PCI-E Gen 2.0 Bug X38/ X48/X58

[JohnZS]

Hi guys

In collaboration with a number of users on the nVidia forums and ASUS forums I am trying to collect data as to what boards are effected by the notorious PCI-E Gen 2.0 bug which appears to be effecting Dual PCI-E 16 Intel Boards and nVidia cards.

The symptoms of the PCI-E Gen 2.0 bug are that your graphics card only operates at PCI-E Gen 1.1 mode.
For example



This GTX 285 is only operating @ PCI-E 1.1 x16 on an X38 based motherboard (ASUS P5E3).

On gigabyte boards the symptoms are slightly different.


Where here the board does present itself as a Gen 2.0 board, the card only runs at Gen 1.1

I and a few other users have been constantly trying to get this issue resolved, but motherboard vendors are not very helpful.

The information we have is

1) PCI-E Gen 2.0 has been disabled on X38/x48/X58 boards with a certain undisclosed, yet confirmed by nVidia SMBIOS bug in Intel BIOSes as of April 2009.

2) Intel DX38BT motherboard with BIOS 2006 is NOT effected by the PCI-E Gen 2.0 bug, apparently this contains a fix.

3) DMI information on ASUS X38 boards is odd, says our PCI-E16 slots are empty?!

4) Intel resolved a Slot mismatch issue on their boards in July 2009 stating it resolved a PCI-E Gen 2.0 issue.

5) Most X48 boards are unaffected, and even fewer X58 boards have this bug.

6) nVidia 9800GX2 and GTX295 do not appear to show this bug as their NF200 chips report the PCI-E link speed direct to the GPU's

Please if you have the time could you take a screenshot of GPU-Z showing your card's PCI-E transfer rate and put this alongside a CPU-Z screenshot showing what motherboard and what BIOS revision you are running.

Many thanks

John

[iTravis]

Hi John, so you're saying that even though GPU-Z shows the bus interface at x16, it only runs at PCI-E 1.1 x16 but not PCI-E 2.0 x16? I have some screenshots and hopefully they help. Does it only affect to Nvidia card but not ATI?

My board is a Maximus Formula X38 that flash to a Rampage Formula. I have the GTX 275, GTX 285, GTX 295 as shown below:

[JohnZS]

Hi John, so you're saying that even though GPU-Z shows the bus interface at x16, it only runs at PCI-E 1.1 x16 but not PCI-E 2.0 x16? I have some screenshots and hopefully they help. Does it only affect to Nvidia card but not ATI?

My board is a Maximus Formula X38 that flash to a Rampage Formula. I have the GTX 275, GTX 285, GTX 295 as shown below:

Thanks for the information iTravis
What we see here is that your GTX 275 and GTX 285 are running in PCI-E 1.1 mode. The GTX 295 corresponds with tests performed on 9800GX and GTX 295 Single PCB whereby the NF200 reports PCI-E 2.0 to the GPU's and the driver.

ATi cards appear unaffected, unless you use HWinfo32 (or some other program that measures PCI-E Bandwidth). For some reason ATi cards do not report information like nVidia cards (such as memory usage, etc)

Thanks for the information, what is concerning is that Rampage Formula does not have this issue, yet you flashed your BIOS to a Rampage Formula and you still have this issue
John

[iTravis]

Well if that's the case the bug is also in P35 chipset as well, here's my 9800GTX+ on an Asus P5K Pro.

[stardust8750]

Well if that's the case the bug is also in P35 chipset as well, here's my 9800GTX+ on an Asus P5K Pro.

[IMG]http://img101.imageshack.us/img101/3051/evga9800gtx.jpg[IMG]

I thought PCI-E 2.0 was released with the x38 chipset

[JohnZS]

I thought PCI-E 2.0 was released with the x38 chipset

Correct, The X38 was Intel's first PCI-E Gen 2.0 motherboard. The P35 is a PCI-E 1.1 motherboard.

P35 Chipset Features

X38 Chipset Features

Thanks for your help.

John

[iTravis]

So how about this? I have this 8800GT before I changed to the 9800GTX+, run on same motherboard.

[JohnZS]

So how about this? I have this 8800GT before I changed to the 9800GTX+, run on same motherboard.

Those are really old drivers
What you see there is a detection of a PCI-E Gen 2.0 capable slot, but not a PCI-E 2.0 capable card.

If you run HWinfo32 it will give you more detailed information

I am awaiting more information from ASUS R&D department, but what I can tell you is that genuine Rampage Forumla's do not exhibit this issue, yet Maximum Formula's and Maximum Formula's flashed to Rampage Forumula's do have this problem

[YukonTrooper]

No laughing at my GPU clocks. I get PSOD even with minor overclocks.

[JohnZS]

Thank you very much for your contribution YukonTrooper
It appears your motherboard is not effected by this annoying bug.
We are still no closer to finding a fix, what we have noticed is that the DMI slot structure is all wrong!? This information has been passed onto ASUS, so who knows....fingers crossed a fix is in sight.
John

[chispy]

eVGA 790i Ultra:





By chispy at 2010-04-02

[Chickenfeed]

Those are really old drivers
What you see there is a detection of a PCI-E Gen 2.0 capable slot, but not a PCI-E 2.0 capable card.

The strange thing is... his 8800gt shot was when it was in the same P35 board ( PCIe 1.1 as mentioned ) yet it reports PCIE 2.0 ( by the way the 8800gt is PCIE 2.0, it was Nvidias first If I recall ) Rather bizarre...

EDIT : I looked into it and some people seem to believe that a newer revision of the P35 chipset supports PCIE2.0 but unofficially... Wether this is true or not is beyond me. Just thought this was interesting.

[overclocking101]

my EVGA P55 FTW does this when running 2 cards, it reports the pci-e 2.0 X16 slot but only runs it at pci-e X16@ X8 the X8 is normal but it should say X8 2.0 it doesnt, but with one card it shows pci-e 2.0X16@ X16 2.0

[JohnZS]

Thanks for taking the time to test and reply guys

chispy

I have noticed that your card is running @ x4 Gen 2.0. Have you by any chance got dirt on the slots or the card? Or are you running Tri SLi or something?!?

It is unusual to see an nForce 790 based motherboard exhibit such a strange problem.

Chickenfeed
Ahh thanks for that, this would explain the card and slot mismatch on your GPU-Z screenshot. A PCI-E 2.0 card running on a PCI-E 1.1 slot. I have not heard about some boards being PCI-E 2.0, however what you say would make sense if they were very late production models out at the same time as the X38 and X48 boards, but just before the P45 was released?!

Thanks for the information overclocking101
Similar behaviour has been noticed on a ASUS Maximus III Extreme, however that was running PCI-E Gen 1.1 with just one card in the system (@ x16) and then x8 with two cards in the system. Perhaps your eVGA board is half fixed?

I would keep an eye out for BIOS updates.

As far as I am aware P45 boards are immune to this issue, however if any P45 readers are out there, please do provide screenshots confirming or contradicting what we have found so far with P45 boards. Anymore X38 and X48 users will be most welcome..

Thanks

John

[death]

hi all ive fixed the prob and its simple to do but ide rather not post it so john if u get the time pm me ill tell u how to fix this and get u pci e to 2.0

[YukonTrooper]

hi all ive fixed the prob and its simple to do but ide rather not post it so john if u get the time pm me ill tell u how to fix this and get u pci e to 2.0

I'm not affected by this bug, but why would you decide not to post the fix for those that could use it? Unless the fix involves messing around with high voltages, I'm a little bewildered.

[death]

ok ill post it
well ide been doing alot of reading about this prob that i have and loads of other have and then found out about the sli patch that nv drivers see u board as a x58 and

well we all no about the sli patch that lets none sli board do sli long as they have 2 x 16 pcie lanes well what it does it make the nv drivers think u got an x58 board well ive got a p5e x38 board and before i did this like 10 mins ago my pcie was at v1.1 after installing the patch to make the drivers think i have a x58 i now have v2.0 pcie

before


after


the difference is amazing over 2k more from v1.1 to v2.0 i hope this is going to make lots happy
the prog is here

[death]

what it is is amazing it really is that simple to do cant post it on nv forums as it bet it will get deleted as soon as its read but there u go ever one i can say ive fixed a prob that been about for years


and i no but i do feel veryy happppppyyy

[VoodooProphetII]

What I found with my Asus Crosshair 3 motherboard and the GTX480 I'm using, that when I run GPUz it will show PCI-E 2.0 x16 @ x16 2.0 but only for like 30 seconds and then it will change to PCI-E x16 @ x16.

Brillian plan death. I just don't think it will work on my motherboard....... or will it?

EDIT.... OOPS I just saw that my board is on the list of those the pacth were tested on. GREAT. Thanks, I will try it out.

[death]

well the patch was ment to let just about ever board with 2x 16 pcie slots have sli but in the same way it made it it has allso fixed the bug that nvidia put in there drivers to stop x38 board haveing the right specs

[donmarkoni]

THERE IS NO BUG!

GPU-Z STOPS DISPLAYING 2.0 WHEN GPU DOWNCLOCKS, AS VoodooProphetII NOTICED.

What I found with my Asus Crosshair 3 motherboard and the GTX480 I'm using, that when I run GPUz it will show PCI-E 2.0 x16 @ x16 2.0 but only for like 30 seconds and then it will change to PCI-E x16 @ x16.

HERE ARE THE PICTURES:

Immediately after start:


Few seconds later - downclock and no 2.0:


Again few soconds later - more downclock, stays the same:

[JohnZS]

Just to clarify things, it is not a bug that nVidia put into the drivers, but an intentional block. Apparently (and this is from nVidia themselves when I asked them about this Gen 1 issue) it is intentional to avoid a potential system hang!

Hi John,
I am sorry however there is nothing we can do in this case. The Intel x38 chipset was Intel's first generation PCI-E 2.0 implementation and unfortunately has some PCI-E 2.0 bugs, one that causes the graphics card to hang when changing P states. The only Intel PCI-E 2.0 chipset which shows these bugs is the Intel x38 chipset. If you see any other Intel chipset reporting PCI-E 1.0 speeds with our GPUs, please let me know. The best way to check is to run a 3D program in the background as this will change the P state of the graphics card from PCI-E 1.0 to PCI-E 2.0. This throttling is done to conserve power when the user is running basic Windows tasks

However, once again well done death
It is about time PCI-E 2.0 was re-enabled properly

EDIT
For those of you doubting the presence of this bug, you are confusing things. Basically by design the cards GT200 or newer (with forceware 190+) should revert back to PCI-E 1.1 in the 2d mode i.e. 300Mhz (or whatever your lowest Pstate clocks are). HOWEVER when you run a 3d application it should change to Gen 2.0 full link width! This change does not happen on motherboards/systems effected by the Gen 1.1 bug. They ALWAYS remain in Gen 1.1 mode, regardless of how long you run furmark/3dmark/*insert game name here*

John

[iTravis]

Apparently it's still an issue with Nvidia cards on X38 chipset, from GTX 275, 285, not 295 though, to GTX 470 and GTX 480. I run FurMark and it still doesn't show PCI-E 2.0 x16.

[JohnZS]

I do not know the full ins and outs, but apparently the whole issue is a form of crash prevention by nVidia. They believe that when switching to an L1 ADSPM state the X38 chipset has a bug/glitch which can cause a hard freeze. FYI The 9800GX2 and GTX 295, only stay in the L0 state (something to do with their design). Therefore all GPU's with multiple P states (G200 and newer) are effected by this lock on the X38 chipset. Apparently the X48 is NOT effected so in theory Gen 2.0 should work correctly on those boards.
I am still waiting to hear from ASUS about this as Gen 2.0 is broken on X38, therefore they need to fix it, or at least replace with equivalent X48 based board.
It would be interesting to know if the work around posted causes any stability issues
John

[VoodooProphetII]

donmarkoni thank you.... I've just tested my mobo and graphics card and it looks like you are correct with your assessment. GPUz would show 1.1 after a few seconds and then when I run OCCT to stress the GPU and the test starts it would switch back to 2.0

Thanks a lot.