PCI Express 2.0 vs 3.0 GPU Gaming Performance Review @ [H]

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,545
PCI Express 2.0 vs 3.0 GPU Gaming Performance Review - Wondering about upgrading to the new Ivy Bridge CPU and PCI Express 3.0 platform? Curious to know if you'll be gaining or losing performance? We compare single, dual-GPU, triple-GPU, single, and multiple display configurations on Ivy Bridge PCIe 3.0 and Sandy Bridge on PCIe 2.0 platforms.
 
Awesome! Been waiting for a review like this for a while.. Kudos!
 
Good article.
The results are what I'd expect, similar to your findings regarding PCIe x8 vs PCIe x16, but it's always good to see theories put to the test.

Although it's probably not worth spending any time on, it would also be interesting to see PCIe 3.0 x8 versus PCIe 2.0 x16.
 
This must have been alot of work.

Good comparison and great idea. Looks like my ASUS Formula 4 has nothing over the Formula 3.:mad:
 
Mondo awesome article. I had been kind of wondering about this since I'm on X58 SLI. Seems, form all the data, there's really no point in moving up just yet. Thanks for the accurate work!
 
A tremendous amount of work went into this by Brent, not to mention a few days of editing. Big kudos to him for getting through it all. He did a great job.
 
Holy Poop Batman. Did Brent even celebrate 4th of July?

Great work guys.....thats alot of testing...wowsa.

Curious as to why the 7970 would get better performance in Batman once tesselation wasnt used...interesting.
 
Nice review. Did you guys consider using the same test bed and just disable 3.0 for 2.0 tests and then enable 3.0 for 3.0 tests to take the IPC out of the equation? Just a thought.
 
Cool review. Learned quite a bit.

Man, Batman just keeps throwing curveballs when it comes to AMD data.
 
This has to be one of the most comprehension reviews I've read. Very detailed and chock full of rumor crushing data. Makes me feel good about my really old mobo accepting a newer Video card.
 
I find it interesting that the max FPS are higher on nearly all setups for PCI-E 2.0 although the average is lower. Very good review.
 
Excellent article, I've been hoping for something like this from you guys. I'd imagine the 2x 680 SLI results translate fairly well to a single GTX 690 as well, or should special considerations be taken for the saturation of a single lane with a dual GPU card?
 
Great job on the testing and I can't really fathom the amount of time and dedication went into this article. Tell Brent to hurry and get well from his suntan he's probably getting right now so we can lock him back in the basement for the next round of testing.

:)
 
As has been proven time and time again, PCI-E 2.0 X16 isn't saturated yet. I'd be interested to see a dual GPU card on a single slot though, as you might see a bigger difference with that setup.
 
Nice review. Did you guys consider using the same test bed and just disable 3.0 for 2.0 tests and then enable 3.0 for 3.0 tests to take the IPC out of the equation? Just a thought.

Yes I did but I thought it would be a bigger benefit to our overall reader base to do it this way when it comes to making purchasing decisions. A purely "scientific" look would be fun, but terribly expensive and we have other priorities.

Excellent article, I've been hoping for something like this from you guys. I'd imagine the 2x 680 SLI results translate fairly well to a single GTX 690 as well, or should special considerations be taken for the saturation of a single lane with a dual GPU card?

I think you would even see less of a difference with on-card SLI comparatively.
 
Absolutely great stuff. Thanks for checking those things for us, like you tested 8/8 16/16 lanes diffference. You save us cash that way :) seems 3.0 pice is still more of marketing gimmick, then real advantage. Wonder though how long it will take to bottleneck 2.0, and if 780 gtx will do it
 
I think objective FPS measures aren't going to show anything.

The one thing I want to see is if frametimes/microstutter have smoothed out with 3.0 vs 2.0...if yall want to send me a couple 680s and 7970s, I'll do the testing myself. :D
 
Whew looks like I'm safe with my 2.0 :) especially since I doubt both of my 6870's are tapping most of that throughput.
 
Although it's probably not worth spending any time on, it would also be interesting to see PCIe 3.0 x8 versus PCIe 2.0 x16.

Yeah, it would be interesting to see if the lower overhead for PCIe 3.0 makes any difference, or if it is really just a bandwidth issue.
 
Interesting, but it's just not clear whether the benefits are coming from PCIe 3.0 or Ivy Bridge.

I currently own a Z68 motherboard. My motherboard is compatible with Ivy Bridge but not PCIe 3.0. I'd love to know exactly what I would be missing if I were running Ivy at PCIe 2.0. Without knowing how much of the performance improvement is coming from Ivy Bridge, I'm not sure how this really presents any useful information about PCIe 3.0.

Really this article is comparing P67/Sandy Bridge vs Z77/Ivy Bridge as whole platforms. That's fine, and certainly useful if you're looking to make that transition, but not very useful if you're just looking to compare PCIe 2.0 vs. 3.0. Unfortunately that's the title of the article...
 
Great article. Given the IPC gains on IB, I think it's safe to say that there is no tangible increase from PCI-e 3.0 to 2.0 (the gains in certain games are within the margin of IPC gains, unless one claims those games are purely GPU constrained at the settings tested).

Fixed, thanks. - Kyle
 
Last edited by a moderator:
Interesting, but it's just not clear whether the benefits are coming from PCIe 3.0 or Ivy Bridge.

I currently own a Z68 motherboard. My motherboard is compatible with Ivy Bridge but not PCIe 3.0. I'd love to know exactly what I would be missing if I were running Ivy at PCIe 2.0. Without knowing how much of the performance improvement is coming from Ivy Bridge, I'm not sure how this really presents any useful information about PCIe 3.0.

Really this article is comparing P67/Sandy Bridge vs Z77/Ivy Bridge as whole platforms. That's fine, and certainly useful if you're looking to make that transition, but not very useful if you're just looking to compare PCIe 2.0 vs. 3.0. Unfortunately that's the title of the article...

This too, wouldn't it have been possible to run the Ivy processor in the P67 board to control for IPC differences?
 
This too, wouldn't it have been possible to run the Ivy processor in the P67 board to control for IPC differences?

Fairly certain that depends on whether or not the motherboard manufacturer updated the BIOS to support Ivy. Mine does, but my board also supports PCI-e 3
 
Great article guys.

Now we just have the vram issue left to put to rest. Unless I missed that one. ;)
 
You guys can merit the scientific values of doing it this way or doing it that way, but what we do on a daily basis is supply our readers with free information to help them make informed purchasing decisions. That is our focus and that is what it will mostly remain.
 
Look at games that typicaly perform better with cpu scaling, say skyrim.
See ivy bridge perform extremely better in skyrim @ same cpu frequency than sandy.
Eureka guys!! ipc makes da game faster! :eek:
 
Interesting, but it's just not clear whether the benefits are coming from PCIe 3.0 or Ivy Bridge.

I currently own a Z68 motherboard. My motherboard is compatible with Ivy Bridge but not PCIe 3.0. I'd love to know exactly what I would be missing if I were running Ivy at PCIe 2.0. Without knowing how much of the performance improvement is coming from Ivy Bridge, I'm not sure how this really presents any useful information about PCIe 3.0.

Really this article is comparing P67/Sandy Bridge vs Z77/Ivy Bridge as whole platforms. That's fine, and certainly useful if you're looking to make that transition, but not very useful if you're just looking to compare PCIe 2.0 vs. 3.0. Unfortunately that's the title of the article...

If there was a significant difference in performance, your point would hold more water.

However, given that the difference was so little, and some of it coming from Ivy Bridge IPC Improvements, it pretty much nails the point, REGARDLESS of testing method, PCI-E 3.0 makes very little of any difference in 1 2 or 3 card setups.

If you want to know IBs 2.0 performance that's another story.
 
Well done. Thanks for taking the time and working so hard on this PCI 2 vs 3 issue. I would've thought for sure that there would've been a little bit better performance increase in PCI 3 but, guess not. So, I'm now very curious why that is?

I'm reminded of this article:

A spokesperson from Nvidia:

“Nvidia is a key contributor to the industry’s development of PCI Express 3.0, which is expected to have twice the data throughput of the current generation (2.0). Whenever there is a major increase in bandwidth like that, applications emerge that take advantage of it. This will benefit consumers and professionals with increased graphics and computing performance from notebooks, desktops, workstations, and servers ...that have a GPU.”

http://www.tomshardware.com/reviews/pci-express-3.0-pci-sig,2695-5.html

I'm also reminded of the fiasco between MSI and Gigabyte:

MSI Calls Out Gigabyte for "Not True PCIe 3.0"

Gigabyte Sets Record Straight on PCIe 3.0 Support

I wonder what the differences would be in Adobe CS6 rather than just games, but perhaps the differences would be even less noticeable?

Is this a case whereby the GPU's &/or motherboards or perhaps CPU's simply aren't taking full advantage of PCIe 3.0 ??? Is there a chance that the next new round of GPU's will perform better in PCI 3? I'm just shocked that there's hardly a performance increase at all and wonder why.
 
Tremendous work guys and a very sincere thank you! I know how much work is involved in this type of testing.
 
Excellent article. Thorough work from Brent as usual. I was thinking of a Z77 board for my next upgrade but might opt for a now cheaper Z68 mobo instead. Nothing but the [H]ard facts, that's what I like about this site :)
 
Great review. My side question is, I see you used a P67 mobo for 2.0 testing. I have an ASUS P8Z68-V/Gen3 mobo which in has supposedly has two PCIe 3.0 slots (according to the newegg specs). Is my gtx 670 using 2.0 or 3.0? Is there a way to test this or is it tied to what kind of CPU I'm using - sandy vs ivy? Please excuse my ignorance
 
Great review. My side question is, I see you used a P67 mobo for 2.0 testing. I have an ASUS P8Z68-V/Gen3 mobo which in has supposedly has two PCIe 3.0 slots (according to the newegg specs). Is my gtx 670 using 2.0 or 3.0? Is there a way to test this or is it tied to what kind of CPU I'm using - sandy vs ivy? Please excuse my ignorance

Since your CPU only has 16 PCIe 2.0 lanes, your GTX 670 will be @ x16 2.0. You will need an Ivy Bridge CPU that is PCIe 3.0 compliant - by this I mean a i5 or i7 CPU since all i3 models are PCIe 2.0.
 
Back
Top