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

Fermi Tessellation Performance Comparison - Page 4

post #31 of 105
Thread Starter 
Quote:
Originally Posted by thrasherht View Post
op, you should build a google spreadsheet, include the username, video card, stock vs OC, all FPS numbers, and then the percentage of drop.
Ugh, too lazy. TBH I wasn't expecting so many samples, but if there is a LOT of interest I'll probably have to.

Quote:
Originally Posted by tx-jose View Post
well since I have a GTX460 guess my results aren't necessary
Post em! This is to compare all Fermi cards!

Quote:
Originally Posted by ACHILEE5 View Post
I'm back
*snip*

Then ran in Sli to see Sli scaling
*snip*
Hmm, so 31.97% single performance loss and 31% in SLI for the 480. I guess that confirms that SLI doesn't skew the results.

Quote:
Originally Posted by KonigGeist View Post
I have a 580 and a couple of 460s that I can run in a little while if they're needed.

Here is my 580:
*snip*

It looks like a 34% loss.
Indeed. What drivers are you using? It seems odd that the 580s are underperforming compared to the 470/570 here.

Quote:
Originally Posted by Tex1954 View Post
Of course your results are wanted!

I have two systems, one has 460, other has 560 Ti. I'll post my results in this post when done.

GTX560 Ti 950MHz, single card.
*snip*

GTX460 794MHz, single card.

*snip*

Funny thing, the 1090T system with the 460's is headless and I normally use logmein to talk to it... but put a real 1080p monitor on it to bench and it showed the logmein driver... LOL! But, I did not use Logmein to get the data... hooked up the kybd/mouse/monitor to it to run...


LOL, I was amazed at these results at first, but then I saw your first run is using moderate tessellation instead of having it disabled. Please post a run with tess disabled altogether for an apples to apples comparison.

Out of curiosity which drivers are you using?

We need more results in the interest of science!
Edited by Booty Warrior - 7/31/11 at 11:40pm
post #32 of 105
There might be a bigger difference if you add AA to the test. Fermi cards use CUDA cores for tessellation. If you add AA to it with tessellation, the cards with more CUDA cores may perform better than cards with less CUDA cores.
post #33 of 105
Thread Starter 
Quote:
Originally Posted by Riou View Post
There might be a bigger difference if you add AA to the test. Fermi cards use CUDA cores for tessellation. If you add AA to it with tessellation, the cards with more CUDA cores may perform better than cards with less CUDA cores.
Possibly, though I imagine that will actually give tighter results since AA/AF will drag down the non tess numbers substantially.
post #34 of 105
Another thing that popped into my head was that while the 580 is quite a bit faster than say a 470 clock for clock it doesn't have a lot more tess units.

Perhaps because the tess units themselves are the bottlenecks here the higher end gtx 580s lose more performance because without tess they have more performance to give, but with tess enabled the render power isn't quite as balanced as the slower cards with nearly as many tess units..

Though still clearly faster Does that make sense?

I would like to see a 580 do a sub 900Mhz core run so we can compare the percent difference in performance without tess and with extreme tess to see if that actually plays out as I think it will (clock for clock).

Edit (again): I went off Konig's post with his 580 - though I don't know what clocks he used.

Anyways in the no tess bench he was 10.5% faster, in the extreme tess bench he was 8.5% faster. So I would assume because the 580 doesn't have as much "more" tess power as it does render power compared to a 470 the performance hit the 580s takes will be greater than what the slower 470s/570s/480s can achieve due to the 470 having less render power but only 2 less tess units and the 570/480 actually being closer both in render power and tess unit count.
Edited by BallaTheFeared - 8/1/11 at 12:14am
    
CPUMotherboardGraphicsGraphics
Intel Core i5 2500K P8P67 PRO NVIDIA GeForce GTX 470 NVIDIA GeForce GTX 470 
GraphicsRAMRAMRAM
NVIDIA GeForce 9800 GT G-Skill A-Data G-Skill 
RAMHard DriveOptical DriveOS
A-Data Crucial M4 64GB + 1TB F3 Spinpoint $155 LS/DL DVD RW $?? Windows 8 64-bit "Epic Registry" Edition 
MonitorPowerCase
ASUS 21.5 1920x1080 2ms $135 CORSAIR HX850 $120 Mother Earth $free 
  hide details  
Reply
    
CPUMotherboardGraphicsGraphics
Intel Core i5 2500K P8P67 PRO NVIDIA GeForce GTX 470 NVIDIA GeForce GTX 470 
GraphicsRAMRAMRAM
NVIDIA GeForce 9800 GT G-Skill A-Data G-Skill 
RAMHard DriveOptical DriveOS
A-Data Crucial M4 64GB + 1TB F3 Spinpoint $155 LS/DL DVD RW $?? Windows 8 64-bit "Epic Registry" Edition 
MonitorPowerCase
ASUS 21.5 1920x1080 2ms $135 CORSAIR HX850 $120 Mother Earth $free 
  hide details  
Reply
post #35 of 105
Here's results of my GeForce GTX 560 Ti. It's an EVGA FBP, so it's relatively lower clocks (850MHz core, can't remember the other two). It's in order from tessellation off, to normal, to extreme.





Quote:
Originally Posted by Booty Warrior View Post
Out of curiosity which drivers are you using?
They are using 266.66, which are the first drivers to support the GeForce GTX 560 Ti (and support only that card). You can see what drivers are being used in the screenshots.
post #36 of 105
I'll try to get some screenshots after I get off work
KhAoS v1
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5 2500K @ 4.7ghz-1.392v ASUS P8P67 EVO MSI GTX 570 Twin Frozr II G.Skill Ripjaws X 4gb 1600 
Hard DriveOptical DriveOSMonitor
1x WD 1tb, 1x OCZ Vertex 2 50gb HP 24x DVD/CD Burner Windows 7 Home Premium 64 bit 23" ASUS 1080p 
PowerCaseMouse
Corsair 850TX HAF 932 CM Storm Sentinel 
  hide details  
Reply
KhAoS v1
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5 2500K @ 4.7ghz-1.392v ASUS P8P67 EVO MSI GTX 570 Twin Frozr II G.Skill Ripjaws X 4gb 1600 
Hard DriveOptical DriveOSMonitor
1x WD 1tb, 1x OCZ Vertex 2 50gb HP 24x DVD/CD Burner Windows 7 Home Premium 64 bit 23" ASUS 1080p 
PowerCaseMouse
Corsair 850TX HAF 932 CM Storm Sentinel 
  hide details  
Reply
post #37 of 105
Thread Starter 
Quote:
Originally Posted by BallaTheFeared View Post
Another thing that popped into my head was that while the 580 is quite a bit faster than say a 470 clock for clock it doesn't have a lot more tess units.

Perhaps because the tess units themselves are the bottlenecks here the higher end gtx 580s lose more performance because without tess they have more performance to give, but with tess enabled the render power isn't quite as balanced as the slower cards with nearly as many tess units..

Though still clearly faster Does that make sense?
Hmm, I see what you're getting at. But then, the 580 and 480 aren't architecturally much different.

Both have 48 ROPs. The 580 has its last SM unlocked which nets it 4 more texture units than the 480, and the extra CUDA cores/tess engine, but other than that, the only thing really separating them are higher clocks on the 580 (similar to the comparison between the 560 ti and the 460).

If I'm not mistaken, looking at those two, it doesn't seem like the ratio of rendering power to tessellation performance would be out of balance.

From what we have here though the 480 takes a ~ 31% performance hit while the 580 still takes over 34%.

Quote:
Originally Posted by Princess Garnet View Post
Here's results of my GeForce GTX 560 Ti. It's an EVGA FBP, so it's relatively lower clocks (850MHz core, can't remember the other two). It's in order from tessellation off, to normal, to extreme.
Hmm your results are pretty much in line with mine, though you took slightly less of a performance hit going from none to extreme (~39.3%). I wonder if clock speeds do play more of a role in this than I thought.
post #38 of 105
Here ya go, man.

Everything on defaults, 1080p resolution, SLI 470's at 775/1790. Comparison between 'disabled' and 'extreme'.

On my rig, the % loss going from Disabled to Extreme arrived at via your maf systems is:

1 - (64.1/93.7) = 31.59%



Edited by brettjv - 8/1/11 at 2:12am
    
CPUMotherboardGraphicsRAM
xeon X5675 6-core @ 4.1ghz (1.29v, 20x205 +ht ) rampage iii extreme msi rx470 gaming X (the $159 budget king) 3 x 2gb corsair xms3 pc12800 (9-9-9-24-1T@1600MHz) 
Hard DriveOptical DriveCoolingOS
hynix 250gb ssd (boot), 2tb deskstar (apps),1tb... plextor px-712sa - still the best optical drive... corsair h8o v2 aio W10 home 
MonitorPowerCaseAudio
asus vw266h 25.5" (1920x1200) abs sl (enermax revolution) * single 70A rail 850w silverstone rv-03 XFi Titanium 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
xeon X5675 6-core @ 4.1ghz (1.29v, 20x205 +ht ) rampage iii extreme msi rx470 gaming X (the $159 budget king) 3 x 2gb corsair xms3 pc12800 (9-9-9-24-1T@1600MHz) 
Hard DriveOptical DriveCoolingOS
hynix 250gb ssd (boot), 2tb deskstar (apps),1tb... plextor px-712sa - still the best optical drive... corsair h8o v2 aio W10 home 
MonitorPowerCaseAudio
asus vw266h 25.5" (1920x1200) abs sl (enermax revolution) * single 70A rail 850w silverstone rv-03 XFi Titanium 
  hide details  
Reply
post #39 of 105
Thread Starter 
So averaging the results thus far (from no tessellation to extreme):
  1. GTX 580 (GF110) 16 Tess Units = ~35% performance loss...
  2. GTX 480 (GF100) / GTX 570 (GF110) 15 Tess Units = ~31% performance loss
  3. GTX 470 (GF100) 14 Tess Units = ~32% performance loss
  4. GTX 465 (GF100) 11 Tess Units = ?
  5. GTX 560 Ti (GF114) 8 Tess Units = ~40% performance loss
  6. GTX 460 (GF104) / GTX 560 (GF114) 7 Tess Units = ~41% performance loss (thanks tex!)
  7. GTS 450 (GF106) / GTX 550 Ti (GF116) 4 Tess Units = ?

Need some 460s/465s to represent! And more 580s please.

Also Brett, what's your take on the 580 results? Balla's explanation (ie. the difference in rendering power is greater than the difference in tessellation power) seems plausible, but then ACHILEE5's 480 results seem to contradict this.
Edited by Booty Warrior - 8/1/11 at 8:17am
post #40 of 105
Quote:
Originally Posted by Booty Warrior View Post
Hmm your results are pretty much in line with mine, though you took slightly less of a performance hit going from none to extreme (~39.3%). I wonder if clock speeds do play more of a role in this than I thought.
Well, remember, I said I have the FPB version which is only 850MHz core, and everything is stock, so my clocks are among the lowest for a GeForce GTX 560 Ti. Looking up your card, I see it actually uses the nVidia reference clocks, so it's slightly lower at 822MHz. If your cards are at stock speed, that may be it.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: NVIDIA
Overclock.net › Forums › Graphics Cards › NVIDIA › Fermi Tessellation Performance Comparison