Originally Posted by Advice Trinity by Knoxville
And the article is currently being slashdotted.
Some git called Kez submitted it, damn him.
PHP Code:$s = new signature();
$s->sarcasm()->intellect()->font('Courier New')->display();
Flog him!
(\__/)
(='.'=)
(")_(")
bin him !
AMD Athlon 4400X2 @ 2.565PenisextentionMhz
Dual Layer, Gold Plated, LED Power,Dual Golden OMG IT MAKES MY CodPiece BIGGER 1-1-1-1 DDR62.3 @ 1222.3433Mhz
5 X 400GB Porn Array
X1800XT Dildo enchanged 3D Version, 512MegaLongJohn
Oh, did I mention.....I like sheep.....
WWW.MrsBurley.CO.UK now updated
Eat him!
Home cinema: Toshiba 42XV555DB Full HD LCD | Onkyo TX-SR705 | NAD C352 | Monitor Audio Bronze B2 | Monitor Audio Bronze C | Monitor Audio Bronze BFX | Yamaha NSC120 | BK Monolith sub | Toshiba HD-EP35 HD-DVD | Samsung BD-P1400 BluRay Player | Pioneer DV-575 | Squeezebox3 | Virgin Media V+ Box
PC: Asus P5B | Core2duo 2.13GHz | 2GB DDR2 PC6400 | Inno3d iChill 7900GS | Auzentech X-Plosion 7.1 | 250GB | 500GB | NEC DVDRW | Dual AG Neovo 19"
HTPC: | Core2Duo E6420 2.13GHz | 2GB DDR2 | 250GBx2 | Radeon X1300 | Terratec Aureon 7.1 | Windows MCE 2005
Laptop: 1.5GHz Centrino | 512MB | 60GB | 15" Wide TFT | Wifi | DVDRW
dont forget the tomato sauce
AMD Athlon 4400X2 @ 2.565PenisextentionMhz
Dual Layer, Gold Plated, LED Power,Dual Golden OMG IT MAKES MY CodPiece BIGGER 1-1-1-1 DDR62.3 @ 1222.3433Mhz
5 X 400GB Porn Array
X1800XT Dildo enchanged 3D Version, 512MegaLongJohn
Oh, did I mention.....I like sheep.....
WWW.MrsBurley.CO.UK now updated
vey, vey interestion...
from a rendering perspective in terms of 3d modelling for broadcast - split frame is generally faster - by orders of magnitude, provided the photon tracing is calculated in advance (due to differences in calculations depending on architecture - i.e. pIII different to PIV, Xeon and PIV Xeon... and of course Athlons for the render farm) can reduce your aggregate per frame render time from 36 hours +, to about 18-20 depending on the number of additional processors that can be assigned to a job. i.e. each processor takes 16th of the screen area (if not smaller) and calculates it's own partial render accordingly - then the manager simply stitches them back together as full res. targas.
you end up gaining an average of 90% speed increase. Only when you've got some very complicated photon tracing do you find that it slows down considerably - but it's still much, much faster.
On the other hand, if you're rendering alternate frames - you've got the added complication of photon tracing and bump-mapping offsets - where this still needs to be calculated accordingly. It's great seeing this applied, as the SLI interpretations we've played with from SGI have been particularly amusing (when you've got 128 GPU's on one board - and you only see a 30% speed increase due to it rendering alternate frames)
great article tho
There are currently 1 users browsing this thread. (0 members and 1 guests)
Posting Permissions
LinkBack URL
About LinkBacks
Reply With Quote