UE4 - intel 7700K or AMD 1800X?

I want to buy new PC so i want to know which processor will speed up more UE4 tasks?
Thanks :slight_smile:

you think anyone have tested 1800x with UE ? ))
probably you need to wait at least month to have some mature comment on that
but it’s very good if 1800 is a real competitor for 6900k

From what I’ve seen from benchmarks so far, the 1800X is not as good as a straight gaming processor as the 4C/8T 7700K, but for heavily multi-threaded tasks it can match and sometimes exceed the 8C/16T 6900K. In case of UE4 development, with relies on plenty of multi-threaded tasks like compiling C++ code, baking lightmaps or compiling shaders, the 1800X might do very well and provide excellent value.

thanks ! :slight_smile:

Yeah, for UE4 where you might spend a lot of time with baking lighting or compiling code, you want to have as many cores as possible. The new Ryzen CPUs are the best you can get for that.

Just a quick note for those on a budget (or anyone who wants to save some cash), if you overclock the 1700 (no x) it appears to clock just as well as the two X chips.
Perhaps wait for more 1700 reviews where they overclock but it looks like a steal.
1800X & 1700X also don’t come with a cooler, the 1700 does.

Also to OC the X’s you have to turn off XFR anyway so you aren’t missing out on anything.

I haven’t read anything about binning yet though, so take it with a grain of salt.
I’ll probs be picking up the 1700 soon, nice upgrade from my trusty i5 750(!).

Edit: If I come across more 1700 OC reviews I’ll try to remember to post a link.

That be great!

Is Blueprint compilation, project loadup (cache generation etc), launching PIE and shader compiling able to fully utilize 16 Threads? Especially the information about blueprint compilation would be useful.

No idea what you mean with “launching PIE”, shader compiling will use all 16 threads, blueprint compilation will likely only use 1 thread. C++ compilation will use 16 threads.

I’d guess PIE = play in editor, but even with my old system I’ve never noticed it take much (if any) time.
Like John said it’ll be great for:
Compiling Shaders, compiling UE4/C++ code, running lightmass and some other external programs like World Machine.

For actually playing/testing your game these won’t offer the best performance.

Some links:

If I were to guess based on my experience working in large Blueprints, I would say that BP compilation uses 1 thread. Otherwise, Epic has a lot of work to do when it comes to compilation times.

People say AMD 1800X not good for gaming, is it true? :slight_smile:

More threads makes lighting bakes go faster, assuming you have the RAM to match.
You might really want to look at the Core i7 6950X if that’s your main limitation :slight_smile:
Then again, if that’s your main limitation, a blade center running 16 blades of multi-core CPUs, each running a Swarm renderer agent, would accelerate that even better :slight_smile:
(and, then, you could actually set that up on Azure or Amazon, and only rent it for the times when you really need the fast render speed)

I’d love to see a write up on the costs and how to take advantage of Azure or Amazon for lightmass. It would be interesting to see the cost analysis and when it would be better to build your own small render boxes/farm.

They’re still fine for gaming, just not as fast as the fastest Intel processors.

Games generally heavily rely on a single main game thread, so single-core performance is still the most important factor for the gaming performance of a CPU. AMD has not quite matched the IPC (instructions per clock) of Intel yet, though they’ve made major strides in that regard, and the Ryzen 7 processors are clocked relatively low because of their high core count. That all equates to lower single-core performance on the 1700/1800 Ryzen processors compared to something like the 7700K from Intel, which results in lesser gaming performance when the CPU is the bottleneck (typically at resolutions of 1080p and below). However, once you reduce the CPU bottleneck by jumping to 1440p resolution or higher, the difference between the 1800X and 7700K becomes much less pronounced. Also, we have yet to see what the quad-core and hexa-core Ryzen 3 and Ryzen 5 processors will look like, especially in terms of clock speed. It’s likely those will be clocked higher, resulting in better gaming scores.

Anyway, the general consensus is that the 1800X is best for content creators who rely heavily on multi-threaded tasks. So video encoding/streaming, 3D rendering, or code compilation and light baking with UE4, those are all good reasons to own an 8-core/16-thread processor. And AMD is now providing great value for those users with their Ryzen chips. Besides that, they also perform very respectably in games, albeit not top of the line.

In other words, there’s a difference between “not good” and “not the best”. People often seems to mix those two up.

I vote for intel:)

For multi-threaded tasks, the Ryzen chips blow Intel out of the water. If I was building a new workstation for UE right now I wouldn’t hesitate. Probably just get a 1700 and overclock it as they seem to clock pretty easy up to 1800X performance levels.

If you want 8 core Intel you’re looking at the 5960X which is big $$$.

Personally, I would not rely on overclocking anything for something I need to actually work.
And, even for gaming these days, having things work reliably is actually more important than getting the extra 3 fps to me…

Regarding rendering on Azure, I specced it out a while back, but didn’t get to the point of actually trying it for real.
16 Windows instances with 8 cores, 64 GB of RAM (A8m v2) for 10 hours, 100 GB of networking, 100 GB of VPN traffic, comes out to $156.
You’d need to sysprep a new VM image that has Swarm installed (and presumably the UE just to get all the dependencies right, even though the UE itself is not needed by Swarm)
Then, join your computer to that VPN, point the rendering setup at that Swarm, and go! (one can hope :slight_smile:

The 10 hour estimate comes from a hobby project that only does final bakes a few times a month. There are 744 hours per month, and 176 working hours per month, approximately.
(Other typical setups might include an automated full release bake/build every monday-friday morning at 1am for a couple of hours.)

Aaanyway, if anyone takes that to the next step, let’s hear about it!

It’s not like the old days where you had to spend hours swapping jumpers and running tests :slight_smile: These days you just click a button in the bios and it’s done, within warranty.

The 330 USD Ryzen 1700 matches and even beats in some cases the 1000+ USD 6900k at stock clocks, clock it a little and it’s a no brainer.

Except for when you get a random blue screen once in a while, usually when doing the most work (so the load is the highest.)

Getting a completely stable 20% overclock with 15 minutes of work is pretty straight forward on pretty much any modern processor. You’ll save that 15 minutes in reduced baking times real quick.