Hi,
My EVGA GTX 1070 SC has been running fine. There are still no visible issues.
 
When I originally set it up years ago, I verified that all the connections were as expected, and I was running in x16 PCIe 3.0 mode in a x16 PCIe 3.0.  I am using the primary 16 PCIe 3.0 graphics slot on a HP Z420 workstation, Xeon E5-2760 @ 2.6 GHz (Sandy Bridge EP), 8 cores, 48 MB memory, 2 NVME SSDs, 2 SATA SSDs, Windows 10 PRO, latest update.  (There is also a second PCIe 3.0 x16 slot in this workstation that I am using for the NVME SSDs.)
 
I am running dual monitors. An 32" HP Omen monitor running at 2560 x 1440 @ 75 Hz and a Samsung SyncMaster 245T at 1920 x 1200 @ 60 Hz.  These are both the native resolutions for the displays.
 
I just happened to run the HWINFO tool, and I noticed that the EVGA graphics card was listed as a PCIe 2.0 x16 connection.  XWINFO also reported that the "Device/Port Type" is"Legacy PCI Express Endpoint".
 
My other PCIe devices are listed as "PCI Express Endpoint".  No "Legacy".
 
The Nvidia control panel, system information also shows "Bus: PCI Express x16 Gen 2", confirming the report from HWINFO.
 
I don't know when this degraded to PCIe 2.0 mode.
I was running an Nvidia driver from January 2020, so I upgraded to the latest Nvidia driver version 451.48 - No change.
 
I went into my system BIOS and set the slot to explicitly PCI-e 3.0 max speed (rather than auto), no difference.
I am on the latest system BIOS from HP for this system version "J61 V03.96". I am running the latest Win10 Pro version.
 
I have had no visible issues with the EVGA GTX 1070 SC card ... except that the Nvidia control panel and HWINFO are reporting that the card is connecting in PCIe 2.0 x16 mode, not PCIe 3.0 x16 mode.
 
HWINFO also reports that the "Device/Port Type" is"Legacy PCI Express Endpoint", not "PCI Express Endpoint".
 
Has anyone seen this connecting at PCIE 2.0 mode on a PCIe 3.0 slot issue?
 
For my graphic usage, I don't think I am doing anything stressful enough that this lower connection speed would make a difference.
 
Any ideas?
 
Thank you for your help.
 
db808
 

Hi arestavo,
 
Thank you for your very quick reply. But I don't think the patch should apply to my situation.
 
Sandy Bridge supports 40 lanes of PCIe 3.0 from the CPU chip itself, via the LGA 2011 socket.  The C602 chipset (sibling to the X79) provides an additional 8 lanes of PCIe 2.0 and other low-speed interfaces.  I am NOT connecting the GTX 1070 SC to the C602, but the Sandy Bridge-EP slots.
 
I don't understand why the patch would be needed for the PRIMARY PCIe slots which are sourced from the Sandy Bridge-EP CPU chip itself.  I would imagine that some motherboard vendors that wanted to support quad-GPU configurations without an on-board PCIe switch, could use the C602 or X79-provided PCIe lanes ... but would be limited to only 8 lanes of PCIe 2.0 (without chipset overclocking).
 
I have two other devices (Samsung NVME SSDs), that ARE properly connecting in PCIe 3.0 mode on the other two PCIe slots that are driven from the Sandy Bridge-EP CPU directly.   I am NOT connecting the GTX 1070 to the slots driven by the C602 chipset.
 
So I should not have to force PCIe 3.0 mode on Nvidia, unless Nvidia has problems with PCIe 3.0 as implemented on the Sandy Bridge, which was Intel's first PCIe 3.0 implementation.
 
I don't know why HWINFO is reporting the GTX 1070 SC as a "Legacy PCI Express Endpoint".  That is how the Nvidia is responding to the PCIe slot initialization.  The Samsung PCIe 3.0 devices are responding as "PCI Express Endpoint"
 
Thank you for you help. 
 
The patch does not seem to make sense to me in this case. The GTX 1070 SC is connecting on the Sandy Bridge-EP direct PCIe 3.0 slots ... not the C602/X79 chipset based PCIe slots.
 
Regards,
 
db808
 

https://www.nvidia.com/download/driverResults.aspx/103913/en-us
 
See if that drivers fixes it.

Just want to see if the driver is the issue since you didn't have the issue years ago.

The X79 platform did not get accredited by Nvidia for PCIE 3.0 on X79/Sandy Bridge-E CPUs (yours is one, even though it is the server variant). This was not the case for Ivy Bridge-E CPUs (which is one of the reasons why I upgraded, years ago, from a 1650V1 to a 1680V2).
 
That's the reason why the patch that I linked exists - PCIE 3.0 on X79 motherboards with Sandy Bridge-E CPUs isn't actually supported by Nvidia drivers natively and caused a bit of an uproar at the time (and at one point, still needed a BIOS update to even allow PCIE 3.0 at all), me being one of the people that was rather miffed for the deceptive advertising. But it was a call that Nvidia made for its drivers because of the performance problems with different X79 motherboards and various Sandy Bridge-E CPUs when they ran at PCIE 3.0.
 
So, I suggest trying it. If you check the instructions, you can easily revert it from command line if you find it to be unstable.
 
Direct quote from Nvidia: 
 
"Though we cannot officially support or guarantee 8GT/s speeds for X79/SNB-E, you can manually enable faster 8GT/s bus speeds (at your own risk) using the following executable: force-enable-gen3.exe"
 
"If you would like to revert back to 5GT/s bus speeds you can do so by running the following command from the DOS command line: “force-enable-gen3.exe –revert”"
 
Again, your PCIE 3.0 for Nvidia cards on an X79 motherboard with a Sandy Bridge-E processor is FORCED to PCIE 2.0 for Nvidia cards by the Nvidia drivers. Which is why that force enable PCIE 3.0 executable exists.