Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Corei7 or Corei5, which one is better for VHDL programming?

Status
Not open for further replies.

ZX_SPECTRUM128

Newbie level 5
Joined
May 9, 2017
Messages
9
Helped
0
Reputation
0
Reaction score
0
Trophy points
1
Activity points
92
Hello everyone
i am thinking about buying a new desktop. and i am confused what processor should i chose to be more appropriate for VHDL development? i5 or i7?
thanks
 

VHDL doesn't have a processor requirement. A text editor will work with either processor equally well.

A simulator is single threaded and will work with either processor equally well, thought the larger cache in an i7 will improve performance somewhat. Basically you want the fastest processor clock rate to speed up simulations.

The tools that implement the VHDL into the vendor's bit stream will likely be a improved with an i7 as they have more cores and many of the implementation tools are multi-threaded and can therefore take advantage of those extra cores in the i7.

Just make sure you have 16 GB of memory unless you never have plans on working with a large FPGA part (i.e. you are using the full version of the tools not the (free) web version).
 

you will keep an i7 longer than an i5,
but don't forget to buy a ssd drive, the speed is mainly here. i5 or i7 don't matter much (and very not much for vhdl programming ...)
if you don't have the money, better buy an i5 with a ssd than an i7 with a mechanical drive ...
 

thanks for the replay
from your answer i understand that i5 is sufficient, but VHDL is concurrent language where every thing, when elaborated, would be converted into a concurrent process and those processes would be working simultaneously doesn't that means that a multi-treading processor like i7 would be better?
don't you think that the hyper-threading feature of i7 would be more appropriate ?
thanks
 

but VHDL simulator is based on concurrent processes don't you think that the hyper-threading feature of i7 is more appropriate ?
 

but VHDL simulator is based on concurrent processes don't you think that the hyper-threading feature of i7 is more appropriate ?
You shouldn't be worrying much about this. The software you are using really matters.
I use VHDL and Xilinx Vivado, and in there the tool allows you to choose the no. of threads.

A real-life eg:
My laptop PC @ home is i7 with 8GB RAM with normal HDD.
My work-desktop is i5 with 8GB RAM with SSD.
The same design (a real life large design project) works faster at my office work-desktop than on my home laptop PC.

Bottom line: If u have money i7 + SSD else i5 + SSD, RAM 8GB will suffice, but 16GB much better.
 

in reality, i5 and i7 have 4 processors cores.
hyperthreading on the i7 makes the system think there are 8 procs, they are simulated by software.
in reality, you won't see much difference between the two.
the rest of the architecture of the computer (ssd or not, graphic card power, ram, ) will make most of the difference
and I insist on it but the ssd is mandatory on a 2017 computer, it would be non sense to omit a ssd drive.
 

but VHDL simulator is based on concurrent processes don't you think that the hyper-threading feature of i7 is more appropriate ?

I've never noticed any of the VHDL/Verilog simulators being multi-threaded. I've used Modelsim, Vivado, Incisive, and Aldec (though not recently). None of them appear to have a multi-threaded simulation kernel. At leas they never use more than 1 core on the processor when I've checked while running a simulation (though Modelsim does seem to use a second core for updating the wave windows when running with the GUI).

The best way to improve simulation performance is use an SSD drive (like others have mentioned), get the fastest processor (clock frequency) (overclock the thing like crazy) and lots of RAM (you never want the simulation to page swap to disk).
 

ok so you are advising me to invest my money in higher clock frequency processor.
i was wondering; why ???!!!
simulation in VHDL seems to be performed in patch. i mean it is just like compiling a code in any other language; compilation is just something that take place when you finish coding and hit the compile bottom. so instead of taking 800 ms on a fast processor i don't mind it if it took 1500 or 2500 ms on a slow processor
the same thing with VHDL simulation it is not like that it is interactive and i can watch how the design is being simulated and how the circuit is processing inputs and computing outputs. but rather simulation ,like C or JAVA compilation, happens in one patch and takes only seconds.
so why to get fast processor at the first place, i won't die if i waited extra 1-2 seconds each time i simulate, elaborate. my code

could you please clear up this point for me and explain why it is not so if i am not right
actually this point confuses me a lot so please explain if i am missing something about VHDL simulation
thank you
 

I have had simulations that run for hours, and I have had regression simulations for ASICs that would take a week or more to run on a single very very fast workstation. To run a simulation takes many instructions cycles on a computer, that is just a fact of life.

Don't try and correlate a software C compiler to a HDL simulator they aren't even close to being the same thing.

But hey don't let me dissuade you, go right ahead and believe that a 10,000,000 gate ASIC simulated for 100ms takes only seconds to run...heck my tiny design right now (< 20,000 gates) takes nearly 1 minute to compile in modelsim on a Xeon workstation and takes nearly a 30 seconds to load and run for ~275 us (the actual time the processor runs the simulation image is 4 seconds).

Another example was a design that ran for ~200 ms and was a 75% full Kintex 325T, the simulation of those testcases would be about 40 minutes on a $12K Xeon workstation.

Of course Modelsim (the real version) is a relatively slow simulator (Vivado's simulator is worse ;-)), if you want fast get Incisive or VCS. Of course that assumes you have $$$ lying around doing nothing.
 

i got it, so in choosing a processor for VHDL it si al about the clock speed since you are sure that most simulators are not multithreaded.
so beside choosing a processor, do you think that a workstation would be more appropriate than a regular pc ?
 

i got it, so in choosing a processor for VHDL it si al about the clock speed since you are sure that most simulators are not multithreaded.
so beside choosing a processor, do you think that a workstation would be more appropriate than a regular pc ?

Of course a workstation is better, at least in the sense of thermal management, maximum amount of memeory, and memory performance. The last "real" workstation I used was a Z800 dual Xeon (48 threads, 12 cores x2 processors) 48GB of quad channel DDR3, ECC memory. That was a nice machine, I could run all of the builds for each of the FPGA configurations (6 total) that were in our system all at once in parallel with only a small 10% increase compared to a single FPGA build. Along with running simulations on the next version of the design (or multiple simulations).
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top