Go Back   nV News Forums > Software Forums > Gaming Central

Newegg Daily Deals

Reply
 
Thread Tools
Old 10-25-05, 04:33 AM   #49
Mr. Nice
Registered User
 
Join Date: Jul 2005
Location: UK
Posts: 43
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

So, will 3mood be getting the better offset trick as well then?
Mr. Nice is offline   Reply With Quote
Old 10-25-05, 04:53 AM   #50
tertsi
www.fornvidiots.net
 
Join Date: May 2003
Posts: 340
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

Quote:
Originally Posted by Mr. Nice
So, will 3mood be getting the better offset trick as well then?
Shortly - Yes.
tertsi is offline   Reply With Quote
Old 10-25-05, 08:50 AM   #51
pliflier
Registered User
 
Join Date: Jul 2004
Posts: 54
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

Quote:
Originally Posted by tertsi
Some fragment program (aka pixel shader) performance analysis with Normal Offset Mapping Super Ultra ....

-------------------- G70 --------------------
Target: GeForce 7800 GTX (G70) :: Unified Compiler: v77.72
Cycles: 38.00 :: R Regs Used: 4 :: R Regs Max Index (0 based): 3
Pixel throughput (assuming 1 cycle texture lookup) 271.58 MP/s

-------------------- NV40 --------------------
Target: GeForce 6800 Ultra (NV40) :: Unified Compiler: v77.72
Cycles: 41.00 :: R Regs Used: 4 :: R Regs Max Index (0 based): 3
Pixel throughput (assuming 1 cycle texture lookup) 156.10 MP/s

-------------------- NV43-GT --------------------
Target: GeForce 6600 GT (NV43-GT) :: Unified Compiler: v77.72
Cycles: 41.00 :: R Regs Used: 4 :: R Regs Max Index (0 based): 3
Pixel throughput (assuming 1 cycle texture lookup) 97.56 MP/s

-------------------- NV40-12 --------------------
Target: GeForce 6800 (NV40-12) :: Unified Compiler: v77.72
Cycles: 41.00 :: R Regs Used: 4 :: R Regs Max Index (0 based): 3
Pixel throughput (assuming 1 cycle texture lookup) 95.12 MP/s

... and same fragment program without Normal Offset Mapping Super Ultra ....

-------------------- G70 --------------------
Target: GeForce 7800 GTX (G70) :: Unified Compiler: v77.72
Cycles: 20.00 :: R Regs Used: 4 :: R Regs Max Index (0 based): 3
Pixel throughput (assuming 1 cycle texture lookup) 516.00 MP/s

-------------------- NV40 --------------------
Target: GeForce 6800 Ultra (NV40) :: Unified Compiler: v77.72
Cycles: 24.00 :: R Regs Used: 4 :: R Regs Max Index (0 based): 3
Pixel throughput (assuming 1 cycle texture lookup) 266.67 MP/s

-------------------- NV43-GT --------------------
Target: GeForce 6600 GT (NV43-GT) :: Unified Compiler: v77.72
Cycles: 24.00 :: R Regs Used: 4 :: R Regs Max Index (0 based): 3
Pixel throughput (assuming 1 cycle texture lookup) 166.67 MP/s

-------------------- NV40-12 --------------------
Target: GeForce 6800 (NV40-12) :: Unified Compiler: v77.72
Cycles: 24.00 :: R Regs Used: 4 :: R Regs Max Index (0 based): 3
Pixel throughput (assuming 1 cycle texture lookup) 162.50 MP/s
Does those numbers indicate that enable Super Ultra offset mapping will decrease performance by around 40%?
pliflier is offline   Reply With Quote
Old 10-25-05, 11:52 AM   #52
Mr. Nice
Registered User
 
Join Date: Jul 2005
Location: UK
Posts: 43
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

Quote:
Originally Posted by pliflier
Does those numbers indicate that enable Super Ultra offset mapping will decrease performance by around 40%?
Sort of, for a start (AFK) those MP/s figures only take into account shader calculation time, and ignores memory latency, or simply lack of memory bandwidth (note the "1 cycle texture lookup" assumption, which is only true if the memory read request goes through far enough in advance for the data to be "on hand" when its needed (in CPU terms, this means the data is in the L1-cache, I don't know the exact details when it comes to GPU memory architecture)).
Also, they are only measures for the fragment in interaction.vxp, which only deals with lighting & texturing (both diffuse and specular), not everything else the GPU has to do (not least sorting out light/shadow casting, ie, which surfaces each light will fall on).
Both these factors mean that that "40%" decrease in practical terms is an upper bound for the possible impact on performance. Also note that is by comparison with no offset mapping at all (I think). I would like to see similar figures comparing it to the current "ultra" offset mapping.
Mr. Nice is offline   Reply With Quote
Old 10-25-05, 02:33 PM   #53
tertsi
www.fornvidiots.net
 
Join Date: May 2003
Posts: 340
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

yes, both offset mappings (high and ultra) uses the same method for offset BUT ultra mode also reconstruct the z vector (5-6 instructions).

3mood (all maxed) - no offset mapping
-------------------- G70 --------------------
Target: GeForce 7800 GTX (G70) :: Unified Compiler: v77.72
Cycles: 15.00 :: R Regs Used: 3 :: R Regs Max Index (0 based): 2
Pixel throughput (assuming 1 cycle texture lookup) 688.00 MP/s

3mood (all maxed) - normal offset map (high)
-------------------- G70 --------------------
Target: GeForce 7800 GTX (G70) :: Unified Compiler: v77.72
Cycles: 17.00 :: R Regs Used: 4 :: R Regs Max Index (0 based): 3
Pixel throughput (assuming 1 cycle texture lookup) 607.06 MP/s

3mood (all maxed) - normal offset mapping (ultra)
-------------------- G70 --------------------
Target: GeForce 7800 GTX (G70) :: Unified Compiler: v77.72
Cycles: 21.00 :: R Regs Used: 4 :: R Regs Max Index (0 based): 3
Pixel throughput (assuming 1 cycle texture lookup) 491.43 MP/s

3mood (all maxed) - normal offset mapping (super ultra)
-------------------- G70 --------------------
Target: GeForce 7800 GTX (G70) :: Unified Compiler: v77.72
Cycles: 37.25 :: R Regs Used: 5 :: R Regs Max Index (0 based): 4
Pixel throughput (assuming 1 cycle texture lookup) 278.92 MP/s

3mood (all maxed) - relief mapping (NOTE! borked - texture coords didn't work with relief !!!!!!
-------------------- G70 --------------------
Target: GeForce 7800 GTX (G70) :: Unified Compiler: v77.72
Cycles: 55.50 :: R Regs Used: 6 :: R Regs Max Index (0 based): 5
Pixel throughput (assuming 1 cycle texture lookup) 187.64 MP/s

3mood (all maxed) - relief mapping
-------------------- G70 --------------------
Target: GeForce 7800 GTX (G70) :: Unified Compiler: v77.72
Cycles: 357.14 :: R Regs Used: 18 :: R Regs Max Index (0 based): 17
Pixel throughput (assuming 1 cycle texture lookup) 28.91 MP/s
tertsi is offline   Reply With Quote
Old 10-25-05, 05:22 PM   #54
tertsi
www.fornvidiots.net
 
Join Date: May 2003
Posts: 340
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

Doom 3 offset mapping (super ultra) and relief mapping benches with single 7800GTX (HQ + 10x7 + AA 2x/AF 8x) ...

timedemo demo1
52.8 fps - offset mapping (super ultra)
14.2 fps - relief mapping
tertsi is offline   Reply With Quote
Old 10-25-05, 05:57 PM   #55
Mr. Nice
Registered User
 
Join Date: Jul 2005
Location: UK
Posts: 43
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

Quote:
Originally Posted by tertsi
52.8 fps - offset mapping (super ultra)
14.2 fps - relief mapping
. Me thinks that issues with distributing the needed height map textures are the least of relief mappings problems then .
Mr. Nice is offline   Reply With Quote
Old 10-26-05, 04:24 AM   #56
Skinner
Registered User
 
Skinner's Avatar
 
Join Date: Aug 2002
Location: Rotterdam, Netherlands
Posts: 1,128
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

Any timeframe for this update? Keep up the good work
__________________
*Intel i7@3,8gHz*Asus P6T Deluxe* 2x Sapphire HD7970 3GB Crossfire *6 GB Corsair Dominator 1600C8*OCZ Vertex 120 GB SSD*adaptec 19160U160*Intel X25 80 GB Quantum Atlas 15KII SCSI U160 147Gig* WD Raptor 300 GB*Apple 24" LED Cinema *X-Fi Titanium*Logitech Z5500*Coolmaster RP 1000W*W7 64 Home *
Skinner is offline   Reply With Quote

Old 10-26-05, 10:41 AM   #57
saweetnesstrev
Registered User
 
Join Date: Jan 2005
Posts: 568
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

tertsi, you have configs for 5800 fx and 5600 fx but wheres 5700 ultra fx? I have a 5700 ultra what should i do? Can you make me my own config tyvm
saweetnesstrev is offline   Reply With Quote
Old 10-26-05, 10:44 AM   #58
ssj4goku887
Xtreme G Author
 
Join Date: Jul 2004
Location: New Jersey, USA
Posts: 507
Send a message via AIM to ssj4goku887 Send a message via MSN to ssj4goku887
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

im sorry to be the bearer of bad news but the performance settings DO decrease image quality... especialy fp16 and fp12. Im not trying to flame or anything your work is greatly appreciated and you take the time to do this which is great but i felt that you had to know... in quake 4 a 7800 would be using fp32....
__________________
Tweaks R Us - Home of the Xtreme G Drivers
ssj4goku887 is offline   Reply With Quote
Old 10-26-05, 11:46 AM   #59
Razor3DC
Registered User
 
Join Date: Oct 2003
Location: Germany
Posts: 30
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

Quote:
Originally Posted by tertsi
timedemo demo1
Hi tertsi,

what about this "demo1"?
Is this an official timedemo?

Per default it doesn't exist...

greets

Razor
Razor3DC is offline   Reply With Quote
Old 10-26-05, 11:48 AM   #60
Mr. Nice
Registered User
 
Join Date: Jul 2005
Location: UK
Posts: 43
Default Re: Quak4 - Quake 4 tuning tool for NVIDIA G70/NV40/NV30 Series GPU's

Quote:
Originally Posted by ssj4goku887
im sorry to be the bearer of bad news but the performance settings DO decrease image quality... especialy fp16 and fp12.
Yeah your right, I said as much months back in the 3mood thread about fp16. On my 6800GT the performance advantage was quite negligible as well for this option.
As a point of order though, it's not fp12, that's fx12.

On another couple of the settings, both doom3 and quake4 actually use precision hint fastest in the standard interaction.vxp already, and quake 4 by default already does what the "swizzle" option does (which is mask the final result by xyz, ie skips writing out the redundant w channel). Indeed, quake 4's interaction.vxp is replete with quite aggressive write masks (as opposed to doom3's, which contains none). Finally, quake 4 already does the specular with maths, not a texture lookup, and indeed it does it with one less instruction then quak4/3mood . Quak4/3mood subtract's 0.75, squares it, then multiplies by 16, 3 instructions. Quake 4 uses a single MAD instruction to multiple by 4 then subtract 3, and then squares it, which is mathematically equivalent (multiplying by 4 before squaring is the same as multiplying by 16 afterwards, since 4 squared is 16, and similarly 3 is subtracted instead of 0.75 since 4x0.75=3).
Mr. Nice is offline   Reply With Quote
Reply


Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


All times are GMT -5. The time now is 07:14 AM.


Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.