Overclock.net - Overclocking.net
     
 
Home Gallery Reviews Blogs Register Today's Posts Mark Forums Read Members List


Go Back   Overclock.net - Overclocking.net > Industry News > Hardware News

Reply
 
LinkBack Thread Tools
Old 04-20-08   #21 (permalink)
Ever seen a mouse pop?
 
kingsnake2's Avatar
 
intel nvidia

Join Date: Apr 2007
Location: Texas
Posts: 2,125

Folding Team Rank: 146
Team Name: Victorious Secret
Trader Rating: 4
Default

Quote:
Originally Posted by Traeumt View Post
who needs that guys
game/cgi developers who use high resolution
__________________

System: Bigger than my desk!!
CPU
Q6600 @ stock :(
Motherboard
GA-P35-DS3L Rev. 2.0
Memory
8Gb GSkill
Graphics Card
Powercooler 4870
Hard Drive
500Gb WD+640Gb WD
Sound Card
X-Fi Xtreme Audio (beats onboard)
Power Supply
650 watt OP650
Case
CM Stacker 830
CPU cooling
Tuniq Tower
GPU cooling
stock
OS
Vista Home Premium 64x
Monitor
1900x1200 24" +140x1050 20"
1 Million+ Folding at Home points
kingsnake2 is offline I fold for Overclock.net Overclocked Account   Reply With Quote
Old 04-20-08   #22 (permalink)
Networking Nut
 
metala's Avatar
 
intel ati

Join Date: Aug 2006
Location: Blagoevgrad, Bulgaria
Posts: 1,143
Blog Entries: 2

Rep: 48 metala is acknowledged by some
Unique Rep: 43
Trader Rating: 0
Default

Quote:
Originally Posted by Coma View Post
Basically, that means for irrational numbers like pi, double precision calculates up to 64 numbers after the "dot". I think gaming ATI cards already have double precision, but IIRC NVIDIA cards are still only 32bit float.
Wrong!!!
IEEE 754 structure: sign , exponent, mantissa(or fraction)

Single precision IEEE 754 is 32-bit.
1bit sign (0 positive, 1 negative)
8bits exponent (e.g. 10^3, 3 is the exponent)
23bits mantissa (e.g. 34556)
Result: 1.34556 * 10^3 = 1345.56

Double precision:
1bit sign
11bits exponent
52bits mantissa

Lets find out the maximum decimal numbers that can mantissa store.
It will be log10(2^52) = 52*log10(2) = 52 *0,30103 ~ 15,65356 ~ 15
That means that mantissa can store 15 + 1 digit numbers, i.e. 16 digits after the point ("dot").

PS. Only ATI RV670 supports double-precision computing for now.. (I compiled a script from my CAL and didn't worked on my 2600 XT)
Quote:
Originally Posted by double_matmult.cpp
// Double precision is only supported on 670 or better. Just
// indicate success if we're not on supported hardware.
if (attribs.target != CAL_TARGET_670)
{
return 0;
}
__________________
"Alcohol doesn't solve problems."
Milk doesn't solve problems either, but nobody mentions it.

Debian worshipper, uxterm is man's best friend.
My: Clubs, Photos , Photos 2 , Phlame - PHP Web Server, C Compatible Preprocessor for PHP, Web Desktop Environment
Quote:
Originally Posted by mct75 View Post
I get tired of seeing every AMD thread ragged to death by intel users saying AMD sucks.
I use an AMD processor but I respect intel users enough to not go into their threads and harass them to death.

System: Obsidian
CPU
Core 2 Duo P8400
Motherboard
Toshiba Satelite Pro A300
Memory
2 x 2 GB @ 800MHz
Graphics Card
Mobility Radeon HD 3650
Hard Drive
250GB Toshiba
Sound Card
Realtek HD
Power Supply
120W (19V 6.32A)
Case
Toshiba Satelite Pro A300
OS
Debian squeeze/sid amd64 + Vista 32bit
Monitor
15,4" TruBright 1280x800

Last edited by metala : 04-20-08 at 06:57 PM
metala is offline   Reply With Quote
Old 04-20-08   #23 (permalink)
Folding Fanatic
 
nategr8ns's Avatar
 
amd nvidia

Join Date: Dec 2007
Location: Maine
Posts: 6,196

Rep: 195 nategr8ns is acknowledged by manynategr8ns is acknowledged by many
Unique Rep: 159
Folding Team Rank: 281
Trader Rating: 3
Default

Quote:
Originally Posted by moward View Post
Coma - graphics cards have zero effect on the quality of CGI films, CGI animation is not rendered in real time, hence the quality is set in the 3D/compositing software.
and the animators

Hmm lets see here: Overtime and hard work or home-by-dinner and less polygons and triangles (meaning round things are less round, stuff is blocker). As a consumer I want harder work, but the film company probably doesn't want to have to pay for quality work, so they find a good balance between cheap designers and good quality effects.

if that makes no sense, thats ok. I don't understand it either, been outside painting for too long and I'm cold.

System: Designed for Optyvation
CPU
Opteron 175 LCB9E
Motherboard
DFI NF4 Ultra-D
Memory
2x1GB GSkill 500mhz HZs
Graphics Card
EVGA 8800GT SC 512
Hard Drive
160GB + 120GB
Power Supply
Ultra X-Finity 600w
Case
AeroCool AeroEngine II <3
CPU cooling
ACFreezer 64 Pro
GPU cooling
AC Accelero S1 w/Dual Tri-Cool 120mm (low)
OS
XP Home
Monitor
Samsung 220WM 22" + LG f-Engine 17" (dual screen)
Overclock.net - 2009 Chimp Challenge Champions
nategr8ns is offline I fold for Overclock.net   Reply With Quote
Old 04-20-08   #24 (permalink)
WaterCooler
 
intel nvidia

Join Date: Sep 2006
Location: Glendale, CA
Posts: 2,559

Rep: 123 tweakboy is acknowledged by manytweakboy is acknowledged by many
Unique Rep: 112
Trader Rating: 0
Lightbulb

These cards are for peepz using 3D rendering like Studio Max or Maya , It's a graphic workstation, not a graphics game card. The price is soo high on them is because the cards allow you to do things related to animation or rendering or autocad which the gamer cards can't do, or do a poor job of,

There's soo many different stuff involved in say SMax like these guys have to make animation for movies and blah blah and theres like thousands of acyrnms or I mean ,,, graphic terms ,,, and it uses OpenGL ,,,,,, good thread!!!
__________________
System: Born 11/21/07 DAW & Memorex External DVD RAM
CPU
Intel Core 2 Q6600 GO @ 3760 Mhz@1.504v 0.0 vdrop
Motherboard
ASUS P5K @ 1880 Mhz FSB @1.7v Pencil Voltmod 470x8
Memory
G.SKILL PC8000 8GB 2GBx4@ 940 Mhz@2.10v@5-5-4-12
Graphics Card
EVGA 8800 GT SC @740/1850/1085 Mhz@1.1v
Hard Drive
Tt iCage SATA WDC 320GB & 120GB Barracuda 80GB
Sound Card
ASUS Xonar D2/PM & M-AUDIO BX8a Deluxe Monitors
Power Supply
Tt ToughPower Modularized Cable Managment@700w
Case
Black Full Tt Big WaterSE Tt P500 Pump & TMG2 Rad
CPU cooling
Tt Copper WaterBlock
GPU cooling
EVGA Stock
OS
Windows 7 Ultimate x64 RTM 7600
Monitor
24" Sony FW900@ 2304x1440@80hz/23"Dell W2300 HDTV
tweakboy is offline   Reply With Quote
Old 04-20-08   #25 (permalink)
4.0ghz
 
Coma's Avatar
 
intel nvidia

Join Date: Jun 2007
Posts: 7,472

Rep: 455 Coma is a proven memberComa is a proven memberComa is a proven memberComa is a proven memberComa is a proven member
Unique Rep: 328
Trader Rating: 0
Default

Quote:
Originally Posted by moward View Post
Coma - graphics cards have zero effect on the quality of CGI films, CGI animation is not rendered in real time, hence the quality is set in the 3D/compositing software.
When it comes to texture resolutions and stuff like that, yeah... but not when it comes to AA and AF. Gaming cards are limited in how high they can go.

The real difference is in resolutions, and slower algorithms used to render (higher AA/AF settings, and more - the way things are calculated.

I don't know exactly how things go on but I'm assuming based on other things, such as encoding video.

x264's ME (motion estimating) Subpixel Refinement has 7 modes - I have 2 profiles: one that uses mode 5, Multi QPel and one that uses mode 7, RDO Level 2.

The profile that uses Multi QPel is almost 3 times faster than the RDO profile, because RDO is more precise, which is exactly the same as here.

Going by this, it's reasonable to say that the premium you pay for workstation cards is the extra RAM... and drivers that are slower and perform more precise calculations.
__________________
System: Akiyama Mio
CPU
E6420 @ stock, 0.98v
Motherboard
Asus P5N-E SLI
Memory
2x1GB OCZ Platinum @ 800MHz 4-4-4-12 1T, 1.9v
Graphics Card
BFG 8800GT 512MB
Hard Drive
WD 250GB, 320GB SATA/3, 16MB Cache
Power Supply
Corsair 520HX
Case
NZXT Apollo Black
CPU cooling
Stock
OS
Ubuntu 9.04 x86 & XP x86
Monitor
Asus VW222U
Coma is offline Overclocked Account   Reply With Quote
Old 04-20-08   #26 (permalink)
4.0ghz
 
Coma's Avatar
 
intel nvidia

Join Date: Jun 2007
Posts: 7,472

Rep: 455 Coma is a proven memberComa is a proven memberComa is a proven memberComa is a proven memberComa is a proven member
Unique Rep: 328
Trader Rating: 0
Default

Quote:
Originally Posted by metala View Post
Wrong!!!
I was going based on the assumption that "double" of 32 is 64.
In any case, I'm pretty sure ATI said before that their FireStream 9170 supports 64bit precision calculations, but my memory has failed me before. If you're bored enough, fish the thread and check.
__________________
System: Akiyama Mio
CPU
E6420 @ stock, 0.98v
Motherboard
Asus P5N-E SLI
Memory
2x1GB OCZ Platinum @ 800MHz 4-4-4-12 1T, 1.9v
Graphics Card
BFG 8800GT 512MB
Hard Drive
WD 250GB, 320GB SATA/3, 16MB Cache
Power Supply
Corsair 520HX
Case
NZXT Apollo Black
CPU cooling
Stock
OS
Ubuntu 9.04 x86 & XP x86
Monitor
Asus VW222U
Coma is offline Overclocked Account   Reply With Quote
Old 04-20-08   #27 (permalink)
Overclocker
 
DuDeInThEmOoN42's Avatar
 
intel ati

Join Date: Dec 2007
Location: Minnesota
Posts: 458

Rep: 11 DuDeInThEmOoN42 Unknown
Unique Rep: 10
Trader Rating: 0
Default

Quote:
Originally Posted by metala View Post
Wrong!!!
IEEE 754 structure: sign , exponent, mantissa(or fraction)

Single precision IEEE 754 is 32-bit.
1bit sign (0 positive, 1 negative)
8bits exponent (e.g. 10^3, 3 is the exponent)
23bits mantissa (e.g. 34556)
Result: 1.34556 * 10^3 = 1345.56

Double precision:
1bit sign
11bits exponent
52bits mantissa

Lets find out the maximum decimal numbers that can mantissa store.
It will be log10(2^52) = 52*log10(2) = 52 *0,30103 ~ 15,65356 ~ 15
That means that mantissa can store 15 + 1 digit numbers, i.e. 16 digits after the point ("dot").

PS. Only ATI RV670 supports double-precision computing for now.. (I compiled a script from my CAL and didn't worked on my 2600 XT)
I don't know what you're saying, but I agree.
__________________
When in doubt, overclock

Quote:
Originally Posted by pauldovi View Post
Nvidia...

"The Way it............. BSOD
CPU-Z Validation

GPU-Z Validation

System: LED Zeppelin
CPU
Q6600 G0 @ 3.6GHz - 1.4v
Motherboard
Intel DX38BT
Memory
4GB Corsair DDR3 1333MHz
Graphics Card
VisionTek 4870 X2
Hard Drive
Seagate Barracuda 500GB 32MB Cache 7200 RPM
Sound Card
Onboard 7.1 IDT
Power Supply
750W Xigmatek +4 12V Rails
Case
Apevia X-Jupiter G Type
CPU cooling
Thermalright TRUE 120 w/ 2x Ultra Kaze
GPU cooling
Stock - 50% Fan Speed
OS
Vista Home Premium x64
Monitor
Samsung 906CW 19" LCD
DuDeInThEmOoN42 is offline   Reply With Quote
Old 04-21-08   #28 (permalink)
With great difficulty
 
rabidgnome229's Avatar
 
intel nvidia

Join Date: Feb 2006
Location: Pittsburgh
Posts: 5,210

Rep: 614 rabidgnome229 is becoming famousrabidgnome229 is becoming famousrabidgnome229 is becoming famousrabidgnome229 is becoming famousrabidgnome229 is becoming famousrabidgnome229 is becoming famous
Unique Rep: 370
FAQs Submitted: 6
Trader Rating: 5
Default

Quote:
Originally Posted by Coma View Post
Ihatethedukes: The proof is in any CGI movie you see. Ever see aliasing in them?
That's because they use these cards, which are pretty much the same except they use much higher antialiasing/anisotropic filtering, and some even offer double precision (64bit) float calculation.

Basically, that means for irrational numbers like pi, double precision calculates up to 64 numbers after the "dot". I think gaming ATI cards already have double precision, but IIRC NVIDIA cards are still only 32bit float.
Your post was very good, but I just want to jump in here and say that that is not quite how floating point works. The 64 bits of a double precision floating point number are broken into 1 sign bit, 11 exponent bits, and 52 fraction bits. The number is in the form +1.[frac]x2^[exp] for normalized values, where frac and exp are numbers derived from the fraction and exponent bits. This yields 52 places, but they are binary digits rather than decimal digits. This makes the set of representable values a bit counterintuitive. Also, a 64 bit float stores 52 digits, but they are not necessarily the 52 places after the decimal point.



Wikipedia has a good explanation of how floating point numbers are represented for anybody interested. It's relatively simple but complicated enough to make for a bad post
__________________
System: It goes to eleven
CPU
E6300
Motherboard
DS3
Memory
2GB XMS2 DDR2-800
Graphics Card
EVGA 8600GTS
Hard Drive
1.294 TB
Sound Card
Audigy 2 ZS
Power Supply
Corsair 520HX
Case
Lian-Li v1000B Plus
CPU cooling
TTBT
GPU cooling
Thermalright V2
OS
Arch Linux/XP
Monitor
Samsung 226bw

Last edited by rabidgnome229 : 04-21-08 at 12:51 AM
rabidgnome229 is offline Overclocked Account   Reply With Quote
Old 04-21-08   #29 (permalink)
4.0ghz
 
Coma's Avatar
 
intel nvidia

Join Date: Jun 2007
Posts: 7,472

Rep: 455 Coma is a proven memberComa is a proven memberComa is a proven memberComa is a proven memberComa is a proven member
Unique Rep: 328
Trader Rating: 0
Default

So, I'm just going to link to your post instead of trying to explain it myself D:
__________________
System: Akiyama Mio
CPU
E6420 @ stock, 0.98v
Motherboard
Asus P5N-E SLI
Memory
2x1GB OCZ Platinum @ 800MHz 4-4-4-12 1T, 1.9v
Graphics Card
BFG 8800GT 512MB
Hard Drive
WD 250GB, 320GB SATA/3, 16MB Cache
Power Supply
Corsair 520HX
Case
NZXT Apollo Black
CPU cooling
Stock
OS
Ubuntu 9.04 x86 & XP x86
Monitor
Asus VW222U
Coma is offline Overclocked Account   Reply With Quote
Reply


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools



All times are GMT -5. The time now is 08:10 AM.


Overclock.net is a Carbon Neutral Site Creative Commons License

Terms of Service / Forum Rules | Privacy Policy | DMCA Info | Advertising | Become an Official Vendor
Copyright © 2009 Shogun Interactive Development. Most rights reserved.
Page generated in 0.16555 seconds with 8 queries