Programmable Devices
CPLDs, FPGAs, SoC FPGAs, Configuration, and Transceivers
20642 Discussions

DDR4 HiLo in Stratix 10 MX

s10M
New Contributor I
2,489 Views

Hello there,

We recently acquired a new Stratix 10 MX devkit - 1SM21BHU2F53E1VG. 

I am currently testing different interfaces of the device using the Board Test System (BTS) that is provided by Intel's Stratix 10 MX FPGA Installer Package - found online on the Intel S10 MX website.

All interfaces seem to work well except for the DDR4 HiLo test. When the board is programmed with the example design, the status monitor for the test shows a failure. Further investigating signals in the EMIF shows that the calibration with the device fails. Basically, it seems that I cannot talk to the DDR4 device. 

This is odd because the other example designs, including the DDR4 DIMM test, passes and works as expected. 

The DDR4 HiLo daughter card seems to be Micron:MT40A512M16JY-075E:B.

I would really appreciate Intel's (or the community's) assistance with this.
Is it that the example designs provided by the manufacturer are faulty or am I perhaps overlooking something?

Thanks a lot

 

 

1 Solution
Deshi_Intel
Moderator
2,373 Views

HI,


I have consulted both Intel marketing and Intel Engineering team regarding your issue


The feedback that I got is typically as long as the HILO card is firmly inserted into the board slot then should be fine. Else if there is loose pin contact then user may encounter DDR4 calibration failure.

  • I recalled you told me you have verified that your HILO card is properly/tightly plug into on board HILO connector, right ?


For DDR4 Hilo card replacement, Intel marketing commented can directly ship DDR4 hilo replacement card to you.

  • Can you share with me your shipment detail ?
  • Shipment attention name, address, phone contact and email contact
  • You can use forum private message function to share with me shipment detail if you don't feel comfortable to share detail in this public forum post


Thanks.


Regards,

dlim


View solution in original post

0 Kudos
13 Replies
Deshi_Intel
Moderator
2,481 Views

Hi,


BTS software is customized software that only works with specific Quartus version only. Can you ensure you are using the correct Quartus version as specified in below dev kit link ?


Also, can you help to check to ensure there is no physical damage on DDR4 HILO card, particularly the pins and then properly insert/plug back into on board HILO socket, then rerun BTS again.


Thanks.


Regards,

dlim


s10M
New Contributor I
2,473 Views

Hello @Deshi_Intel,

Thank you for your reply. I appreciate it. 

We are currently using the correct Quartus version as specified by the BTS. All the designs in the BTS seem to work fine except the DDR4 HiLo one.

I did verify the daughtercard itself and it seems to be firmly connected into the board (with no visible damage).

Could you kindly confirm with me that the DDR4 HiLo design in the BTS does in fact work in your (or your engineer's) MX devkit? 

It seems like this problem was experienced by another user as well (https://community.intel.com/t5/Programmable-Devices/DDR4-HILO-in-S10MX-failed-initialization/m-p/1194331#M76403) which seemed to be unresolved as well.

Thank you for your help. I appreciate it. 

Deshi_Intel
Moderator
2,446 Views

HI,


Thanks for your feedback.


Let me try to look around to search for S10 MX dev kit board to try out the BTS sof file myself.


If it works on my side then maybe your DDR4 HILO card is faulty then I can initiate RMA process for you to get new replacement HILO card.


Thanks.


Regards,

dlim


0 Kudos
s10M
New Contributor I
2,443 Views

Perfect - thank you so much for your assistance. Much appreciated.

Looking forward to hearing from you soon. 

0 Kudos
Deshi_Intel
Moderator
2,431 Views

Hi,

I have verified DDR4 Hilo test at my S10 MX board and it's passing as shown in attached BTS GUI screenshot.

I don't have the exact Quartus v18.1.1 so I tried out with Quartus v19.1 and somehow it still works

So, I suspect your DDR4 Hilo card is faulty. Can you help to confirm following ?

  • Can you at least program BTS pre-compiled sof "bts_ddr4_hilo.sof" using Quartus programmer tool ? This is to ensure your MX dev kit board is good while likely only DDR4 Hilo card is bad 
  • Also, can you share with me the BTS screen shot showing failure on DDR4 Hilo test ?

Thanks.

Regards,

dlim 

  

 

0 Kudos
s10M
New Contributor I
2,425 Views

Hi @Deshi_Intel ,

Thank you for the check. 

After re-plugging the HILO card, we realized that the board is now detecting the HILO card, but it seems that the calibration for the card is failing. 

As per your request, I have programmed my board with the original  BTS pre-compiled "bts_ddr4_hilo.sof". The programming completes with no failure. The attached screenshot is what I see when I launch the BTS with this design loaded. 

Just for information, my MX device is 1SM21BHU2F53E1VG, and the BTS design I have used is in the "Production 8GB" MX FPGA Installer Package (https://www.intel.com/content/dam/altera-www/global/en_US/support/boards-kits/stratix10/mx_fpga/stratix10MX_1sm21bhu2f53_fpga_revB_v18.1.1b263_v1.0.zip)

Could you kindly confirm this is the same package design that you used to test your MX i.e. you used the "Production 8GB" Hilo Design rather than the "Production 16GB" Hilo Design? (I have noticed the two are slightly different with the way the EMIF is parameterized) 

Thank you once again. I appreciate your help. 

0 Kudos
Deshi_Intel
Moderator
2,416 Views

Hi,


Based on my understanding, the 8GB vs 16GB memory density difference is applicable to HBM2 memory, not DDR4 Hilo memory density.


You can see the difference in ordering info table. The FPGA OPN is difference as well


The board that I am using is same like yours which is 1SM21BHU2F53E1VG (8GB HBM2)


Can you provide me your S10 MX dev kit board serial number so that I can validate whether your board is still under warranty claim ?


In parallel, I will find out further within Intel internally on the RMA claim process as well.


Thanks.


Regards,

dlim



0 Kudos
s10M
New Contributor I
2,409 Views

Hello @Deshi_Intel ,

Yes - you are correct regarding the difference between the two boards. However, when I looked at the HiLo example designs for each of the board, the EMIF is parameterized as UDIMM in the 16GB HBM2, and COMPONENT in the 8GB HBM2. Since the DDR4 Hilo is not different in the two boards, I was confused (and curious) about which design you used for your test. 

Since you are using the same board as me, I guess your EMIF is parameterized as  COMPONENT.

For your reference, my board's serial number is 10SMXPCIe000398 and K54160-001. I should mention that the board works well: all the designs we have tested on it so far have passed. It is just the HiLo design that is failing in the calibration stage. This could point to a problem with the design and/or the daughter card. 

Could I ask you what the DDR4 HiLo card you are using? 
The FBGA and Component Marking in my daughter card is 7LB45 D9TNZ (which according to https://www.micron.com/support/tools-and-utilities/fbga tells me the part number).

Thanks again for the help and support you are providing to resolve this issue.

 

0 Kudos
Deshi_Intel
Moderator
2,402 Views

Hi,


Frankly speaking, I don't pay attention to the 16GB installation package hilo example design setting since I am using 8GB installation package. (I am using 8GB HBM2 board anyway)


I am using \\stratix10MX_1sm21bhu2f53_fpga_revB_v18.1.1b263_v1.0\examples\board_test_system\image\PRD\bts_ddr4_hilo.sof directly without reviewing the DDR4 IP setting.


I can see my DDR4 Hilo card on board DDR4 SDRAM marking info is "7LB45 D9TNZ" as well


The selection of DIMM vs component is just for user own acknowledgement purpose to pay attention to board layout design routing phase later. It doesn't alter DDR4 IP setting.


I have validated your board is still under warranty claim.

  • However, there is one issue here. I just found out from Intel Marketing Micron already EOL existing version of DDR4 Hilo card
  • I am now checking with Intel marketing on what's the replacement Hilo card and how does it difference from original DDR4 Hilo card


Stay tuned. Will keep you posted once I hear back more from Intel marketing.


Thanks.


Regards,

dlim


0 Kudos
Deshi_Intel
Moderator
2,374 Views

HI,


I have consulted both Intel marketing and Intel Engineering team regarding your issue


The feedback that I got is typically as long as the HILO card is firmly inserted into the board slot then should be fine. Else if there is loose pin contact then user may encounter DDR4 calibration failure.

  • I recalled you told me you have verified that your HILO card is properly/tightly plug into on board HILO connector, right ?


For DDR4 Hilo card replacement, Intel marketing commented can directly ship DDR4 hilo replacement card to you.

  • Can you share with me your shipment detail ?
  • Shipment attention name, address, phone contact and email contact
  • You can use forum private message function to share with me shipment detail if you don't feel comfortable to share detail in this public forum post


Thanks.


Regards,

dlim


0 Kudos
Deshi_Intel
Moderator
2,349 Views

Hi,


I have not hear back from you for sometime.


Do you still want DDR4 Hilo replacement card ?


Can you share with me your shipment address for Intel to replace DDR4 Hilo card for you ?


Thanks.


Regards,

dlim


0 Kudos
s10M
New Contributor I
2,333 Views

Hello @Deshi_Intel ,

Thank you for the information and the help. 

We have tried using another DDR4 Hilo daughter card that we had available to us, and as you had experienced, the design and the interface works well! 

So it seems that the original DDR4 daughter card may have been faulty.

I am still in conversation with my boss regarding the next steps on obtaining a replacement and shipment details. I will let you know our next plan of action.

Thank you, once again for your help. Your and Intel's support is much appreciated. 

0 Kudos
Deshi_Intel
Moderator
2,238 Views

HI,


I am setting this forum case to closure since we have moved the discussion in private message mode.


Let's continue to close loop in private message.


Thanks.


Regards,

dlim


0 Kudos
Reply