Overclock.net › Forums › Industry News › Hardware News › [dvHW] NVIDIA: CUDA to deliver 100x performance gains
New Posts  All Forums:Forum Nav:

[dvHW] NVIDIA: CUDA to deliver 100x performance gains

post #1 of 19
Thread Starter 
Quote:
Speaking at The Future of 3D Graphics in San Francisco last week, NVIDIA chief scientist David Kirk promised NVIDIA's GPU-based parallel computing will be able to boost the performance of real-world applications by a hundredfold:
Kirk, delivering a chalk talk on "The Future of 3D Graphics' in San Francisco Friday, touted the advantages of GPU-based parallel computing for powering applications related to oil and gas exploration, computational finance and other computational modeling projects, as well as for faster, more powerful hybrid rendering within the graphics discipline itself. Because GPU computing is already a "data parallel process," the work of breaking apart computing problems into smaller sets of instructions to be carried out concurrently is more easily done on GPUs than on multi-core CPUs, Kirk said.

Describing a kind of Moore's Law on steroids, he promised 100x performance gains in real-world applications, just as soon as people take advantage of Nvidia's General-Purpose computing on GPUs (GPGPU) initiatives that have resulted in some 50 million Nvidia GPUs already shipped that are capable of running the CUDA programming language for parallel computing.

"This is truly the democratization of supercomputing. We ship a million parallel units a week," Kirk said.

CUDA or Compute Unified Device Architecture, is a C programming language developed by Santa Clara, Calif.-based Nvidia that allows GPGPU programmers to code algorithms for execution on graphics processors. Currently, it's possible to run CUDA on Nvidia's GeForce desktop chipsets, as well as its Quadro workstation and Tesla high-performance compute products, and according to Kirk the graphics chipmaker recently released an SDK for the Macintosh operating system
Source
post #2 of 19
Good stuff, but I hope they have the power to back those statements up.
Rig
(13 items)
 
  
CPUMotherboardGraphicsRAM
Q6600 G0 Gigabyte GA-EP45-UD3P Sapphire 5870 Cypress XT EF6 2GB 4GB (2x2GB) G.SKILL 1066 
Hard DriveOptical DriveOSMonitor
32GB Corsair SSD + 2x 320 + 1x 250 Lightscribe drive Windows 7 Ultimate Hanns-G HG-216DPO 
KeyboardPowerCaseMouse
Logitech G15 Gaming Keyboard OCZ GameXStream 850W HAF 932 Logitech G5 Laser Gaming Mouse 
Mouse Pad
Allsop 
  hide details  
Reply
Rig
(13 items)
 
  
CPUMotherboardGraphicsRAM
Q6600 G0 Gigabyte GA-EP45-UD3P Sapphire 5870 Cypress XT EF6 2GB 4GB (2x2GB) G.SKILL 1066 
Hard DriveOptical DriveOSMonitor
32GB Corsair SSD + 2x 320 + 1x 250 Lightscribe drive Windows 7 Ultimate Hanns-G HG-216DPO 
KeyboardPowerCaseMouse
Logitech G15 Gaming Keyboard OCZ GameXStream 850W HAF 932 Logitech G5 Laser Gaming Mouse 
Mouse Pad
Allsop 
  hide details  
Reply
post #3 of 19
They probably exaggerated, a little bit.
    
CPUMotherboardGraphicsRAM
Intel Core i7 5820k @4.5GHZ Asus Rampage V Extreme  EVGA 980ti classified 64GB Corsair Vengeance 3200MHZ 
Hard DriveOptical DriveCoolingOS
[Samsung 950 Pro 512GB][2X Samsung 840 250GB Ra... 16X LITESCRIDE DVD-RW EK Double Res, Double Pump, GPU, CPU combo Windows 10 x64 
MonitorKeyboardPowerCase
55IN 4K  Ducky Shine Corsair RM 1000i Modular STH10 
Mouse
Corsair m95 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Intel Core i7 5820k @4.5GHZ Asus Rampage V Extreme  EVGA 980ti classified 64GB Corsair Vengeance 3200MHZ 
Hard DriveOptical DriveCoolingOS
[Samsung 950 Pro 512GB][2X Samsung 840 250GB Ra... 16X LITESCRIDE DVD-RW EK Double Res, Double Pump, GPU, CPU combo Windows 10 x64 
MonitorKeyboardPowerCase
55IN 4K  Ducky Shine Corsair RM 1000i Modular STH10 
Mouse
Corsair m95 
  hide details  
Reply
post #4 of 19
So they're developing the Cuda language for faster processing power or am I understanding this wrong?
    
CPUMotherboardGraphicsRAM
Intel C2Q 9550 @ 4.00 Ghz Gigabyte Ga EP45-UD3P XFX Radeon HD4890 TeamXtreemDDR2 800(2GB)+GSkill PI-B DDR2 800(4GB) 
Hard DriveOptical DriveOSMonitor
WD 320GB SATA,WD 640GB SATA,WD 80GB IDE,WD 250GB LG DVD-RW;LG Blu-ray/DVD-RW Windows 7 (x64) Pro Samsung SyncMaster T240 
PowerCase
Corsair 750W Antec 900 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Intel C2Q 9550 @ 4.00 Ghz Gigabyte Ga EP45-UD3P XFX Radeon HD4890 TeamXtreemDDR2 800(2GB)+GSkill PI-B DDR2 800(4GB) 
Hard DriveOptical DriveOSMonitor
WD 320GB SATA,WD 640GB SATA,WD 80GB IDE,WD 250GB LG DVD-RW;LG Blu-ray/DVD-RW Windows 7 (x64) Pro Samsung SyncMaster T240 
PowerCase
Corsair 750W Antec 900 
  hide details  
Reply
post #5 of 19
less talking and more producing sounds good though, with all that power i'll probably be able to heat a small meal on it
post #6 of 19
They added a chip on all 8 series cards and above, and it decodes the CUDA and runs it on the core. Very powerful stuff.

From what I have seen in stories, 100x isn't that exaggerated, sometimes (via an array of cards) it is an understatement.
    
CPUMotherboardGraphicsRAM
C2D T7100 1.8 ghz (undervolted) ummm... Dell Intel X3100 2 x 1gb 667mhz 
Hard DriveOptical DriveOSMonitor
Fujitsu 7200 RPM 120gb CD-RW/DVD dual boot Vista business 1440x900 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
C2D T7100 1.8 ghz (undervolted) ummm... Dell Intel X3100 2 x 1gb 667mhz 
Hard DriveOptical DriveOSMonitor
Fujitsu 7200 RPM 120gb CD-RW/DVD dual boot Vista business 1440x900 
  hide details  
Reply
post #7 of 19
What applications take CUDA into consideration?

Do certain games' physics use CUDA to process them or are they just touting theoretical numbers and things that we will never use nor see?

I have never really understood the whole "CUDA thing".
MY XFX l0ng Pwn3r
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core 2 Q6600 @ 4.05Ghz Lapped XFX 680i LT 2 x 8800GT 775core/2000mem 4Gb Patriot Viper 
Hard DriveOptical DriveOSMonitor
1 x 400Gb, 1 x 320Gb, 1 x 160Gb 2 x LG DVD-RW Windows Vista Ultimate 64bit Samsung 226bw 
KeyboardPowerCaseMouse
Ideazon Z-Board Ultra X Pro 750 Ultra Razer DeathAdder 
Mouse Pad
AttackPAD Pro 
  hide details  
Reply
MY XFX l0ng Pwn3r
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core 2 Q6600 @ 4.05Ghz Lapped XFX 680i LT 2 x 8800GT 775core/2000mem 4Gb Patriot Viper 
Hard DriveOptical DriveOSMonitor
1 x 400Gb, 1 x 320Gb, 1 x 160Gb 2 x LG DVD-RW Windows Vista Ultimate 64bit Samsung 226bw 
KeyboardPowerCaseMouse
Ideazon Z-Board Ultra X Pro 750 Ultra Razer DeathAdder 
Mouse Pad
AttackPAD Pro 
  hide details  
Reply
post #8 of 19
So can I use it for anything?
BoS Terminal
(15 items)
 
  
CPUMotherboardGraphicsRAM
i5 4440 ASRock Z87E-ITX MSI OC GTX 760 4GB x 2 G.Skill RipJaw 
Hard DriveCoolingOSMonitor
Crucial 240GB SSD Gelid Silent Fans Windows 8.1 55" LED Samsung 3D 
KeyboardPowerCaseMouse
Logitech K400 Corsiar RM650 CM Elite 130 m ITX Logitech G700S 
Audio
Logitech G930 or 5.1 Home Theater Surround 
  hide details  
Reply
BoS Terminal
(15 items)
 
  
CPUMotherboardGraphicsRAM
i5 4440 ASRock Z87E-ITX MSI OC GTX 760 4GB x 2 G.Skill RipJaw 
Hard DriveCoolingOSMonitor
Crucial 240GB SSD Gelid Silent Fans Windows 8.1 55" LED Samsung 3D 
KeyboardPowerCaseMouse
Logitech K400 Corsiar RM650 CM Elite 130 m ITX Logitech G700S 
Audio
Logitech G930 or 5.1 Home Theater Surround 
  hide details  
Reply
post #9 of 19
Quote:
Originally Posted by dualhYbrid View Post
So can I use it for anything?
Precisely.

The only issue is that you have to be able to cut the task up hundreds of times. Programs like Superpi cannot be done in multiple sections, they can only be done one step at a time, so it would not benefit from GPGPU.

However the one story I am thinking of is a medical scan, it normally takes something like 3-5 days, however implementing Tesla (a special server of GPU's) and of course implementing CUDA code (you cannot just use the regular code) they cut the time down to several minutes.

Currently I don't think any applications we will be playing with use CUDA yet, however you can download demo's of some very impressive physics stuff off Nvidia's website.

I was hoping folding at home would start using CUDA, however it seems their aided by ATI, and are focusing entirely on ATI development while claiming they won't use CUDA because it would lock them into one brand.
    
CPUMotherboardGraphicsRAM
C2D T7100 1.8 ghz (undervolted) ummm... Dell Intel X3100 2 x 1gb 667mhz 
Hard DriveOptical DriveOSMonitor
Fujitsu 7200 RPM 120gb CD-RW/DVD dual boot Vista business 1440x900 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
C2D T7100 1.8 ghz (undervolted) ummm... Dell Intel X3100 2 x 1gb 667mhz 
Hard DriveOptical DriveOSMonitor
Fujitsu 7200 RPM 120gb CD-RW/DVD dual boot Vista business 1440x900 
  hide details  
Reply
post #10 of 19
CUDA theoretically offers very high performance, perhaps maybe even 100 fold. However realistically a CUDA program must be highly optimized to even be faster at all. CUDA programming is difficult, picky, and only works on a small range of applications.

Furthermore, CUDA cannot handle high precision data types, like the double.

Quote:
Originally Posted by trueg50 View Post
Precisely.

The only issue is that you have to be able to cut the task up hundreds of times. Programs like Superpi cannot be done in multiple sections, they can only be done one step at a time, so it would not benefit from GPGPU.

However the one story I am thinking of is a medical scan, it normally takes something like 3-5 days, however implementing Tesla (a special server of GPU's) and of course implementing CUDA code (you cannot just use the regular code) they cut the time down to several minutes.

Currently I don't think any applications we will be playing with use CUDA yet, however you can download demo's of some very impressive physics stuff off Nvidia's website.

I was hoping folding at home would start using CUDA, however it seems their aided by ATI, and are focusing entirely on ATI development while claiming they won't use CUDA because it would lock them into one brand.
CUDA would work very well on calculating PI. The calculation of PI requires finding the sum of many Taylor Series for e^x, ln(x), and exponential(x). This could benefit greatly from parallel computing.
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [dvHW] NVIDIA: CUDA to deliver 100x performance gains