Overclock.net › Forums › Graphics Cards › NVIDIA › Huh *** Sli increase less than 10%
New Posts  All Forums:Forum Nav:

Huh *** Sli increase less than 10% - Page 2

post #11 of 31
Use vantage or heaven benches, you will see the difference then
    
CPUMotherboardGraphicsRAM
I5 2500K @4.5 ASROCK Z68 Extreme4 gen3 VisionTek 7970 CF GSKILL Ripjaws X 16gb 
Hard DriveOptical DriveCoolingOS
Crucial M4 128g/ WD Caviar Black 1tb ASUS DVD Corsair H100 w/4 Cougar Vortex fans Winders 7 pro 
MonitorPowerCaseMouse
ASUS VS248 Corsair HX850 CM HAF932 advanced Logitech G500s 
AudioOther
Turtle Beach DPX 21 Headset Thrustmaster T300rs wheel 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
I5 2500K @4.5 ASROCK Z68 Extreme4 gen3 VisionTek 7970 CF GSKILL Ripjaws X 16gb 
Hard DriveOptical DriveCoolingOS
Crucial M4 128g/ WD Caviar Black 1tb ASUS DVD Corsair H100 w/4 Cougar Vortex fans Winders 7 pro 
MonitorPowerCaseMouse
ASUS VS248 Corsair HX850 CM HAF932 advanced Logitech G500s 
AudioOther
Turtle Beach DPX 21 Headset Thrustmaster T300rs wheel 
  hide details  
Reply
post #12 of 31
If only I had a nickel for every time I've answered a question that went along the lines of "I got way better gpu(s) than I had before but my 3dMark06 score barely went up zomg what is going on?!?"

It's called a CPU bottleneck.

Edit: Actually, come to think of it ... 19K with a single GTX275? That actually doesn't seem right ... that's way too high I think. The SLI score seems a lot more normal.

Quote:
Originally Posted by mimic58 View Post
ok well it ran fine with sli disabled.. Could it be my overclock settings in riva tuner??

I will try it without hybrid LethalRise750
Don't use Rivatuner anymore. Use Precision or Afterburner. And like Lethal said, you don't want anything to do w/Hybrid SLI. Turn it off, and never think about it again.
Edited by brettjv - 12/21/10 at 9:36pm
    
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 #13 of 31
Thread Starter 
Quote:
Originally Posted by raptor5150 View Post
Use vantage or heaven benches, you will see the difference then
Nice bit of software that heaven benches , i much prefer it to 3dmark

I'v done away with the hybrid and asigned physics to the cpu instead and am seeing almost 100% between sli enabled and disabled on heaven

Think that 3dmark 6 is just junk
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Phenom 2 x6 1090t 4ghz :-) Asus Crosshair 2 GTX 275 SLI O/C Crosshair dominator 4g 1066mhz 
Hard DriveOptical DriveOSMonitor
Raid 0 x4 SATA2 7200rpm LG BLueray BD_RW windows 7 ultimate (64bit) 42" LG LCD 
KeyboardPowerCase
MCE wireless touchpad Silverstone 1kw 80amp rail Xclio A380 plus 
  hide details  
Reply
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Phenom 2 x6 1090t 4ghz :-) Asus Crosshair 2 GTX 275 SLI O/C Crosshair dominator 4g 1066mhz 
Hard DriveOptical DriveOSMonitor
Raid 0 x4 SATA2 7200rpm LG BLueray BD_RW windows 7 ultimate (64bit) 42" LG LCD 
KeyboardPowerCase
MCE wireless touchpad Silverstone 1kw 80amp rail Xclio A380 plus 
  hide details  
Reply
post #14 of 31
Thread Starter 
Quote:
Originally Posted by brettjv View Post
If only I had a nickel for every time I've answered a question that went along the lines of "I got way better gpu(s) than I had before but my 3dMark06 score barely went up zomg what is going on?!?"

It's called a CPU bottleneck.

Edit: Actually, come to think of it ... 19K with a single GTX275? That actually doesn't seem right ... that's way too high I think. The SLI score seems a lot more normal.



Don't use Rivatuner anymore. Use Precision or Afterburner. And like Lethal said, you don't want anything to do w/Hybrid SLI. Turn it off, and never think about it again.

Iv ditched it now, seems to be like you say a waste of time .. i will check out Precision & Afterburner

You reacon im bottlenecking at the cpu??

Edit: only showing 47% cpu load during bench mark?

Edited by mimic58 - 12/21/10 at 10:17pm
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Phenom 2 x6 1090t 4ghz :-) Asus Crosshair 2 GTX 275 SLI O/C Crosshair dominator 4g 1066mhz 
Hard DriveOptical DriveOSMonitor
Raid 0 x4 SATA2 7200rpm LG BLueray BD_RW windows 7 ultimate (64bit) 42" LG LCD 
KeyboardPowerCase
MCE wireless touchpad Silverstone 1kw 80amp rail Xclio A380 plus 
  hide details  
Reply
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Phenom 2 x6 1090t 4ghz :-) Asus Crosshair 2 GTX 275 SLI O/C Crosshair dominator 4g 1066mhz 
Hard DriveOptical DriveOSMonitor
Raid 0 x4 SATA2 7200rpm LG BLueray BD_RW windows 7 ultimate (64bit) 42" LG LCD 
KeyboardPowerCase
MCE wireless touchpad Silverstone 1kw 80amp rail Xclio A380 plus 
  hide details  
Reply
post #15 of 31
Quote:
Originally Posted by mimic58 View Post
Iv ditched it now, seems to be like you say a waste of time .. i will check out Precision & Afterburner

You reacon im bottlenecking at the cpu??

Edit: only showing 47% load during a bench mark?
On 3dMark06, when run at the standard settings, yes, with your system in SLI there will be a significant number of frames over the course of the 4 graphics tests where the CPU will be limiting factor to the framerate.

If you actually look at the details of your results from both runs (1 vs 2 cards), you should discover that the only test that went up in FPS significantly was test 3, Canyon Flight. That is because the other 3 (particularly tests 1 and 2) are being heavily CPU-limited.

And 47% of CPU usage is very close to 100% ... if the particular application can only make use of 2 of your 4 cores ... which is indeed the case with 3dMark06. There was no such thing as quads when that bench came out.
Edited by brettjv - 12/21/10 at 10:24pm
    
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 #16 of 31
Thread Starter 
Quote:
Originally Posted by brettjv View Post
On 3dMark06, when run at the standard settings, yes, with your system in SLI there will be a significant number of frames over the course of the 4 graphics tests where the CPU will be limiting factor to the framerate.

If you actually look at the details of your results from both runs (1 vs 2 cards), you should discover that the only test that went up in FPS significantly was test 3, Canyon Flight. That is because the other 3 (particularly tests 1 and 2) are being heavily CPU-limited.

And 47% of CPU usage is very close to 100% ... if the particular application can only make use of 2 of your 4 cores ... which is indeed the case with 3dMark06. There was no such thing as quads when that bench came out.
Hmm... I see , so is this cpu quick enough to get the full potencial out of the cards on an application thats written to use all 4 cores?? or will it still bottleneck before the cards limits hit?

Edit: there mustbe a way i can squeez a bit more performance out of this thing, Even if its just a lil
Edited by mimic58 - 12/21/10 at 10:36pm
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Phenom 2 x6 1090t 4ghz :-) Asus Crosshair 2 GTX 275 SLI O/C Crosshair dominator 4g 1066mhz 
Hard DriveOptical DriveOSMonitor
Raid 0 x4 SATA2 7200rpm LG BLueray BD_RW windows 7 ultimate (64bit) 42" LG LCD 
KeyboardPowerCase
MCE wireless touchpad Silverstone 1kw 80amp rail Xclio A380 plus 
  hide details  
Reply
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Phenom 2 x6 1090t 4ghz :-) Asus Crosshair 2 GTX 275 SLI O/C Crosshair dominator 4g 1066mhz 
Hard DriveOptical DriveOSMonitor
Raid 0 x4 SATA2 7200rpm LG BLueray BD_RW windows 7 ultimate (64bit) 42" LG LCD 
KeyboardPowerCase
MCE wireless touchpad Silverstone 1kw 80amp rail Xclio A380 plus 
  hide details  
Reply
post #17 of 31
Quote:
Originally Posted by mimic58 View Post
Hmm... I see , so is this cpu quick enough to get the full potencial out of the cards on an application thats written to use all 4 cores?? or will it still bottleneck before the cards limits hit?

Edit: there mustbe a way i can squeez a bit more performance out of this thing, Even if its just a lil
The issue with 3dMark06 is not really so much that it only uses 2 cores, although it would help if it could use all 4 of course. Rather, it is that it's quite old, and the standard settings (1280x1024, no AA, 4xAF, medium graphics settings) were calibrated around the single GPU setups of it's day.

Keep in mind, when that bench came out, nobody had SLI, and it was tough to break 4000 score with the existing high-end setups.

Nowadays, a high end GPU setup (like yours) easily can run over 5x as fast.

So the actual issues here are that:
A) the FPS on this test at such easy-to-run graphical settings is now very high relative to what the test was designed to run at, and
B) three of the 4 tests (not so much Canyon Flight, but the other 3) are fairly 'CPU-dependent' to begin with. By 'CPU-dependency' I'm referring to 'the relative amount of work required of the CPU by the GPU for every frame rendered'.

This means that by running 3dMark06 at 'stock' settings, you've created a 'testing scenario' where even the most modern CPU's can't keep up with the demands made of them by the GPU's for data. That's because on this test, it's just too easy for modern GPU's to run at these settings.

However, if you owned the full version of this test, and you went in and ran it at, say, 1920x1080, and cranked up every setting like AA/AF and whatnot, it would challenge your GPU enough that the FPS would slow down significantly (probably be cut down to like 25% of what you see now). You will have then created a different 'testing scenario', wherein your CPU would no longer have any issue keeping up with the data requests being made by the GPU's.

Bottom-line, when it comes to which part, CPU or GPU, will act as the limiting performance factor ... it's all about the specific test you're running. What is the inherent CPU-dependency of the test/bench/game, and what settings (resolution/AA/etc) are you running it at?

In general, the *lower* the IQ (image quality) settings, then the higher the FPS should be, because the GPU has less work to do per frame rendered, right? However, the higher the FPS, the more likely it becomes that the CPU will limit performance, because more work is being requested of it in the same amount of time.

Now, the point in the continuum of FPS at which the CPU limits performance describes the CPU-dependency of the test.

For example, on my system, if I test with Crysis, I will start to see CPU bottlenecking at around 60fps. This is a pretty cpu-dependent game, because that is a pretty low FPS to have cpu-limitation already kicking in, esp. given my proc is pretty badass. However, if I ran a different game, say ... FEAR for example ... I don't see signs of a CPU BN until my FPS gets up into the 200+ range. So I say this is a much less CPU-dependent test.

You follow what I'm saying here?
Edited by brettjv - 12/21/10 at 11:37pm
    
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 #18 of 31
Thread Starter 
Quote:
Originally Posted by brettjv View Post
The issue with 3dMark06 is not really so much that it only uses 2 cores, although it would help if it could use all 4 of course. Rather, it is that it's quite old, and the standard settings (1280x1024, no AA, 4xAF, medium graphics settings) were calibrated around the single GPU setups of it's day.

Keep in mind, when that bench came out, nobody had SLI, and it was tough to break 4000 score with the existing high-end setups.

Nowadays, a high end GPU setup (like yours) easily can run over 5x as fast.

So the actual issues here are that:
A) the FPS on this test at such easy-to-run graphical settings is now very high relative to what the test was designed to run at, and
B) three of the 4 tests (not so much Canyon Flight, but the other 3) are fairly 'CPU-dependent' to begin with. By 'CPU-dependency' I'm referring to 'the relative amount of work required of the CPU by the GPU for every frame rendered'.

This means that by running 3dMark06 at 'stock' settings, you've created a 'testing scenario' where even the most modern CPU's can't keep up with the demands made of them by the GPU's for data. That's because on this test, it's just too easy for modern GPU's to run at these settings.

However, if you owned the full version of this test, and you went in and ran it at, say, 1920x1080, and cranked up every setting like AA/AF and whatnot, it would challenge your GPU enough that the FPS would slow down significantly (probably be cut down to like 25% of what you see now). You will have then created a different 'testing scenario', wherein your CPU would no longer have any issue keeping up with the data requests being made by the GPU's.

Bottom-line, when it comes to which part, CPU or GPU, will act as the limiting performance factor ... it's all about the specific test you're running. What is the inherent CPU-dependency of the test/bench/game, and what settings (resolution/AA/etc) are you running it at?

In general, the *lower* the IQ (image quality) settings, then the higher the FPS should be, because the GPU has less work to do per frame rendered, right? However, the higher the FPS, the more likely it becomes that the CPU will limit performance, because more work is being requested of it in the same amount of time.

Now, the point in the continuum of FPS at which the CPU limits performance describes the CPU-dependency of the test.

For example, on my system, if I test with Crysis, I will start to see CPU bottlenecking at around 60fps. This is a pretty cpu-dependent game, because that is a pretty low FPS to have cpu-limitation already kicking in, esp. given my proc is pretty badass. However, if I ran a different game, say ... FEAR for example ... I don't see signs of a CPU BN until my FPS gets up into the 200+ range. So I say this is a much less CPU-dependent test.

You follow what I'm saying here?
yes it does make sence now, Thanks for the full explanation I couldnt understand why i was getting such a small difference between the two marks but now i do... obviously i cannot expect a oc/965 to match a oc/930 but that said id still like to get as much as i can out of it, and wanted to be sure my sli setup was correctly configured as realiscaly i cant get anymore speed out of the cpu and memory.. I was using riva tuner to push the cards a lil harder but i did see much of an increase on the mark either but based on what you have said that is to be expected. it just apeared inicialy that something was drastically wrong with my sli rig lol

pehaps this is just the limit for this system now

Doesnt look to shabby though



Edited by mimic58 - 12/22/10 at 7:47am
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Phenom 2 x6 1090t 4ghz :-) Asus Crosshair 2 GTX 275 SLI O/C Crosshair dominator 4g 1066mhz 
Hard DriveOptical DriveOSMonitor
Raid 0 x4 SATA2 7200rpm LG BLueray BD_RW windows 7 ultimate (64bit) 42" LG LCD 
KeyboardPowerCase
MCE wireless touchpad Silverstone 1kw 80amp rail Xclio A380 plus 
  hide details  
Reply
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Phenom 2 x6 1090t 4ghz :-) Asus Crosshair 2 GTX 275 SLI O/C Crosshair dominator 4g 1066mhz 
Hard DriveOptical DriveOSMonitor
Raid 0 x4 SATA2 7200rpm LG BLueray BD_RW windows 7 ultimate (64bit) 42" LG LCD 
KeyboardPowerCase
MCE wireless touchpad Silverstone 1kw 80amp rail Xclio A380 plus 
  hide details  
Reply
post #19 of 31
Enjoy those 275s!
With a little OC these things scream.
Mighty Mini
(13 items)
 
  
CPUMotherboardGraphicsRAM
2500k Z68 Gene Asus 970 Strix 8gigs Corsair Vengeance 
Hard DriveOptical DriveOSMonitor
Kingston HyperX 120gig Seagate Cuda 1TB W10 Acer 23" 1080p 
KeyboardPowerCaseMouse
G15 HX1000W NZXT Vulcan G9X 
  hide details  
Reply
Mighty Mini
(13 items)
 
  
CPUMotherboardGraphicsRAM
2500k Z68 Gene Asus 970 Strix 8gigs Corsair Vengeance 
Hard DriveOptical DriveOSMonitor
Kingston HyperX 120gig Seagate Cuda 1TB W10 Acer 23" 1080p 
KeyboardPowerCaseMouse
G15 HX1000W NZXT Vulcan G9X 
  hide details  
Reply
post #20 of 31
Thread Starter 
Quote:
Originally Posted by Lost-boi View Post
Enjoy those 275s!
With a little OC these things scream.
Indeed what o/c settings are you running??

Im currently at

Core 648
Shader 1476
memory 1242

Would be nice to peg that 2g 295 i only need a snitch to do it
Edited by mimic58 - 12/22/10 at 8:23am
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Phenom 2 x6 1090t 4ghz :-) Asus Crosshair 2 GTX 275 SLI O/C Crosshair dominator 4g 1066mhz 
Hard DriveOptical DriveOSMonitor
Raid 0 x4 SATA2 7200rpm LG BLueray BD_RW windows 7 ultimate (64bit) 42" LG LCD 
KeyboardPowerCase
MCE wireless touchpad Silverstone 1kw 80amp rail Xclio A380 plus 
  hide details  
Reply
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Phenom 2 x6 1090t 4ghz :-) Asus Crosshair 2 GTX 275 SLI O/C Crosshair dominator 4g 1066mhz 
Hard DriveOptical DriveOSMonitor
Raid 0 x4 SATA2 7200rpm LG BLueray BD_RW windows 7 ultimate (64bit) 42" LG LCD 
KeyboardPowerCase
MCE wireless touchpad Silverstone 1kw 80amp rail Xclio A380 plus 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: NVIDIA
Overclock.net › Forums › Graphics Cards › NVIDIA › Huh *** Sli increase less than 10%