Overclock.net › Forums › Graphics Cards › NVIDIA › Fermi Tessellation Performance Comparison
New Posts  All Forums:Forum Nav:

Fermi Tessellation Performance Comparison

post #1 of 105
Thread Starter 
So, in another thread, I was linked to an AnandTech review of the 400 series cards which found some interesting results when it came to tessellation performance. According to their tests:

Quote:
Originally Posted by Anandtech
"With the DirectX 11 Detail Tessellation sample program, we’re primarily looking at whether we can throw a high enough tessellation load at the GPU to overwhelm its tessellation abilities and bring it to its knees. In this case we cannot, as the GTX 460 scales from tessellation factor 7 to tessellation factor 11 by basically the same rate as the GTX 480 and GTX 465. "
http://www.anandtech.com/show/3809/n...he-200-king/16

They also found similar results when reviewing the GTS 450. The 450 has 4 dedicated tessellation units and the 460 has 7. Meanwhile the 456 has 11, the 470 has 14 and the 480 has 15 units respectively. Based on this, it appears that whatever performance benefit there is in having extra units, it doesn't scale linearly.

So I figured since we have so many Fermi users floating around here, we could figure out just how much of a difference there really is in tessellation performance.

Using Heaven 2.5, since that seems to be the heaviest tessellation-based bench available atm, I benched with tess disabled, on normal and then on extreme (all other settings were left at their defaults) and compared the results to figure out the average performance penalty.

I used a single, stock clocked 560 Ti @1080p as a reference w/ForceWare 275.50 (the 560 Ti has 8 dedicated tess units).
Tess Off: Average FPS = 50.1
Tess Normal: Average FPS = 39.6
Tess Extreme: Average FPS = 29.6

1 - (39.6/50.1) = 0.2096
Meaning going from no tessellation to "normal" incurred a ~21% performance penalty.
1 - (29.6/39.6) = 0.2525
Further, going from normal to extreme incurred a relative 25% performance penalty.
1 - (29.6/50.1) = 0.4091
And the overall performance penalty from no tessellation to extreme was ~40.9%.

I'd like to compare my results with other Fermi chips, GF100s/104s/106s/110s etc to find out how performance scales relative to tessellation units and to find out if the gap is truly tangible.

So for anyone interested, please post up 1080/1200p Heaven 2.5 benches at each of the 3 tessellation setting to compare. Clock speeds shouldn't matter in theory, so long as they are constant throughout all 3 of your runs.

For reference, the number of tessellation units per card are as follows:
  1. GTX 580 = 16 Units
  2. GTX 570/GTX 480 = 15 Units
  3. GTX 470 = 14 Units
  4. GTX 465 = 11 Units
  5. GTX 560 Ti = 8 Units
  6. GTX 560/GTX 460 = 7 Units
  7. GTS 450/GTX 550 Ti = 4 Units



Edited by Booty Warrior - 7/31/11 at 2:32pm
post #2 of 105
when I get on my computer later, I will bench and let you know how the 580 does.
post #3 of 105
Thread Starter 
Quote:
Originally Posted by thrasherht View Post
when I get on my computer later, I will bench and let you know how the 580 does.
Awesome
post #4 of 105
what about SLI, go you want single card or SLI.
Main Rig
(15 items)
 
Work Rig
(9 items)
 
 
CPUMotherboardGraphicsRAM
5820k Asus X99 Deluxe MSI GTX980 Ti Gaming/EK GPU Block Corsair 16GB 
Hard DriveCoolingOSMonitor
M4 128GB,C300 128GB, F3 1TB  480mm RAD, 655/top, 250mm RES, EK LX CPU Block windows 7 PRO 64bit Asus PB278Q 
PowerCase
Enermax Platimax 1500W Cosmos S 
CPUMotherboardGraphicsRAM
Core i7 2600k ASUS P8Z77-V MSI GT 610 1GB Corsair 1333mhz 16GB 
Hard DriveCoolingOSPower
Intel 330 60GB Cooler Master Hyper 212 EVO W7 Corsair CX500M 
Case
Cooler Master Elite 334U 
  hide details  
Reply
Main Rig
(15 items)
 
Work Rig
(9 items)
 
 
CPUMotherboardGraphicsRAM
5820k Asus X99 Deluxe MSI GTX980 Ti Gaming/EK GPU Block Corsair 16GB 
Hard DriveCoolingOSMonitor
M4 128GB,C300 128GB, F3 1TB  480mm RAD, 655/top, 250mm RES, EK LX CPU Block windows 7 PRO 64bit Asus PB278Q 
PowerCase
Enermax Platimax 1500W Cosmos S 
CPUMotherboardGraphicsRAM
Core i7 2600k ASUS P8Z77-V MSI GT 610 1GB Corsair 1333mhz 16GB 
Hard DriveCoolingOSPower
Intel 330 60GB Cooler Master Hyper 212 EVO W7 Corsair CX500M 
Case
Cooler Master Elite 334U 
  hide details  
Reply
post #5 of 105
Just gotta install and do it, and I'll post up my results.
post #6 of 105
Thread Starter 
Quote:
Originally Posted by driftingforlife View Post
what about SLI, go you want single card or SLI.
Hmm, in theory it shouldn't matter either way, but just to eliminate any extra variables lets go with single cards, for now.
post #7 of 105
I just installed Heaven last night for some stability testing. I'll post up screens in a bit for ya.
post #8 of 105
Quote:
Originally Posted by Booty Warrior View Post
Hmm, in theory it shouldn't matter either way, but just to eliminate any extra variables lets go with single cards, for now.
ok, will do it tomoz.
Main Rig
(15 items)
 
Work Rig
(9 items)
 
 
CPUMotherboardGraphicsRAM
5820k Asus X99 Deluxe MSI GTX980 Ti Gaming/EK GPU Block Corsair 16GB 
Hard DriveCoolingOSMonitor
M4 128GB,C300 128GB, F3 1TB  480mm RAD, 655/top, 250mm RES, EK LX CPU Block windows 7 PRO 64bit Asus PB278Q 
PowerCase
Enermax Platimax 1500W Cosmos S 
CPUMotherboardGraphicsRAM
Core i7 2600k ASUS P8Z77-V MSI GT 610 1GB Corsair 1333mhz 16GB 
Hard DriveCoolingOSPower
Intel 330 60GB Cooler Master Hyper 212 EVO W7 Corsair CX500M 
Case
Cooler Master Elite 334U 
  hide details  
Reply
Main Rig
(15 items)
 
Work Rig
(9 items)
 
 
CPUMotherboardGraphicsRAM
5820k Asus X99 Deluxe MSI GTX980 Ti Gaming/EK GPU Block Corsair 16GB 
Hard DriveCoolingOSMonitor
M4 128GB,C300 128GB, F3 1TB  480mm RAD, 655/top, 250mm RES, EK LX CPU Block windows 7 PRO 64bit Asus PB278Q 
PowerCase
Enermax Platimax 1500W Cosmos S 
CPUMotherboardGraphicsRAM
Core i7 2600k ASUS P8Z77-V MSI GT 610 1GB Corsair 1333mhz 16GB 
Hard DriveCoolingOSPower
Intel 330 60GB Cooler Master Hyper 212 EVO W7 Corsair CX500M 
Case
Cooler Master Elite 334U 
  hide details  
Reply
post #9 of 105
Hit up google spreadsheet on the op so we can store and see all the data in one place.
2500k
(15 items)
 
  
CPUMotherboardGraphicsRAM
2500k @ 4.6 1.27V Asus p67 Pro SLI 560 TI Ripjaws X 1600 2x4GB 
Hard DriveCoolingOSMonitor
Crucial C300 + Spinpoint F3 Noctua N-D14 Win 7 x64 LG 1920x1080 
KeyboardPowerCaseMouse
Das Professional NZXT 750w Hale90 Antec 600 M60 
Audio
ATH AD700 
  hide details  
Reply
2500k
(15 items)
 
  
CPUMotherboardGraphicsRAM
2500k @ 4.6 1.27V Asus p67 Pro SLI 560 TI Ripjaws X 1600 2x4GB 
Hard DriveCoolingOSMonitor
Crucial C300 + Spinpoint F3 Noctua N-D14 Win 7 x64 LG 1920x1080 
KeyboardPowerCaseMouse
Das Professional NZXT 750w Hale90 Antec 600 M60 
Audio
ATH AD700 
  hide details  
Reply
post #10 of 105
Disabled AA and AF for a wider FPS range. (Hadn't noticed default settings in OP, hope it's alright anyway)


New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: NVIDIA
Overclock.net › Forums › Graphics Cards › NVIDIA › Fermi Tessellation Performance Comparison