1.27 poor core optimization?

User avatar
Tarso
Beta tester
Posts: 8243
Joined: 04 Apr 2016 11:03
Location: -5.865761, -35.216600
Contact:

Re: 1.27 poor core optimization?

#21 Post by Tarso » 07 Apr 2017 23:33

What`s you VGA card?

I`m getting straight and solid 60fps out and 45-50 in city with just a FX-6100 and a RX470 card on ultra 1080p 200% scaling
teckerBP-Z3
Posts: 4
Joined: 11 Feb 2017 05:23

Re: 1.27 poor core optimization?

#22 Post by teckerBP-Z3 » 07 Apr 2017 23:45

I'm running dual GTX 980s in SLI. My GPUs don't do hardly anything unless I'm on a menu where the FPS jumps up to the 400 - 800 range. If I'm driving though, they do so little that the fans don't even spool up. Seems a bit backwards to me, but what do I know?

Also, I run 100% scaling with MLAA turned off, and instead run an nVidia Inspector profile that dials in 4x MSAA with 4x supersampling and 16x anisotropic filtering. It looks a bit better and netted me an additional 10 fps prior to 1.27. It seems to still give me an additional 5 - 7 fps after 1.27, but it's not enough to make the game playable in cities.
User avatar
Max
SCS Software
Posts: 6480
Joined: 26 Nov 2012 10:00
Location: SCS, Prague

Re: 1.27 poor core optimization?

#23 Post by Max » 08 Apr 2017 06:55

just note, the game is NOT locked on single core. only one of the threads is.
we have done it to prevent migrating our main thread from core to core as these switches cost unnecessary time. this one thread runs of first core now, others can be on any core. it helps in most cases. issue may come if you have another application or service that wants to be on that same core (aka do the same thing as we did).
currently we are collecting info about result. so far majority has more fps than before.
Nobody can tell you anything about it. Those who know about it aren't allowed to talk. And those who talk know nothing.
teckerBP-Z3
Posts: 4
Joined: 11 Feb 2017 05:23

Re: 1.27 poor core optimization?

#24 Post by teckerBP-Z3 » 09 Apr 2017 05:23

Could locking down the thread be compounding the already known FPS issues in games with Ryzen's Neural Net Prediction and Smart Prefetch? Supposedly the latest BIOS updates solved this issue for the vast majority of games, and I can attest to that, seeing a noticeable FPS improvement in all the other games I play.

However, wouldn't locking the thread down technically be counterproductive on a Ryzen chip? In essence, you're basically forbidding the chip from utilizing the two aforementioned optimization technologies to spread out the work load how it sees fit?

I was also wondering just how important core clock speed would be now with the thread being locked down. Could the 1+ GHz less clock speed most high core count chips see be a viable factor now?

Also, is there anything I can do to help determine the cause of this issue? Would a log file from a Ryzen system be beneficial?
User avatar
TwinShadow
Posts: 892
Joined: 18 Jun 2014 22:28

Re: 1.27 poor core optimization?

#25 Post by TwinShadow » 09 Apr 2017 05:37

I think setting the processor affinity to some other core other than Core 0 for the ets2.exe process (and ATS), may help a little bit with some AMD CPU's, and maybe Intel too. But then again, AMD is known to still have crap when it comes to single-thread performance unlike Intel. (AMD do beat Intel on multi-thread though in some cases)

Now, me personally, I've never had an issue running both games on all high settings on my old Core 2 Duo system with 8GB of RAM (DDR2 mind you), a GTX 550 Ti (later 750 Ti), on a WD Black HDD.

I've also not had issues with my current build, i7 6700k, GTX 1070, and 32GB of RAM. Same HDD btw.


Oh, and I never messed with the affinity setting either on my CPU for the games. FPS are still great even in cities which are more than playable for me.
teckerBP-Z3
Posts: 4
Joined: 11 Feb 2017 05:23

Re: 1.27 poor core optimization?

#26 Post by teckerBP-Z3 » 10 Apr 2017 21:07

I don't know if you've kept up with recent tech news lately Twin, but Ryzen single core performance is on par with Haswell- and Broadwell-E and steadily improving with each BIOS update. We're a long way from the AMD FX processors of old.

I did implement your suggestion to set affinity for ets2.exe away from Core 0 (and Core 1, just to be safe). While FPS gains were only marginal (maybe 1 - 2 fps), forcing the game to use the other cores has completely solved the stuttering issues I was experiencing. The game is now playable again, so, despite your obvious disdain for my decision to go with Ryzen, thank you.
User avatar
EricF
Posts: 300
Joined: 05 Feb 2017 22:16
Location: Great Lakes area, USA

Re: 1.27 poor core optimization?

#27 Post by EricF » 11 Apr 2017 11:33

When I've experimented with core affinity, I've always stayed off of core 0 -- just on the outside chance that some other badly programmed software's resident thread will stubbornly attempt to grab it and cause unnecessary contention. Not sure if it truly makes a difference on modern Windows, which does a pretty good job of distributing loads across cores. But it never hurt.
System Specs for Beta and performance comparisons:
Ryzen 9 5900x - 32GB RAM - NVMe SSD - RTX2080 Super - 3440x1440 Ultrawide - TrackIR - T248 Wheel/Pedals
User avatar
TwinShadow
Posts: 892
Joined: 18 Jun 2014 22:28

Re: 1.27 poor core optimization?

#28 Post by TwinShadow » 12 Apr 2017 04:00

teckerBP-Z3 wrote:I don't know if you've kept up with recent tech news lately Twin, but Ryzen single core performance is on par with Haswell- and Broadwell-E and steadily improving with each BIOS update.
I won't argue that Ryzen is far better than AMD's previous attempts, but they still have a ways to go to properly compete with Intel. Single-threaded performance still sucks in some cases and AMD is only competing on platforms that are literally at least 3 years old.

When comparing to Skylake and Kaby Lake? Yeah, Intel will still blow them out on single-thread. Multi-thread? Now that's where AMD can be good at. Matter of fact, their server variants are far better than Intel's.
User avatar
shalafi
Posts: 17
Joined: 22 Jan 2014 19:11
Location: Bratislava, Slovakia

Re: 1.27 poor core optimization?

#29 Post by shalafi » 21 Apr 2017 18:25

ok, riddle me this: what is the game doing in the main menu that would utilise a 3925MHz core to 100%? Framerate is capped at 75fps, GPU usage is 4%.
Attachments
eurotrucks2_2017_04_21_20_11_38_278crop.png
eurotrucks2_2017_04_21_20_11_38_278crop.png (117.72 KiB) Viewed 648 times
User avatar
Reinhard
Posts: 4712
Joined: 20 Dec 2012 16:46
Location: Berlin, Germany

Re: 1.27 poor core optimization?

#30 Post by Reinhard » 22 Apr 2017 07:26

No idea. All I can say is that my system's CPU usage is around 5% at that point in the game's menu. The "eurotrucks2.exe" game process at 0%. Quite a difference.

Did you check if it is really the "eurotrucks2.exe" process eating all the cake?
Post Reply

Return to “Technical Problems”

Who is online

Users browsing this forum: No registered users and 5 guests