Hello everyone, i bought a workstation with an AMD Threadripper 2990wx (64 thread) and ram 128 gb ddr4. During simulation it is equal to a 8 thread Intel I7. Why is it slow during simulation? Thanks for the answere.

It depends on the situation, but there are always diminishing returns with number of threads and simulations. The single threaded parts of the simulation won't benefit from more threads, so increasing the thread count will eventually hit a wall. If the intel part has higher single threaded performance, it may end up being faster for certain simulation types. Threadripper complicates the problem more with its non uniform memory access–not all cores have the same speed of memory access.

What type of simulation are you running, and what operating system are you using? There is a known issue with threadripper on windows that severely limits performance when using all threads.

jsmack
It depends on the situation, but there are always diminishing returns with number of threads and simulations. The single threaded parts of the simulation won't benefit from more threads, so increasing the thread count will eventually hit a wall. If the intel part has higher single threaded performance, it may end up being faster for certain simulation types. Threadripper complicates the problem more with its non uniform memory access–not all cores have the same speed of memory access.

What type of simulation are you running, and what operating system are you using? There is a known issue with threadripper on windows that severely limits performance when using all threads.

Hi, in response to your question : I use Windows 10 Pro for Workstations. I'm not trying to make a precise simulation, but I'm trying to understand why a workstation of this power … with 128GB of RAM … does not differ from the 8 Intel I7 threads …
Thanksss

Hi, in response to your question : I use Windows 10 Pro for Workstations. I'm not trying to make a precise simulation, but I'm trying to understand why a workstation of this power … with 128GB of RAM … does not differ from the 8 Intel I7 threads …

..as jsmack said..it might be because there are a single threaded processes occurring which limits the benefit of more threads.

Perhaps you could post a hip of your sim with times it took for your system. Someone with experience of applying sims on different systems might be able to tell you if your times are as what would be expected or not.

Those processors are known to have memory bandwidth latency issues via infinity fabric, as mentioned. Also, as others pointed out, you can circumvent the largest issue, if you deal with Windows problem with multi-core iteration allocation.

I'm not a Windows person, but I recall it having something to do with setting the first core count to start at 1, instead of 0… or perhaps visa versa. Heard there is signifcant jumps in performance once you get Windows bugginess dealt with. Still, to my knowledge, infinity fabric has some bandwidth latency, particularly on the 2990WX as two of the four dies are handicapped.

Fix the Windows thing, and you should see vast improvement.