This category is for discussion of different hardware options for running Itasca software. Pros and cons of different processors, architectures and cloud offerings should be discussed here.
Hi
I wanted to know the best CPU configuration to run FLAC, 3DEC and PFC software to work on engineeering projects.
Hello Itascans and fellow users.
I’m trying to establish the specification for a new numerical modelling “cruncher” and was hoping to get some advice from those with experience benchmarking or running larger models on similar hardware.
The What computer hardware does Itasca recommend for its software? webpage provides some useful guidance on what to prioritise, particularly the recommendations around high CPU clock speed/core count and maximising RAM bandwidth by populating all available memory channels. However, I’m still left with a few questions and unfortunately I haven’t had the opportunity to test many “crunchers” in the past, nor do I have the luxury of trying different configurations now before buying.
Due to company procurement/commercial arrangements, the choice is likely to come down to a compromise between CPU performance and memory bandwidth. The two main options I’m currently considering are:
Option 1 – AMD Threadripper PRO
- Dell Precision 7875 Tower
- AMD Ryzen Threadripper PRO 9975WX - 32 cores / 64 threads, 4.0 GHz base / up to 5.4 GHz boost, 32 MB L2 / 128 MB L3 cache
- 512 GB ECC RDIMM - 8 x 64 GB, all eight memory channels populated. Unfortunately, memory limited to DDR5-5200, with a possibility of DDR5-5600.
Option 2 – Intel Xeon
- Dell Pro Precision 9 T4 (PW9T4260) or T6 (PW9T6260)
- Intel Xeon 676X - 32 cores / 64 threads, 2.8 GHz base / up to 4.9 GHz boost, 128 MB L2 / 144 MB L3 cache
- 512 GB ECC RDIMM - 8 x 64 GB or 16 x 32 GB, all eight or sixteen memory channels populated. DDR5-6400 supported.
The remaining components would be selected to complement the CPU and RAM, with GPU performance being a secondary consideration.
The main trade-off is the higher CPU frequencies of the Threadripper versus the greater memory bandwidth and cache of the Xeon. I would be particularly interested in any experience or guidance on the following:
- How important is CPU/per-core performance relative to RAM bandwidth? If having to compromise on one, is there generally greater benefit in prioritising higher CPU performance or greater memory bandwidth?
- Is there a useful rule of thumb for balancing CPU core count with RAM capacity? Does the CPU clock speed and/or RAM bandwidth play into the equation?
If anyone has benchmarked either of these processors, or similar current-generation Threadripper PRO or Xeon processors, with 3DEC, FLAC3D or other Itasca software, those results would also be very useful.
Any experience or advice would be greatly appreciated. Thank you!
@jhazzard @ddegagne tagging you both as you’ve previously shared some useful insights on hardware and performance.
You can look at the benchmark page here to see if anyone has tried similar configurations. I know that our own consultants generally use the threadrippers.