GeForce GTX 680 Release Driver Limits PCI-Express to Gen 2.0 on X79/SNB-E Systems

[btarunr]

Users of systems running Core i7 processors in the LGA2011 package and X79 chipset motherboards will find that their GeForce GTX 680 will not run at PCI-Express 3.0 x16 mode, using launch drivers. This is because NVIDIA decided against implementing Gen 3.0 support for the new GPU on X79/SNB-E systems, at the very last moment. Pre-launch drivers, including the driver that ships on discs with the graphics cards, will let the GPU run at Gen 3.0 mode on some X79/SNB-E systems, however the post-launch stable drivers (such as GeForce 301.10 WHQL), won't.

The specifications page of GeForce GTX 680 on GeForce.com has been updated to mention this:

GeForce GTX 680 supports PCI Express 3.0. The Intel X79/SNB-E PCI Express 2.0 platform is only currently supported up to 5GT/s (PCIE 2.0) bus speeds even though some motherboard manufacturers have enabled higher 8GT/s speeds.

Sources in the PC motherboard industry we spoke with, confirmed this change. PCI-Express 3.0 x16, for now, might only run on upcoming "Ivy Bridge" Core systems, running on motherboards with PCI-Express 3.0 compliant components. NVIDIA could be working to fix the issue.

Update 3/23, 21:56
NVIDIA courteously responded to our article, with a statement. Here's the statement verbatim:

While X79/SNB-E is a native Gen2 platform, some motherboard manufacturers have enabled Gen3 speeds. With our GTX 680 launch drivers, we will only be supporting Gen2 speeds on X79/SNB-E while we work on validating X79/SNB-E at these faster speeds. Native Gen3 chipsets (like Ivy Bridge) will still run at full Gen3 speeds with our launch drivers.

GeForce GTX 680 supports PCI Express 3.0. It operates properly within the SIG PCI Express Specification and has been validated on multiple upcoming PCI Express 3.0 platforms. Some motherboard manufacturers have released updated SBIOS to enable the Intel X79/SNB-E PCI Express 2.0 platform to run at up to 8GT/s bus speeds. NVIDIA is currently working to validate X79/SNB-E with GTX 680 at these speeds with the goal of enabling 8GT/s via a future software update. Until this validation is complete, the GTX 680 will operate at PCIE 2.0 speeds on X79/SNB-E-based motherboards with the latest web drivers.

This statement confirms our assertions made in the article. NVIDIA's statement on the GeForce.com product page originally formed the basis of this article, which we pasted verbatim, while preparing the article. Motherboard manufacturers have extensively marketed their LGA2011 products as natively supporting PCI-Express Gen 3.0 (8 GT/s). What's more, the datasheet of Intel Core i7 LGA2011 processors (refer 1.2.2) clearly mentions the PCI-Express root complex as supporting 8 GT/s. AMD Southern Islands GPUs have supported PCI-Express Gen 3.0 on LGA2011 platforms since day one.

[MeanBruce]

So glad I didn’t buy into the X79 platform, the chipset is so incomplete. Was going to go with a 3820 and Rampage 4 Extreme, then thought I would regret my decision, so happy I waited for the 3770K (still waiting) and the Maximus 5 Extreme which will support PCIe 3.0 and maybe even thunderbolt connectivity, let’s see an X79 board do that!


...was X79 rushed? Even the owners complain the chipset is not feature laden, maybe X89.

[btarunr]

I think this issue is due to discrepancy with validation. When Intel was testing and validating Gen 3.0 on SnB-E, only AMD had working samples of Gen 3.0 compliant GPUs, the validation went through, and Intel mentioned that only "some" PCI-E Gen 3.0 devices may run on Gen 3.0 mode, on SnB-E systems.

NVIDIA is probably a little late with being 100% sure that its Gen 3.0 mode will work with SnB-E PCIe roots. When it is, it could release a driver update than brings back Gen 3.0 support.

[MeanBruce]

WOW, that post just magically appeared, thoroughly impressed, thanks Btarunr…

[btarunr]

Originally Posted by MeanBruce

WOW, that post just magically appeared, thoroughly impressed, thanks Btarunr…

Yeah, we're in the magic business.

[buggalugs]

Sounds a little dodgy to me. Sounds like X79 validation would have taken time, maybe weeks, so they thought dont worry about X79 its like 3% of the market just get the card on the market before more people buy a 7970.

[Live OR Die]

Sad but wont stop me buying one i can see support coming later on.

[btarunr]

Tests have shown that Gen 2.0 x16 doesn't bottleneck HD 7970. It shouldn't bottleneck a GTX 680, either. Where this issue could bite NVIDIA is performance at x8. PCI-Express 3.0 x8 is a better bet than PCI-Express 2.0 x8.

[Aquinus]

Originally Posted by MeanBruce

So glad I didn’t buy into the X79 platform, the chipset is so incomplete. Was going to go with a 3820 and Rampage 4 Extreme, then thought I would regret my decision, so happy I waited for the 3770K (still waiting) and the Maximus 5 Extreme which will support PCIe 3.0 and maybe even thunderbolt connectivity, let’s see an X79 board do that!


...was X79 rushed? Even the owners complain the chipset is not feature laden, maybe X89.

I don't know, I feel that my P9X79 Deluxe is pretty feature filled. I have my 3820 running a healthy 24/7 4.75ghz overclock. It is definitely one of the nicest motherboards I've used. I don't know where ASUS would fit more stuff on this board.

http://www.asus.com/Motherboards/Int.../P9X79_DELUXE/

[eidairaman1]

Originally Posted by btarunr

Tests have shown that Gen 2.0 x16 doesn't bottleneck HD 7970. It shouldn't bottleneck a GTX 680, either.

they are only doing this to expand the lanes before they become congested and to spur sales

[hhumas]

surprising

[Dj-ElectriC]

Something that has no impact on performance isn't working on something that is just out and will be updated soon.

Great to know btarunr.

[btarunr]

Originally Posted by Dj-ElectriC

Something that has no impact on performance

In x8, it might, and that might impact 3-way and 4-way SLI performance.

We know that PCI-Express 3.0 x16 vs. PCI-Express 2.0 x16 has no difference. That's barely something to worry about. But we know PCI-Express 2.0 x8 bottlenecks HD 7970, so it should bottleneck GTX 680, too. In 3-way and 4-way SLI, the cards are reconfigured to run at x8. Since GTX 680 doesn't do Gen 3.0 on SnB-E, those cards will be running at PCI-Express 2.0 x8. Had they ran at PCI-Express 3.0 x8, they would end up with bandwidth comparable to PCI-Express 2.0 x16.

[GC_PaNzerFIN]

Doesn't affect me with Z68/gen3 board. Nothing can stop me from getting my totally useless pci-e 3.0 if I slap in IB CPU.

[Aquinus]

Originally Posted by btarunr

In x8, it might, and that might impact 3-way and 4-way SLI performance.

We know that PCI-Express 3.0 x16 vs. PCI-Express 2.0 x16 has no difference. That's barely something to worry about. But we know PCI-Express 2.0 x8 bottlenecks HD 7970, so it should bottleneck GTX 680, too. In 3-way and 4-way SLI, the cards are reconfigured to run at x8. Since GTX 680 doesn't do Gen 3.0 on SnB-E, those cards will be running at PCI-Express 2.0 x8. Had they ran at PCI-Express 3.0 x8, they would end up with bandwidth comparable to PCI-Express 2.0 x16.

You can run two full 2.0 x16 lanes on LGA2011, that doesn't help you with 1155 though.

[btarunr]

Originally Posted by Aquinus

You can run two full 2.0 x16 lanes on LGA2011, that doesn't help you with 1155 though.

You can, but I'm talking about 3-way and 4-way SLI scenarios, where the two x16 links from the CPU are split into x8 links.


1155 isn't part of the problem. This issue only affects X79/SnB-E/LGA2011 platforms.

[Aquinus]

Originally Posted by btarunr

1155 isn't part of the problem. This issue only affects X79/SnB-E/LGA2011 platforms.

I do understand what you're saying. It's the 3.0 not kicking in that could make them bottleneck. I'm just saying since 1155 doesn't have 3.0 (yet), it will run into a bottleneck. It's the same problem for 3 or 4 GPUs on SB-E. I'm not disagreeing though, but 1155 would run into a similar problem, not that it doesn't exist.

[rpsgc]

Nothing new here... just NVIDIA being NVIDIA.

[cadaveca]

Originally Posted by btarunr

You can, but I'm talking about 3-way and 4-way SLI scenarios, where the two x16 links from the CPU are split into x8 links.


1155 isn't part of the problem. This issue only affects X79/SnB-E/LGA2011 platforms.

SNB-E has 40 lanes on CPU, so Tri-SLi should get x16, x16, and x8.

[Live OR Die]

Got this reply by one of the guys from asus that work on there bios's, So that would be why.

well there was an old driver that made it 3.0 for x79 but i think becos once oc bclk to 105 then the vga cannot handle so u bsod when entering os so then the new drivers forced it to gen 2 on x79

[Jiraiya]

Here

http://www.4gamer.net/games/022/G002210/20120323002/

[btarunr]

Find NVIDIA confirming us in its statement

[15th Warlock]

I don't care, Nvidia didn't support surround on X79 when it released and they supported it a week or so after it was released, if they say Gen 3 support is coming then I believe them, I don't think my RIVE board will bottleneck my 680s either way...

To me this is a non issue.

[newtekie1]

Intel says SB-E/X79 supports 8GT/s, but you'll notice that Intel doesn't say PCI-E 3.0 anywhere in its marketing or whitesheets on the processor and chipset. They didn't get the platform certified for PCI-E 3.0, so it either lacks some feature that PCI-SIG requires or they just didn't bother certifying it. Either way, there is no guarantee that PCI-E 3.0 devices will run at PCI-E 3.0 speeds, and that is no one's but Intel's fault.

Plus, I doubt there is really any performance lost from not using PCI-E 3.0, so who really cares?

[ViperXTR]

hmm, if i remember correctly in the old bench of the HD 7900 series, impact of PCI-E 3.0 vs 2.0 has very little effect on gaming, but its noticeable on GPGPU tasks that requires massive amount of data to be delivered in the PCIE lane, i wonder if this cripples the GK104 more with its inferior GPGPU performance.