Question Upgrade suggestions for structural engineering desktop

Oct 18, 2024
5
0
10
Hi people, this might be quite the niche case but I am confident some of you guys here are just the people to ask.

Our company has a desktop machine for design and engineering work with the following specs

Intel Xeon Gold 6230 20 core/39 threads x 2 (dual core)

HP 81C7 mobo Intel C622(Lewisburg-2) PCH

192 GB ddr4 6 channel (1463x2) 21 21 21 47 timings

DUAL RTX 2080 ti TY102-300A 11gb ggdr6 352 bit

samsung 970 evo plus 1 tb

We use it to run design and engineering software such as STARCCM+ for simulations, Autocad, Rhinoceros which for the most part run adequately but

one of our projects involves a Rhino 7 model that is very unoptimized and resource heavy

It appears like it bottlenecks at using only a single processor of the CPU and no I see no resource use when it comes to the GPU.

Is there a viable CPU upgrade path for our machine to upgrade to higher frequency CPU's or should we look into acquiring a new build altogether?

We want it to be able to handle some of the heaviest loads in revit, rhinoceros, navisworks if possible.
 
Hi people, this might be quite the niche case but I am confident some of you guys here are just the people to ask.

Our company has a desktop machine for design and engineering work with the following specs

Intel Xeon Gold 6230 20 core/39 threads x 2 (dual core)

HP 81C7 mobo Intel C622(Lewisburg-2) PCH

192 GB ddr4 6 channel (1463x2) 21 21 21 47 timings

DUAL RTX 2080 ti TY102-300A 11gb ggdr6 352 bit

samsung 970 evo plus 1 tb

We use it to run design and engineering software such as STARCCM+ for simulations, Autocad, Rhinoceros which for the most part run adequately but

one of our projects involves a Rhino 7 model that is very unoptimized and resource heavy

It appears like it bottlenecks at using only a single processor of the CPU and no I see no resource use when it comes to the GPU.

Is there a viable CPU upgrade path for our machine to upgrade to higher frequency CPU's or should we look into acquiring a new build altogether?

We want it to be able to handle some of the heaviest loads in revit, rhinoceros, navisworks if possible.
First I think your CPU is a Xeon Silver 6230 rather than a 6320. That is a second gen Xeon scalable. If you need more CPU speed, you generally have fewer cores. For example, your 6230 has a base clock of 2.1Ghz and a max clock of 3.9Ghz. A Xeon Gold 6256 has a base of 3.6Ghz and max of 4.5Ghz. It also has 12 cores rather than 20. The C622 motherboard is limited to first and second gen Xeon Scalable CPUs. The Intel ARK -- https://ark.intel.com/content/www/u...3/2nd-gen-intel-xeon-scalable-processors.html can show you the option. Your motherboard was from an HP Z8 G4 workstation (saying that would have been helpful). HP may only support specific CPU options.
 
First I think your CPU is a Xeon Silver 6230 rather than a 6320. That is a second gen Xeon scalable. If you need more CPU speed, you generally have fewer cores. For example, your 6230 has a base clock of 2.1Ghz and a max clock of 3.9Ghz. A Xeon Gold 6256 has a base of 3.6Ghz and max of 4.5Ghz. It also has 12 cores rather than 20. The C622 motherboard is limited to first and second gen Xeon Scalable CPUs. The Intel ARK -- https://ark.intel.com/content/www/u...3/2nd-gen-intel-xeon-scalable-processors.html can show you the option. Your motherboard was from an HP Z8 G4 workstation (saying that would have been helpful). HP may only support specific CPU options.
Correct it is a Z8 G4 workstaion buuut I am 100% sure it is a Xeon Gold as it was 40 threads. We are considering keeping this machine for rendering and acquiring a new one for structural engineering, have anything in mind that will be relatively future proof?
 
Correct it is a Z8 G4 workstaion buuut I am 100% sure it is a Xeon Gold as it was 40 threads. We are considering keeping this machine for rendering and acquiring a new one for structural engineering, have anything in mind that will be relatively future proof?
In terms of something new I would look at AMD Threadripper Pro 7000 series or 4th or 5th Gen Epyc if it has to be Intel then 4th or 5th Gen Xeon. Do you have the ability to put this in a rack mount instead of a desktop? If so that opens up a lot of possibilities for different setups.
 
In terms of something new I would look at AMD Threadripper Pro 7000 series or 4th or 5th Gen Epyc if it has to be Intel then 4th or 5th Gen Xeon. Do you have the ability to put this in a rack mount instead of a desktop? If so that opens up a lot of possibilities for different setups.
We do inded have some space on the rack left, are there any decent ones that come in rack mounted form?
 
Correct it is a Z8 G4 workstaion buuut I am 100% sure it is a Xeon Gold as it was 40 threads. We are considering keeping this machine for rendering and acquiring a new one for structural engineering, have anything in mind that will be relatively future proof?
The 6230 is 20 cores and 40 threads. But you say your software can't use that many threads. This is where understanding your problem comes in. Fewer higher clocked cores may be more productive. Disabling hyperthreading in the BIOS might actually improve your performance. You might benefit from upgrading to a newer version of Rhino. Have you posted to the community on the Rhino website? They are probably much more expert than a generic tech website.
There is no one-size-fits-all answer.
 
The 6230 is 20 cores and 40 threads. But you say your software can't use that many threads. This is where understanding your problem comes in. Fewer higher clocked cores may be more productive. Disabling hyperthreading in the BIOS might actually improve your performance. You might benefit from upgrading to a newer version of Rhino. Have you posted to the community on the Rhino website? They are probably much more expert than a generic tech website.
There is no one-size-fits-all answer.
I have, the issue is that my issue has two sides a hardware side that extends to the software side.
They can answer stuff that related mainly but not only to the software side and the most solid hardware advice I found, well right here from you tbqh. So I am trying to compile opinions from both sides to make a decision.
 
I have, the issue is that my issue has two sides a hardware side that extends to the software side.
They can answer stuff that related mainly but not only to the software side and the most solid hardware advice I found, well right here from you tbqh. So I am trying to compile opinions from both sides to make a decision.
The Rhino community would be able to address the relative scalability of the software and any enhancements in Rhino 8 vs 7. That is where I would start. Or in ways to optimize your model. Fixing the software is the best answer. Hardware can only compensate some for bad software.
 
The Rhino community would be able to address the relative scalability of the software and any enhancements in Rhino 8 vs 7. That is where I would start. Or in ways to optimize your model. Fixing the software is the best answer. Hardware can only compensate some for bad software.
Ι agree but it is extremely improbable we will get a more optimized model and near impossible our problem will push Rhino to reprogram their software into a multithreaded one, so seeing as how our relatively low frequency machine struggles BUT I saw it performing quite better on a normal workstation with a higher clock cpu I thought maybe the hardware people could be onto something here.
 
We do inded have some space on the rack left, are there any decent ones that come in rack mounted form?
There are a lot of rack mountable workstations or servers with GPU slots. For example the SuperMicro AS - 2015HS-TNR can have up to 4 Dual Slot GPUs installed in it and still have network connectivity via the AIOM slot (the AS - 2025HS-TNR is the dual socket version). Since your software doesn't seem to be very well threaded, going with a higher clocked CPU is ideal AND you could still go dual socket if need be. From AMD the Epyc 9175F is a good example. It is a 16c/32t CPU with a 4.2GHz base clock and 5GHz boost clock. Since it has a high TDP (320W), I wouldn't be shocked if it ran near the max boost most of the time.

On the Intel side it is a bit harder to find the extrememly high clocked CPUs anymore. The 5th Gen and 6900 series top out at 4.2GHz with a 3.9GHz base clock on an 8c/16t CPU and the 6900 series is 72c/144t minimum for the CPU. Going down to the 4th Gen you have the Xeon W5-2555X which is 14c/28t 3.3GHz base and 4.8GHz boost CPU. That is a single socket only CPU. These CPUs are going to have more per core performance than your Cascade Lake CPUs right now, but the Zen 5 core Epyc 9175F is still quite a bit faster core for core over anything Intel offers right now.

Threadripper Pro is only on the Zen 4 cores. The 9175F will be equally as fast if not faster than the 16c/32t 7955WX Threadripper Pro.
 
Ι agree but it is extremely improbable we will get a more optimized model and near impossible our problem will push Rhino to reprogram their software into a multithreaded one, so seeing as how our relatively low frequency machine struggles BUT I saw it performing quite better on a normal workstation with a higher clock cpu I thought maybe the hardware people could be onto something here.
I would say the solution then is a new workstation with a single fast CPU.
 
  • Like
Reactions: Lutfij
Ι agree but it is extremely improbable we will get a more optimized model and near impossible our problem will push Rhino to reprogram their software into a multithreaded one, so seeing as how our relatively low frequency machine struggles BUT I saw it performing quite better on a normal workstation with a higher clock cpu I thought maybe the hardware people could be onto something here.
I don't know how Rhino is licensed, but if you can get a temporary license to try some different hardware. Maybe run on AWS or Google cloud.
The other thing is to look for Rhino benchmarks that have been done by others. That will give you some insight, but won't transfer perfectly since your model is different than a benchmark model.
How many total threads are utilized when running Rhino? 1, 4, 10? It might be the case that a desktop CPU, that is optimized for single thread performance would be the best option. A Ryzen 7800X3D might give you as much performance as your Xeon. A desktop CPU has MUCH less memory bandwidth, so your might benefit from clock speed and get hammered by memory bandwidth. I don't know.