Question to the Developers: In Unity 5 64 Bit, Will Multiple Cores Be Used For Single Ships?

[Majorjim!]

Hmm. According to http://physxinfo.com/news/11327/multithreaded-performance-scaling-in-physx-sdk/ even with a single core, rigidbody simulation in PhysX 3.3 is significantly faster. So with any luck we'll get a big performance boost on large ships, multi-core or not.

We will indeed get an improvement over 1.0+ versions. Having said that, performance dropped massively from version 0.90 so it had better be a big boost to match the excellent performance of previous versions.

We will have to wait and see.

[EladDv]

plus dont forget the U5 GUI tools which will take a massive overhead from the CPU

[Majorjim!]

plus dont forget the U5 GUI tools which will take a massive overhead from the CPU

Yes and hopefully put an end to the horrid GUI GC stutter that many people are experiencing. I look forward to this patch with great hope.

[Parkaboy]

Fot the record, a sexcore would definitely be more popular.

[OddFunction]

I gotta say, I am loving how convinced everyone is of their speculation when

a) Most of us have never programmed so much as a coin counting program

Of those programmers, even fewer work directly with Unity

and

c) Squad themselves refuse to release benchmarks and/or specifics

Yet everyone already knows exactly how its going to go? Anybody wanna do a group-buy in of lottery tickets?

[Alshain]

I gotta say, I am loving how convinced everyone is of their speculation when

a) Most of us have never programmed so much as a coin counting program

Of those programmers, even fewer work directly with Unity

and

c) Squad themselves refuse to release benchmarks and/or specifics

Yet everyone already knows exactly how its going to go? Anybody wanna do a group-buy in of lottery tickets?

I had to go back and read the thread. Are you reading the same one? I don't think I saw a single person that sounded like they believe they knew exactly how it was going to go. You even said it yourself, it's speculation.

[OddFunction]

I had to go back and read the thread. Are you reading the same one? I don't think I saw a single person that sounded like they believe they knew exactly how it was going to go. You even said it yourself, it's speculation.

I lack the articulation to adequately explain myself. It's not specifically this thread, but more all the related threads since the announcement. I've seen the following confirmed as fact by various users at some point or another:

-Unity 5 will significantly raise FPS

-Unity 5 will have negligible effects on FPS

-Unity 5 will have physix distributed across multiple cores

-Unity 5 will not have physix distributed across multiple cores

-Unity 5 will have a single craft distributed across multiple cores

-Unity 5 will not have a single craft distributed across multiple cores

-Unity 5 will be 64 bit

-Unity 5 will not be 64bit

So you're right, my post seems a bit out of left field in this thread specifically, but perhaps not when considering the wider conversation.

My bad

[keeper]

In The most recent Squadcast Max said that one ship would be spread across multiple cores of your processor.

Is that true? I hope this means we can rack up those part counts!

BEcause reading that has made my day, i read it outloud and it is like liquid gold oozing into my own ears.

Truly badass.

[OddFunction]

Is that true? I hope this means we can rack up those part counts!

BEcause reading that has made my day, i read it outloud and it is like liquid gold oozing into my own ears.

Truly badass.

Right? Like literally the best thing that can happen to this game.

[selfish_meme]

I gotta say, I am loving how convinced everyone is of their speculation when

a) Most of us have never programmed so much as a coin counting program

Of those programmers, even fewer work directly with Unity

and

c) Squad themselves refuse to release benchmarks and/or specifics

Yet everyone already knows exactly how its going to go? Anybody wanna do a group-buy in of lottery tickets?

My posts have what Squad said, not my speculations.

[keeper]

Right? Like literally the best thing that can happen to this game.

Well, I'd say so! I havn't bumped into any/many "bugs"!

- Individual GameObjects are distributed among cores, and each is run on a single thread. In this case each ship is still a single thread, but multiple nearby ships could be run faster.

- Individual rigidbodies are distributed among cores, and an additional thread is in charge of synchronizing information from each of them. In this case, parts within a ship can run on different cores, so a single large ship could be run faster.

- Different types of jobs are handled by different cores, e.g. all the lights are on one thread, all the collisions are on another thread, all the shadows are on a third, all the scripted forces on a fourth, etc. or something of the like. In this case, again, a single ship could be run much faster.

All of this sounds fantastic. I hope all three are simultaneously implemented immediately.

[Jouni]

Lol, well it really doesn't matter. But computer tech doesn't use Hexad, partly because it conflicts with Hexadecimal which is used heavily in computer tech. It always uses the tuples (sometimes using the alternate "dual" instead of "double"). Dual core, Quad core, Double precision floating point, AMD was developing a Duodeca core Opteron CPU at one time, I never heard if they finished it.

It should be 'sexacore', if anybody cared about consistency. After all, we have quad-cores instead of tetracores, so we should also have sexacores instead of hexacores. Mixing Greek and Latin is a cardinal sin.

[SAI Peregrinus]

It should be 'sexacore', if anybody cared about consistency. After all, we have quad-cores instead of tetracores, so we should also have sexacores instead of hexacores. Mixing Greek and Latin is a cardinal sin.

So what do you think of television? It's half Greek and half Latin!

WRT this thread and speculation, I'd expect single ships to still be single-threaded for the collision calculations. Thermals can be split off, of course. But I'm a hardware guy, so unless you want physics simulations on an FPGA or ASIC I'm not really the right person to ask. That said, I have no idea how I'd make hardware to accelerate the kinds of collider physics KSP needs, it's not really parallelizable. That means going for really high clock speeds, and that's really annoying (at best) to design.

[NecroBones]

It should be 'sexacore', if anybody cared about consistency. After all, we have quad-cores instead of tetracores, so we should also have sexacores instead of hexacores. Mixing Greek and Latin is a cardinal sin.

But knowing the industry, they'll go with the cool/marketable buzzword, rather than what is more consistent. I suspect no marketing department will go for "sexcore" or "sexacore", but "hexacore" sounds "cool" and can be thrown around as the next buzzword.

[problemecium]

Hmm. If that SQUADcast transcript is accurate / Max's claim is accurate (not that I mistrust anybody, it's just a grapevine effect), then I sure hope that it doesn't break saves, because Kidonia could REALLY use an FPS boost lately.

[Alshain]

I did not expect that little tease to hijack this thread. They are really just calling them "6-Core", both AMD and Intel. Back to topic now.

[_Aramchek_]

Well there goes all the hype for Unity 5. If the FPS issues with larger crafts doesn't go away eventually I don't think I will have any renewed interest in the game. Sorry. Just being honest.

The fix to fps issues, is, get a better cpu, this is how pc gaming works. My core system is two years old and I laugh at the idea of a 700 part ship being monstrous..so it's pretty easy to fix those fps issues..even on a budget. Get an intel cpu, boom..improved fps.

And on that note, anyone who has a fairly current intel cpu isn't going to be put off by "bad fps" because they won't be experiencing "bad fps".

[peachoftree]

The fix to fps issues, is, get a better cpu, this is how pc gaming works. My core system is two years old and I laugh at the idea of a 700 part ship being monstrous..so it's pretty easy to fix those fps issues..even on a budget. Get an intel cpu, boom..improved fps.

And on that note, anyone who has a fairly current intel cpu isn't going to be put off by "bad fps" because they won't be experiencing "bad fps".

Define "fairly current". Is it core 2000-series and up or even newer like a core-4000?

[problemecium]

I have a 3rd Gen. Core i7 and KSP runs pretty well as long as I don't go full Whackjob. So you don't have to go terribly modern I guess.

[severedsolo]

Define "fairly current". Is it core 2000-series and up or even newer like a core-4000?

While maybe not "current" in computer standards, my i5-2500k can handle anything I throw at it, KSP included. Sure it lags a bit at high end, but I suspect that's more to do with my millions of mods.

[qromodynmc]

I think a cpu with high clock speed would play with ksp nicely. My i7-3610m lags when i pass over 300-400 parts, probably related to being mobile cpu and low mhz (2.2, 3ghz with turbo but i have to close turbo sometimes,its getting so hot.)

[_Aramchek_]

Define "fairly current". Is it core 2000-series and up or even newer like a core-4000?

Anything in any of those cpu families, should be within 10% -/+ of each other. It's a 3570k..of which similar cpu's can be had very cheaply second hand.

[TMS]

Anything in any of those cpu families, should be within 10% -/+ of each other. It's a 3570k..of which similar cpu's can be had very cheaply second hand.

Pretty much. Things haven't really progressed that much over the past few years.

I have a three year old i7 3770, which is a quad core with hyper-threading. As such, it shows eight cores under properties. I'm wondering if I might actually feel some CPU benefit with those "pseudo cores"?

[Streetwind]

(Well, since this thread is utterly off topic already anyway...)

There are reasons why the latest couple CPU generations from Intel haven't really scaled up in the high end.

Every CPU architecture has a "sweet spot", where it has the best power:performance ratio. If you push it harder than that, power cost (and with it, heat) increases faster than performance; if you drive it more gently, performance drops off faster than power cost. Up until Nehalem/Sandy Bridge, all of Intel's CPU architectures were fully performance oriented, with the sweet spots up near the 100W TDP mark.

Then mobile computing happened in earnest, and Intel realized that they could no longer compete with half-assed solutions like the original Atom that was so bad that it made people want to buy the higher power CPUs. In a mobile device, that is not an option. Intel needed their flagship architectures to scale down. But with their sweet spot set that high... they wouldn't scale down properly. At the same time, they couldn't just go and dismantle their core business by discontinuing the flagship architecture line in favor of an all-new low power concept.

So since then, every new Intel flagship architecture had its sweet spot nudged down a bit, gradually, generation after generation. As a result, the top of the line models only post a few percent more performance than the previous generation, and no architecture has been as overclock-friendly as Sandy Bridge. Running increasingly far away from their sweet spot, these CPUs scale worse and worse as you try to push them higher. But at the low end, in the 15 to 35 watts TDP range, there's been a small miracle happening for Intel, with each generation a massive improvement over the previous because they run closer and closer to the sweet spot. They are even able to scale as low as 5W TDP nowadays, something that Sandy Bridge is physically incapable of without running at a snail's pace.

As for the actual performance difference between Sandy Bridge and today's architectures at the high end... people tend to underestimate it a bit. Even if each architecture step only brings a couple percent, that can compound over the years. Upon the release of the first Skylake CPUs earlier this month, my go-to tech website (in German, apologies) tested every top-end model from the i7-2600K to the i7-6700K. They found a 40% performance advantage for the Skylake chip over Sandy Bridge in applications and workstation tasks, and a 15% advantage in gaming under typical graphics-limited scenarios. This compounded into a 31% overall advantage across the entire benchmark suite. And it did this while consuming roughly 10% less power under non-synthetic full load - despite running far away from its sweet spot.

(The Cinebench single-core test credited a 52% advantage to the Skylake chip with 15% less power draw, by the way, which may be of special interest to KSP players.)

Other site's test results on different benchmark suites may vary, but the point is: while the differences between one generation and the next are indeed minor, there is still a tangible advantage to be had when upgrading from a CPU several generations behind.

Edited August 23, 2015 by Streetwind

[_stilgar_]

I'm wondering if I might actually feel some CPU benefit with those "pseudo cores"?

One core with HT has the performance of about 1.5 cores (it varies, sometimes you can get even 1.9, sometimes it's just 1.1). So, quad core with HT is like six core without it.

Of course, you can only feel it if the program you are trying to run is multithreaded.