Overclock.net › Forums › Industry News › Hardware News › [[H]ard|OCP] NVIDIA 3-Way SLI and AMD Tri-Fire Redux
New Posts  All Forums:Forum Nav:

[[H]ard|OCP] NVIDIA 3-Way SLI and AMD Tri-Fire Redux - Page 10

post #91 of 168
Quote:
Originally Posted by ryboto View Post
I was going to correct my post. You can draw the conclusion that the Nvidia setup was originally bottlenecked, but given the results, no performance comparisons between the brands can be made. You said that all people wanted was a proper review, seemed like you were implying this was proper, but I misread.
Yeah, what I meant was that those 2 issues, the CPU bottleneck and the x4 lane issue was addressed for the 580's. In the previous review those bottlenecks were pretty obvious.
http://hardocp.com/article/2011/04/2...ifire_review/2

Just look at F1 for example: 2 way = 43.5 and 3 way = 47.8? When I see that, I see an obvious bottleneck.

About my conclusion ( I don't really have one ) because now that those 2 things have been addressed on the Nvidia side, it's puzzling how the AMD cards had very small gains in 3 games ( 6.7% was their largest gain, then 6.5% and 3.4% ) and of course, the 2 games that saw a negative impact.

So obviously, all we can do is speculate, and I don't have an AMD set up to replicate their results...

The only thing that I said was that I doubt the NF200 chip had anything to do with it. ( this is speculation, but more like an educated guess )

If I were to speculate about the AMD cards, I'd say that they're not as CPU dependent as the 3 580's.

They also were never bandwidth starved in either review running x16 x16, another possible for such little gains.

One a side note, I wish they tested MORE games.
Edited by 2010rig - 5/3/11 at 7:02pm
2010rig
(14 items)
 
Galaxy S3
(8 items)
 
 
CPUMotherboardGraphicsRAM
X5660 @ 4.5  ASUS P6X58D-E 980TI? 12GB OCZ Platinum - 7-7-7-21 
Hard DriveCoolingOSMonitor
1 80GB SSD x25m - 3TB F3 + F4 NH-D14 Windows 7 Ultimate LG 47LH55 
KeyboardPowerCaseMouse
Natural Wireless Keyboard Corsair 750HX CM 690 II Advanced MX 518 
CPUGraphicsRAMHard Drive
Snapdragon S4 Dual core 1500mhz Adreno 225 Samsung 2GB 16GB Onboard Flash 
OSMonitorPowerCase
Android 4.4.2 - CM11 4.8" AMOLED 1280x720 2100 mAh battery Otterbox Defender 
  hide details  
Reply
2010rig
(14 items)
 
Galaxy S3
(8 items)
 
 
CPUMotherboardGraphicsRAM
X5660 @ 4.5  ASUS P6X58D-E 980TI? 12GB OCZ Platinum - 7-7-7-21 
Hard DriveCoolingOSMonitor
1 80GB SSD x25m - 3TB F3 + F4 NH-D14 Windows 7 Ultimate LG 47LH55 
KeyboardPowerCaseMouse
Natural Wireless Keyboard Corsair 750HX CM 690 II Advanced MX 518 
CPUGraphicsRAMHard Drive
Snapdragon S4 Dual core 1500mhz Adreno 225 Samsung 2GB 16GB Onboard Flash 
OSMonitorPowerCase
Android 4.4.2 - CM11 4.8" AMOLED 1280x720 2100 mAh battery Otterbox Defender 
  hide details  
Reply
post #92 of 168
I really dont understand why they used a 2600k, I thought that HT had a negative impact on FPS (or not impact at all) but its still true that HT can reduce your FPS, so when you are a Benchmarking site ... whats the point to use a CPU that may lower your resust or give false result. They should have use a 2500k.
post #93 of 168
BS
When was that? On Pentium 4?

and they did turn off HT

and they used 2600k because it OCs better then 2500k
post #94 of 168
Quote:
Originally Posted by 2010rig View Post
Yeah, what I meant was that those 2 issues, the CPU bottleneck and the x4 lane issue was addressed for the 580's. In the previous review those bottlenecks were pretty obvious.
http://hardocp.com/article/2011/04/2...ifire_review/2

Just look at F1 for example: 2 way = 43.5 and 3 way = 47.8? When I see that, I see an obvious bottleneck.

About my conclusion ( I don't really have one ) because now that those 2 things have been addressed on the Nvidia side, it's puzzling how the AMD cards had very small gains in 3 games ( 6.7% was their largest gain, then 6.5% and 3.4% ) and of course, the 2 games that saw a negative impact.

So obviously, all we can do is speculate, and I don't have an AMD set up to replicate their results...

The only thing that I said was that I doubt the NF200 chip had anything to do with it. ( this is speculation, but more like an educated guess )

If I were to speculate about the AMD cards, I'd say that they're not as CPU dependent as the 3 580's.

They also were never bandwidth starved in either review running x16 x16, another possible for such little gains.

One a side note, I wish they tested MORE games.
I'm leaning toward the info we have about the board, that with only two cards present the board defaults to x8x8, which makes more sense to me. To truly test if the Nvidia setup utilizes more CPU time, we'd have to also look at power consumption of the two systems as well as CPU usage. Otherwise, we still can't draw a conclusion as to which setup was more CPU limited than the other. The lower lane bandwidth makes some sense if you think about it. The additional CPU power counteracts the lower PCIE lane bandwidth in some titles, but in the ones that are much more GPU dependent, you see a negative impact.
mitx
(16 items)
 
  
CPUMotherboardGraphicsRAM
Intel i5 3570K 4.5ghz ASRock Z77E-ITX Sapphire R9 290x Tri-X 1200/1500 Crucial Ballistix Tactical LP 2x8 GB 8-8-8-24-1T 
Hard DriveHard DriveHard DriveHard Drive
Samsung 840 Pro 256gb Western Digital Velociraptor Western Digital WD15EARX Western Digital WD15EADS 
CoolingOSMonitorKeyboard
Silverstone NT06-Pro Windows 10 x64 LG 29UM67 Logitech G710+ 
PowerCaseMouseAudio
Silverstone SG08 600W PSU Silverstone SG08B Logitech G700 Pioneer SE-A1000 
  hide details  
Reply
mitx
(16 items)
 
  
CPUMotherboardGraphicsRAM
Intel i5 3570K 4.5ghz ASRock Z77E-ITX Sapphire R9 290x Tri-X 1200/1500 Crucial Ballistix Tactical LP 2x8 GB 8-8-8-24-1T 
Hard DriveHard DriveHard DriveHard Drive
Samsung 840 Pro 256gb Western Digital Velociraptor Western Digital WD15EARX Western Digital WD15EADS 
CoolingOSMonitorKeyboard
Silverstone NT06-Pro Windows 10 x64 LG 29UM67 Logitech G710+ 
PowerCaseMouseAudio
Silverstone SG08 600W PSU Silverstone SG08B Logitech G700 Pioneer SE-A1000 
  hide details  
Reply
post #95 of 168
I question the intelligence of the people doing these benchs. Did they really say they didn't expect a cpu bottleneck with 3 way sli/xfire of the fastest cards on the market on a generation old cpu at 3.6Ghz? Hell I get slight bottlenecking in certain maps on bc2 with my 920@4Ghz and sli 470's. I mean I realize they were using extremely high resolutions but still.
Edited by scyy - 5/3/11 at 7:29pm
3930k
(20 items)
 
  
Reply
3930k
(20 items)
 
  
Reply
post #96 of 168
Quote:
Originally Posted by ryboto View Post
I'm leaning toward the info we have about the board, that with only two cards present the board defaults to x8x8, which makes more sense to me. To truly test if the Nvidia setup utilizes more CPU time, we'd have to also look at power consumption of the two systems as well as CPU usage. Otherwise, we still can't draw a conclusion as to which setup was more CPU limited than the other. The lower lane bandwidth makes some sense if you think about it. The additional CPU power counteracts the lower PCIE lane bandwidth in some titles, but in the ones that are much more GPU dependent, you see a negative impact.
If you only use 2 PCIe slots, the WS Revolution uses dual 16x16x not 8x8x.
post #97 of 168
Quote:
Originally Posted by ryboto View Post
I'm leaning toward the info we have about the board, that with only two cards present the board defaults to x8x8, which makes more sense to me. To truly test if the Nvidia setup utilizes more CPU time, we'd have to also look at power consumption of the two systems as well as CPU usage. Otherwise, we still can't draw a conclusion as to which setup was more CPU limited than the other. The lower lane bandwidth makes some sense if you think about it. The additional CPU power counteracts the lower PCIE lane bandwidth in some titles, but in the ones that are much more GPU dependent, you see a negative impact.
You know, you're probably right.

If the board defaults to x8 x8, the only noticeable impact ( although very small ) is at this high resolution:
http://www.hardocp.com/article/2010/...6x16_vs_x8x8/2

I'm going to do more research on that board, because with only 2 cards, they should be running at x16 x16, at least, according to ASUS that is:

Quote:
Designed for true power users, the P8P67 WS Revolution uses a built-in NF200 controller that enhances bandwidth availability between the board and the four graphics card expansion slots. This is ideal for NVIDIA GeForce SLI™ and AMD CrossFireX™, as the new motherboard can easily handle 2-Way SLI in dual PCI Express X16, while 3-Way SLI works in dual PCI Express X8 and one X16 link. This translates into a 26.4% performance increase compared to regular P67 motherboards with similar SLI configurations, as measured by 3DMark06. For CrossFireX, the P8P67 WS Revolution supports up to quad GPUs builds in PCI Express X8 links, unlike standard motherboards, which lack the bandwidth to do so.

Edited by 2010rig - 5/3/11 at 7:30pm
2010rig
(14 items)
 
Galaxy S3
(8 items)
 
 
CPUMotherboardGraphicsRAM
X5660 @ 4.5  ASUS P6X58D-E 980TI? 12GB OCZ Platinum - 7-7-7-21 
Hard DriveCoolingOSMonitor
1 80GB SSD x25m - 3TB F3 + F4 NH-D14 Windows 7 Ultimate LG 47LH55 
KeyboardPowerCaseMouse
Natural Wireless Keyboard Corsair 750HX CM 690 II Advanced MX 518 
CPUGraphicsRAMHard Drive
Snapdragon S4 Dual core 1500mhz Adreno 225 Samsung 2GB 16GB Onboard Flash 
OSMonitorPowerCase
Android 4.4.2 - CM11 4.8" AMOLED 1280x720 2100 mAh battery Otterbox Defender 
  hide details  
Reply
2010rig
(14 items)
 
Galaxy S3
(8 items)
 
 
CPUMotherboardGraphicsRAM
X5660 @ 4.5  ASUS P6X58D-E 980TI? 12GB OCZ Platinum - 7-7-7-21 
Hard DriveCoolingOSMonitor
1 80GB SSD x25m - 3TB F3 + F4 NH-D14 Windows 7 Ultimate LG 47LH55 
KeyboardPowerCaseMouse
Natural Wireless Keyboard Corsair 750HX CM 690 II Advanced MX 518 
CPUGraphicsRAMHard Drive
Snapdragon S4 Dual core 1500mhz Adreno 225 Samsung 2GB 16GB Onboard Flash 
OSMonitorPowerCase
Android 4.4.2 - CM11 4.8" AMOLED 1280x720 2100 mAh battery Otterbox Defender 
  hide details  
Reply
post #98 of 168
research if NF200 mobo gives x16/x16 on P67?

please post back results of your extensive research
post #99 of 168
It's good to be the king?
    
CPUMotherboardGraphicsRAM
Intel i7 920 D0 stepping @ 4.+GHz ASUS Rampage III Extreme + EK-FB-R3E water block MSI GTX 480s in 3-Way SLI Mushkin Enhanced Ridgeback DDR3 1600 6GB (3 x 2GB) 
Hard DriveOptical DriveOSMonitor
Crucial RealSSD C300 256GB ASUS CD/DVD burner Windows 7 Home Premium 64-bit (Triple) ASUS VW266H 19x12 each @ 5760 x 1200 
KeyboardPowerCaseMouse
Logitech Comfort Wave 450 Black (Dual) Corsair HX1000s Mountain Mods Ascension (modded) Logitech MX518 
Mouse Pad
(Dual) XTracPads Ripper XXL 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Intel i7 920 D0 stepping @ 4.+GHz ASUS Rampage III Extreme + EK-FB-R3E water block MSI GTX 480s in 3-Way SLI Mushkin Enhanced Ridgeback DDR3 1600 6GB (3 x 2GB) 
Hard DriveOptical DriveOSMonitor
Crucial RealSSD C300 256GB ASUS CD/DVD burner Windows 7 Home Premium 64-bit (Triple) ASUS VW266H 19x12 each @ 5760 x 1200 
KeyboardPowerCaseMouse
Logitech Comfort Wave 450 Black (Dual) Corsair HX1000s Mountain Mods Ascension (modded) Logitech MX518 
Mouse Pad
(Dual) XTracPads Ripper XXL 
  hide details  
Reply
post #100 of 168
Well according to this review, they're using the Asus P8P67 WS Revolution:
http://www.tomshardware.com/reviews/...fire,2910.html

You can look up all the gaming results and notice dual card set ups running at x16 x16, same as HardOCP.

The AMD cards should've been running at x16 x16, unless they really are running at x8 x8 and weren't aware?

Also, an interesting quote:

Quote:
Even at 2560x1600, the NF200’s only performance benefit appears to be that it enables three-way SLI on P67 motherboards.

Edited by 2010rig - 5/3/11 at 8:08pm
2010rig
(14 items)
 
Galaxy S3
(8 items)
 
 
CPUMotherboardGraphicsRAM
X5660 @ 4.5  ASUS P6X58D-E 980TI? 12GB OCZ Platinum - 7-7-7-21 
Hard DriveCoolingOSMonitor
1 80GB SSD x25m - 3TB F3 + F4 NH-D14 Windows 7 Ultimate LG 47LH55 
KeyboardPowerCaseMouse
Natural Wireless Keyboard Corsair 750HX CM 690 II Advanced MX 518 
CPUGraphicsRAMHard Drive
Snapdragon S4 Dual core 1500mhz Adreno 225 Samsung 2GB 16GB Onboard Flash 
OSMonitorPowerCase
Android 4.4.2 - CM11 4.8" AMOLED 1280x720 2100 mAh battery Otterbox Defender 
  hide details  
Reply
2010rig
(14 items)
 
Galaxy S3
(8 items)
 
 
CPUMotherboardGraphicsRAM
X5660 @ 4.5  ASUS P6X58D-E 980TI? 12GB OCZ Platinum - 7-7-7-21 
Hard DriveCoolingOSMonitor
1 80GB SSD x25m - 3TB F3 + F4 NH-D14 Windows 7 Ultimate LG 47LH55 
KeyboardPowerCaseMouse
Natural Wireless Keyboard Corsair 750HX CM 690 II Advanced MX 518 
CPUGraphicsRAMHard Drive
Snapdragon S4 Dual core 1500mhz Adreno 225 Samsung 2GB 16GB Onboard Flash 
OSMonitorPowerCase
Android 4.4.2 - CM11 4.8" AMOLED 1280x720 2100 mAh battery Otterbox Defender 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [[H]ard|OCP] NVIDIA 3-Way SLI and AMD Tri-Fire Redux