SSD for KSP

[Red Iron Crown]

32 minutes ago, kBob said:

Did you see my post right after you first asked the question?  SSD is considerably faster.

Ah sorry, I missed it. I can't find that particular model of HDD, what are its specs? What is your CPU load like during the loading process?

[pandaman]

OK.. my test results...

KSP v1.1.2 x64

WIndows 10 64bit,  quad xeon w3550,  12GB DDR3

I copied my KSP folder onto my HDD and tested both from the links in the folders rather than use desktop shortcuts so all was as even as possible.

Modded install (4 mods - TWP, KAC, KER, EVE)

SSD  1 min  06 sec

HDD  1 min  34 sec

 

Clean Stock install

SSD  1 min  05 sec

HDD  1 min  23 sec

 

I only ran each test once so, allowing for 'human error' with a stopwatch, the stock installs were essentially the same, but the modded install was a bit slower on HDD (I would guess the difference would possibly increase with more mods).

So yes it would appear that the biggest difference in start up speed is not so much due to SSD or HDD, for a stock or low mod install at least.

 

Edited June 21, 2016 by pandaman
saved before I was ready

[kBob]

Just now, Red Iron Crown said:

Ah sorry, I missed it. I can't find that particular model of HDD, what are its specs? What is your CPU load like during the loading process?

It a somewhat older SATA 6 drive.  Haven't looked at the CPU load just timed the total loading times, but it shouldn't be factor since it's the same computer.  The cpu is a 5930k (not over clocked at present) so it handles the computations pretty well  of course an old i5 or i3 would reduce the % improvement as the computation time becomes a bigger factor, but an SSD is still going to help--quite possibly a lot.

[Red Iron Crown]

2 minutes ago, kBob said:

It a somewhat older SATA 6 drive.  Haven't looked at the CPU load just timed the total loading times, but it shouldn't be factor since it's the same computer.  The cpu is a 5930k (not over clocked at present) so it handles the computations pretty well  of course an old i5 or i3 would reduce the % improvement as the computation time becomes a bigger factor, but an SSD is still going to help--quite possibly a lot.

That's a good point. With a low end CPU the difference between SSD and HDD will be less significant as the CPU load dominates the storage load. For a high end CPU the situation should reverse.

[DiRoSVK]

SSD in RAID 0 - if one disk breaks down so you'll lose all your data, but it is a quick read and write, 2x250GB SSD = 500GB
SSD in RAID 1 - a quick read as in RAID 0 and data are protected, 2x250GB SSD = 250GB

 

[Carraux]

For a valid test please make sure that the disk cache is not hitting in. I recommend to make each test after a fresh reboot.

Edited June 22, 2016 by Carraux

[Znath]

It is a bit faster. 

Like some said there's only so much you can do like when you get file-bottlenecked when you start loading dozens of tiny files instead of one big one.  Usually if the game crashed before I got my ssd, I'd give up and leave.  Now it loads fast enough I will just reload the game.

[Carraux]

12 hours ago, Red Iron Crown said:

That's a good point. With a low end CPU the difference between SSD and HDD will be less significant as the CPU load dominates the storage load. For a high end CPU the situation should reverse.

Yeah, that's true.

[kBob]

15 minutes ago, Carraux said:

For a valid test please make sure that the disk cache is not hitting in. I recommend to make each test after a fresh reboot.

I reboot by habit when testing loading speed.  Of course if you're just doing single tests it doesn't matter as these will be from two different install locations it would only be an issue if you were repeating the same load to get an average figure or if you already have run KSP then decide to test from the same install of KSP without flushing the cache (in my case I already knew the no cache load time from my SSD from a post in another thread).

[Carraux]

I made some performance tests on my system.

Specs: Windows 10 with KSP 64 and ~55 mods (GameData approx. 8GB).
Core i7-2600k with 8GB RAM, drive #1 is a RAID 0 of 2xSDD @ 200GB, drive #2 is a RAID 1 of 2xHDD @ 1TB.

HDD with valid MM cache: 3:58
HDD w/o valid MM cache: 6:39

SSD with valid MM cache: 2:42
SSD w/o valid MM cache: 5:20

Please note that the RAID 1 HDD may be up to two times faster than a normal HDD, because it can read from both disks simoutaneously (like RAID 0). So you may add 50-100% more time for a non RAID drive.

Modulemanager sucks up nearly 50% of loading time, when the cache is present. It is interesiting to see the big performance hit when the mm cache is missing.

Edited June 23, 2016 by Carraux