• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Crossfire x16/x4 ?

Brutalfate

New Member
Joined
Sep 12, 2006
Messages
374 (0.06/day)
Location
AUS, NSW, Sydney
Processor Q6600
Motherboard Asus P5k
Cooling Zalman 9500 LED
Memory Kingston HyperX 4GB Kit 1066
Video Card(s) Gigabyte HD5770
Storage 2x Samsung 500GB SataII
Display(s) LG 22" 1680x1050 2ms
Case Tt Soprano
Audio Device(s) Onboard 7.1
Power Supply Coolermaster RealPower M520
Software Windows Vista Ultimate
Hi there, just a quick question before I get my hd3870 tomorrow, if i do decide to go crossfire in the near future, how much bottleneck will I see from the x4 pci-e lane? My board will can only do x16/x4 in a crossfire setup.


I have had a few experiences in the past, where i ran a x1950pro on a x4 lane, and i saw very little differences performance wise, compared to x16.
 

Tatty_Two

Gone Fishing
Joined
Jan 18, 2006
Messages
25,801 (3.87/day)
Location
Worcestershire, UK
Processor Rocket Lake Core i5 11600K @ 5 Ghz with PL tweaks
Motherboard MSI MAG Z490 TOMAHAWK
Cooling Thermalright Peerless Assassin 120SE + 4 Phanteks 140mm case fans
Memory 32GB (4 x 8GB SR) Patriot Viper Steel 4133Mhz DDR4 @ 3600Mhz CL14@1.45v Gear 1
Video Card(s) Asus Dual RTX 4070 OC
Storage WD Blue SN550 1TB M.2 NVME//Crucial MX500 500GB SSD (OS)
Display(s) AOC Q2781PQ 27 inch Ultra Slim 2560 x 1440 IPS
Case Phanteks Enthoo Pro M Windowed - Gunmetal
Audio Device(s) Onboard Realtek ALC1200/SPDIF to Sony AVR @ 5.1
Power Supply Seasonic CORE GM650w Gold Semi modular
Mouse Coolermaster Storm Octane wired
Keyboard Element Gaming Carbon Mk2 Tournament Mech
Software Win 10 Home x64
Hi there, just a quick question before I get my hd3870 tomorrow, if i do decide to go crossfire in the near future, how much bottleneck will I see from the x4 pci-e lane? My board will can only do x16/x4 in a crossfire setup.


I have had a few experiences in the past, where i ran a x1950pro on a x4 lane, and i saw very little differences performance wise, compared to x16.

I read a review on the new HD3870's in Crossfire a few weeks ago (but I cannot find it now), the tests were done on a P35 @16 x 4 and a P38 @ 16 x 16, across the board there was a 15 - 20% performance hit on 16 x 4, it was still pretty fast tho!
 

von kain

New Member
Joined
Sep 13, 2006
Messages
780 (0.12/day)
Location
athens/greece
System Name intel-hate
Processor Amd phenom 2 955 @ 3800
Motherboard msi 790fx-gd70
Cooling coolermaster
Memory 4096 @1600 corsair
Video Card(s) sapphire 5770 @950-1350
Storage 1 seagate 80gb sata2 1 160gb seagate sata
Display(s) 20inch wide lg
Power Supply chieftech 750w
Software windows 7 x64
Benchmark Scores nah
Joined
Sep 22, 2007
Messages
3,527 (0.58/day)
Location
UK
System Name Dream Weaver
Processor Ryzen 5600
Motherboard B450 MSI Gaming Pro Carbon AC
Cooling NH-D14
Memory 16GB Flare X @ 3200MHz CL14
Video Card(s) Sapphire Vega 64 Nitro+
Storage 500GB evo 970 m.2 + 4TB HDDs
Display(s) ASUS 27" 1440p 144hz FreeSync
Case Fractal Design Define R3
Power Supply Corsair 850W
Software Windows 10 x64
Joined
Dec 18, 2005
Messages
8,253 (1.23/day)
System Name money pit..
Processor Intel 9900K 4.8 at 1.152 core voltage minus 0.120 offset
Motherboard Asus rog Strix Z370-F Gaming
Cooling Dark Rock TF air cooler.. Stock vga air coolers with case side fans to help cooling..
Memory 32 gb corsair vengeance 3200
Video Card(s) Palit Gaming Pro OC 2080TI
Storage 150 nvme boot drive partition.. 1T Sandisk sata.. 1T Transend sata.. 1T 970 evo nvme m 2..
Display(s) 27" Asus PG279Q ROG Swift 165Hrz Nvidia G-Sync, IPS.. 2560x1440..
Case Gigabyte mid-tower.. cheap and nothing special..
Audio Device(s) onboard sounds with stereo amp..
Power Supply EVGA 850 watt..
Mouse Logitech G700s
Keyboard Logitech K270
Software Win 10 pro..
Benchmark Scores Firestike 29500.. timepsy 14000..
i have just tried crossfire with 8 x8 plus pci-e 2.. which they say is the same as 16 x 16.. i was so enthralled i sold the second card and crossfire mobo on ebay..

its hard to lose 20% of something that isnt there in the first place..

lets say a possible but perhaps not loss assuming u get a possible but not likely gain in the first place..

either way i have just bought an Abit IP35 Pro 16 x 4 motherboard on the assumption i wont bother with crossfire again in the near future..

trog
 

Tatty_Two

Gone Fishing
Joined
Jan 18, 2006
Messages
25,801 (3.87/day)
Location
Worcestershire, UK
Processor Rocket Lake Core i5 11600K @ 5 Ghz with PL tweaks
Motherboard MSI MAG Z490 TOMAHAWK
Cooling Thermalright Peerless Assassin 120SE + 4 Phanteks 140mm case fans
Memory 32GB (4 x 8GB SR) Patriot Viper Steel 4133Mhz DDR4 @ 3600Mhz CL14@1.45v Gear 1
Video Card(s) Asus Dual RTX 4070 OC
Storage WD Blue SN550 1TB M.2 NVME//Crucial MX500 500GB SSD (OS)
Display(s) AOC Q2781PQ 27 inch Ultra Slim 2560 x 1440 IPS
Case Phanteks Enthoo Pro M Windowed - Gunmetal
Audio Device(s) Onboard Realtek ALC1200/SPDIF to Sony AVR @ 5.1
Power Supply Seasonic CORE GM650w Gold Semi modular
Mouse Coolermaster Storm Octane wired
Keyboard Element Gaming Carbon Mk2 Tournament Mech
Software Win 10 Home x64
http://www.bit-tech.net/hardware/2007/10/12/crossfire_comparison_intel_x38_versus_p35/1

It will affect performance a little bit, but oddly enough according to that article x8/x8 is sometimes faster than x16/x16... How queer

That was true in some cases, but not anymore since the x38, like your review shows, in some cases there is like 20% difference, very much depends on the game and as importantly, what resolution you are playing it, the higher the res.....the more bandwidth you need, if you play at 1280 x 1024 then TBH there is little difference most of the time which is why I personally would not bother....too iffy! Trog.....PCI-E 2.0 does double the bandwidth but as there is only one card in each slot no single card uses the PCI-E 1.1 2.5g bandwidth never mind 2.0 so whilst some manufactures like to tout it as being "theoretically" correct.....it aint!. Just think AGP 8x, we went PCI-E how long ago? 3 or 4 years now, the first card to ever use the equivilent of the full bandwidth availabe from AGP 8x was the 8800GTX!
 
Last edited:
Joined
Dec 18, 2005
Messages
8,253 (1.23/day)
System Name money pit..
Processor Intel 9900K 4.8 at 1.152 core voltage minus 0.120 offset
Motherboard Asus rog Strix Z370-F Gaming
Cooling Dark Rock TF air cooler.. Stock vga air coolers with case side fans to help cooling..
Memory 32 gb corsair vengeance 3200
Video Card(s) Palit Gaming Pro OC 2080TI
Storage 150 nvme boot drive partition.. 1T Sandisk sata.. 1T Transend sata.. 1T 970 evo nvme m 2..
Display(s) 27" Asus PG279Q ROG Swift 165Hrz Nvidia G-Sync, IPS.. 2560x1440..
Case Gigabyte mid-tower.. cheap and nothing special..
Audio Device(s) onboard sounds with stereo amp..
Power Supply EVGA 850 watt..
Mouse Logitech G700s
Keyboard Logitech K270
Software Win 10 pro..
Benchmark Scores Firestike 29500.. timepsy 14000..
That was true in some cases, but not anymore since the x38, like your review shows, in some cases there is like 20% difference, very much depends on the game and as importantly, what resolution you are playing it, the higher the res.....the more bandwidth you need, if you play at 1280 x 1024 then TBH there is little difference most of the time which is why I personally would not bother....too iffy! Trog.....PCI-E 2.0 does double the bandwidth but as there is only one card in each slot no single card uses the PCI-E 1.1 2.5g bandwidth never mind 2.0 so whilst some manufactures like to tout it as being "theoretically" correct.....it aint!. Just think AGP 8x, we went PCI-E how long ago? 3 or 4 years now, the fiurst card to ever use the equivilent of the ull bandwidth availabe from AGP 8x was the 8800GTX!

all very true tatty.. and it pretty much applies to every new piece of "go faster technology" they have come up with since the year dot.. it might well become the norm three years down the line but when announced its bullsh-t

trog
 

HousERaT

New Member
Joined
Apr 8, 2005
Messages
108 (0.02/day)
Processor Q6700
Motherboard Asus Maximus Formula
Cooling Ultra Extreme 120
Memory Ballistix Tracer 8500
Video Card(s) HD 3870 Crossfire
Storage 74G Raptor
Display(s) A Sony big one
Case Kingwin SK523
Power Supply Corsair HX620
Software Yes :p
Hi there, just a quick question before I get my hd3870 tomorrow, if i do decide to go crossfire in the near future, how much bottleneck will I see from the x4 pci-e lane? My board will can only do x16/x4 in a crossfire setup.


I have had a few experiences in the past, where i ran a x1950pro on a x4 lane, and i saw very little differences performance wise, compared to x16.
The difference is there. You can still see some improvement in performance by going from 1 card to 2 cards however as already mentioned there is a noticeable difference between 16x4 and 16x16. Depending on the application the difference is more or less noticeable. For someone like myself who does a lot of benchmarking, running crossfire on a 16x4 motherboard almost defeats the purpose.

Running my 3870 crossfire set on a 16x4 motherboard I was literally throwing away thousands of 3D Marks.
 
Joined
Dec 18, 2005
Messages
8,253 (1.23/day)
System Name money pit..
Processor Intel 9900K 4.8 at 1.152 core voltage minus 0.120 offset
Motherboard Asus rog Strix Z370-F Gaming
Cooling Dark Rock TF air cooler.. Stock vga air coolers with case side fans to help cooling..
Memory 32 gb corsair vengeance 3200
Video Card(s) Palit Gaming Pro OC 2080TI
Storage 150 nvme boot drive partition.. 1T Sandisk sata.. 1T Transend sata.. 1T 970 evo nvme m 2..
Display(s) 27" Asus PG279Q ROG Swift 165Hrz Nvidia G-Sync, IPS.. 2560x1440..
Case Gigabyte mid-tower.. cheap and nothing special..
Audio Device(s) onboard sounds with stereo amp..
Power Supply EVGA 850 watt..
Mouse Logitech G700s
Keyboard Logitech K270
Software Win 10 pro..
Benchmark Scores Firestike 29500.. timepsy 14000..
so i ran it using the latest wizbang 790x chipset motherboard.. how come i didnt get thousands of 3dmarks to throw away in the first place..

we both have amd systems.. we can compare scores.. mine were..

2006 one card 11000.. two cards 13000.. no thousands to throw away there..

2005 one card 18000 two cards 19000.. even less thousands to thow away there..

now if i was running a 16 x 4 mobo i would have put it all down to the x4 factor.. but being as i wasnt i cant..

so what does your amd system with the x 4 factor produce in crossfire 3dmarks.. ????

trog
 

HousERaT

New Member
Joined
Apr 8, 2005
Messages
108 (0.02/day)
Processor Q6700
Motherboard Asus Maximus Formula
Cooling Ultra Extreme 120
Memory Ballistix Tracer 8500
Video Card(s) HD 3870 Crossfire
Storage 74G Raptor
Display(s) A Sony big one
Case Kingwin SK523
Power Supply Corsair HX620
Software Yes :p
The topic is in regards to crossfire in 16x4 vs 16x16.

Until you actually run those two different motherboards you can only guess what I'm talking about. Furthermore if you want to really see the difference you have to have a cpu that can max out two cards which AMD doesn't have at the current time. Any AMD processor right now will surely force your rig to be cpu limited.

My testing was with a P5K and a Maximus Formula with a E6850 running over 4.1Ghz. The best Benchmark that demonstrates the difference is 3DMark03. In this benchmark my Crossfire 16X4 score was 54K and with the 16x16 it was 66K. I don't know how much you know about 3DMark01SE but this is the difference in nature score with the exact same cpu and gpu clocks:
 

Attachments

  • P35vsX38.JPG
    P35vsX38.JPG
    122 KB · Views: 938

jameskelsey

New Member
Joined
Aug 12, 2006
Messages
69 (0.01/day)
Location
Oakland,Mi. USA
Processor i5 750
Motherboard Intel DP55KG
Cooling Water
Memory 4 GB DDR3-1333 HyperX
Video Card(s) Gigabyte HD 6950 2GB 880/1375 1536 shaders stock voltage
Storage WD 500GB Black
Power Supply CORSAIR HX650
so i ran it using the latest wizbang 790x chipset motherboard.. how come i didnt get thousands of 3dmarks to throw away in the first place..

we both have amd systems.. we can compare scores.. mine were..

2006 one card 11000.. two cards 13000.. no thousands to throw away there..

2005 one card 18000 two cards 19000.. even less thousands to thow away there..

now if i was running a 16 x 4 mobo i would have put it all down to the x4 factor.. but being as i wasnt i cant..

so what does your amd system with the x 4 factor produce in crossfire 3dmarks.. ????

trog

If I remember right the 790X has a 16x and a 8x slot.The 790FX has two 16x slots.Would the 16x 8x combo give less performance than the two 16x?
 
Joined
Dec 18, 2005
Messages
8,253 (1.23/day)
System Name money pit..
Processor Intel 9900K 4.8 at 1.152 core voltage minus 0.120 offset
Motherboard Asus rog Strix Z370-F Gaming
Cooling Dark Rock TF air cooler.. Stock vga air coolers with case side fans to help cooling..
Memory 32 gb corsair vengeance 3200
Video Card(s) Palit Gaming Pro OC 2080TI
Storage 150 nvme boot drive partition.. 1T Sandisk sata.. 1T Transend sata.. 1T 970 evo nvme m 2..
Display(s) 27" Asus PG279Q ROG Swift 165Hrz Nvidia G-Sync, IPS.. 2560x1440..
Case Gigabyte mid-tower.. cheap and nothing special..
Audio Device(s) onboard sounds with stereo amp..
Power Supply EVGA 850 watt..
Mouse Logitech G700s
Keyboard Logitech K270
Software Win 10 pro..
Benchmark Scores Firestike 29500.. timepsy 14000..
If I remember right the 790X has a 16x and a 8x slot.The 790FX has two 16x slots.Would the 16x 8x combo give less performance than the two 16x?

they do say 8 x 8 at pci-e 2 (790x) is the same as 16 x 16 normal pci-e.. if what they say is correct the 16 x 16 pci-e 2 (790fx) would be the same as 32 x 32 normal pci.. ????

and forgive me for looking at your system specs houserat which clearly say amd..

so that still leaves us kind of speculating as to the affects of x 4..

perhaps on my system which according to houserat isnt even capable of maximizing a 3870 card it wouldnt make much difference at all..

out of interest i will soon know what difference a fast intel system makes in something like crysis (real life) compared to an amd system.. shame i dont still have my second card else i might know what 16 x 4 does..

trog

ps.. and whatever happens in a bench like 2001 at a million frames per second cant really be used as a gaming example now can it.. ????.. two different worlds again i get confused with.. the world of benching and the world of gaming.. they are fundamentally different..
 
Last edited:

jameskelsey

New Member
Joined
Aug 12, 2006
Messages
69 (0.01/day)
Location
Oakland,Mi. USA
Processor i5 750
Motherboard Intel DP55KG
Cooling Water
Memory 4 GB DDR3-1333 HyperX
Video Card(s) Gigabyte HD 6950 2GB 880/1375 1536 shaders stock voltage
Storage WD 500GB Black
Power Supply CORSAIR HX650
Yea I forgot that it's PCI-E 2.So at 8x 8x on PCI-E 2 you have more than enough bandwith and don't even need 16x in crossfire for any cards out now right?
 

HousERaT

New Member
Joined
Apr 8, 2005
Messages
108 (0.02/day)
Processor Q6700
Motherboard Asus Maximus Formula
Cooling Ultra Extreme 120
Memory Ballistix Tracer 8500
Video Card(s) HD 3870 Crossfire
Storage 74G Raptor
Display(s) A Sony big one
Case Kingwin SK523
Power Supply Corsair HX620
Software Yes :p
Yes,for the most part 8x8 is not fully saturated by current cards. It's when you have the x4 that you start to encounter a bottleneck.
 

jameskelsey

New Member
Joined
Aug 12, 2006
Messages
69 (0.01/day)
Location
Oakland,Mi. USA
Processor i5 750
Motherboard Intel DP55KG
Cooling Water
Memory 4 GB DDR3-1333 HyperX
Video Card(s) Gigabyte HD 6950 2GB 880/1375 1536 shaders stock voltage
Storage WD 500GB Black
Power Supply CORSAIR HX650
Thanks for the info guys.I can save a little money on my next MB by going with the 790X instead of the 790FX
 
Top