Velociraptor or SSD for FSX?

[Noel 4,572]

 

Also, can FSX and FS9 be run on the same computer which is optimized for FSX via the AVSIM Software/Hardware Guide for FSX (on separate drives) or should I just maintain a separate computer for each sim?

 

For my upcoming build (still several months out) I plan on keeping OS & Sim on the same single SSD.  This allows for easy single drive cloning or backups to a HDD.   My understanding is that SSDs, unlike HDDs, don't benefit greatly by putting the OS & Sim on their own separate drives.  I plan to dedicate a single SSD to each set of OS+Sim.  The only extra expense I assume will be in having two licenses for Win 7.   When you have separate drives you have to do synchronized simultaneous cloning whereas not so w/ single drive installations.

[Guest]

SSD on the correct SATA port.  There is still some benefit to keeping the FSX on a separate SSD from the OS.  Even more benefit if you use a different SATA controller.  Just make sure SATA port spec matches the SSD spec (aka SATA III port and SATA III SSD).

 

I would avoid a mechanical drive for FSX.  In fact, you would have better success with a Mechanical drive for the OS and an SSD for FSX and then another SSD for Scenery ... try to keep scenery on a separate drive from FSX.

[Noel 4,572]

 

There is still some benefit to keeping the FSX on a separate SSD from the OS.

 

Can you quantify 'some benefit'?  Give an example of how the two approaches would manifest differences if you know...

[Guest]

 

Can you quantify 'some benefit'?

 

For SATA and PCIe ... serial point to point device ... so if you have FSX on the same SSD as the OS ... so the OS is requesting data and FSX is requesting data ... they both end up in a single queue.

 

Move FSX to a separate SATA SSD from the OS, you now have two separate queues of I/O requests. 

 

Extend that to OS on one SSD, FSX on another SSD, and Scenery on a 3rd SSD ... you now have 3 separate I/O request queues rather than 3 I/O requests in a single queue.

 

Benefit is that FSX doesn't have to wait for the OS I/O request to finish before it can process it's own I/O request.

 

This is obviously over simplified, but it's essentially what happens.  It's a game of maintaining the best possible data flow.

[foibles 0]

For SATA and PCIe ... serial point to point device ... so if you have FSX on the same SSD as the OS ... so the OS is requesting data and FSX is requesting data ... they both end up in a single queue.

 

Move FSX to a separate SATA SSD from the OS, you now have two separate queues of I/O requests. 

 

Extend that to OS on one SSD, FSX on another SSD, and Scenery on a 3rd SSD ... you now have 3 separate I/O request queues rather than 3 I/O requests in a single queue.

 

Benefit is that FSX doesn't have to wait for the OS I/O request to finish before it can process it's own I/O request.

 

This is obviously over simplified, but it's essentially what happens.  It's a game of maintaining the best possible data flow.

 

Where does that help you in the sim though? Does it help in reducing how long it takes to start up the sim? How long it takes to load a flight? Frames per second during play? Less stuttering during play?

 

Lars (Saab340) previously posted his results

https://www.dropbox.com/s/uq6taqp0dlpwhhw/FinalResults2.GIF

comparing different scenarios involving a 5400rpm HDD, a 10,000rpm Velociraptor, an Intel SSD, and an OCZ SSD, as mentioned.

I believe the data showed that FPS did not change regardless of where he placed the OS and FSX. If he put OS and FSX both on the 5400rpm HDD, his FPS were 41 and stutters he was able to measure were 29/min. Same if he put the OS and FSX both on the Velociraptor. Same if he split the OS and FSX between mechanical drive and SSD, or between two SSDs. And except for when he put FSX on the 5400rpm HDD, the time required for texture loading that he could measure (at least within the specific constraints of his experimental setup) did not vary much -- texture loading took 20 or 21 seconds regardless of how FSX and OS were distributed.

 

The only things that changed were the amount of time required to launch the program and then to load individual flights.

 

It's true that I don't know how Lars was assigning the drives in terms of SATA controllers, and it's also true that he was using older SSDs/slower SSDs than are available today. Plus, he didn't tease out the effects of placing scenery files on a separate drive. But his results seem to be reason to be sceptical of the argument that splitting all the sim components up makes a significant difference in the sim experience, once you've got your flight loaded.

 

In fact, another user on one of the other forums said, "I tried Win7 on the 80GB SSD and FSX on a WD Black, then vice versa. Ultimately both on the SSD. Seperated, they performed about the same, better than on the spinner, but, with them both on the SSD for my FSX, there was a vast improvement across the board. Stuttering was greatly reduced and I could actually add a little more eye candy.

I like lots of traffic, detailed clouds, and RW weather. I also have quite a few large photoreal sceneries and am planning more."

http://www.flightsim.com/vbfs/showthread.php?236348-FSX-Scenery-etc-what-install-size-%2860Gb-SSD-Win7-x64%29&p=1602140#post1602140

(I bring up this last point to get back to some of the earlier points made in the thread, not to whether assigning separate SSDs for each component makes a difference.)

[Guest]

Vic,

 

I would not expect SSD to impact Frame rates, certainly not on FSX SP2 and/or acceleration where texture loading was coded for separate threads (did SP1 include the texture loading threaded code also -- think it did, not sure).

 

Without knowing all the details of the various scenarios Lars tested/presented, I can't really comment.  Stuttering could be any number of issues that may or many not be related to I/O performance.  

 

As far as noticeable will depend on how you fly and your FSX.CFG and a lot of other variables (video card memory, cache, etc. etc.) ... pretty hard to make a decisive conclusion, it really is a case of try it and see.  Will having things separated out on two or three SSD's make a huge noticeable difference in texture loads -- impossible to say, but optimizing for the best case scenario is something I would always want to do when it comes to FSX.

 

I was simply suggesting the "best" option for performance, not that it will be "noticeable" defective if that option is NOT used.  The only time I noticed some tile blurries (not loading the high res textures) was when I move my SATA III drive to an SATA II port ... I may have been able to solve that problem with adjustments to my FSX.CFG.

 

I setup my FSX as best as I possibly can and it seems to be running well but I have separate loads depending what aircraft I fly (like PMDG for example).  But a flight in a Heli around Hawaii / Maui / and other islands (with photo scenery and FSDT airports) and I was at 120-140 fps (2560 x 1600, LOD 9.5, 4096 textures, maxes across the board) with some incredible weather/water (REX) and scenery (yes buildings too) and never a hint of a blurry tile even with rapid movement ... even tested with the Hornet and was amazed.

[MikeAdamo 51]

Jeff,

 

I've got a Vraptor drive exclusively for my FSX install and in addition to the other great suggestions on segregating your OS from FSX, make sure to invest in a very good HD defrag util. There are some camps that say there's no benefit to defragmenting your HD but I'm one who believes a clean house is a well run house. Keeping the fragmentation on the HD down and files contiguous will help load times and will mitigate scenery loading as you fly along.

[Noel 4,572]

This is obviously over simplified, but it's essentially what happens.  It's a game of maintaining the best possible data flow.

 

Thanks, and that was already known, i.e. the qualitative reason for this.  I have always had OS and FSX on separate HDDs for those reasons.   Can you address the quantitative piece w/ respect to SDDs?  For example, what is the latency when there are serial or even simultaneous requests?  Does the ram cache resolve the inherent latencies involved such that they really aren't meaningful?  And to move this even closer towards the meaningful (for me) from the theoretical, what is an example that would describe how this would empirically affect FSX 'performance', w/ performance defined as Frame Rate + Image Quality + Video smoothness?  I can imagine the primary domain would be found in video smoothness aspect of performance, and maybe freedom from transient hard freezes in video.  Here's an example:  sometimes I notice a slight video hitch when an ATC .wav file initially loads, that is absent when that file execute subsequently during the FSX session.   I had deduced this to be likely from HDD reads since as I say subsequent plays don't cause the slight hitch.  Others report this behavior.  If this particular issue was solved by putting OS/FSX on separate SDDS, versus a single SDD, then I might consider doing the former.  But I'd like to see this assessed because at this point in my understanding, I would not trade in the simplicity of the single drive approach unless there were compelling performance reasons.  Any practical experience w/ this by chance?

 

Addendum:  Ah, I see some other comments above on this where OS & FSX were on the same drive--sounds like good 'nuf testimony to go that route, and it's pretty good for backups/clones, which is no little thing.  Thanks all for your insights ;o)

[JNS 106]

Wow, some really good comment from all of you.  Seems like I kinda opened up a can of worms.  I like the idea of OS/FSX on one SSD and may try that mainly for ease of backups and then try the other method, too.  Decisions, decisions.

 

Jeff

[Guest]

 

I notice a slight video hitch when an ATC .wav file initially loads, that is absent when that file execute subsequently during the FSX session.

 

You'll need to run a tool like this: http://technet.microsoft.com/en-us/sysinternals/bb896645 to monitor exactly what is happening.  It will also answer some of your other questions with latency, stacked requests, etc.

 

Unless the .wav file was huge and of very high audio sample quality, I would not expect it to cause a wait condition.  Will need to know your OS also as Vista/Win7/Win8 don't have a direct path to the audio drivers, but XP does.  Audio access changed A LOT in Vista onwards, especially 3D audio where you need to use ASIO and OpenAL ... like Creative's ALchemy which translates DS3D into OpenAL calls ... I'd suggest looking at your audio drivers/versions before looking at possible I/O delays from an HD or SSD.

[SAAB340 28]

 

I believe the data showed that FPS did not change regardless of where he placed the OS and FSX. If he put OS and FSX both on the 5400rpm HDD, his FPS were 41 and stutters he was able to measure were 29/min.

Yes, it's true that average FPS did not change between a low performing HDD and an SSD. The storage won't give you better avg FPS. It's also true that the stutters-per-minute value i messured also remained the same, but that value does not take in to consideration how hard each stutter were. I just want to clarify that the 5400rpm drive give a few very hard stutters. And that is very noticeble during gameplay. This can be improved on by having OS and FSX on separate drives.

 

 

It's true that I don't know how Lars was assigning the drives in terms of SATA controllers,

They were running on the SATA II ports off the Intel P55 chipset.

 

 

Plus, he didn't tease out the effects of placing scenery files on a separate drive.

I have results with the OS+FSX installed on a SSD and Photo scenery installed on a Velociraptor there. And that is giving you loadtimes that are a lot closer to SSD only installation than Velociraptor only installation. What I don't have are results with the OS on the SSD and FSX+Scenery on the Velociraptor to compare with.

 

 

Benefit is that FSX doesn't have to wait for the OS I/O request to finish before it can process it's own I/O request.

That's right. And that's why having OS and FSX on separate mechanical drive gives less hard stutters and improved loadtimes/texture loading.

But when we are talking about SSDs they are a huge order of magnitude better at dealing with a lot of IOs over mechanical drives. SSDs are very good att handle queed requests. In order to get maximum throughput form an SSD you pretty much have to up the quedepth. That's why it doesn't matter if you have everyting on a single SSD. A single SSD can handle it with ease. Even thou having separate SSDs for OS and FSX in theory might have some benefit it won't actually show within FSX.

[Noel 4,572]

I'd suggest looking at your audio drivers/versions before looking at possible I/O delays from an HD or SSD.

 

Well, mainly was using this as an example of how OS & FSX on separate SSDs might produce noticeable results.  I guessed that might be it.  Also, back when I had FSX on XP 32bit, I had the same issue.  It's minor though and doesn't really hurt the suspension of disbelief as it were.  But this aside, doesn't the fact there is no delay after the file has loaded into memory from the HDD support the notion it was the HDD read that generated the stutter?  If not, what would cause the stutter not to occur when the file plays subsequently?

 

Robains, do you have any practical examples about how two SSDs outperform one for FSX?  For me that answer is where the rubber meets the road for folks trying to ascertain things like cost:benefit on a design strategy.

[Noel 4,572]

 

You'll need to run a tool like this: http://technet.micro...ernals/bb896645 to monitor exactly what is happening.  It will also answer some of your other questions with latency, stacked requests, etc.

 

That's a good tip thank you Robains--seems like a great idea for overall evaluation of your configuration eh?

[Adam T Lutley 12]

SSD's are continuously dropping in price. If you already have an SSD it seems like a no brainer to me. Faster, no noise, no de fragging to worry about. Will never go back to a "spinner" drive.

[SAAB340 28]

 

And that is very noticeble during gameplay.

I just found this screenshot of 2 graphs showing just that.

 

 

Remember that stutters over 33ms will show even when the vsync tweak is used.

 

There was no improvement in how hard the stuters were when having the OS and FSX on 2 separate SSDs compared to having everything on just one SSD. I don't have that graph at hand right now, but I have tested it thorougly.