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

GeForce GTX 980 PCI-Express Scaling

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
21,527 (3.53/day)
Processor Core i7-8700K
Memory 32 GB
Video Card(s) RTX 3080
Display(s) 30" 2560x1600 + 19" 1280x1024
Software Windows 10 64-bit
PCI-Express x16 3.0 is well established in the market, and the majority of gamers are using the interface. But what happens if you end up in a slot-bandwidth-constrained situation? We are testing NVIDIA's latest GeForce GTX 980 flagship in 17 games, at four resolutions, including 4K, to assess what performance to expect.

Show full review
 
Last edited:
Joined
May 1, 2012
Messages
1,024 (0.32/day)
Location
New Jersey, USA
System Name Current Rig
Processor Intel i7 920 oc to 3.5ghz
Motherboard MSI x58 Platinum
Cooling Cooler Master Seidon 120M
Memory G.Skill 6gb ddr3 1333mhz 7-7-7-18
Video Card(s) Gigabyte GTX 670 OC WindForce 3X GV-N670OC-2GD
Storage Samsung EVO SSD
Display(s) Viewsonic 1440x900 5ms
Case CoolMaster HAF 932
Power Supply Coolmax 950w
Software Win 7 64bit
Thanks for doing an update W1zzard.
 
Joined
Aug 2, 2012
Messages
740 (0.24/day)
Location
Netherlands
System Name TheDeeGee's PC
Processor Intel Core i7 4770K
Motherboard Gigabyte Z87X-UD5H
Cooling Noctua NH-D15S
Memory Crucial Ballistix Tactical LP 16GB
Video Card(s) MSI GTX 1070 AERO OC (Custom Cooler)
Storage Crucial BX500 1TB, Crucial BX500 2TB, TeamGroup CX2 SSD 1TB
Display(s) EIZO CX240
Case Fractal Design Define 7
Audio Device(s) Creative SoundBlaster ZxR
Power Supply Seasonic Prime PX 850
Mouse Logitech G500s
Keyboard Logitech G710+
Software Windows 10 Pro 64-Bit
Interesting to see that 3.0 8x is sometimes faster than 3.0 16x.
 
Joined
Feb 18, 2006
Messages
5,146 (0.94/day)
Location
AZ
System Name Thought I'd be done with this by now
Processor i7 4790K 4.4GHZ turbo currently at 4.6GHZ at 1.16v
Motherboard MSI Z97-G55 SLI
Cooling Scythe Mugen 2 rev B (SCMG-2100), stock on gpu's.
Memory 8GB G.SKILL Ripjaws Z Series DDR3 2400MHZ 10-12-12-31
Video Card(s) EVGA GTX 760 Superclocked replaced HIS R9 290 that was artifacting
Storage 1TB MX300 M.2 OS + Games, 4x ST31000524NS in Raid 10 Storage and Backup, external 2tb backup,
Display(s) BenQ GW2255 surprisingly good screen for the price.
Case Raidmax Scorpio 668
Audio Device(s) onboard HD
Power Supply EVGA 750 GQ
Software Windows 10
Benchmark Scores no one cares anymore lols
idk it seems extremely odd on how dynamic and ryse, wow, and wolfenstein differences are. It really seems like there is a frame limiter detecting the speed of the bus and adjusting the limit accordingly.
 
Joined
May 21, 2008
Messages
4,092 (0.88/day)
Location
Iowa, USA
System Name FUTURE CUBE!
Processor intel Core i5 6600k
Motherboard Gigabyte Z170X-Gaming 7
Cooling Phanteks PH-TC14PE BK
Memory G.Skill TridentZ 3000 Mhz C15 32GB 2x16GB
Video Card(s) Gigabyte Aorus 1080 Ti
Storage 2x M.2 Samsung Evo 250GB/500GB / WD Blue 500GB / 2x RAID1 Toshiba P300 3TB
Display(s) Samsung C24FG70 1080p 144hz Quantum Dot/ASUS VH226H 1080p 21.5"
Case "THE CUBE" Custom built, pure Red Alder wood
Audio Device(s) Creative Sound Core3D/ Logitech Z-2300 200 watts/ Beyerdynamic DT 880
Power Supply Seasonic X Gold 650W
Mouse Logitech G700
Keyboard Logitech G910
Software Windows 10 Pro
Nice review still no real difference.

Wow big boost from WoW MoP to WoW WoD. 161FPS to 231FPS at 1080p
 
Joined
Feb 8, 2012
Messages
2,975 (0.91/day)
Location
Zagreb, Croatia
System Name Windows 10 64-bit Core i7 6700
Processor Intel Core i7 6700
Motherboard Asus Z170M-PLUS
Cooling Corsair AIO
Memory 2 x 8 GB Kingston DDR4 2666
Video Card(s) Gigabyte NVIDIA GeForce GTX 1060 6GB
Storage Western Digital Caviar Blue 1 TB, Seagate Baracuda 1 TB
Display(s) Dell P2414H
Case Corsair Carbide Air 540
Audio Device(s) Realtek HD Audio
Power Supply Corsair TX v2 650W
Mouse Steelseries Sensei
Keyboard CM Storm Quickfire Pro, Cherry MX Reds
Software MS Windows 10 Pro 64-bit
I get the Id Tech 5 engine with it's constant mega texture streaming, but I don't get what Ryse has to move over the PCIE bus other than draw calls. All resources that kind of games need are usually preloaded in VRAM at level loading.
 
Joined
Jun 28, 2014
Messages
2,388 (1.00/day)
Location
Shenandoah Valley, Virginia USA
System Name Home Brewed
Processor i9-7900X and i7-8700K
Motherboard ASUS ROG Rampage VI Extreme & ASUS Prime Z-370 A
Cooling Corsair 280mm AIO & Thermaltake Water 3.0
Memory 64GB DDR4-3000 GSKill RipJaws-V & 32GB DDR4-3466 GEIL Potenza
Video Card(s) 2X-GTX-1080 SLI & 2 GTX-1070Ti 8GB G1 Gaming in SLI
Storage Both have 2TB HDDs for storage, 480GB SSDs for OS, and 240GB SSDs for Steam Games
Display(s) ACER 28" B286HK 4K & Samsung 32" 1080P
Case NZXT Source 540 & Rosewill Rise Chassis
Audio Device(s) onboard
Power Supply Corsair RM1000 & Corsair RM850
Mouse Generic
Keyboard Razer Blackwidow Tournament & Corsair K90
Software Win-10 Professional
Benchmark Scores yes
Thanks for the post W1zzard, it was a good read.
 

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
21,527 (3.53/day)
Processor Core i7-8700K
Memory 32 GB
Video Card(s) RTX 3080
Display(s) 30" 2560x1600 + 19" 1280x1024
Software Windows 10 64-bit
All resources that kind of games need are usually preloaded in VRAM at level loading.
Not anymore

It really seems like there is a frame limiter detecting the speed of the bus and adjusting the limit accordingly.
I see no mechanism how a game could do that (detecting PCIe bandwidth is not trivial, I know from GPU-Z). Also why would a game do that, and why would a gamedev invest time for it :)
 
Joined
Oct 31, 2014
Messages
3 (0.00/day)
Great article.
As someone with a dual-gpu card, a 7990, is there any chance you could benchmark just a few games with a dual-gpu card?
The reason why I'm asking, is that I'm still running a x58 system with an overclocked i7 920 @ 4.2ghz but it's still pcie 2.0.
 
Joined
Jul 7, 2014
Messages
152 (0.06/day)
Location
Columbia, SC
Processor Intel 2500k OCed at 4.6ghz
Motherboard Intel Z77
Cooling Thermalright Macho Rev.A
Memory 8GB G.Skill 2133
Video Card(s) Gigabyte GTX 670 Windforce 3X OCed at 1050mhz base and 1600mhz vram
Storage Mushkin Enhance 256gb SSD, Western Digital 750gb and 3TB HHDs
Display(s) Asus 24" 1080p
Case Lian-Li Mid Tower
Audio Device(s) Mobo sound
Power Supply SeaSonic 560 watt gold
Mouse Logitec 3 button laser mouse
Keyboard Das Keyboard Model S (the blank key model)
Software Windows 8.1 64 bit
Nice article, but it does prove that it's pointless to waste money on a socket 2011 platform for a gaming machine, or, that 3 way 8x SLI on a Z97 platform is going to only show a 0-5% decrease in performance for a CPU, RAM and MOBO setup that costs at least 50% less. In fact, that loss in performance would likely be mitigated by the faster clock speeds native to the Devil's Canyon chips.

Socket 2011 gaming rigs are for people with more money than sense.
 
Last edited:
Joined
Oct 31, 2014
Messages
3 (0.00/day)
Socket 2011 gaming rigs are for people with more money than sense

Strictly speaking about pci-e lanes, yes but those 4 or 8 extra threads of a 2011v3 chip might come in handy in the future, quite a few game engines multithread pretty well already.
 
Joined
Apr 27, 2011
Messages
53 (0.01/day)
I wonder if AMD cards would behave differently? I was thinking of buying the next high end card from them for my pci-e 2.0 board and was just thinking about this a few days ago. If new games like Ryse and Wolfenstein start showing a difference it might be finally time to start planning for an update for my 2500K.
 
Joined
Apr 17, 2014
Messages
222 (0.09/day)
System Name GSYNC
Processor i9-10920X
Motherboard EVGA X299-FTW
Cooling Custom water loop: D5
Memory G.Skill RipJawsZ 16GB 2133mhz 9-11-10-28
Video Card(s) (RTX2080)
Storage OCZ vector, samsung evo 950, Intel M.2 1TB SSD's
Display(s) ROG Swift PG278Q, Acer Z35 and Acer XB270H (NVIDIA G-SYNC)
Case 2x Corsair 450D, Corsair 540
Audio Device(s) sound blaster Z
Power Supply EVGA SuperNOVA 1300 G2 Power
Mouse Logitech proteus G502
Keyboard Corsair K70R cherry red
Software WIN10 Pro (UEFI)
Benchmark Scores bench score are for people who don't game.
Excellent article guys! thanks!
 

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
21,527 (3.53/day)
Processor Core i7-8700K
Memory 32 GB
Video Card(s) RTX 3080
Display(s) 30" 2560x1600 + 19" 1280x1024
Software Windows 10 64-bit
I wonder if AMD cards would behave differently? I was thinking of buying the next high end card from them for my pci-e 2.0 board and was just thinking about this a few days ago. If new games like Ryse and Wolfenstein start showing a difference it might be finally time to start planning for an update for my 2500K.
Great article.
As someone with a dual-gpu card, a 7990, is there any chance you could benchmark just a few games with a dual-gpu card?
The reason why I'm asking, is that I'm still running a x58 system with an overclocked i7 920 @ 4.2ghz but it's still pcie 2.0.
I have no plans for any other PCIe scaling tests, not until new cards are released from AMD.
 
Joined
Sep 27, 2012
Messages
26 (0.01/day)
System Name PC_3770K
Processor Intel 3770K
Motherboard Asus Maximus Gene V
Cooling Cooler Master Hyper 212 EVO
Memory Kingston HyperX 8GB DDR3 @2400MHz
Video Card(s) ZOTAC GeForce GTX 1080 Mini
Storage Samsung 850 Evo 250GB / WD 320GB + SEAGATE 500GB
Display(s) DELL U2713HM 27"@2560*1440
Case Corsair Obsidian 350D MATX
Audio Device(s) Integrated Asus SupremeFX III
Power Supply Seasonic Modular G Series 550W
Mouse Asus Rog Sica
Keyboard Redragon Devarajas Mechanical KB
Software Windows 10 Home x64
First page:
"While PCI-Express 1.0 pushes 250 MB/s per direction, PCI-Express 2.0 pushes 500 MB/s, and PCI-Express 3.0 doubles that to 1 GB/s. While the resulting absolute bandwidth of PCI-Express 3.0 x16, 32 GB/s, might seem like overkill, the ability to push that much data per lane could come to the rescue of configurations such as 8-lanes (x8) and 4-lanes (x4)."


PCI-Express 3.0 at 16x has a ~16GB/s bandwidth, not 32GB/s
 

newtekie1

Semi-Retired Folder
Joined
Nov 22, 2005
Messages
27,499 (4.96/day)
Location
Indiana, USA
Processor Intel Core i7 10850K@5.2GHz
Motherboard AsRock Z470 Taichi
Cooling Corsair H115i Pro w/ Noctua NF-A14 Fans
Memory 32GB DDR4-3600
Video Card(s) ASUS Strix GTX 1080Ti
Storage 500GB SX8200 Pro + 8TB with 1TB SSD Cache
Display(s) QNIX QX2710 1440p@120Hz
Case Fractal Design Define S
Audio Device(s) Onboard is good enough for me
Power Supply eVGA SuperNOVA 1000w G3
Software Windows 10 Pro x64
W1z is right, it's 32GB/s.

1GB/s x 16 x 2 = 32GB/s.

PCI-E is a duplex connection, so each lane is 1GB/s in two directions, so the total bandwidth is 2GB/s per lane(1GB/s in each direction). So the total bandwidth for an x16 3.0 slot is 32GB/s.
 
Last edited:
Joined
Jun 28, 2014
Messages
2,388 (1.00/day)
Location
Shenandoah Valley, Virginia USA
System Name Home Brewed
Processor i9-7900X and i7-8700K
Motherboard ASUS ROG Rampage VI Extreme & ASUS Prime Z-370 A
Cooling Corsair 280mm AIO & Thermaltake Water 3.0
Memory 64GB DDR4-3000 GSKill RipJaws-V & 32GB DDR4-3466 GEIL Potenza
Video Card(s) 2X-GTX-1080 SLI & 2 GTX-1070Ti 8GB G1 Gaming in SLI
Storage Both have 2TB HDDs for storage, 480GB SSDs for OS, and 240GB SSDs for Steam Games
Display(s) ACER 28" B286HK 4K & Samsung 32" 1080P
Case NZXT Source 540 & Rosewill Rise Chassis
Audio Device(s) onboard
Power Supply Corsair RM1000 & Corsair RM850
Mouse Generic
Keyboard Razer Blackwidow Tournament & Corsair K90
Software Win-10 Professional
Benchmark Scores yes
Thank you for this article. It makes me feel more at ease that my i7 2700k which only supports x16 2.0 is not going to restrict current high end cards.

I don't get any restriction with my i7-2600 System. I have a pair of R9-280X-OC cards in it.
I just bought a third 280X-OC card, but I'll have to swap out my motherboard to run Tri-Crossfire with it.
 
Joined
Apr 29, 2014
Messages
3,911 (1.59/day)
Location
Texas
System Name SnowFire / The Reinforcer / Portable?
Processor i7 10700K 5.1ghz (24/7) / 2x Xeon E52650v2 / AMD Ryzen 5 3600X
Motherboard Asus Strix Z490 / Dell Dual Socket (R720) / Asrock X570 ITX
Cooling RX 360mm + 140mm Custom Loop / Dell Stock / Noctua L9i (Yes L9i)
Memory Corsair RGB 16gb DDR4 3000 CL 16 / DDR3 128gb 16 x 8gb / Corsair RGB 3200 16gb
Video Card(s) GTX Titan XP (2025mhz) / Asus GTX 950 (No Power Connector) / GTX 970 (Temp)
Storage Samsung 970 1tb NVME and 2tb HDD x4 RAID 5 / 300gb x8 RAID 5 / 2x Samsung 850 Pro 512gb
Display(s) Acer XG270HU, Samsung G7 Odyssey (1440p 240hz) / HP Omen 1080p 240hz
Case Thermaltake Cube / Dell Poweredge R720 Rack Mount Case / Fractal Design Node 202
Audio Device(s) Realtec ALC1150 (On board)
Power Supply Rosewill Lightning 1300Watt / Dell Stock 750 / Brick / Fractal Design 450 Watt Bronze
Mouse Logitech G5
Keyboard Logitech G19S
Software Windows 10 Pro / Windows Server 2016 / Windows 10 Pro
Wow great article @W1zzard, its nice to see some formal testing updated with a recent card about PCI-E bandwidth. Its such an odd subject to get into because there are not many areas that testing to this extent is done to show people when they question it.
 
Joined
Dec 6, 2005
Messages
10,799 (1.95/day)
Location
Manchester, NH
System Name Working on it ;)
Processor I7-4790K (Stock speeds right now)
Motherboard MSI Z97 U3 Plus
Cooling Be Quiet Pure Rock Air
Memory 16GB 4x4 G.Skill CAS9 2133 Sniper
Video Card(s) GIGABYTE Vega 64 (Non Reference)
Storage Samsung EVO 500GB / 8 Different WDs / QNAP TS-253 8GB NAS with 2x2Tb WD Black
Display(s) 34" LG 34CB88-P 21:9 Curved UltraWide QHD (3440*1440) *FREE_SYNC*
Case Rosewill Challenger
Audio Device(s) Onboard + HD HDMI
Power Supply Corsair HX750 (love it)
Mouse Logitech G5
Keyboard Corsair Strafe RGB & G610 Orion Red
Software Win 10 upgraded from Win 7 Pro
∆ IIRC, the story was pretty much the same with AGP...

:lovetpu:
 
Joined
Sep 29, 2013
Messages
97 (0.04/day)
Processor Intel i7 4960x Ivy-Bridge E @ 4.6 Ghz @ 1.42V
Motherboard x79 AsRock Extreme 11.0
Cooling EK Supremacy Copper Waterblock
Memory 65.5 GBs Corsair Platinum Kit @ 666.7Mhz
Video Card(s) PCIe 3.0 x16 -- Asus GTX Titan Maxwell
Storage Samsung 840 500GBs + OCZ Vertex 4 500GBs 2x 1TB Samsung 850
Audio Device(s) Soundblaster ZXR
Power Supply Corsair 1000W
Mouse Razer Naga
Keyboard Corsair K95
Software Zbrush, 3Dmax, Maya, Softimage, Vue, Sony Vegas Pro, Acid, Soundforge, Adobe Aftereffects, Photoshop
@ wizard,

"The most surprising find to me is the huge performance hit some of the latest games take when running on limited PCIe bandwidth. The real shocker here is certainly Ryse: Son of Rome, based on Crytek's latest CryEngine 4. The game seems to constantly stream large amounts of data between the CPU and GPU, taking a large 10% performance hit by switching to the second-fastest x16 3.0 configuration. At x4 1.1, the slowest setting we tested, performance is torn down to less than a third, while running lower resolutions! Shocking!

Based on id's idTech5 engine, another noteworthy title with large drops in performance is Wolfenstein: The New Order. Virtual Textures certainly look great in-game, providing highly detailed, non-repeating textures, but they also put a significant load on the PCI-Express bus. One key challenge here is to have texture data ready for display in-time. Sometimes too late, it manifests as the dreaded texture pop-in some users have been reporting.

Last but not least, World of Warcraft has received a new rendering engine for its latest expansion Warlords of Draenor. While the game doesn't look much different visually, Blizzard made large changes under the hood, changing to a deferred rendering engine which not only disallows MSAA in-game, but also requires much improved PCI-Express bandwidth."


All you're really proven is there is a slight gain or loss in certain scenarios, but you don't go into depth as to why PCIe 16x 2.0 performs equal to or better than 3.0. Doubt it really matters. Though, I agree with some points of your message, but you aren't really proving any more than a drop or gain in average of what, around 10% in any of the other scenarios. It seems informative, but also a waste of your own time. In addition to that, games that are either MMOs or highly-progressed games like Crysis 3, BF4, Wolfenstein 3D, and others, will make better use of 3.0 over 2.0. One good example of this will probably be Star Citizens in the not to distant future. I would highly suggest using Planetside 2 and EQN for upcoming benches. For the higher resolutions (above 1080p), you'll probably see a higher use in 3.0 if you enabled more AA at 4k resolutions.

Here's an idea. Instead of sitting in the Shrine in World of Warcraft, why don't you conduct test during a Garrosh 25 man fight at Ultra Settings. Tell us what the results are of the PCIe Lane Saturations after that. I would think that would be more vital information than just staring at a wall to stare at the in-game FPS meter to see how high your FPS can get. Also, why don't you measure the same games with 2-way, 3-way, and 4-way SLI. It's not like NVidia has anything to hide right...
 
Joined
Jun 13, 2012
Messages
1,216 (0.39/day)
System Name desktop
Processor i7-4770k
Motherboard Asus z87-plus
Cooling Corsair h80
Memory 32gb G.Skill Ares @ 2400mhz
Video Card(s) EVGA GeForce GTX 1080 SC (ACX 3.0)
Storage 2x Samsung 1tb 850evo SSD, 3x WD drives, 2 seagate
Display(s) Acer Predator XB271HU 27inch IPS G-Sync 165hz
Audio Device(s) Sound Blaster x-FI Platium, Turtle beach Elite pro 2 + superamp.
Power Supply OCZ Z Series 850W (10 years strong)
Mouse Logitech G502 hero
Keyboard Logitech G710+
Interesting to see that 3.0 8x is sometimes faster than 3.0 16x.

I probably within the realm of error you could say something in video slightly diff happened cause the small difference or cpu usage.

I wonder if AMD cards would behave differently? I was thinking of buying the next high end card from them for my pci-e 2.0 board and was just thinking about this a few days ago. If new games like Ryse and Wolfenstein start showing a difference it might be finally time to start planning for an update for my 2500K.

I would probably expect AMD cards would yield pretty much same kinda results.
 
Top