Dev Kit and Dell "Training Error"

Sounds like Altera's problem and not Dell's.

We have very similar problems on Intel S3420GPLC motherboard which is completely different from your AMD-based Dell in just about everything.

PCIe part of Stratix-IV GX devkit is simply broken. If something appears to work for you in some other PC that's probably because the chipset in this other PC doesn't check the replies from devkits as thoroughly as server-class chipsets in S3420GPLC or PE-6950.

BTW, which of PCIe core variants did you test - hard IP or soft IP?

In our experience both are problematic (and almost certainly non-compliant with PCIe standards) but soft IP is less problematic than the hard IP. Soft IP not always trains to desired width (x4 in our case), but at least when it manages to trains correctly it works correctly thereafter. No such luck with hard IP variant that hangs the whole system during majority of attempts to change the power state, including system restart/shutdown.

I wonder whether the PCIe problem is only of devkit or of the whole Stratix-IV GX line. In later case Altera is in deep trouble.

I've been informed that by updated the PCIe megacore, this issue will be resolved.... which has introduced a whole new set of issues.

Given actions and communications (or lack of) from Altera... It's apparent that Altera doesn't stand behind the products they sell (or their word for that matter).

Regardless, I need to try and get this working... has anyone had success in flashing the PCIe megacore into the Stratix IV GX Development Kit? I've tried to take the Stratix IV example Quartus projects for PCIe on the Altera website, following the Altera directions... and have had zero success in programming the flash.

Could someone give me a pointer on how to program the flash on this device?

In fact, after upgrading to Quartus 9.1 SP1 our situation with training is ALOT better.

Hard IP when burned into on-board flash always trains to correct number of lanes. As documented in the IP release notes hard IP has other serious problems. For example, every attempt to disable our device from Windows device manager lead to a fatal PCIe bus error. Luckily, we didn't pick the devkit as a platform by ourselves but got it from the client so it's sort of his own fault. So he would have to live with this crap.

Soft PCIe IP behaves better in that particular regard, but in order to burn it into flash one needs the license that we don't have. If either Altera reps or our client will give us the license for free (and I feel like it's their moral obligation after all the bull**** that we were forced to eat with broken 9.0 and 9.1 versions) we could try that configuration as well.

Overall, I should say that PCIe on Stratix-IV GX devkit story is my worst ever experience with Altera. By far.

Back to your last question - with Quartus 9.1 SP1 programming the flash via Board Update Portal finally works. 'sof2flash -pfl' is as fast as a drunk snail but it produces the working hex file.

See the discussion on this very forum:

http://www.alteraforum.com/forum/showthread.php?t=20577

Unfortunately... I don't have a DHCP network available. Plus, following directions from an Altera SR, the default flash is no longer there.

So I don't have the Board Update portal available to me, just the USB blaster.

Wow... the SR took me down the path of flashing a couple of images.

It seemed to work... but I was wrong. The factory defaults are no longer in the board, the 2 images I attempted to flash via the SR aren't loading (or something, regardless it's not working).

AND... after I follow the Altera instructions for restoring factory defaults.... it fails. So now I have no idea what's loaded... but it doesn't work. Nor can I seem to flash anything else.

WOW. This is the equivalent to a "Hello World" application not running out of the box and the supplier refusing to support it.

This whole ordeal has verified that I made a poor design decision by choosing Altera.

Regardless, we're too far along now to change unless we have no other choice... so...

If ANYONE has had ANY success in flashing this board successfully through the USB. I'd really appreciate the assistance.

Thanks

Okay... thanks to the local FAE, I now have the proper directions on how to flash.

There are magic options in the sof2flash utility that make the flash image compatible with this board. The fix was found under a SR for "...issues with the sof2flash utility...".

The following options are needed for the sof2flash utility before, the flash image will be compatible with this board:

--pfl

--optionbit=0x18000

Yep. SR was over 1 month old... frustration and anger from the customer... all because of two option switches were left out.

Again.... I'm not an FPGA developer...

The pfl flag I tried per the thread but, the --optionbit flag wasn't clear (to me anyway)