Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/984
Article Thumbnail

Unreal Engine 4.16 CPU Comparison: Skylake-X, Kaby Lake-X, Broadwell-E, Skylake, Ryzen 7

Written on July 12, 2017 by Matt Bach
Share:

Introduction

It is very common for video games to be used in hardware reviews, but the development platform for those games are typically completely ignored. Today, we are going to help alleviate this problem by looking at a how a number of CPUs perform in the Unreal Editor for Unreal Engine 4.16. This includes the latest CPUs from both Intel (Skylake-X and Kaby Lake-X) and AMD (Ryzen 7) in addition to CPUs from the slightly older Intel Broadwell-E and Skylake lines.

One thing to note is that although the Skylake-X and Kaby Lake-X CPUs are officially launched, Intel has only released the 4, 6, 8, and 10 core models so far. According to Intel's E3 press release, there will be a 12 core model available in August along with 14, 16, and 18 core models in October. Unreal Engine has a number of tasks that are very well threaded (where a higher core count improves performance) including building lighting and compiling, so this is really only a taste of how the Skylake-X line as a whole might perform in Unreal Editor.

There are a wide variety of tasks we could test in Unreal to see how all these CPUs perform, but in this article we will specifically be looking at the following tasks:

  1. Time to open a project
  2. Time to package a project for Win64
  3. Time to launch the level editor "play" for standalone and mobile preview
  4. FPS in the editor viewport
  5. Time to build lighting (preview, medium, high, and production quality settings)
  6. Time to compile engine from source

If you would like to skip over our test setup and analysis of the individual benchmarks, feel free to jump right to the conclusion section.

Test Setup

Since the new processors from Intel include two different CPU architectures, our testing platform is less straightforward than normal. Most of the new CPUs support DDR4-2666 RAM, although the Intel Core i7 7740X (Kaby Lake-X) CPU only supports four sticks of RAM rather than eight. In addition, the Intel Core i7 7800X (Skylake-X) only supports DDR4-2400 RAM for some odd reason. We typically test with the highest officially supported RAM configuration for each CPU, so components used in our X299 test platform requires an entire table just for itself:

To act as a comparison to the new Intel CPUs, we will also be testing AMD's Ryzen 7 1700X/1800X CPUs as well as the Intel Core i7 7700K (Kaby Lake) and Intel Core i7 6850K/6900K/6950X (Broadwell-E) CPUs. Again, we will be testing with the best officially supported RAM configuration for each CPU.

Inconsistent RAM aside, one thing we do want to point out is how much more affordable the Skylake-X CPUs are compared to the previous generation Broadwell-E CPUs. Where a six core CPU from Intel used to cost over $600, you can now purchase an eight core CPU for roughly the same cost. In fact, the i7 7800X is even a hair less expensive than AMD's Ryzen 7 CPUs. This means that regardless of any performance gains we may see, Skylake-X is at the very least a big deal in terms of cost.

In order to accurately benchmark the different CPUs, we tested the performance across two sample projects that are freely available for download. Our testing will obviously not be 100% accurate for any project you might create, but these two projects are different enough to help us identify a number of performance trends.

Infiltrator Demo [Marketplace]

Epic Zen Garden [Marketplace]

Open Project

Unreal Engine Editor Skylake-X 7900X 7820X 7800X Kaby Lake-X 7740X Ryzen 7 Project Open Benchmark

The time it takes to open a project may not be a terribly exciting way to start our testing, but it is something you do multiple times a day, every single day. Performance here may not be your number one priority, but the less time you spend waiting on your system, the more productive you will be.

The key thing to note in this test is that the lower core count CPUs (which tend to have higher single-threaded performance) are clearly the best for this task. The new Intel Core i7 7740K (Kaby Lake-X) is the fastest overall, although the Intel Core i7 7700K (Kaby Lake) was only a tiny 1.5% slower. The Intel Core i7 7820X was the next fastest, followed by the rest of the Intel CPUs which all performed roughly on par.

The lowest performing were the AMD Ryzen 7 CPUs, although that is somewhat to be expected as those CPUs do not have particular great single-threaded performance.

Package for Win64

Unreal Engine Editor Skylake-X 7900X 7820X 7800X Kaby Lake-X 7740X Ryzen 7 Package for Win64 Benchmark

Packaging (whether it be for Win64, Win32, Android, etc.) is a task that some readers may expect to finish faster on a higher core count CPU, but our results indicate that more cores is definitely not better.

Just like in the previous section, the lower core count CPUs (which have higher single-threaded performance) are again the best for this task. The higher core count CPUs did better than they did when opening a project, but none of them are able to quite match the quad Core CPUs. The new Intel Core i7 7800X and 7820X along with the Core i9 7900X were the closest as they trailed the Core i7 7740X by only 4-8%.

Also once again, the lowest performing CPUs were the AMD Ryzen 7 1700X and 1800X which were about 15-25% slower than the Core i7 7740X.

Launch Level Editor Play

Unreal Engine Editor Skylake-X 7900X 7820X 7800X Kaby Lake-X 7740X Ryzen 7 Level Editor Play Benchmark

The first time you launch the level editor "Play" can take a significant amount of time depending on your project and the mode you are using. For this test, we decided to measure the time it took to get into your game with both the "Mobile Preview (PIE)" and "Standalone Game" modes.

It may be feeling a bit repetitive at this point, but we again saw the best performance with the quad core Intel CPUs. The new Skylake-X CPUs were not far behind, however, with the Core i7 7820X only performing about 4% slower on average. Across the other CPUs we tested, the results are pretty closely in line with what we saw in the previous two sections.

Editor Viewport FPS

Unreal Engine Editor Skylake-X 7900X 7820X 7800X Kaby Lake-X 7740X Ryzen 7 Editor Viewport FPS Benchmark

We considered doing a whole host of FPS testing, but given that the FPS is heavily dependent on the GPU we decided to stick with just the FPS in the editor viewport rather than measuring the FPS in a variety of situations. Note that unlike every other test, this time a higher result is better since we are looking at frames per second rather than the time it takes to complete a task.

Overall, the results for the new Skylake-X CPUs are a bit odd. For the 6-10 Core across both the Broadwell-E and Skylake-X product lines, we saw a clear increase in FPS as we got into the higher core counts. What is weird is that the new Skylake-X gave much lower FPS than the Broadwell-E CPUs. Since the Skylake-X CPUs have out-performed the old Broadwell-E CPUs on virtually every test we have run over the last few months, this looks more like a new architecture bug than anything else. If maximum viewport FPS is a major concern for you, we would recommend holding off on purchasing a Skylake-X CPU for now but we expect these results to rise dramatically in the coming months.

Build Lighting

Unreal Engine Editor Skylake-X 7900X 7820X 7800X Kaby Lake-X 7740X Ryzen 7 Build Lighting 1 Benchmark

Unreal Engine Editor Skylake-X 7900X 7820X 7800X Kaby Lake-X 7740X Ryzen 7 Build Lighting 2 Lightmass Benchmark

Building lighting can take a significant amount of time and since we wanted to see if there was a difference in relative performance depending on the quality setting this one test took longer than all our other tests combines. The short answer is that while the lower quality modes are certainly faster to complete, the relative performance difference between each CPU did not change to a significant degree.

Building lighting is one of the few tasks in Unreal Editor than can take great advantage of having a high number of CPU cores, but oddly this seems to only apply to the Intel CPUs. While the AMD Ryzen CPUs have 8 cores, they were only about to beat the Core i7 7740X quad core CPU by around 5% on average and were roughly half the performance of Intel's 8 core CPUs.

The new Core i7 7800X 6 core was a bit disappointing as it is slower than the old Core i7 6850K 6 core - although given its much lower price the price-to-performance ratio isn't terribly bad. However, the real outstanding CPU in this test is the new Core i7 7820X. Even though it is fairly middle of the road in terms of price, it gave terrific performance and was only slightly slower than the much more expensive Core i9 7900X.

Compile Engine From Source

Unreal Engine Editor Skylake-X 7900X 7820X 7800X Kaby Lake-X 7740X Ryzen 7 Compile Engine from Source Benchmark

While compiling the Unreal Engine from source is technically done outside the Unreal Editor itself, we feel it is still a very valid benchmark for Unreal Engine developers.

Like build lighting, compiling is a task that can benefit greatly from having a high number of CPU cores. There are a few oddities in our results - such as the i7 7740X being slower than the i7 7700K - but overall the results line up pretty much as you might expect. If you spend a decent amount of time compiling, the new Core i7 7820X and Core i9 7900X are terrific and likely the go-to option.

Conclusion

Unreal Editor is an interesting application to test since it has a combination of lightly threaded tasks and heavily threaded ones. Because of this, which CPU is best for you is going to depend on the tasks you typically find yourself waiting on.

Unreal Engine 4.16 Skylake-X 7900X 7820X 7800X Kaby Lake-X 7740X Overall Benchmark Results

To help you decide which CPU you should use, we first normalized our results to the Intel Core i7 6850K which has typically been a great mid-range CPU for game developers. After that, we separated our results into three categories:

  1. Overall results across all tasks
  2. Lightly threaded tasks - open project, package for Win64, launch level editor "play, and viewport FPS
  3. Heavily threaded tasks - build lighting and compile engine from source

Starting with the lightly threaded tasks, it is no surprise that the Core i7 7740X and Core i7 7700K were the top performers. While these CPUs only have four cores, they excel at tasks where it is more important to have faster cores rather than having more cores. The new Core i9 7900X and Core i7 7820X were the next best CPUs coming in at about 10-13% slower than the quad core CPUs.

On the more heavily threaded tasks, the Core i9 7900X was overall the fastest CPU we tested, trailed by the Core i7 7820X and Core i7 6950X which were about 9% slower. Considering that the Core i9 7900X is almost half the cost of the Core i7 6950X and the Core i7 7820X is almost a third the cost, this is terrific from a price-to-performance standpoint! The only bit of disappointment from the new Skylake-X CPUs is the Core i7 7800X which was only a tiny bit faster than the quad core CPUs. However, given its $389 price tag it really is only slightly worse than we expected.

Although the AMD Ryzen CPUs are often touted as being an excellent value for heavily threaded tasks, they really did not do all that well in our Unreal Editor testing. Even the somewhat disappointing Core i7 7800X was still both faster and cheaper than the AMD Ryzen CPUs. And while the Core i7 7820X is certainly a bit more expensive than the AMD Ryzen CPUs, the fact that it was twice as fast for heavily threaded tasks speaks volumes for how much more effective an Intel CPU is for Unreal Editor at the moment. 

Overall, if there was only one CPU we could recommend to Unreal Editor users it would be the Core i7 7820X. It may not be the absolute fastest for any single task, but it is a great balance for anything you might do. If you spend a significant amount of time building lighting or compiling, the Core i9 7900X would also be an excellent choice if you have a bit larger of a budget. On the other hand, if you rarely build lighting or compile the engine from source, either the Core i7 7740X or Core i7 7700K are terrific at most other Unreal Editor tasks while also costing significantly less.

Unreal Engine Workstations

 

Compact
Workstation

Configure

Standard
Workstation

Configure

Tags: Unreal Engine, Unreal Editor, Game Development, Skylake-X, Ryzen 7, Broadwell-E, Kaby Lake-X
Dan_R_C

Matt
It looks like Lightmass in Unreal might not have Embree enabled for the AMD chips which would explain the terrible rendering times for Ryzen.

Posted on 2017-07-17 17:30:10

It might be that the AMD CPUs don't work with Intel's Embree ray-tracing library. I tried to do some research on whether it is Intel only or simply optimized for Intel but couldn't find much information. The log file does show list a line for "Embree Used Memory" when we ran it on the Ryzen system, however, so I'm pretty sure it uses it - just not as well as Intel.

On the other hand, the results for the Ryzen CPUs compared to the Intel CPUs was not that much different than in the compile test. So if Embree is making a difference for Lightmass I'm not sure by how much.

Posted on 2017-07-17 17:47:04
Lars Passic

Awesome article!! I am sending this to my friend who works in Unreal.

Posted on 2017-08-02 17:02:33
Brandon

Did you set "ProcessorCountMultiplier" to 2 in BuildConfiguration.xml before compiling UE4?

Posted on 2017-08-09 00:47:02

We left it at 1 (so just one worker per core). That would be an interesting thing to test next time, however, so I'm going to write it down so we don't forget. I did some research on this, but I couldn't find anyone giving an actual measurement of the performance difference they see by switching this to 2. Just some nebulous "its faster!". Any idea on what the actual performance difference is?

Posted on 2017-08-09 21:02:23
Brandon

It should yield a significant improvement - I think it could almost halve all of your times. It allows you to utilize both threads per core. On my laptop with 4 core 8 thread i7, I think it actually halved.

Posted on 2017-08-10 00:24:32
Cho

In my case, it wasn't I'm using 6900k system, and after enabling it, there was no actual improvements. I tested it several times.

Posted on 2017-08-11 01:33:32
Alessandro “Melvasul” Timpone

AMD CPU and GPU optimisation for this Engine is still terrible...

Posted on 2017-08-13 14:15:39
Eisen

Thanks for your efforts on this review! With all due respect, there's a few points that doesn't quite make sense, I'll point one of the major: You've found Kaby Lake-X (i7 7740X) being outperformed nicely by the i7 7770K on the "compile engine from source test". That can only be justifiable by blotched methodology. You see, unlike Skylake-X that has differences such as the move to L3 victim cache and introduction of the mesh as a substitute to ring for inter-core communication compared to Skylake, Kaby Lake-X (i7 7840X) is purely binned Kaby Lake (i7 7770K) on top of a big heat spreader. It's utterly identical apart from being better clocked. I'd sincerely recommended double checking your test methodology and re-running the benchmarks. They're potentially doing a disservice as-is.

Posted on 2017-08-13 14:22:41
Guest

It's still using a different chipset, and there was issues early on with the kabylake parts on some x299 boards, where it performed worse than the 7700k.

Posted on 2017-08-14 03:48:44
Research_before_write

You wrote: "Again, we will be testing with the best officially supported RAM configuration for each CPU."

This, released on July 12th, together with using a 2400 kit for the X370 Prime-Pro is simply not true. As you can see on Asus' website they last updated their Memory QVL on July 4th:
https://www.asus.com/us/Mot...

In this list there are several 16GB sticks that run on 2666 with 4 slots populated:

Corsair CMD128GX4M8A2666C15(Ver4.31) (8x16GB)
Corsair CMU32GX4M2A2666C16 (2x16GB)
Klevv IM4AGU88N26-FFFHMZ
Crucial BLS16G4D26BFSC.16FBD
HyperX HX426C16FB2K4/64 KHX2666C16/16G

As well as the Ryzen 7 1800X officially supporting up to 2666MHz according to AMD's own website:
https://www.amd.com/en/prod...

Ryzen is scaling very strongly with memory speed as it increases Infinity Fabric speed / decreases latency. Using 2666 RAM instead of 2400 can give 10% performance improvement on this platform.

Posted on 2017-08-13 14:29:30
pillybilly

The "only" problem is that ryzen can barely handle 32gb, these kind of tasks require a lot of ram, 64gb should be the minimum nowadays.

Posted on 2017-08-31 13:06:28

"...We found going above 32 GByte resulting in diminishing returns." quoted from the Epic games docs (even though the page seems old because it says get a GTX 770, I'll not bother getting more than 32GB RAM)
https://wiki.unrealengine.c...

Posted on 2017-09-10 19:54:37
pillybilly

Thanks for the info anyway UE is not the only software here there are other software requiring more than 32GB.

Posted on 2017-09-10 20:00:32

We get this kind of question quite often, so I reached out to our contacts at AMD to get their official word on supported RAM speed for Ryzen. Our contacts at AMD have confirmed that this post https://community.amd.com/c... is accurate as of this time. They also sent us a matching chart to the one on this post https://community.amd.com/c... as far as the officially-supported DRAM configurations.

A lot of people confuse motherboard QVL lists with the officially supported RAM speed from Intel/AMD. This is definitely not the case. QVL lists from Asus/Gigabyte/EVGA/MSI/etc. are really just a list of RAM that they have confirmed works. They don't usually mean much in terms of whether it will be stable and they don't mean that you won't be voiding AMD's warranty by using that RAM. In the first link I posted, AMD says that "Please note that values greater than DDR4-2667 is overclocking. Your mileage may vary (as noted by our big overclocking warning at the end of this blog)." with the note in question being "WARNING: Overclocking memory will void any applicable AMD product warranty, even if such overclocking is enabled via AMD hardware and/or software..." Now, they only say DDR4-2667 here and not DDR4-188/2133/2400, but given that we have confirmation that the officially-supported RAM speed varies between 1866MHz and 2667MHz depending on the memory ranks and number of sticks, I think that is more of summary or simplification than an endorsement of DDR4-2667 across all RAM configurations. Also note that on AMD's product pages they list "Max Memory Speed: 2667MHz" rather than just "Memory Speed" so that also is not an endorsement of DDR4-2667 in all configurations.

In the end, however, I can tell you that in our testing with DDR4-2666 RAM we've only seen about a 5-7% increase in performance. Definitely not nothing, but even if it was 10% the question for us a system integrator is not first and foremost about raw performance. It is about stability and whether the configuration is something we can guarantee to our customers as stable. If you build your own PC and want to use higher frequency RAM that is entirely your choice, but the vast majority of our customers are purchasing from us because they don't want to deal with problems. So anything that increases the risk of bluescreens or other forms of instability is something we take pretty heavy measures to avoid. Technically even DDR4-2400 is a step up from the RAM speed we should have tested, but since we didn't want to have to distinguish between 2 and 4 stick configurations in our testing we felt it was prudent to present somewhat of a best case scenario.

Posted on 2017-09-01 21:34:03
pax256

Be cool to see this tested on threadripper.

Posted on 2017-08-13 16:51:35

AMD got SO FUCKING REKT MUAHAHAHA

Posted on 2017-08-13 22:32:17
Anaron

I found the 14-year-old.

Posted on 2017-08-14 06:28:41
pillybilly

i don't think he is so wrong. A lot of people read a lot of benchmarks and every site shows every time the same benchmarks then in a real situation we see different results. yeah ryzen costs 100$ less and has only few point less than intel in cinebench... than you use the cpu for other stuff and it performs really bad. i need a new pc and i think i will get the 7820x, too many problems with memory on ryzen, pratically only 32gb.

Posted on 2017-08-29 17:48:28
Guest

I would love to see a similar test done with (dont shoot me) Unity. I do work in both engines, and I was really curious about this one. I was just about to ask someone to compile and convert to mobile a project I was working on, which took about 45 minutes on my setup. It would be great if something similar was done with Unity.

Posted on 2017-08-13 23:21:00

Unity was actually first on our list before Unreal, but we had problems getting our hands on real world projects that would stress the systems. If things take too short of a time to finish, it makes it really hard to show accurate performance differences between components. If you (or anyone else reading this!) think you might be able to help, shoot me an email at labs@pugetsystems.com . Just an FYI that I'm actually out of the office next week though so it may take a bit for me to get back to you.

Posted on 2017-08-14 01:19:19
Guest

Oh definitely will do. I'll be in pretty busy this coming week, but I should be able to throw something together over the weekend.

Posted on 2017-08-14 03:46:53
HAF-Blade

Oh my god. This is the test I have been waiting for. Thank you for that. 7820x will be ordered today.

Posted on 2017-08-16 09:19:38
Niko Nikolov

Im not sure what went wrong with the amd tests,but in the last post of this thread in the UE4 forums:

https://forums.unrealengine...

You coud clearly see a ryzen 1700 at stock clocks doing the zen garden light building at high quality and he gets 103.5 seconds vs the 188 you got with the amd 1700x.

At this point the result is the same as the 6900k.I suspect something went wrong with your setup(ram speed/old driver/having the windows creators update) the reasons coud be many.

There is a reason why people applaud the new amd cpus and is not only the price.Its,true,you have to do some tweaking in order to make it work at its best>starting with the ram speed but the ryzen and even new threadripper are even better option now.

For 800$ you get a 12 core 3.5ghz all most server grade cpu with ecc memory and 60pci lanes for gpu rendering!The closest thing you can get for that price from intel,and it will not howd that good,is a ES last gen xeon cpu.

Posted on 2017-08-18 09:13:29
pillybilly

Ryzen has too many memory and pce lines limitations and the threadripper is too expensive and power hungry for many users.

Posted on 2017-08-29 20:50:17
PMcKenzie2004

Light mass is getting multi bounce support in version 4.18, any chance of further tests with this new version? Excellent article, there aren't a lot of UE4 benchmarks out there.

Posted on 2017-09-30 11:59:37
Chris Rock

Any chance you can update this to include the rest of the i9 chips ( 12, 14, 16, and 18 core variants ), as well as adding the 2 ThreadRipper chips? Thank you for being one of the very few system integrator/builders to do this level of testing.

Posted on 2017-10-05 16:57:34

That's on our list, but we have a couple tradeshows coming up that we are focusing our testing around. Hopefully we can update our Unreal testing with the Core i9 and Threadripper CPUs soon, but it may actually be a couple of months.

Posted on 2017-10-05 17:31:40
Chris Rock

Matt, any chance you will be revisiting Unreal Engine testing like what was done in this article?

Posted on 2019-03-26 22:51:43

At some point, yes. I hate to admit it, but Unreal/Unity has taken a bit of a backseat recently compared to many of the other software packages we target. We really need to change that since these game engines are no longer just about gaming anymore - they are being used for VR experiences, architecture, engineering, and even for making full-blown movies.

The technical side of automating the benchmarks and making them consistent is the one of the issues we have (this testing took wayyyy to long to complete). Plus, it is really hard to get our hands on real-world projects that are actually CPU/GPU intensive. Since we are not game devs ourselves, it isn't like we can make great, real-world projects from scratch. And with things like RTX raytracing being added constantly, we really need to find a partner to work with who is excited about helping us develop our testing.

Posted on 2019-03-26 23:02:55
polysix

8700k overclocked to 4.9 - 5ghz with 3333mhz o/c ram. Should be tested here now! ;)

Posted on 2017-10-23 10:12:19
rozan

Something is seriously wrong with this test. At best it's outdated, at worst it's falsified. My times were much shorter than theirs on the latest engine version, so just to be sure I installed the older 4.16.3 (4.16.2 is unavailable, hotfix difference) and Zen Garden lightmass production took 262 seconds for "Processing mappings" in the Swarm Agent window, or around 5 minutes for the whole thing, from clicking the button to finish (edit: measured it properly, it's 4:40 whole).

So even if we count the longer time, my 8 core Ryzen 1700 is beaten only by Intel's 10 core processors or their 8 core processor, all of them $1000 or higher, if the prices listed here are still correct.

Edit: I missed the 7820X. It's currently $559 on Amazon, while my Ryzen 1700 $285.

Posted on 2017-11-15 21:41:17

We've had some reports of our results being way slower than what people are seeing on their personal systems, but usually it is due to a difference in configuration settings. You could try resetting both Unreal and Windows back to complete defaults and see what you get. Also, make sure your CPU is running at stock speeds. Even if you haven't overclocked, a lot of motherboard manufacturers are starting to do small overclocks at "stock" BIOS settings to try to make their motherboards look faster than their competitors. With all cores loaded, I believe the Ryzen 1700 should be running at either 3.1 or 3.2GHz.

We do all our testing with as stock of settings as possible, and just a few small changes in the settings or a small overclock can make a huge impact on the time it takes to run some of these tests. I don't think it would change the relative results much (so if you tested two or more of the CPUs we used on your system you should see similar performance deltas), but the actual time in seconds can be drastically changed. A result of 5 minutes (about 300 seconds) is about 30% faster than I would expect with a Ryzen 1700, but that difference could easily be covered by a combination of different hardware (DDR-3200MHz RAM for example could make up almost half of that), Windows/Unreal settings, and a small overclock. One thing to note though, you wouldn't have to go to a $1000 Intel CPU to get higher performance. Even with the faster times you are seeing, the Core i7 7820X (~$599) would be significantly faster than your CPU.

Posted on 2017-11-15 22:03:02
rozan

Thanks for the reply.

Indeed I missed the 7820X, sorry about that. I edited my original post. Still, Ryzen 1700 is much cheaper.

Of course my Ryzen is not overclocked, I would've mentioned it otherwise. The cores run at 3200 MHz under stress. Indeed I have 3200 MHz RAM (16GB). It's faster than your RAM, yes, but I read online that Ryzen benefits from faster memory, so that's what I bought.

I assembled this PC 5 days ago, it has the latest BIOS, the latest Windows updates and drivers, but I don't think it's unfair, with Ryzen being a new platform it seems more relevant to test with the latest updates.

I think it would make sense for you to redo the benchmark, even with fewer processors (just one Ryzen and the 7820X, for example), and fewer tests. Skip the FPS and loading tests, just do one or two which are the most reliable, describe the procedure so that people can compare. Test with all latest updates (but Unreal 4.17.3 for example, because 4.18 is still new).

If results are more favorable for Ryzen than the last time, then edit this article and post a link to the new benchmark at the top. I think it is important because a lot of people are coming here from Google and basing their decision on your tests, and it would be unfair to write Ryzen off because of immature BIOS or incompatible RAM.

Edit: measured the start-finish time properly with a stopwatch: 4:40.

Posted on 2017-11-15 22:48:24
waroth

Check out the times on this thread: https://forums.unrealengine...
Zen lighting build time for 1700 is as fast as 103 seconds, threadripper is a minute

Posted on 2017-11-16 05:17:28
Marek Orzechowski

I have ryzen 1800. Zen garden preview 18.6 sec
Zen garden production 3:36 so......... (test in unreal4.16)

Almost 200 sec faster than on this site(pugetsystems.com)..... no comment

Posted on 2018-01-16 12:16:45
Matthew Papadopoulos

Now Im just waiting for the benchmarks with Ryzen 2 processors.

Posted on 2018-04-27 14:38:03
WhiteSkyMage

Quite eye opening. I thought High Core Count = better for everything UE4... and no, not really..

Posted on 2018-10-14 06:45:30
Gooseman

Matt, can you do another one of these reviews using the new Threadrippers ( 2950x / 2990wx ) and the new Intel 9980xe

These articles are incredible useful to me. The majority of tech sites only focus on 3d apps, photoshop , etc... In fact, this is the only article I've seen that attempts to shed some light on game dev.

Posted on 2018-11-14 02:21:36

Game Dev is definitely something we want to get more into, but it turns out to be fairly challenging to benchmark repeatedly and accurately. One of the best things you could do to help is toss an email to info@pugetsystems.com telling us what you would find the most useful for us to test. Code compile time? Viewport FPS? Something else?

Posted on 2018-11-14 17:19:43
Gooseman

Ok, I'll fire off some suggestions to that email. thanks for your time Matt !!

Posted on 2018-11-14 22:39:52

We got your email - thanks! I may reply directly to you with some questions, when we get to working on game dev stuff some more :)

Posted on 2018-11-16 21:53:57
Gooseman

ooh awesome! More than happy to help answer anything

Posted on 2018-11-16 22:20:15
TheParrotPapa

My results for Epic Zen Garden Production Quality Lighting in UE 4.16.3 on Ryzen 1600X - DDR4 3000 16-18-18-39 2T GDM-off BGS-off dual rank ram 1.25v

352 seconds

Quite a bit faster than your measurements. Also not the fastest ram by a long fetch.

Posted on 2018-12-09 04:25:43