Overclock.net › Forums › Industry News › Hardware News › [Toms] Micro-Stuttering And GPU Scaling In CrossFire And SLI
New Posts  All Forums:Forum Nav:

[Toms] Micro-Stuttering And GPU Scaling In CrossFire And SLI - Page 2

post #11 of 177
read? just base your opinion on the reactions of fellow OCN'ers.
Shhh!
(14 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k Asus Maximus IV Gene-Z MSI GTX1080 Gaming X Corsair XMS3 2GB x 4 
Hard DriveCoolingMonitorPower
Crucial M4 256GB Vertex 4 512GB 1 TB Black Caviar Noctua NH-D14 Achieva Shimian 27" 2560x1440 Kingwin Lazer Platinum LZP-550 550W  
Case
Silverstone SG09 
  hide details  
Reply
Shhh!
(14 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k Asus Maximus IV Gene-Z MSI GTX1080 Gaming X Corsair XMS3 2GB x 4 
Hard DriveCoolingMonitorPower
Crucial M4 256GB Vertex 4 512GB 1 TB Black Caviar Noctua NH-D14 Achieva Shimian 27" 2560x1440 Kingwin Lazer Platinum LZP-550 550W  
Case
Silverstone SG09 
  hide details  
Reply
post #12 of 177
That was a good read. I was thinking about CrossFire on my card but I think I will pass, I need 3 cards to reduce it or go nvidia for SLI for reduced micro-stutter.

I wish they tested 3 and 4 way crossfire and sli on the high end cards.
post #13 of 177
Enjoyed reading it.

Never would have thought that at 3x microstuttering is not only reduced but almost nonexistent. Such a surprise.
post #14 of 177
Crazy. Still, I agree. Why not use a more traditional "trifire" package? E.g. the 69-- series or the 58-- series, as the lack of the second CFX connector makes me wonder if all of this is fine. Doesn't the 6870x2 use a Hydra Chip?
Little Messy
(16 items)
 
The 1720 (Laptop)
(13 items)
 
 
CPUMotherboardGraphicsRAM
Phenom II x4 955 (@ 3.8) MSI 890GXM-G65 ASUS HD 6850 (@ 975/1125) G.Skill RipjawX 
Hard DriveHard DriveOptical DriveCooling
Samsung Eco Green Samsung F3 Sony DVD+-RW Writer CoolerMaster Hyper 212+ 
OSOSMonitorKeyboard
Windows 7 Home Premium Arch Linux Dell ST2420 Microsoft Reclusa 
PowerCaseMouse
OCZ StealthXStream II 600W NZXT Vulcan Logitech MX518 
CPUMotherboardGraphicsRAM
Core 2 Duo T8300 Dell's own 8600M 4.0GB Hyundai Electronics DDR2 
Hard DriveOS
250GB Hitachi / 250GB Samsung Windows Vista 32 bit 
  hide details  
Reply
Little Messy
(16 items)
 
The 1720 (Laptop)
(13 items)
 
 
CPUMotherboardGraphicsRAM
Phenom II x4 955 (@ 3.8) MSI 890GXM-G65 ASUS HD 6850 (@ 975/1125) G.Skill RipjawX 
Hard DriveHard DriveOptical DriveCooling
Samsung Eco Green Samsung F3 Sony DVD+-RW Writer CoolerMaster Hyper 212+ 
OSOSMonitorKeyboard
Windows 7 Home Premium Arch Linux Dell ST2420 Microsoft Reclusa 
PowerCaseMouse
OCZ StealthXStream II 600W NZXT Vulcan Logitech MX518 
CPUMotherboardGraphicsRAM
Core 2 Duo T8300 Dell's own 8600M 4.0GB Hyundai Electronics DDR2 
Hard DriveOS
250GB Hitachi / 250GB Samsung Windows Vista 32 bit 
  hide details  
Reply
post #15 of 177
this is interesting read, but after reading it you still don't know how much would microstuttering bother you
the 3 crossfire setup is really a nice example however how it should work
the article states that microstuttering is visible when frames per second go around 30, but with high fps it isn't visible

so let's say 2 cards on a game like metro2033 for example
give 27 minimum fps, average 55, high 80
that would mean you see microstuttering when frames drop to round 30fps

lets say another game, battlefield bad company 2
2 cards give minimum 70, average 100, high 140fps
that would mean that with that game you wouldn't see any microstuttering, not because it isn't there, but because the framerate is high enough for you not to notice it???
LEGA
(15 items)
 
  
CPUMotherboardGraphicsRAM
Phenom II 945 ASUS M4A785TD-V EVO Sapphire 5850 4 GB G.Skill Ripjaws 1866mhz 
Hard DriveHard DriveCoolingOS
Samsung spinpoint F1 1000GB  Samsung spinpoint F3 1000GB  Zalman CNPS10X Performa win7 ultimate x64 
MonitorKeyboardPowerCase
Viewsonic VG2427wm 24'' Genius KB-G255 520W seasonic Chieftec Dragon 
MouseMouse Pad
Gigabyte m8000x Wooden table with smooth finish 
  hide details  
Reply
LEGA
(15 items)
 
  
CPUMotherboardGraphicsRAM
Phenom II 945 ASUS M4A785TD-V EVO Sapphire 5850 4 GB G.Skill Ripjaws 1866mhz 
Hard DriveHard DriveCoolingOS
Samsung spinpoint F1 1000GB  Samsung spinpoint F3 1000GB  Zalman CNPS10X Performa win7 ultimate x64 
MonitorKeyboardPowerCase
Viewsonic VG2427wm 24'' Genius KB-G255 520W seasonic Chieftec Dragon 
MouseMouse Pad
Gigabyte m8000x Wooden table with smooth finish 
  hide details  
Reply
post #16 of 177
Quote:
Originally Posted by Korlus;14672008 
Crazy. Still, I agree. Why not use a more traditional "trifire" package? E.g. the 69-- series or the 58-- series, as the lack of the second CFX connector makes me wonder if all of this is fine. Doesn't the 6870x2 use a Hydra Chip?

It does. On one of the first few pages when they took 30 frames from metro the 6870 x2 had more ms than 2 6870s leadingcthem to believe the lucid bridge was a worse performer.
post #17 of 177
Quote:
Originally Posted by Deathclaw;14672019 
this is interesting read, but after reading it you still don't know how much would microstuttering bother you
the 3 crossfire setup is really a nice example however how it should work
the article states that microstuttering is visible when frames per second go around 30, but with high fps it isn't visible

so let's say 2 cards on a game like metro2033 for example
give 27 minimum fps, average 55, high 80
that would mean you see microstuttering when frames drop to round 30fps

lets say another game, battlefield bad company 2
2 cards give minimum 70, average 100, high 140fps
that would mean that with that game you wouldn't see any microstuttering, not because it isn't there, but because the framerate is high enough for you not to notice it???

That's not what it states at all. It states that at lower fps ms is worse and 30fps might be the worst example of stuttering. However even at blistering speeds it never truly disappears
post #18 of 177
Quote:
Originally Posted by Blameless;14671691 
I wonder if AMD dual-gpu solutions are more vulnerable to PCI-E latency/bandwidth, or have an issue with the NForce 200 chip. I'd like to see multiple board platforms examined.

It would appear so since it has to transfer everything to the first card via the bridge and PCIe where as the Nvidia bridge can talk directly to each card.

If you throw in multi-monitor setups it gets even worse. ATI has to transfer frame buffer from card 4 through card 3, then through card 2 which also has to carry data from card 3 plus itself. Card 2 is sending data over the bridge from 3 different cards all at the same time to card 1.
Nvidia can just send data from card 4 straight to card 2 and card 3 can talk straight to card 1.

569px-R700_interconnect.png


At high resolutions the ATI bridge appears to be a bottleneck. I haven't seen anyone else with 4 6970s at these high resolutions to compare his crashing beyond a certain mexapixel issue:
Quote:
"AMD readily admits. The core issue is the fact that the dedicated CrossFire interconnect used for passing completed frames between cards has "only" enough bandwidth to sustain a 2560x1600 display resolution. Even three 1080p displays will exceed its capacity. The alternative is to transfer frame buffer data via PCI Express, which is what AMD does when necessary. Using PCIe works, but it can limit performance scaling somewhat—don't expect to see the near-linear scaling"

If I simply lowered the Eyefinity resolution to a certain point, the Tri-Quad crossfire would magically start working again.
http://www.xtremesystems.org/forums/showthread.php?266367-3x-30-quot-Portrait-Crossfire-Eyefinity-vs-SLI-Surround-Showdown&p=4737558&viewfull=1#post4737558


Quote:
Originally Posted by Deathclaw;14672019 
2 cards give minimum 70, average 100, high 140fps
that would mean that with that game you wouldn't see any microstuttering, not because it isn't there, but because the framerate is high enough for you not to notice it???

Yes, to eliminate micro-stuttering you need to ensure your minimum fps never goes below your refresh rate.
Edited by PoopaScoopa - 8/23/11 at 6:15pm
post #19 of 177
never noticed it on SLI or CF or eyefinity.
post #20 of 177
Quote:
Originally Posted by bce22;14672100 
That's not what it states at all. It states that at lower fps ms is worse and 30fps might be the worst example of stuttering. However even at blistering speeds it never truly disappears

so, how is microstuttering for you and your dual gpu setup?
do you see it?
when and where?

what can be seen from graphs, the single card also has ups and downs, what is normal otherwise there wouldn't be min, avg, max framerate, but only framerate

the problem with 2 cards of course is them sync-ing and working together
and microstuttering occurs when frames they output aren't synced
therefore the huge ups and down's in frames

but on one card you don't see difference when your framerate drops from max 150 to minimum 100

when you have 2 cards, you have one card rendering faster, other rendering slower and you don't have same framerate from both, and then you get those huge ups and downs in graphs

and by my logic, 3 cards have smaller jumps because each card is given less stress, and therefore can render pretty much the same framerate, but that is the same reason why you only get 20% performance over 2 same cards, because the cards are working without much stress

so by my logic if you give them less to do then you would lower those ups and downs

example
game with max settings
1 gpu min 28 avg 40 max 60

2gpu cfx/sli
-1 rendering 28
-2 rendering 44
(result microstuttering)

lowered settings in same game

1gpu min 40 avg 50 max 70

2gpu cfx/sli
-1 rendering 40
-2 rendering 47
(less microstuttering)

very low setting in same game v-sync on
1gpu min 60 avg 60 max 60

2gpu cfx/sli
-1 rendering 60
-2 rendering 60
(no microstuttering)

the article does state that the microstuttering is less if you have more powerfull graphic card, and that would support my theory
more powerfull graphic card would be the same as using less powerfull card but lowered settings in game
LEGA
(15 items)
 
  
CPUMotherboardGraphicsRAM
Phenom II 945 ASUS M4A785TD-V EVO Sapphire 5850 4 GB G.Skill Ripjaws 1866mhz 
Hard DriveHard DriveCoolingOS
Samsung spinpoint F1 1000GB  Samsung spinpoint F3 1000GB  Zalman CNPS10X Performa win7 ultimate x64 
MonitorKeyboardPowerCase
Viewsonic VG2427wm 24'' Genius KB-G255 520W seasonic Chieftec Dragon 
MouseMouse Pad
Gigabyte m8000x Wooden table with smooth finish 
  hide details  
Reply
LEGA
(15 items)
 
  
CPUMotherboardGraphicsRAM
Phenom II 945 ASUS M4A785TD-V EVO Sapphire 5850 4 GB G.Skill Ripjaws 1866mhz 
Hard DriveHard DriveCoolingOS
Samsung spinpoint F1 1000GB  Samsung spinpoint F3 1000GB  Zalman CNPS10X Performa win7 ultimate x64 
MonitorKeyboardPowerCase
Viewsonic VG2427wm 24'' Genius KB-G255 520W seasonic Chieftec Dragon 
MouseMouse Pad
Gigabyte m8000x Wooden table with smooth finish 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [Toms] Micro-Stuttering And GPU Scaling In CrossFire And SLI