How will NS2 run on Ryzen?
James1986
Join Date: 2007-03-15 Member: 60366Members
Join Date: 2007-03-15 Member: 60366Members
Hi all,
Just wondering how NS2 is likely to run on Rzyen as I'm hoping to win a comp soon! Is NS2 GPU based, or would it benefit from a nicely multithreaded CPU?
Kind regards,
T4
Help me win the AMD Ryzen Last Clip Standing competion, no signup required, just scroll down and vote! woobox.com/yog2u7/gallery/qNjgbPUXEY8 #noshame
Just wondering how NS2 is likely to run on Rzyen as I'm hoping to win a comp soon! Is NS2 GPU based, or would it benefit from a nicely multithreaded CPU?
Kind regards,
T4
Help me win the AMD Ryzen Last Clip Standing competion, no signup required, just scroll down and vote! woobox.com/yog2u7/gallery/qNjgbPUXEY8 #noshame
Comments
I'll compare to
AMD 8350 4.8ghz
AMD 1055xt 3.9ghz
intel xeon 5650 4.ghz
intel i7920 @4ghz
NS2 is generally a single core performance based game, seriously 4.5ghz is generally the sweet spot for intel CPU >2009, amd on the other hand falls behind quite a bit. We'll see.
And with NS2 at a maximum of DX11, which is buggy and has worse performance as well. Even that DX version is not designed for proper multithreading, add to this the Lua design. If NS2 is still running on an older version of C++ (pre C++ 11), it also has to do some magic low<->high level code dancing to multithread, getting those dreaded diminishing returns.
Some small corrections: First of all Ns2 is compiled using the C++11 standard. Second of all it's time to bust the myth that Lua is the root of all evil when it comes to NS2's perfomance. Lua code running with LuaJIT can be as fast if not faster than any C code these days. But as with all coding languages you can easily write rather slow code no matter what if you are not really focusing on writing great performing code.
Most of the NS2 game code was written with a focus of being functional as soon as possible and not to run fast. In that regard it often does not play well with the LuaJIT compiler and runs a lot slower than it could and should be. That's the very reason we are still able to cut the run time of certain segments to a 10th and more with recent updates.
However it is true that the NS2 game code wise is not designed to support parallel update routines and would require a major redesign to do so. Most of the modern game engines enforce parallel designs to some degree. And that's why more and more titles get released with a somewhat descent multi threading design.
But the major CPU efficiency issue today for NS2 is the massive amount of data that gets and needs to be networked. Causing tremendous world state update times while everything else is waiting for the world state update to be finished. Looking at how other game engines solve these kind of issues I'm pretty certain there is still a lot that can be optimized in this field. And hopefully we one day can get to it.
I think there should be an option to toggle simple models to switch all entities and players models to alternate versions with way less polygons and simplified animations.
Will it benefit from more then 4 threads? NO! - i checked.
Got a 1800X here, does it run better then my old FX9370@4.8ghz? Yes!
[Make sure you pin the NS2 threads to one compute complex unless using the adware disguised as an operating system called Windows 10.]
ps. Vulcan support when?
But my 6 years old intel 2500k is still faster in programs that need single thread perfomance.
So for NS2 intel is still the best choice.
If you build a pure gaming PC, a i7 or i5 is still the way to go.
The consoles are the leading platforms for gaming, so the question is:
Can a 4 core 4,5 ghz PC compete against a 2,1ghz 8 core Jaguar (PS4 Neo)
Im pretty sure it can.
Current mid class PCs are already faster than the PS4 Neo.
Thats why im sure that you dont need a 8 core CPU for PC-gaming the next 4 years.
For Video editing or transcoding a Ryzen is a nice cheap alternative to Intel.
Never mind this is going off-topic
2500K is a legendary best in any case!
You know, this really feels like good ol' days with NS pushing GoldSrc beyond what it was designed for and newer PC's kinda being able to brute force it. But GoldSrc was bottlenecked in how much it could render, push through the pipes etc... Spark seems limited by tickrate and bandwidth mostly, so even with us being able to bruteforce it with newer tech. Spark seems to be the bottleneck right now (once again
Oh how much dejavu can we handle, it's too much I telzz ya
Hehe dem dev type peeps are kinda pushing Unity beyond what it can do as well
Kind regards,
T4
Help me win the AMD Ryzen Last Clip Standing competion, no signup required, just scroll down and vote! woobox.com/yog2u7/gallery/qNjgbPUXEY8 #noshame
2) AMD seems to be scoring similarly in single threaded benchmarks to Intel's Sandy Bridge. I would expect 3.4Ghz + Ryzen would do about as well as the 2500k/2600k. That means Ryzen would do fairly well with NS2, but newer intel cpu's would do even better.
3) Intel CPU's have made marginal improvements in performance since sandy bridge. They do perform better, but the performance jump is less than prior generations. Intel has been focusing on power efficiency.
You mean the meh entity tickrate which I cant care less about?
Or the main loop which for must servers right now runs on *checks* 130 to 160?
A main loop of 160 sounds plenty fast to me.
If game mods have shown anything it is that stuff like cpu load can be much less on spark if the game simply doesnt need to calculate a gazilion entities in combination with players.
Sounds more of a game rather than a engine problem to me.
"AMD’s big failure with its previous FX processors was single-threaded performance and it just so happens that games are particularly sensitive to just that.
In Cinebench’s single-threaded mode, the Ryzen 7 1700X is between 10 and 20 per cent slower than various gaming-relevant Intel CPUs, with about 10% of that due to a slight comparative deficiency in Ryzen’s architecture and the other 10% thanks to clockspeed."
https://www.rockpapershotgun.com/2017/03/02/hands-on-with-amds-fab-new-ryzen-cpu/
Thats what i said above. Its noticable faster than a FX but still slower than the Intel ones.
Looks like the Ryzen X1800 is on the same level like my i5 2500k @ default speed, wich is a huge improvement over the AMD FX.
https://www.cpubenchmark.net/singleThread.html
But:
I can OC my to 4,6ghz while the Ryzen goes up to only 4ghz with luck 4,1ghz.
So in the end: My 6 years old i5 is still faster than the Ryzen in single thread operations.
Well not much has been gained in terms of raw power, the CPU's just went into the lower power consumption and more efficiency for an average of ~10-15% more speed compared to Sandy and Ivy. Instead of the usual huge speed increase we were used to, between generations. Alongside less actual progress to get higher speeds, due to that magical 5Ghz barrier and 10 nm cores/gates not giving enough resistance to higher voltages for stable clockspeeds.
An overclocked Sandy is still nothing to be trifled with
That article cites the i5-7600k skylake as a gaming relevant cpu. 10-20% slower than various "gaming-relevant Intel CPUs" like the i5-7600k would put it right around Sandy Bridge. That is faster than AMD's FX CPU, and is a huge step forward in single threaded performance for AMD. AMD FX cpu's were comparable to C2D quad cpu's in terms of single threaded performance. Jumping from that to Sandy Bridge level performance is a massive leap.
You are both saying the same thing, which is similar to what I said.
Don't forget that these AMD cpu's are have outstanding power efficiency to performance too. AMD may not have the best "gaming" performance as intels latest CPU's, it isn't bad. They are amazing CPU's for most things.
AMD's cheaper CPU's like the $189 1500x quad core will have practically the same gaming performance as the 1800x. The competing Intel CPU is the overclockable i3-7350K. AMD is very competetive at this price bracket.
Indeed @SquishpokePOOPFACE, indeed \o/
\o\ \o/ /o\ /o/ ಠ_ಠ o"
Tick in ns2 means not that much.
Update rate is however as advertised