Nvidia's answer to x79 and PCI-e 3.0

[daniel66]

Played GTA4 last night . After 5 minutes or so i got a nice BSOD with Gen3 enabled. Restarted tryed again, the same.
Changed To Gen2 in BIOS and everything is OK.

Now, in my opinion all are guilty here:

Intel for (to the limit of)false advertsing.
ASUS-I bet all my money that ASUS knew about this and continued to advertise GEN 3 slots. I can see right now, the PR guys in the office:
-Who the F*** cares , we will blame Intel for this. We dont want the rivals to have something extra.We will do it too, bend the truth a little.

The buyer because he believes all the s**** that is on a box. Also reviewers must not be trusted. They show you that you can OC to this limit,
you run this and that, but the truth is that few people will achieve what they saw.
The best place for info is a forum, people. Also do not buy when the product appear in the store, just watch your product for a few months and see what problems will show.

Not sure about Nvidia, the cards work well in actual Z77 GEN3(certified as per spec) slots, no?

Now, on the good side, i knew about this but still bought a R4F. But i didnt bought it for Gen 3 i bought it for a six core CPU.

And i have to say it's working ok, it was working from the start.

[mdzcpa]

SB-E is not advertised as 3.0 Compliant. The data sheet CLEARY says PCIE 2.0.
I point to the Data Sheet 1.2.2:

• Support for PCI Express* 2.0 (5.0 GT/s), PCI Express* (2.5 GT/s), and capable of
up to PCI Express* 8.0 GT/s.

• Up to 40 lanes of PCI Express* interconnect for general purpose PCI Express
devices capable of up to 8.0 GT/s speeds that are configurable for up to 10
independent ports.


Furthermore EVERY review of Sandy Bridge on release noted no PCIE 3.0 compliance.

So what you have above is Intel's way of saying support "might" be capable for PCIE 3.0 speeds. Nothing more.

As for Asus, I beleive the RIVE is 3.0 compliant. It supports the AMD 3.0 cards just fine. The conflict lies between SB-E and Nvidia in my opinion. SB-E was simply not fully baked. And once IB-E arrives we will know for sure if the RIVE is fully compatible as we will finally have a compliant CPU to test it with. The main issue here is that PCIE 3.0 was simply not fully implemented on SB-E.

[Praz]

It supports the AMD 3.0 cards just fine.

This isn't fully accurate. Users are also reporting issues with ATI cards on various motherboards. And on motherboards that have the option of locking the PCIe slot to 2.0 the issues go away. The main difference between the two companies is nVidia decided to disable 3.0 support by default.

[mdzcpa]

I am not referring to "various motherboards." I am referring to the RIVE. There may be an issue here or there as there are with any new builds and/or weak CPUs, but by and large, there are a bunch of folks happily running the AMD 7900 series (solo and in CF) on the RIVE. Threads all around the net substantiate that. It is a night and day difference between AMD and Nvidia PCIE 3.0 cards on the RIVE. Of the 6 builds I've completed on the RIVE, all of them run AMD 7900 series cards in 3.0 with flying colors. That doesn't prove anything, but it is a meaningful trend.

As far as conflicts on "various" other motherboards, I would honestly look to the CPU for the conflict. Support for PCIE 3.0 begins there. And SB-E was not validated for obvious reasons. I believe AMD simply worked within the specification envelope of SB-E to strived to get their cards to work. Nvidia decided it wasn't worth it. For a mobo to be PCIE compliant is not that big of an issue. Simply traces as Raja pointed out. I wouldn't be surprised if the RIVE and most other X79 mobos work just fine when an IB-E is dropped in. Maybe not...but likely.

The bottom line is that I'm not buying into all the "false advertising" hoopla. We knew from launch that SB-E was NOT PCIE 3.0 compliant. It was hard not know that. As far as mobos advertising complaince, I suspect most indeed are. The lack of support lies in the CPU and no 2 chips are exactly alike. Therefore most do but some don't when it comes to the AMD cards. For Nvidia they simply removed the hit and miss game by disabling support. No conspiracy here. Just the time honored tradition of aggravating early adopters that choose not to read up on the issues before they buy.

[Praz]

The AMD forums have a large number of posts with these same issues. But yeah the real issue here is some of the processors being marginal as far as signaling thresholds. I agree with most not having issues though. I think I wrote the same a couple of pages back.

[Wrxguy2012]

Hi all,

Nice forum, I just wanted to share my experience on the topic.

I recently put together a build about 1.5 month ago using x79 Sabertooth board and (1) gtx 690. Getting the board stable and setting up raid ssd for boot drive took about a week and was the most challenging and difficult build I have done. I have been building high end machines for the last 22 years. The single 690 performed flawlessly on the pcie gen 3.0 without the need to install any patches. I was very happy.

I use this machine for Primegrid for grid computing and I wanted to add another 690 to give the system even more horsepower for CUDA computing. I installed the card in the second x16 slot and connected the SLI bridge. The hardware looks very very impressive and I was proud to be the owner of the system. So I powered on the computer and windows 7 did its thing with installing drivers for the second 690. I saw that the Nvidia control panel showed the second card and I configured it to be quad SLI and clicked apply. It went through its changes with the flashing screens and showed it was indeed quad SLI so I was happy. Then the problems started. Every second, my computer would pause for a brief moment and then resume. So if you are moving the mouse pointer around, every second it would get stuck and then continue moving. At first I thought I got a virus but then realized it was hardware / driver issues. After hours and many reboots later I still the same problem, but worse. It would take minutes to do anything on the machine. I tried disabling SLI and still the same problem. I changed the drivers to the Beta version, and no change.

I then started searching forums on my mobile phone and then made two changes: Connecting the monitor to the single DVI port on the first card and changing to gen2 on the pcie configuration in the BIOS. This helped greatly but I don't know yet if it is stable. I will know in the next few days. I tried running some preliminary benchmarks and I don't think adding the second card helped much in benchmarks but when I ran Boinc (to compute Primegrid) it seemed to run fine on the 2 new gpus for a total of 4 running cuda apps. I am praying that it is stable and will know in the next few days. I have also noticed that after having quad SLI, one gpu core temp is always much higher than the other 3. I tried switching around the cards and same problem existed. Not sure what is going on there, and yes, I have really good air cooling in my case using industrial dc fans.


These are my wishes that I can dream about but I know they will not come true anytime soon:
I really wish I can run them at gen3 but will have to settle for gen2 on the SB-E x79.
I wish for Nvidia to release better drivers for quad SLI.
I wish for Nvidia not to artificially reduce floating and double floating point operation performance on their consumer gpus to push their workstation cards.
I wish companies would stop bringing products to market until they are ready. The wait is well worth it.
Companies need to own up to their mistakes if they have made some and offer timely solutions to their customers.
Companies need to inform their customers when there are incompatibilities that they are aware of instead of them having to spend sleepless nights struggling to figure out what is wrong.

So how do I fight back? Although when it all works together it really works very nicely, but because I had to struggle to solve these issues in their products, I give them bad reviews on popular websites so that others will know when they research to make a purchase. If someone reads my negative review and decides not to purchase, then i've done my part. The second way that I fight back is that if there are other attractive options available, then I go with them instead of problem companies.

I'm disappointed and have lost trust in companies. They won't care either, they have 2 billion new customers in China and India. Business is booming for them.

Just my two cents.

[Kazutopia]

Hi all,

I have downloaded the .exe and I'm not sure that it is working.

I can see both my cards as gen3 in the nvidia control panel but still says pcie2.0 in gpuz.

Any special instruction with this file? just double click, click run, and reboot, right?

bios is 1404, drivers are 301.42 and cards are gtx680's

by the way. one of cards is showing 16x in slot 1 and the other is showing 8x in slot 3. How can I get the 2nd card to show 16x?

Thanks

[camila]

I runs two GTX 690 SLI and I really happy with their performance. In addition, I've never encountered any problem when I am gaming or benchmarking. These GTX 690s run at PCI 3.0 and they work flawlessly even though their performance is not scale well insome games at quad SLI mode. I hope that nvidia will improve their driver next time. Honestly, I am happy with the SLI performance or single one GTX 690, not only their temperature but also their power consumption. One thing I don't understand that why nvidia only enables the GTX 690 at PCI 3.0 by default but not for GTX 680 and 670 graphic cards on the X79 platform.

[Arne Saknussemm]

by the way. one of cards is showing 16x in slot 1 and the other is showing 8x in slot 3. How can I get the 2nd card to show 16x?

Just to be clear.....The second PCIe x16 slot is the 4th PCIe slot from the top (the third red one from the top). This numbering is all over the place depending what page you look at in the manual! that's why I ask...


This makes "interesting" reading.... http://www.intel.com/content/dam/doc...ion-update.pdf

It is dated this month.There are a whole host of PCIe issues that are BIOS solvable I wonder if they are already in the BIOS or are coming in future updates? @ASUS?

[Kazutopia]

Just to be clear.....The second PCIe x16 slot is the 4th PCIe slot from the top (the third red one from the top). This numbering is all over the place depending what page you look at in the manual! that's why I ask...


This makes "interesting" reading.... http://www.intel.com/content/dam/doc...ion-update.pdf

It is dated this month.There are a whole host of PCIe issues that are BIOS solvable I wonder if they are already in the BIOS or are coming in future updates?

Thanks for the input Arne.

Yes the 2nd card is in the 3rd red slot. It says this slot can be either 8x or 16x. I have tried to set both slots to "16x native occupied" in the advanced tab to no avail, still posts as 16x and 8x in the nvidia control panel.

Any input for this or the pcie 3.0 trouble I'm having is greatly appreciated.

Thanks