 142
Quote:
Originally Posted by prava View Post

Certainly, but forget about it being released in a Geforce product. They want it to be big and probably exclusive for gpgpu, as they are introducing it into a compute conference (GTC).

So we'd still be waiting for the GeForce version?
post #93 of 142
Quote:
Originally Posted by Mhill2029 View Post

If GK110 does indeed come out (here's hoping) it will sit between the GTX680 and GTX690 in terms of performance. I believe the 700 Series will be reserved for Maxwell most likely.
GK110 will be the top dog in terms of Single GPU performance, the GTX690 will keep the overall top dog crown of the Kepler line up.

Assuming GK110 (GTX 685?) is coming out soon, when is Maxwell (780???) coming out? When the AMD 8000 series comes out in December? It can't be, cause this is all seems so rushed and short lived:

March - GTX 680

August? - GTX 685

December/early 2013 - GTX 780 to compete with the 8790?
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
[Intel Core i7] [*T] [Intel iGPU] [*] 
Hard DriveOSMonitorKeyboard
[*] [Windows 7 Professional 64-bit] [Samsung SyncMaster P2770FH 27" 1080P] [Logitech MK320 cordless] 
PowerCaseMouse
[Corsair AX1200] [Corsair Obsidian 800D] [Logitech MK300 cordless] 
  hide details  
Reply
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
[Intel Core i7] [*T] [Intel iGPU] [*] 
Hard DriveOSMonitorKeyboard
[*] [Windows 7 Professional 64-bit] [Samsung SyncMaster P2770FH 27" 1080P] [Logitech MK320 cordless] 
PowerCaseMouse
[Corsair AX1200] [Corsair Obsidian 800D] [Logitech MK300 cordless] 
  hide details  
Reply
post #94 of 142
Quote:
Originally Posted by Alatar View Post

They've released similar GPUs as GeForce cards before. Actually afaik they've never had a tesla exclusive product and it wouldn't even make sense to have one since they're going to end up with chips that are too leaky etc. for GPGPU use so it makes sense to release those as GeForce.
Also with around 7 billion transistors the big kepler chip would have a similar die size as the previous NV flagship GPUs like GF100, GF110, GT200, GT200b, etc.
I just can't see them releasing a tesla-only product.

Sure they've also released similar GPUs as Geforce cards before, but...

a) They are introducing a Tesla card in may 15th
b) Such card will be released by the end of 2012
c) They've just released GTX690
d) GK104 performs like crap in GPGPU

http://blogs.nvidia.com/2012/05/nvidia-ceo-headline-gtc-2012-keynotes/

So, do you still believe that the chip that they will be introducing will actually launch at Geforce cards? Keep in mind that we have never seen a situation like this, and that the professional cards that NVIDIA owns features a HUGE gross margin (look at the raw numbers that compare 2011 vs 2012, in fiscal years). Altogether, with the fact that the professional segment is growing at a very nice pace, makes me wonder if the chip they will show will ever launch as a Geforce card because it could be designed with only GPGPU in mind...and that wouldn't be a bad idea at all when you realise that using different chips for different targets, if they are large enough, brings a lot of efficiency to the table (as it is, GK104 is super efficient at raster tasks...but is crappy at gpgpu ones...).
My System
(16 items)
 
  
CPUMotherboardGraphicsGraphics
i5 2500k MSI P67 GD65 GTX460 1GB GTX460 1GB 
RAMHard DriveCoolingOS
G. Skill Ripjaws Crucial C300 Prolimatech Megahalems + 2 x GT AP15 Win 7 x64 
MonitorKeyboardPowerCase
Dell U2311H Noppoo Choc mini Corsair HX620 Microcool Banchetto rev. 2.0 black aluminium 
MouseMouse PadAudio
Zowie EC1 Zowie Fiio E7 + E9 / Sennheiser HD 25 1 II 
  hide details  
Reply
My System
(16 items)
 
  
CPUMotherboardGraphicsGraphics
i5 2500k MSI P67 GD65 GTX460 1GB GTX460 1GB 
RAMHard DriveCoolingOS
G. Skill Ripjaws Crucial C300 Prolimatech Megahalems + 2 x GT AP15 Win 7 x64 
MonitorKeyboardPowerCase
Dell U2311H Noppoo Choc mini Corsair HX620 Microcool Banchetto rev. 2.0 black aluminium 
MouseMouse PadAudio
Zowie EC1 Zowie Fiio E7 + E9 / Sennheiser HD 25 1 II 
  hide details  
Reply
post #95 of 142
Quote:
Originally Posted by prava View Post

Sure they've also released similar GPUs as Geforce cards before, but...
a) They are introducing a Tesla card in may 15th
b) Such card will be released by the end of 2012
c) They've just released GTX690
d) GK104 performs like crap in GPGPU
http://blogs.nvidia.com/2012/05/nvidia-ceo-headline-gtc-2012-keynotes/
So, do you still believe that the chip that they will be introducing will actually launch at Geforce cards? Keep in mind that we have never seen a situation like this, and that the professional cards that NVIDIA owns features a HUGE gross margin (look at the raw numbers that compare 2011 vs 2012, in fiscal years). Altogether, with the fact that the professional segment is growing at a very nice pace, makes me wonder if the chip they will show will ever launch as a Geforce card because it could be designed with only GPGPU in mind...and that wouldn't be a bad idea at all when you realise that using different chips for different targets, if they are large enough, brings a lot of efficiency to the table (as it is, GK104 is super efficient at raster tasks...but is crappy at gpgpu ones...).

But you forget the fact that they absolutely MUST strike a nice balance between the two for their Quadro offerings. They will not let their professional graphics cards suffer in either area, so they really can't design a compute only chip unless they design 3 different high end chips for each generation, one for gaming, one for general workstation use, and one for HPC supercomputing. A highly unlikely situation.

What is more likely is that GK110 will be much like GF100/110 in that the chip will be much more compute focused, with hardware scheduler and other such compute oriented things, and kinda brute force it's way through graphical tasks. It won't be elegant as GK104 for GPU tasks, but it will be mighty powerful.
Poison Apple
(13 items)
 
The one Free-man
(15 items)
 
 
CPUMotherboardGraphicsRAM
Core i7 3820 @4.65GHz and climbing! ASRock X79 Extreme 6 Gigabyte GTX480 1.5GB @902/1804/1848 32GB Mushkin Blackline DDR3 1600 
Hard DriveCoolingCoolingCooling
60GB Mushkin Chronos SSD Watercool Heatkiller 3.0 EK FC 480 GTX XSPC RX240 
CoolingOSMonitorPower
Swiftech MCR320 Windows 7 Enterprise 3x Samsung SyncMaster 225BW PC Power and Cooling Silencer 950W 
Case
Apple PowerMac G5 ATX conversion 
CPUMotherboardGraphicsRAM
Core i7 860 4.3ghz 1.44v Asus P7P55D PRO Gigabyte GTX480 1.5GB 16gb G.skill Ripjaws DDR3 1333 
Hard DriveHard DriveOptical DriveCooling
Mushkin Chronos 60GB SSD Seagate Barracuda 1TB 7200RPM HP Lightscribe DVD-RW Xigmatek Dark Knight 
OSMonitorPowerCase
WINDOWS 7 ENTERPRISE ASUS VH222H 21.5", 2X SAMSUNG SyncMaster 225bw 22" Rosewill Performance series 600W Lian Li PC-7B 
Mouse
Microsoft Wireless Intellimouse Explorer 2.0 
  hide details  
Reply
Poison Apple
(13 items)
 
The one Free-man
(15 items)
 
 
CPUMotherboardGraphicsRAM
Core i7 3820 @4.65GHz and climbing! ASRock X79 Extreme 6 Gigabyte GTX480 1.5GB @902/1804/1848 32GB Mushkin Blackline DDR3 1600 
Hard DriveCoolingCoolingCooling
60GB Mushkin Chronos SSD Watercool Heatkiller 3.0 EK FC 480 GTX XSPC RX240 
CoolingOSMonitorPower
Swiftech MCR320 Windows 7 Enterprise 3x Samsung SyncMaster 225BW PC Power and Cooling Silencer 950W 
Case
Apple PowerMac G5 ATX conversion 
CPUMotherboardGraphicsRAM
Core i7 860 4.3ghz 1.44v Asus P7P55D PRO Gigabyte GTX480 1.5GB 16gb G.skill Ripjaws DDR3 1333 
Hard DriveHard DriveOptical DriveCooling
Mushkin Chronos 60GB SSD Seagate Barracuda 1TB 7200RPM HP Lightscribe DVD-RW Xigmatek Dark Knight 
OSMonitorPowerCase
WINDOWS 7 ENTERPRISE ASUS VH222H 21.5", 2X SAMSUNG SyncMaster 225bw 22" Rosewill Performance series 600W Lian Li PC-7B 
Mouse
Microsoft Wireless Intellimouse Explorer 2.0 
  hide details  
Reply
post #96 of 142
Quote:
Originally Posted by XPC View Post

But you forget the fact that they absolutely MUST strike a nice balance between the two for their Quadro offerings. They will not let their professional graphics cards suffer in either area, so they really can't design a compute only chip unless they design 3 different high end chips for each generation, one for gaming, one for general workstation use, and one for HPC supercomputing. A highly unlikely situation.
What is more likely is that GK110 will be much like GF100/110 in that the chip will be much more compute focused, with hardware scheduler and other such compute oriented things, and kinda brute force it's way through graphical tasks. It won't be elegant as GK104 for GPU tasks, but it will be mighty powerful.

Firstly, the Geforce range are designed for graphically intensive tasks such as gaming. They are not designed with professional GPGPU applications in mind.

Secondly, QuadroFX cards are not, and will not, ever be intended for gaming/graphical uses. They are workstation cards. They are intended for GPGPU applications such as render farms and for use in professional applications of CUDA accelerated software such as Photoshop.

Thirdly, Tesla cards are intended for ultra high-end workstations and primarily, for supercomputer applications.

So your argument is totally invalid. I have never heard of anyone who buys a QuadroFX card then turns around and says "I wish gaming performance was better!". Why? Because when you buy a QuadroFX card, you know what its designed for, workstation applications. Likewise, you want to game, get a Geforce card. You want to build a supercomputer, go get some Tesla's.

Nvidia don't need to make multi-discipline cards. It appears from what we have seen from Kepler so far that Nvidia have decided to target the specific applications with all the performance in the key areas, cutting out the crap that isn't needed. Hence why the GTX680 is great for gaming but sucks for GPGPU applications. Its a gaming card, not a workstation card. Just like Kepler based QuadroFX cards will be GPGPU centric with minimal gaming performance. Why? Because you don't buy a QuadroFX card for gaming, so why cut back GPGPU performance to give the card performance in an area where it is not designed to be used? Quite simply, you wouldn't.
Current Rig
(10 items)
 
  
CPUMotherboardGraphicsRAM
i5 2500k MSI Z77A-GD65 Sapphire RX 480 Nitro+ OC G.Skill Ripjaw X 2133MHz 
Hard DriveCoolingOSPower
OCZ Vertex 2 Antec Kuhler 620 Windows 10 Silverstone ST1200 
Case
Corsair 200R 
  hide details  
Reply
Current Rig
(10 items)
 
  
CPUMotherboardGraphicsRAM
i5 2500k MSI Z77A-GD65 Sapphire RX 480 Nitro+ OC G.Skill Ripjaw X 2133MHz 
Hard DriveCoolingOSPower
OCZ Vertex 2 Antec Kuhler 620 Windows 10 Silverstone ST1200 
Case
Corsair 200R 
  hide details  
Reply
post #97 of 142
Quote:
Originally Posted by Vengeance47 View Post

Firstly, the Geforce range are designed for graphically intensive tasks such as gaming. They are not designed with professional GPGPU applications in mind.
Secondly, QuadroFX cards are not, and will not, ever be intended for gaming/graphical uses. They are workstation cards. They are intended for GPGPU applications such as render farms and for use in professional applications of CUDA accelerated software such as Photoshop.
Thirdly, Tesla cards are intended for ultra high-end workstations and primarily, for supercomputer applications.
So your argument is totally invalid. I have never heard of anyone who buys a QuadroFX card then turns around and says "I wish gaming performance was better!". Why? Because when you buy a QuadroFX card, you know what its designed for, workstation applications. Likewise, you want to game, get a Geforce card. You want to build a supercomputer, go get some Tesla's.
Nvidia don't need to make multi-discipline cards. It appears from what we have seen from Kepler so far that Nvidia have decided to target the specific applications with all the performance in the key areas, cutting out the crap that isn't needed. Hence why the GTX680 is great for gaming but sucks for GPGPU applications. Its a gaming card, not a workstation card. Just like Kepler based QuadroFX cards will be GPGPU centric with minimal gaming performance. Why? Because you don't buy a QuadroFX card for gaming, so why cut back GPGPU performance to give the card performance in an area where it is not designed to be used? Quite simply, you wouldn't.

Here's the thing. Do you think Nvidia would be dumb enough to lose market share to AMD for the people that want to game and also do Folding@Home, considering NVida is the fastest at F@H? I don't think so. There will be a Geforce card that is good at both gaming and GPGPU. It won't game as good as GK104 or compute as good as Quadro, but it will exist. Maybe not on May15th, but end of this year or early next you bet ya.
post #98 of 142
Quote:
Originally Posted by i7monkey View Post

Assuming GK110 (GTX 685?) is coming out soon, when is Maxwell (780???) coming out? When the AMD 8000 series comes out in December? It can't be, cause this is all seems so rushed and short lived:
March - GTX 680
August? - GTX 685
December/early 2013 - GTX 780 to compete with the 8790?

Maxwell is probably mid to late 2013, unless they totally trash the consumer version of Big Kepler and are pouring resources into Maxwell now.
post #99 of 142
Quote:
Originally Posted by sticks435 View Post

Here's the thing. Do you think Nvidia would be dumb enough to lose market share to AMD for the people that want to game and also do Folding@Home, considering NVida is the fastest at F@H? I don't think so. There will be a Geforce card that is good at both gaming and GPGPU. It won't game as good as GK104 or compute as good as Quadro, but it will exist. Maybe not on May15th, but end of this year or early next you bet ya.

In the scheme of things, F@H will have little impact on Nvidia's bottom line. Heck, even here on OCN the amount of people who actually fold would be what, less than 5%? This is on an enthusiast website too. And enthusiasts are very much in the minority, we account for only a tiny fraction of the broader community that game on PC's.

All the money is in low power, portable applications like Tegra and in ultra high-end computing where they make huge margins (ie. Tesla). Remember, Nvidia and AMD don't have a big profit margin on their desktop GPU's. Compare the profit margin from a $500 GTX680 to the profit margin on a $100,000 Tesla card......no comparison mate

I honestly don't think Nvidia could care less about losing some people changing to AMD because their F@H score is lower on a Kepler GPU. Especially if it means that their QuadroFX and Tesla cards are that extra 5% better. That extra 5% could make them several hundreds of millions of dollars in revenue from corporate clients. So the impact felt by a few people switching due to F@H performance would be like a drop in the ocean compared to the gains in corporate sector revnue.
Current Rig
(10 items)
 
  
CPUMotherboardGraphicsRAM
i5 2500k MSI Z77A-GD65 Sapphire RX 480 Nitro+ OC G.Skill Ripjaw X 2133MHz 
Hard DriveCoolingOSPower
OCZ Vertex 2 Antec Kuhler 620 Windows 10 Silverstone ST1200 
Case
Corsair 200R 
  hide details  
Reply
Current Rig
(10 items)
 
  
CPUMotherboardGraphicsRAM
i5 2500k MSI Z77A-GD65 Sapphire RX 480 Nitro+ OC G.Skill Ripjaw X 2133MHz 
Hard DriveCoolingOSPower
OCZ Vertex 2 Antec Kuhler 620 Windows 10 Silverstone ST1200 
Case
Corsair 200R 
  hide details  
Reply
post #100 of 142
Quote:
Originally Posted by Vengeance47 View Post

Warning: Spoiler! (Click to show)
In the scheme of things, F@H will have little impact on Nvidia's bottom line. Heck, even here on OCN the amount of people who actually fold would be what, less than 5%? This is on an enthusiast website too. And enthusiasts are very much in the minority, we account for only a tiny fraction of the broader community that game on PC's.
All the money is in low power, portable applications like Tegra and in ultra high-end computing where they make huge margins (ie. Tesla). Remember, Nvidia and AMD don't have a big profit margin on their desktop GPU's. Compare the profit margin from a $500 GTX680 to the profit margin on a $100,000 Tesla card......no comparison mate
I honestly don't think Nvidia could care less about losing some people changing to AMD because their F@H score is lower on a Kepler GPU. Especially if it means that their QuadroFX and Tesla cards are that extra 5% better. That extra 5% could make them several hundreds of millions of dollars in revenue from corporate clients. So the impact felt by a few people switching due to F@H performance would be like a drop in the ocean compared to the gains in corporate sector revnue.

Good points. I dunno, I still see NV having to much pride and dare we say arrogance to let it slide. Guess we'll know soon enough. lol
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [MT] CEO Will Announce Big GPU News On May 15: NVIDIA