Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
10 years ago

Quartus Prime Arria 10 synthesis and PC memory utilization

I'm just starting to play with Quartus Prime (15.1.2)

I have a legacy Stratix V design that I've ported to Arria 10.

After kicking off synthesis, Quartus eventually spawns 7 instances of quartus_syn.exe which mostly allocate around 17 GB RAM (as listed in the "Commit" column in Windows resource monitor, which I believe represents total virtual memory allocated), but one of the seven allocates 42 GB, so the total is around 140 GB. My PC is Win7, 12 core @ 3.3 GHz, 32 GB RAM, with an additional 120 GB (!) pagefile reserved on the drive containing the Quartus program.

Altera recommends up to 48 GB RAM for Arria 10. I'm using all the default synthesis options.

Is it normal for Quartus Prime to use up this much memory and spawn this many processes during synthesis? I didn't have any trouble routing this design on Stratix V with Quartus 13.1. The design does contain a very large qsys component, which I've also regenerated.

Are there synthesis options that can reduce memory utilization in Quartus Prime?

Ultimately we plan to buy a beefier machine to do FPGA routes and I want to get a sense of real memory requirements...so that's my focus for now.

Thanks,

Tom

2 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor

    --- Quote Start ---

    I'm just starting to play with Quartus Prime (15.1.2)

    --- Quote End ---

    Tom- I'd love to hear more about this too. I've been looking at network/cloud-based solutions, as well, because anything is better than needing a supercomputer for routing/synthesis.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor

    I tried reducing the number of processor cores (default setting is to use all available cores) and that seemed to help. I reduced it from 12 cores to 4. Synthesis completed about 25% faster and used much less memory (51 GB peak virtual memory), and also spawned fewer processes (5 instead of 7). Interesting...