CPU Performance Database

[DMagic]

Also something to note is that with some mods in KSP I noticed an unbearable FPS drop. My tanker station with a bunch of crap docked to it went from ~20fps down to ~10fps. I think Kethane was the culprit but I can't prove it because I removed another mod or two at the same time when it got bad and my framerates went back to normal.

Yeah, some mods can really kill performance. I've had 500 part crafts that just crawl along at 5 FPS. But I've never noticed mods having a significant effect when they aren't being used. I'm sure there are some partless plugin mods that noticeably affect performance, and ideally everyone would run this test on a clean install. But I know that's not going to happen, so I'm not going to worry too much about variations in performance due to mods. The more results I get, the more things like that will be smoothed out.

The RAM bandwidth issue has come up before (single, dual, or triple channel), but I've yet to see anyone actually test it out, so I'd be interested to see if you find anything.

Thanks Demodus, I updated the front page. Those high end mobile CPUs are really powerful.

[Blaster]

Here's my chart of Benchies.

System Specs:

Biostar TA970

AMD Phenom IIX2 "B55" (a 555 with 2 more cores unlocked) 3.2 GHz Quad Core*

G.Skill Ripjaws DDR3 1600 8 GB Single Stick, Single Channel, running at DDR3 1333 speeds

MSI Radeon 7770 Graphics Card

WD Caviar Black 640GB Hard Drive

Microsoft Windows 8.1 Pro x64 OS

Download Results

See post further down

Edited January 25, 2014 by Blaster

[Blaster]

I then installed the AMD Catalyst 13.12 driver, did an Overdrive Tune, rebooted, and ran it again with the CPU Overclocked to 3.8 GHz.

Download OC Results

See post further down

What really bottlenecks these systems, and especially for the AMD systems, is the lack of Multithread support. That FX-8350 should provide outstanding performance if it had all 8 cores utilized, and not just 1.

Using the AMD Graphics Activity meter, I find the 7770 doesn't go higher than 60 percent utilization at any time, and a lot of the graphics slowdown in hangar and launchpad is mainly CPU related.

Edited January 25, 2014 by Blaster

[DMagic]

That's odd, your second run is a little slower than the first. It begins about the same, but then starts to lag behind about half way through. A 20% OC like that should give you a pretty good boost. Something weird must have happened there.

[_Aramchek_]

That's odd, your second run is a little slower than the first. It begins about the same, but then starts to lag behind about half way through. A 20% OC like that should give you a pretty good boost. Something weird must have happened there.

Blaster, what kind of cpu cooling do you have, and what are your temperatures like?

That's the first thing I'd ever look at in that situation.

[gilflo]

Hi

I did not read the 21 pages, but what is the conclusion of the study?

What are your basic recommandations, how to set the different parameters, depending on CPU, video cards, etc....?

Thank you

[Blaster]

Blaster, what kind of cpu cooling do you have, and what are your temperatures like?

That's the first thing I'd ever look at in that situation.

And that's what I'm looking at checking.

[Blaster]

Running Speedfan, I'm not hitting any temperature limits. Hottest it ever got was 126F on overclock (just over 40C for metrics).

However, I changed the graphics settings to this:

Which rendered a significant change in performance. Here is the new results, Overclocked to 3.8, then stock at 3.2

Download Overclocked Bench

Download Stock Bench

In case you can't scroll up for some reason, here's the System Specs:

Biostar TA970

AMD Phenom IIX2 "B55" (a 555 with 2 more cores unlocked) 3.2 GHz Quad Core*

G.Skill Ripjaws DDR3 1600 8 GB Single Stick, Single Channel, running at DDR3 1333 speeds

MSI Radeon 7770 Graphics Card

WD Caviar Black 640GB Hard Drive

Microsoft Windows 8.1 Pro x64 OS

Edited January 25, 2014 by Blaster

[DMagic]

Thanks for running all of those tests Blaster. That's a surprisingly big boost in performance from turning all of the graphics settings down. The HD 7770 is fairly powerful, but I guess it's not always enough for KSP.

Hi

I did not read the 21 pages, but what is the conclusion of the study?

What are your basic recommandations, how to set the different parameters, depending on CPU, video cards, etc....?

Thank you

I don't have many recommendations about settings, there isn't much to change. There are a few things in particular that are know to affect performance though:

The terrain lag issue is a pretty common one that affects all planets with an ocean, there are some ways to work around that, which reduces most of the impact. Lowering the texture and rendering settings can also help sometimes.

Some people have issues with anti-aliasing on AMD cards, that's something to keep in mind.

Mods and memory usage is another matter, one not really related to performance. Too many mods and you'll run out of memory and the game will crash. There are some methods in the links below about addressing that.

There are also a few mods that affect performance more directly. Mechjeb's delta-v window in particular can really drag down your performance if it's open. There are probably others, too.

Ubizor's part welder mod is another option if you're specifically looking for part-count related performance fixes.

There are a few threads more specifically related to performance and settings:

http://forum.kerbalspaceprogram.com/threads/60230-I-would-like-to-ask-you-to-configure-your-computer-for-advanced-players

http://forum.kerbalspaceprogram.com/threads/61798-Optimization

The best way to increase KSP's performance though, is a better CPU. A recent generation i3 or i5 from Intel is probably your best bet depending on how much you want to spend. Of course you have to be able to upgrade your computer for that to be a possibility though.

The best advice I can give is less hardware related and more gameplay oriented; use fewer parts and practice more efficient design (which is, admittedly, not always helpful advice, but fewer parts will pretty much always lead to better performance).

Edited January 25, 2014 by DMagic

[gilflo]

Thank's Magic for those advises.

[ThePsuedoMonkey]

Got another system tested:

Samsung R580, Intel I5 M430 @2.27GHz, NVidia GeForce 310M (1GB shared), 1366x768 windowed.

https://drive.google.com/file/d/0B6828vVl_-ZoTWtvMU16MVJrTWs/edit?usp=sharing

[DMagic]

Thanks, ThePsuedoMonkey. Were you using the old, pre-0.23 version of the rocket? Your performance levels look ok, but the total run time seems too high for that kind of framerate.

I can still use the results for the CPU comparison chart (whenever I actually put it up again) because it doesn't affect performance during the first stage, it just makes it difficult to compare the framerate graph to others.

[teal'c]

Here are the results of my benchmark using the new rocket in .23. At the moment I'm using a Core 2 Quad Q9550. Going to upgrade to i7 4770k soon, will post a benchmark then.

<a title="Download your file from SendSpace" href="http://www.sendspace.com/file/9m91jw" target="_blank">Intel Core 2 Quad Q9550 FRAPS.csv</a>

<a title="Download your file from SendSpace" href="http://www.sendspace.com/file/nrniv4" target="_blank">Intel Core 2 Quad Q9550 frametimes.csv</a>

Edited February 11, 2014 by teal'c

[Tiakatt]

This is a great project, thanks for doing this. Here's my stuff

FPS:

https://drive.google.com/file/d/0Byklrb1gybyDTkdnOFhaQjhLamM/edit?usp=sharing

Frametimes:

https://drive.google.com/file/d/0Byklrb1gybyDYkdrajE1WGVJYjQ/edit?usp=sharing

CPU-Z

[DMagic]

Thanks for all of the results, I've added them to the first page.

I also added back the CPU comparison chart. I left off the FX8350 that might have been running the FRAPS video recorder and the i7 3770 that was running on the integrated GPU, but otherwise everything lines up pretty well. Everything in the chart is the same as from before 0.23, and the description and sources for most of the benchmark values is below it on the first page.

[ThePsuedoMonkey]

Thanks, ThePsuedoMonkey. Were you using the old, pre-0.23 version of the rocket? Your performance levels look ok, but the total run time seems too high for that kind of framerate.

I can still use the results for the CPU comparison chart (whenever I actually put it up again) because it doesn't affect performance during the first stage, it just makes it difficult to compare the framerate graph to others.

Yep, I used the old version of the rocket (the same as the first runs I did). I flew it until all fuel was burned up, so that is probably why the flight time is high. I meant to include the staging times in the text file, but I'll put 'em here instead.

S1-0:39.42, S2-3:20.30, S3-4:18.23, S4-4:56.82, S5-5:41.97, S6-7:40.33, S7-8:55.37, S8-9:07.98, S9-9.59.98, S10-11:16.92, S11-13:42.81, S12-16:12.49

[teal'c]

As promised here's my benchmark of the i7 4770k using the new rocket in .23:

<a title="Download your file from SendSpace" href="http://www.sendspace.com/file/g8l6xp" target="_blank">Download i7 4770k @ 3.5 GHz fps.csv</a>

<a title="Download your file from SendSpace" href="http://www.sendspace.com/file/b9keml" target="_blank">Download i7 4770k @ 3.5 GHz frametimes.csv</a>

<a title="Download your file from SendSpace" href="http://www.sendspace.com/file/imdlw2" target="_blank">Download i7 4770k OC 4.3 GHz fps.csv</a>

<a title="Download your file from SendSpace" href="http://www.sendspace.com/file/axpq4u" target="_blank">Download i7 4770k OC 4.3 GHz frametimes.csv</a>

<a title="Download your file from SendSpace" href="http://www.sendspace.com/file/70xul3" target="_blank">Download i7 4770k OC 4.6 GHz fps.csv</a>

<a title="Download your file from SendSpace" href="http://www.sendspace.com/file/lr2xnp" target="_blank">Download i7 4770k OC 4.6 GHz frametimes.csv</a>

[DMagic]

Thanks for those teal'c, multiple results at different clock speeds are great because they give the clearest indication of their effect on performance, I don't have to worry about settings or GPUs, or any other complicating factors.

[Sirine]

My Dell Inspiron Laptop

Spec and KSP Setting.

Javascript is disabled. View full album

Result.

http://www./download/gimiiurmge9j476/KSP_2014-02-17_14-50-29-93_minmaxavg.csv

http://www./download/ts4g1sf6v07hdo4/KSP_2014-02-17_14-50-29-93_fps_.csv

http://www./download/7mwbc2c85v5oww2/KSP_2014-02-17_14-50-29-93_frametimes.csv

Graph.

---

My Dell Vostro 1400 Laptop

Spec and KSP Setting.

Javascript is disabled. View full album

Result.

With V-Sync

http://www./download/8jwde87avpgj35c/KSP_2014-02-17_15-52-01-76_minmaxavg.csv

http://www./download/38bdcgshk3844mc/KSP_2014-02-17_15-52-01-76_fps.csv

http://www./download/8z8kucb25sqyflh/KSP_2014-02-17_15-52-01-76_frametimes.csv

Without V-Sync

http://www./download/ogqd12dbd5etph7/KSP_2014-02-17_16-29-29-04_minmaxavg.csv

http://www./download/53ae1b57m5uaw95/KSP_2014-02-17_16-29-29-04_fps.csv

http://www./download/3cf587z44q7iw16/KSP_2014-02-17_16-29-29-04_frametimes.csv

Graph.

With V-Sync

Without V-Sync

Edited February 17, 2014 by Sirine

[DMagic]

Thanks for these results Sirine. At some point I'm going to stop neglecting this thread and figure out a good way of incorporating all of these frametime results into my analysis. Maybe checking if CPU power correlates with frametime variance, or the Leonidas3 framerate, or maybe plotting the frametime difference between consecutive frames. There are several options for using all of that data.

[Sirine]

While launching your test rocket in the beginning of 1st minutes.

For what I have notice, base on your data collected, so far, none have achieve a 30fps at 600 parts.

The best is i7-3630QM 2.4GHz.

http://ark.intel.com/products/family/75023/4th-Generation-Intel-Core-i7-Processors/desktop

Intel® Core™ i7-4770R Processor (6M Cache, up to 3.90 GHz) Hope to see this in the result sheets.

Edited February 21, 2014 by Sirine

[Blaster]

I just replaced my old dinosaur CPU, the Phenom II B55, with an AMD FX-6350. The Upgrade was a drop in replacement on my board, which has the AMD 970 chipset. The Triple Deuce setup on the 6350 is more than enough parallel processing power for my needs, and the $129.99 pricetag on Newegg with a free game makes it a good buy. So here's the new performance benchmark after dropping in the new CPU.

Download Benchmark

The system is the same, the only differences are:

AMD FX-6350, 3.9 GHz Six-Core (3x2) Processor with 4.2 GHz Turbo Boost

8 GB G.Skill Ripjaws DDR3 1600 (now running at 1600 and not 1333)

Overall, the system feels more responsive with the FX-6350 compared to the old Phenom II B55

Edited February 26, 2014 by Blaster
Cop y Pasta only pasta'ed half the copy

[DMagic]

Thanks for the results Blaster. They actually look really similar to the FX-4350, which isn't too surprising considering that KSP won't actually do anything with the 2 extra cores.

While launching your test rocket in the beginning of 1st minutes.

For what I have notice, base on your data collected, so far, none have achieve a 30fps at 600 parts.

The best is i7-3630QM 2.4GHz.

http://ark.intel.com/products/family/75023/4th-Generation-Intel-Core-i7-Processors/desktop

Intel® Core™ i7-4770R Processor (6M Cache, up to 3.90 GHz) Hope to see this in the result sheets.

Yep, it seems unlikely that we'll see anyone break 30FPS with this rocket until when get some more performance improvements. Top-end, single-core performance on Intel CPUs hasn't changed much in the past few years, and it's unlikely to do so anytime soon. Though I'm actually curious to see how a newer, really high end mobile Intel CPU would do. Their most recent ultrabook CPUs actually perform really well, and all within a really constrained power level.

That i7 4770R isn't actually any better than the regular 4770. The reason it's so expensive is that it has Intel's most powerful integrated GPU, I actually don't think you can buy those parts retail, they only come in OEM formats like the BRIX Pro. One of those tiny little things would probably make a really nice KSP computer.

[Wurmi]

I made also a Test.

https://drive.google.com/file/d/0BxmXhA1O4msaWkpwTGRidmlZZHc/edit?usp=sharing

https://drive.google.com/file/d/0BxmXhA1O4msaandrV1NwRUQxREU/edit?usp=sharing

Macbook pro 15" mid 2012

intel core i7-3615QM 2.3Ghz (Turbo till 3.2Ghz)

16GB Ram

nvidida geforce 650m

Windows 7 x64

Edited February 27, 2014 by Wurmi

[DMagic]

Thanks Wurmi, I updated all the charts. That seems like a pretty powerful machine, it's the best of all the mobile results I have.