Hi fellas, it looks like I will be getting a new PC recently. https://unity3d.com/unity/system-requirements
This link is not helping me much. Right now I am using one, having low to medium range, with this specification:
Core i5-4570
16GB ram
GTX 650 Ti
Windows 10-64bit
No SSD
The light-map baking, takes too much time to bear, comparing the former “autodesk beast light-mapping”. A compilation of UE4 from the source, takes around 30 to 40 minutes. A webgl build or il2cpp takes ridiculously long time to finish, 10 to 30 minutes :(. Now I wish to stop this madness and thus decided to upgrade.
I should be happy for at least 3-4 years next. The estimated budget is around 2k USD. So suggest me some.
I have a 4000$ computer with bi-xeon 4ghz(24 logics cores)and the lightmap bake system with Unity5 is very very long too.
My previous computer had a simple i7 4 cores($800 computer), and sincerely the difference is not impressive.
Don’t waste your money, just buy a SSD and at least a i7 core for your 1150socket
I strongly suggest using SSD drives, since that will significantly boost performance. If you want to upgrade your existing PC, switching to an SSD drive, 32GB of RAM, and a GTX 1070 would yield a huge performance boost.
Switching from your current i5 to an i7 would basically only get you hyperthreading support. Hyperthreading improves performances, but not as much as additional cores will do. With hyperthreading, your PC will show 8 “procs” in the OS, but it won’t run much better than your i5 runs in most tasks. An i7 will still be a 4 core CPU. Hyperthreading offers a relatively mild performance boost.
As for 16GB vs 32GB, I have seen a few times in both Unity and UE4 where light baking certain scenes needed more than 16GB of RAM. Plus the added RAM can be used for file caching by the OS, which can be helpful when compiling applications and games.
I would not hold my breath waiting for AMD to ship a competitive CPU. The new Zen based AMD CPUs sound impressive, but AMD’s track record in recent years has been consistently poor. AMD hypes high and then under delivers. If I had to guess, I would say AMD’s Zen CPU will probably ship late and perform measurably slower than Intel’s current CPUs. AMD’s Zen CPU will probably not even match the performance of previous generation Intel i7 CPUs.
My point is if you upgrade is not an urgent need then it could be worth your while to wait for AMD’s Zen. If it does challenge intel on price/performance then expect Intel to drop their prices. If not then you can probably save up more for a better upgrade.
A Gtx1080 or a gtx650 will change nothing.
And I prefer to develop a game with a gtx650 to optimize my game to stay above 30fps.
Start to buy a Samsung 850Pro, check if your 16Go are enough or not.
hmm, SSD will be added for sure. Its sad that enlighten can’t utilize GPU power for baking. The baking stage is terrible and essentially worsen the whole game development pipeline Its the weakest link in the chain Several times I have thought of building a plugin or something like that, for blender. So that I can back n forth DCC packages and game engines. The idea is to leverage the GPU accelerated lighting on those DCC packages. I would bake away the light as texture and put it into unity by a single button click. Its so much pain to generate them within unity
It is one of the prime reasons, for upgrade. As pointed out by menfou , its really sad that upgrading wont help in this area. Is there any server side solution any of you can think of? For example I would put the whole unity project into the cloud and wait a few minutes and then done with lighting?
I can wait few more months for AMD but I honestly do not want to. I have also thought of getting mac pro. Would be nice to run both windows and macOS, I could cover the platforms. but it is overpriced for me
About RAM, going from DDR3 to DDR4, how much does it worth? Is ECC ram that necessary for me?
Another thing that will improve performance is a RAM cache. About a year ago I maxed out the RAM on my motherboard (32GB) and installed SuperCache. It’s a big improvement.
On my next computer I’m going to get a motherboard than can handle 128GB RAM and dedicate most of it to a RAM cache.
Has anyone actually tried bench-marking this stuff?
I am suddenly interested in knowing the performance differences between building/baking/etc on different configurations.
Even if the difference isn’t a lot, I would rather optimize my hardware for Unity than for playing games…
That CPU has 64-72 cores, but runs at a relatively low clock speed of 1.3GHz-1.5GHz. It would only be useful for situations where you knew all of the cores could be used effectively. Unfortunately, many tasks we do day to day are going to benefit more from high clock speed than from high core count. The Intel Xeon Phi 7290 might be great for running a bunch of virtual machines in a datacenter, but I am guessing Unity will likely run better on an Intel i7-6700K.
Did you notice the Cores come in tiles, 2 cores and 2 VPU’s per tile, the VPU’s are 512 bit SIMD processor for doing batched maths very fast e.g. 8 double or 16 floating point ops per cycle. I’d betting they could tear through GI calculations in a fraction of the time of a regular CPU can. Assuming your GI calculator was written and compiled to run on Knights Landing hardware.
Yeah, the Knights Landing hardware looks impressive, and I don’t doubt some software will eventually get tuned for it. But I highly doubt most software we use day to day will benefit from it. For example, I don’t expect to see the Unity editor or Enlighten get optimized for Knights Landing.
I would imagine that up to a point calculating scene GI is not as restricted to processor speed as it is to number of cores and bandwidth. And who wouldn’t love to be able to send GI off to a cheap ARM server farm and get it back in when it’s done, whilst continuing work on you desktop machine.
Individual core speed is just as important as the number of cores in your processor. Just as one example the Intel i3-6320 is only a dual core processor yet it’s able to compete with the eight core AMD FX-8350 because each of its cores are that fast.
Right now I am using one, having low to medium range, with this specification:
