CPU Performance Database

[Steve-S]

Thanks Steve, I added this to the older Intel CPU graph. Yours performs similar to the Q6600 at 2.4GHz.

That's weird about the docking. I tried this once and it worked ok. My method, I think, was to EVA a Kerbal over, detach the docking port from the adaptor first, then activate the stack separator, open the docking port and dock using the command module. It worked fine for me. However, I have had problems before when detaching a docking port in-flight like this. It seems that it can cause the docking port to get stuck. They behave like you describe when this has happened to me.

You can check if this is the case by opening the persistence file and searching for "state = acquire" without the quotes. If you find a docking node with that line you can just change it to "state = ready" and it should fix the problem. I've had this happen many times, but I've never been able to reliably replicate it on a stock install. Hopefully you've stumbled onto a way to make this happen. I'll look into this later and if I can recreate it then I'll submit a bug report. This is one of the docking port bugs that's been giving me problems for a long time, so it would be great if this could help find a solution.

Most likely I did something wrong on trying to get the lander's docking port available. I did get the adapter to release but could not get any option to open the docking port on the lander. I'm going to try again with a turn & 10K and see what happens. Most likely I missed something as I've not done a docking yet. I've only had one successful rendezvous and then simply pushed him back into re-entry.

[DMagic]

Most likely I did something wrong on trying to get the lander's docking port available. I did get the adapter to release but could not get any option to open the docking port on the lander. I'm going to try again with a turn & 10K and see what happens. Most likely I missed something as I've not done a docking yet. I've only had one successful rendezvous and then simply pushed him back into re-entry.

I tried it out both ways, separating the lander docking port first, or blowing the stack separator, then separating the lander docking port, and both worked for me. Once the lander's port is free you shouldn't have to do anything more with it. The command module has the RCS units so obviously you have to use that to dock; you just have to open the shielded docking port on top and it should work.

And just to try it out, I confirmed that you can fly to the Mun, land, take off, rendezvous with the command module, and get back to Kerbin. I didn't fly very efficiently, but even still I had just enough fuel for all of that.

[Steve-S]

I tried it out both ways, separating the lander docking port first, or blowing the stack separator, then separating the lander docking port, and both worked for me. Once the lander's port is free you shouldn't have to do anything more with it. The command module has the RCS units so obviously you have to use that to dock; you just have to open the shielded docking port on top and it should work.

And just to try it out, I confirmed that you can fly to the Mun, land, take off, rendezvous with the command module, and get back to Kerbin. I didn't fly very efficiently, but even still I had just enough fuel for all of that.

I had not separated the adapter. Had to EVA a chap over to handle a few things then got the docking to work fine. I was reasonably sure I would not make Mun so just did an orbital docking around Kerban and then re-entered. Landed the lander, IIRC, 5 or 6 Klicks from the spaceport. Not so close with the capsule. :-) What flight profile did you use to get to the Mun? I turned 45* on 090 at about 10K and circularized at about 100k. Had only the poodle's on the CMM & MLM at that point and was looking at a long burn. Actually I had a small amount of the previous stage left, I remember now, but it was not enough to get to Mun on it's own and I was concerned I would not be able to get back.

[DMagic]

What flight profile did you use to get to the Mun? I turned 45* on 090 at about 10K and circularized at about 100k. Had only the poodle's on the CMM & MLM at that point and was looking at a long burn. Actually I had a small amount of the previous stage left, I remember now, but it was not enough to get to Mun on it's own and I was concerned I would not be able to get back.

I wasn't planning on landing on the Mun and returning, I was just going to test out docking. So I had a pretty sloppy launch and ended up in an elliptical orbit at around 300*150km, with almost full tanks in the lander and orbiter.

I didn't think that I would have enough, but I tried the landing anyway. And even though I somehow got into a retrograde orbit around the Mun I just managed to get down and back again. With the lander full, or almost full, of fuel you should have no problem landing and taking off again. At that point it actually requires very little fuel to break Munar orbit and get back to Kerbin, you can even use RCS if you have to.

[minerman30]

Because this entire thread is really TL;DR, it there any definitive answer as to what allows higher part count in a computer? I'm looking to build one and I want to know what I need for good high-part framerates.

[DMagic]

Because this entire thread is really TL;DR, it there any definitive answer as to what allows higher part count in a computer?

A high end CPU. Other components can have an effect, but the CPU is by far the most important component, and the one that is most directly affected by part count.

The GPU and graphics settings can also affect performance, but that seems to be mostly related to terrain settings, light counts and maybe a few other settings. The actual parts on your crafts aren't too complicated and shouldn't really stress the GPU by themselves. If you want to test this just launch a big ship and point the camera straight up while monitoring GPU usage (MSI Afterburner, or something like it works for this) and compare with GPU usage when looking at the horizon, especially the ocean.

That said, it seems from the results that I have, that there isn't much of a difference once you get above 4GHz or so on a decent Intel i5. Just look at the first few graphs in the second post.

If you want to OC then one of the K part i5 CPUs would be your best bet (a 3570K, or 4670K). I haven't seen many i7 CPUs, but you'd be better off staying away from them. The Hyperthreading they offer doesn't really affect KSP and the increased cost probably isn't worth it (unless you have some other use for them). You can compare CaptainKorhonen's results with those in the second post and see that there isn't much of an improvement.

TL;DR:

An Intel i5 CPU - 2500K, 3570K, or 4670K and OC it.

A high clockspeed, non-K part if you aren't going to OC.

This is, of course, assuming you're talking about a desktop.

Edited August 5, 2013 by DMagic

[ThePsuedoMonkey]

Here's my attempt to summarize the data here into one graph:

https://docs.google.com/file/d/0B6828vVl_-ZoOWN3NXVNbkJkUFU/edit?usp=sharing

I know this is not as precise but it will probably help those whose CPU info isn't already here. Unfortunately there really isn't a good way to compare overclocked CPUs this way. If you want to help fill out the data, I've provided the download link above as well.

Edited August 6, 2013 by ThePsuedoMonkey

[minerman30]

A high end CPU. Other components can have an effect, but the CPU is by far the most important component, and the one that is most directly affected by part count.

The GPU and graphics settings can also affect performance, but that seems to be mostly related to terrain settings, light counts and maybe a few other settings. The actual parts on your crafts aren't too complicated and shouldn't really stress the GPU by themselves. If you want to test this just launch a big ship and point the camera straight up while monitoring GPU usage (MSI Afterburner, or something like it works for this) and compare with GPU usage when looking at the horizon, especially the ocean.

That said, it seems from the results that I have, that there isn't much of a difference once you get above 4GHz or so on a decent Intel i5. Just look at the first few graphs in the second post.

If you want to OC then one of the K part i5 CPUs would be your best bet (a 3570K, or 4670K). I haven't seen many i7 CPUs, but you'd be better off staying away from them. The Hyperthreading they offer doesn't really affect KSP and the increased cost probably isn't worth it (unless you have some other use for them). You can compare CaptainKorhonen's results with those in the second post and see that there isn't much of an improvement.

TL;DR:

An Intel i5 CPU - 2500K, 3570K, or 4670K and OC it.

A high clockspeed, non-K part if you aren't going to OC.

This is, of course, assuming you're talking about a desktop.

So, cycle speed, right? Thanks. Or do you mean the actual K number?

[DMagic]

Here's my attempt to summarize the data here into one graph:

Wow, very cool, thanks Psuedo. The last three on your list are mine by the way (w/AMD 7850 GPU), they are all stock, I'm just reporting the turboboost speed.

Passmark has numbers for single threaded benchmarks too. I'm wondering if those match up even better. I'll try to take a look at them later.

Edit: The single threaded numbers don't seem to match up very well, the 4650U and 3470 are almost the same, which doesn't seem to reflect KSP performance at all.

And I'll add this to the 2nd post, if you don't mind.

So, cycle speed, right? Thanks. Or do you mean the actual K number?

Well basically, yes, but you can't just compare clockspeed and say one CPU is better than the other. There is a huge range of performance for different CPU's at the same speed.

And I'm not sure what you mean by K number. Those are just the CPU model numbers, and K means they are unlocked and can be OC'd. But generally the higher the model number the better.

Edited August 6, 2013 by DMagic

[_Aramchek_]

You need an intel i5/i7 running at as fast a speed as possible.

[ThePsuedoMonkey]

Wow, very cool, thanks Psuedo. The last three on your list are mine by the way (w/AMD 7850 GPU), they are all stock, I'm just reporting the turboboost speed.

And I'll add this to the 2nd post, if you don't mind.

-snip-

Sweet. I've changed those CPUs to the stock data series, which means the only real holes left are the data for the stock i5-3570K and i7-2600K. If the overclocked CPUs also ran the Passmark CPU suite, we could just compare them to that and wouldn't need to treat them separately on the graph. As always, let me know if you find any errors/anomalies in my data.

Edited August 6, 2013 by ThePsuedoMonkey

[Jesrad]

There really seems to be a big difference between Mobile and Desktop CPUs. That's very interesting. Thank you for all this hard work, this is an awesome thread !

[earth]

I've been reading the thread, and i must say you have gathered very solid data! What strikes me the most is how KSP takes absolutely no advantage from the number of cores, and that clockspeed is the only factor. this saddens me quite a bit..

[DMagic]

There really seems to be a big difference between Mobile and Desktop CPUs. That's very interesting. Thank you for all this hard work, this is an awesome thread !

I think the model numbers could be part of why this might be surprising. An i7 4650U (basically a top-of-the-line ultrabook CPU) sounds like it should be better than an i5 3470. But the two are really no where near each other. Of course, the 3470 is a 77W part while the 4650U is a 15W part, so this is really no surprise. But TDP, and even clockspeed to some extent, are not immediately obvious when looking at just the model number.

A good resource for this kind of basic CPU information is Intel's database (this is usually the first result in a Google search for any specific Intel CPU): http://ark.intel.com/

I've been reading the thread, and i must say you have gathered very solid data! What strikes me the most is how KSP takes absolutely no advantage from the number of cores, and that clockspeed is the only factor. this saddens me quite a bit..

Thanks. And yes, having four or six cores doesn't really have any effect on KSP. Although, I imagine that a single core CPU would do a lot worse than an equivalent dual core CPU, but there aren't too many of those around to test.

If the overclocked CPUs also ran the Passmark CPU suite, we could just compare them to that and wouldn't need to treat them separately on the graph.

I put the plot up on the first page with a link back to your post. Some of these CPU's are common enough that it should be easy to find PassMark results at a range of different clockspeeds. It would be best to have the numbers from each user, but scores from other people might at least give us an idea of the range they should be in. I'll try to look into it later today.

[loknar]

Awesome thread.

I'll post my own benchmarks when I'm free - but basically I'm in a very sad situation with a new GPU, 3-way RAID0 and fast RAM; but old CPU. I have no plans to upgrade my CPU soon hence I need to optimize my settings and game strategy.

I saw that overclocking has great impact, especially when your FPS drops very low.

I plan to benchmark your rocket with stock, overclock and also (as discussed here: http://forum.kerbalspaceprogram.com/showthread.php/43153-Low-FPS-at-Space-Stations/page3 ) test the effect of adding +50 antennas vs +50 thermometers +50 solar panels vs +50 struts vs +50 winglets and etc to check how each part affect performance. If some parts are just decorative or absolutely necessary I will adapt my designs.

I don't have time to perform so many test (partly because my pc is slow) so I appreciate any help.

[Slugy]

I've done some more results for my Q6600 with cores disabled via msconfig boot options - running on a single core machine makes a much bigger difference than you might expect!

Labels on Graph included this time ;-)

Edit:

Passmark results for 3.3GHz all cores:

CPU Mark

This Computer 4585

CPU - Integer Math

This Computer 7245

CPU - Floating Point Math

This Computer 4227

CPU - Prime Numbers

This Computer 14.6

CPU - Extended Instructions (SSE)

This Computer 11.8

CPU - Compression

This Computer 7401

CPU - Encryption

This Computer 921

CPU - Physics

This Computer 285.5

CPU - Sorting

This Computer 4217

CPU - Single Threaded

This Computer 1315

Looking at the passmark results above I wonder if the single threaded result might be the best indicator of KSP performance (assuming the CPU has at least 2 cores).

Edited August 7, 2013 by Slugy

[DMagic]

Looking at the passmark results above I wonder if the single threaded result might be the best indicator of KSP performance (assuming the CPU has at least 2 cores).

I thought so too, but I'm not so sure after looking at the numbers from this database: http://www.cpubenchmark.net/singleThread.html

The single threaded results are all over the place. The mobile CPU's especially, seem to get scores much higher than you would expect, and close to much more powerful desktop CPUs. And the i7 920 and 980X (basically consumer versions of Xeons) score very low.

Your results definitely confirm my suspicions about how a single core CPU would perform. It's a little surprising that it's that much of a difference though.

I added your results and superm18's to the front page, and I reorganized the charts a bit.

[Slugy]

Disclaimer - I'm not a big fan of Passmark, testing the program you are interested in is always best

But ....

it is a very large database of results, although being all user submitted and (roughly) averaged may cause problems - one of which is that the top few chips in the chart are all based on a few results, and should be taken with a large pinch of salt imo.

The few mobile CPU results I looked at don't seem too far out, allowing for the effect of turbo boost (which might be less relevant if running KSP for a hour or three...) e.g. i7-4900MQ and 4670k are nearly the same, i7-3840QM is at 1941, 3570 is at 1994, and 3570k at 2011

The Xeon W3520 and i920 score basically the same at 1182 and 1160.

The newer Xeon E3-1275 and 4770k have a slight difference at 2125 and 2252, but that 6% could easily be the difference between machines that tend to have fast memory in small amounts and servers with a lot more memory running somewhat slower. As well as only 4 results for the Xeon.

More results with passmark needed to see!

[_Aramchek_]

it is a very large database of results, although being all user submitted and (roughly) averaged may cause problems - one of which is that the top few chips in the chart are all based on a few results, and should be taken with a large pinch of salt imo.

Why is that? The ivy/sandy bridge results are all in line with each other and personally I know I could run the "benchmark" ship run a thousand times and get the same results.

[DMagic]

Why is that? The ivy/sandy bridge results are all in line with each other and personally I know I could run the "benchmark" ship run a thousand times and get the same results.

I think he means the Passmark database, which can be a little questionable sometimes (aside from the entire idea of standardized benchmarks being a little iffy). The same chips run by different people can give a surprisingly wide range of results.

[Slugy]

I think he means the Passmark database, which can be a little questionable sometimes (aside from the entire idea of standardized benchmarks being a little iffy). The same chips run by different people can give a surprisingly wide range of results.

Exactly, Just look at the 2 results for the "top performing" i7-4930MX : 10296 and 8897, pretty big range there!

To illustrate for [email protected]

Average CPU mark Passmark 2973, mine 3432 - 15% is a reasonable difference...

Single thread Passmark 920, mine 961 - 4% is not that much different.

And that's with (fast) DDR2, some will be using DDR3, some will be using pretty slow DDR2.

[loknar]

I've done some more results for my Q6600 with cores disabled via msconfig boot options - running on a single core machine makes a much bigger difference than you might expect!

Edit:

Passmark results for 3.3GHz all cores:

Looking at the passmark results above I wonder if the single threaded result might be the best indicator of KSP performance (assuming the CPU has at least 2 cores).

Wait this graph is passmark results not KSP right?

The conclusion I get is: Single-thread passmark can help choose which CPU is best for KSP before buying it. I kinda suspect an i5 i7 is much better at KSP than a more expensive Qxxxx but please confirm that so I can plan an upgrade.

Edited August 8, 2013 by loknar

[DMagic]

The graph above is framerate results using my rocket in KSP. The numbers after that are Slugy's Passmark scores, but they don't have anything to do with the graph above it.

I wouldn't rely only on the Passmark scores when looking for a CPU for KSP, but they do seem to be a decent indicator of what kind of performance you'll get. ThePsuedoMonkey's plot on the first page shows this.

Pretty much any desktop i5 or i7 CPU will be pretty good. The Q CPU's (Core 2 Quad) are much older and I wouldn't buy one of those unless you have the right motherboard for it and don't want to upgrade.

The 3000 and 4000 series of i5 and i7 (and i3) CPU's should be easy to find at retail prices; getting the right motherboard is another thing to keep in mind. And pretty much any of them should perform ok. The lowest end i3's might not be quite as good though.

Mobile CPU's are another story. They use similar model numbers (i5 2467M vs. i5 2500) but don't have anywhere near the performance of a desktop CPU, keep that in mind.

[loknar]

Understood. Thx Slugy. And can I borrow your Q6600?

I'm looking at CPU+mobo+ram upgrade; it ain't happenin' this year. Got a nice new ATI GPU though. When I increase/decrease purely graphical stuff on screen it has 0 visible performance impact! But I digress...

I learned in this thread that the time IRL and time in game are totally different when you have low FPS and that still shocks me. Last night I spent 1 hour docking the *damn* corner in my station - in game time it was probably 4 minutes Thx for the info dmagic.

Edited August 8, 2013 by loknar

[BAPBAP]

So most of KSP perfomance issues related to CPU power, not GPU?

My current rig is pretty outdated - dual core pentium 2.5 GHz, but i have spare 4-core 8-thread 2.66 GHz HP Proliant server. Should i expect some perfomance boost due to move KSP to this platform?