Thinking about upgrading to 7820x or maybe Threadripper…
Was curious if anyone has done any UE4 work using Threadripper?
Care to share any performance thoughts?
i’m also interested in the Theadripper performance, please, someone can share their experience with this new c.p.u.?
Although this isn’t strictly for Threadripper, you can assume the performance will scale similarly from Ryzen
I had seen that - those results are less than what I was hoping for Ryzen. 
Anyone else have any experience with Threadripper and UE4?
One person has contradictory results: https://forums.unrealengine.com/comm…38#post1046938
In their result it shows Ryzen 1700 lightmass getting a build time closer to 6900k (8Core) on Epiz Zen Garden. That’s more in line with what you’d expect given what we know about ryzen’s overall performance.
I wonder if there were some settings that limited the pugetsystems benchmarking for ryzen.
A threadripper vs 7820x benchmark would be really nice.
I posted my threadripper scores. Sample is only 1, so take it with a grain of salt.
Get a 7820X instead, cheaper and faster for unreal
It also costs twice as much as the 7820X though.
What is your ram speed set to?It is known that the faster the ram (for lets say a 1700x)the better it does the job.Same is with the threadripper which i have my eyes set on.
The person with the ryzen 1700 having the same score as a 6900k, i suspect he is using fast ram.
https://forums.unrealengine.com/comm…38#post1046938
If you havent tweaked the ram in the bios,try to set it to 2666 or 2800 and repeat the test.I suspect it will improve quite a bit.
Zen wins in some areas, intel wins in others … typical.
We knew out the gate that the Zen would “lose” to Intel for gaming. AMD Zens dont run graphics, the GPU does.
Either way, i think for UE4, intel would be the better choice here. I am not an intel fan, but, per case, intel.
I just sended a message to the person with the amd 1700 asking if he is using higher frequency ram.I still think its a ram thing and the 1950x coud score more.
I added a new post with results from my “fixed” RAM speed settings. The results are a lot better and beat all of the other scores in the thread as far as I can see.
Just logged on and saw your message.
Right now my RAM is set to 3200Mhz, and i think it was that way when i did the lightmass test too. Dont have much time right now (newborn here since a few days and i neeeeeed sleep…), but i probably can post my timings at some point, but its definitely c16 timing.
If you are going to fill your RAM and need even higher transfer rates under load, consider using 2x8 instead of 1x16 sticks if you want to stay at 16gb (like me). Dual channel is simply a bit faster again than single channel (and note that ryzen doesnt support quad channel ram).
Other than that, my 1700 was on stock clocks during that test. I usually have it overclocked to safe 3,7Ghz with the stock cooler (shouldnt go really higher than 3.8 with that one, but awesome for a stock cooler).
Cheers,
I’m also investigating the option to go Threadripper for the cores, I found something that I think everyone considering to go to TR SHOULD know lol
it helped me make the obvious decision to move over
It is a rare occurrence when forum readers are blessed by a thread necro. I’ll have to buy in as well. Some of these build times are killing me.