He_162
-
Posts
918 -
Joined
-
Last visited
6 Followers
Recent Profile Visitors
.png.e83a738b9a21f7a0dd1d71fb6143e2b9.thumb.png.dde29df0960c0576b81372f58780d936.png)
Single Status Update
-
About your question on Ryzen and KSP.
More influential is the enhancements to the SSE pipeline which is used for physics on KSP 1.2 and later.
The DDR4 memory should also see a LOT of added performance in the GC cycles, and the cache won't hurt. However, short of about 256MB of cache, you won't see massive improvement from the cache itself.
Finally, I didn't want to mention this in front of the Intel fan-boys because they'd claim I was BSing (despite direct notes from the Linux kernel staff having to do the same thing) but we won't see the full performance from Ryzen cpus for about 6 months because the Operating Systems need to update their thread balancing systems to better distribute for maximum performance on desktop and maximum power saving on laptops. This will likely give a 10-20% performance increase on desktops and an equal power saving on laptops (sacrificing some of that performance increase, of course.)Hopefully most motherboards will properly report the power-save data so that it runs efficiently (nothing is worse than having to run in the always-maximum "performance" mode that actually cripples the turbo feature on the CPU.)
- Show previous comments 2 more
-
I wasn't talking about the power-plan software. There is only so much AMD can adjust with that. I was talking more about how Windows distributes various programs over the various cores.
It currently is unaware of the layout of a Ryzen system, and thus does not know the best way to distribute activity to optimally use both cache and power. -
Actually, windows knows how to do that, the scheduler isn't an issue with performance, but I can set in the bios one core to run faster than the others for a short amount of time with "P states"
So before, with my 1200 ton lifter, I would get around 3 FPS
Now I am getting lets say, 13 - 15 FPS.That's a 5x improvement in FPS because of ram speed, onboard cache amounts, etc.
-
I know that Linux actually saw that 10-20% in the following months. It still is recieving updates. The new SchedUtils power manager on Linux is particularly seeing gains. However, it too needs updates to reduce latency. I really think
When my Ryzen Quad Core (8 thread) motherboard was working I saw a 25%-45% performance gain over my faster clock speed 6 core FX processor, especially on large craft. This was even when under-clocking it drastically. (Damn that bad bios, though. Now it locks in an endless boot cycle.)
