I've seen some confusion around SSD's not performing or making much of a difference in one's system or a difference in FSX, XP10, P3D, etc.

 

So here are some things that might help clarify:

 

SATA is 1.5GB/s

SATA II is 3GB/s

SATA III is 6GB/s

 

Motherboards often have a mix of SATA ports, they are NOT all the same -- some are SATA II and some are SATA III and some maybe even be SATA. You'll need to check with your motherboard manufacturer/manual as it should clearly define which ports support the various different SATA specifications. Does it make a difference in performance which port you use for your SSD, yes, absolutely.

 

For example I have two OCZ Vector 256GB SATA III SSDs. I plugged one (remember these are identical drives both supporting SATA III) into an SATA II port on my motherboard and the other into a SATA III port on my motherboard and ran SiSoftware Sandra Pro's Disk I/O tests.

 

OCZ Vector plugged into my SATA II port on my motherboard:

 

Benchmark Results

Drive Score : 247.15MB/s (246.68MB/s - 248.87MB/s)

Random Access Time : 153µs (153µs - 3.74ms)

Performance per Thread

Drive Score : 10.3MB/s (10.28MB/s - 10.37MB/s)

 

OCZ Vector plugged into my SATA III port on my motherboard:

 

Benchmark Results

Drive Score : 527.61MB/s (495.5MB/s - 534.72MB/s)

Random Access Time : 18µs (18µs - 542µs)

Performance per Thread

Drive Score : 22MB/s (20.65MB/s - 22.28MB/s)

 

As you can see, a pretty staggering difference in performance for the same drive just in a different port.

 

Now some of you may not have SATA ports so you go out an buy a SATA controller ... these can range from $30 to $1000, but often they get limited to a 1X or 2X or 4X PCIe configuration which can negate the benefit of SATA III (6GB/s) control (especially if the controller has more than one SATA III port.

 

And to muddy the waters up even more, SATA III uses a slightly different cable type and connector ... they will (backwards compatible) fit all SATA ports, but the cable specification is different for SATA III.

 

So the moral of this story is consistent with what I've always maintained ... a computer's real world performance is a sum of all it's parts not any single component.

 

Rob.

 

P.S. Here are performance numbers from an external USB 3.0 3TB HDD (7200rpm mechanica drive by seagate):

 

 

Benchmark Results

Drive Score : 153.53MB/s (97.78MB/s - 191.58MB/s)

Random Access Time : 16ms (16ms - 285.41ms)

 

Performance per Thread

Drive Score : 12.8MB/s (8.15MB/s - 16MB/s)

Thanks for posting that Rob.

 

I was actually just searching the AVSIM forums for an answer to that very question....6 Gb/s SATA port versus 3 Gb/s SATA port.

 

Its actually partly your fault, yours and Firehawk44's.  You guys kept posting those really nice screenshots of your photoscenery flights in a few different threads in the FSX forum....now I'm hooked

 

I tried MSEv2 for Florida and really liked it.  At low altitude I preferred my UTX/GEX combination...but for 5000-10000 feet and above, I really liked that photoscenery.  So I went on to pick up Florida, Alabama, Mississippi, Tennessee, Georgia, Lousiana, Kentucky, and Illinois.  I'm not out of space yet...but its forcing me to start thinking about space now and I'm starting to think maybe I should pick up another SSD.

 

My system runs an ASUS Z77 Deluse Mobo which has 2 6 Gb/s SATA ports, 2 more Marvel 6 Gb/s SATA ports, and 4 3 Gb/s SATA ports.

 

I've got my Main OS & Programs 512 GB SSD plugged into one of the 6 Gb/s SATA ports, and my FSX "stuff" (FSX, Addons, Scenery, etc) 512 GB SSD plugged into the other 6 Gb/s port.  The two Marvel 6 Gb/s ports are taken up by a 3 TB HDD cached to a small 64 GB SSD using the built in ASUS caching which only runs through those 2 Marvel ports.

 

I've got a Blu Ray Writer plugged into one of the remaining 3 Gb/s Ports, and a hot-swappable drive bay plugged into another (There is no actual drive in that bay, its empty, but I have the hot swap bay plugged in and configured in case I ever need it).

 

So that leaves 2 3 Gb/s ports.

 

I don't know very much about SATA speeds, what is a noticeable difference and what isn't, and how they work...so I was concerned that 1) Hooking a new SSD up to a 3 Gb/s port would be too slow and 2) Because I don't know much about the way the ports work, I was worried it was a lowest common denominator thing and that by hooking an SSD up to one of those 3 Gb/s ports I'd somehow be limiting the speed of my 6 Gb/s port SSD's as well.

 

So I'm not sure where the heck I'm going to put more photoscenery as I get it.  My FSX SSD still has alot of space but eventually thats going to start to run down.  I'm guessing its probably a bad idea to put it on my 3 TB HDD because I'm guessing it would be alot slower and maybe dog my FSX performance.

The good thing about PCIe is that it's not "shared" device - serial endpoint ... dedicated physical path from the device to the controller ... so you don't penalize performance across other devices.  However, the PCIe controller will have a shared interconnection with the rest of the system but not affected by running a slower device.

 

PCI is a shared parallel bus, but the Asus Z77 is using PCIe (no PCI) and the Intel Z77 chipset (2 SATA III) with ASMedia controller for 2 SATA III and 2 eSATA III.  So in theory you have 6 SATA III ports if you wanted to use the eSATA ports.

 

For SSD speeds, you'll need to see what the drive claims and it's spec support.  Key is that if you have a SATA III device (SSD), you want to put it in a SATA III port.  If you put an SATA III SSD capable of 500+ MB/s and put it in an SATA II port, you'll likely get 1/2 of it's potential transfer speed (around 200-250 MB/s).

 

Be careful if you RAID an SSD ... be sure to confirm the SSD will work in a RAID configuration with the controller.  I know my OCZ Vectors do NOT like to be setup in a RAID configuration ... they become unstable and unreliable.

 

I would NOT put scenery on a mechanical drive ... it's not so much the transfer speed but the access time that hurts when you are running FSX.  SSD random access is several orders of magnitude better than mechanical drives.

 

All my internal drivers are fast SSDs (SATA III 6GB/s) with the BD burner on a SATA II port (3GB/s).  I have two external USB 3.0 3TB drives that are connected via USB 3.0 ports ... I use these for "offline" storage and backups ... they are surprisingly fast but not fast enough for FSX.

 

I'm at 2TB of SSD's right now ... you can always add a 4 port SATA III controller ($60) when you run out of motherboard SATA III ports.  Or for more extreme, go with Areca RAID SATA III controller -- $500 - $1200 ... but be sure to check that the controller works with your SSDs.

 

When I tested my SATA III SSD in a 3GB/s SATA II port and loaded it with some scenery ... I did get a few blurries in FSX (not bad -- just a few here and there with rapid transitions, but not as good as when I used SATA III 6GB/s ports which were blurry free) -- but this could be just a matter of me needing to adjust my FSX.CFG -- however I'm at FIBER_FRAME_TIME_FRACTION=0.99 which as I understand it is about as high as you can go in terms of sharing CPU/render time vs loading textures - basically for every one millisecond spent rendering then one millisecond is spent loading scenery (default is 0.33 I believe which is 3ms to render and 1ms to scenery texture load).

It's actually preferrable to use Intel SATA II ports over 3rd party (Marvel) SATA III ports for SSD's (even for SATA III units).

Both for performance reasons and because you don't get TRIM on Marvel controllers.

 

Also, cheap PCIe SATA III adapters are all crap

 

SATA II only limits large sequential transfers significantly. Random access is much more important

 

It's actually preferrable to use Intel SATA II ports over 3rd party (Marvel) SATA III ports for SSD's (even for SATA III units).
Both for performance reasons and because you don't get TRIM on Marvel controllers.

 

Not in my experience - as you can see from the testing I did above, Random Access took a big hit also when using SATA II port (keeping in my the same SSD was used in both tests).

 

SATA II Random Access Time : 153µs (153µs - 3.74ms)

SATA III Random Access Time : 18µs (18µs - 542µs)

 

TRIM or GC is a function of the SSD controller built into the SSD, not the SATA controller.

 

PCIe SATA III controllers aren't all crap, certainly no more crap that what's built onto the motherboard ... both aren't "great" controllers but they get the job done.  If you want really good SATA controllers then look into something from Areca -- these are server grade components that are incredibly fast with their own Cache and processors (yes multiple) so your main CPU does less work.

 

But back to nominal cost effective products, I did notice a difference in FSX using SATA II port vs. SATA III port on an SSD that is SATA III.

TRIM or GC is a function of the SSD controller built into the SSD, not the SATA controller.

 

TRIM requires support from the SATA controller's driver in order to work. The current Intel, AMD and standard MS SATA drivers all pass the TRIM command through. Old drivers from these companies and some third parties won't.

 

GC, on the other hand, is entirely internal to the drive.

Ok, but still, it's better to stick with Intel ports. Marvel 9128 or your ASMedia controllers perform worse and no, you get no TRIM with those like goates said. Some GC algorithms work well enough that it may not be a huge deal, and you can always do some secure erase every now and then, but SATA III is the least of the factors

here. Not sure how you got those random access results, but for the most part, it goes unaffected

I used SiSoftware Sandra Personal 2013 SP1.a - Physical Disks benchmarks and Fils System Bandwidth tests.  Testing was done on the same system using the same SSD with the only difference being SATA II vs. SATA III ports on the motherboard.  I also have another HighPoint Rocket 640L SATA III card that performs a little better than Intel ports or ASMedia controller.

 

 

Ok, but still, it's better to stick with Intel ports. Marvel 9128 or your ASMedia controllers perform worse and no, you get no TRIM with those like goates said.

 

I guess we'll just have to disagree.  As far as difference in FSX, minimal, I noticed some occasional delays it loading hi-res textures (aka REX and/or UTX) ... very minor when on a SATA II port.

 

TRIM on newer SSDs is completely irrelevant, but again TRIM is not a function of the SATA controller.

but again TRIM is not a function of the SATA controller.

 

Sorry but this is not even debatable. Anyway, if you're happy with your setup, so be it

May I suggest this reading on TRIM http://en.wikipedia.org/wiki/TRIM

 

The TRIM command is initiated by the OS which is then passed on by the SATA controller to the SSD device and the SSD device will either perform the TRIM (it if supports TRIM) or perform an equivalent operation or ignore the TRIM completely.  TRIM is a command specification of the ATA interface, it is NOT the function of the SATA controller to perform the TRIM, only to pass the TRIM command onto the SSD device.  SSD devices do implement TRIM different.

 

Perhaps my interpretation of "Function of" is different than yours.  But any controller that supports the SATA specification should in theory support TRIM as it's a part of the SATA specification, but the actual implementation of TRIM is handle by the SSD ... and hence a "function of" the SSD device.

Perhaps my interpretation of "Function of" is different than yours.  But any controller that supports the SATA specification should in theory support TRIM as it's a part of the SATA specification, but the actual implementation of TRIM is handle by the SSD ... and hence a "function of" the SSD device.

 

Yes, the actual work associated with the TRIM happens inside the SSD, but it does require support from the OS and SATA controller drivers to run. Windows XP does not support TRIM itself, thus it will never initiate the command on its own, for example. Windows 7, however, will initiate the TRIM command every time you empty the Recycle Bin. Another example is that Intel's drivers support the TRIM command for RAID0 arrays with their 7 series chipsets, but not with the previous 6 series. This is why I was saying that TRIM support is not just dependent on the SSD itself.

Whoah....alot of all that stuff was way over my head :blink:

 

So, with regards to FSX scenery in my situation described above...what is my best solution?

 

Since the time I posted that, I've discovered that I seem to have some "blurry" issues (Which I'll post about in the FSX Forum instead of here in the hardware forum) but I need to at least touch on it here as it does have some bearing on the hardware question I asked.

 

I've noticed that even on my SATA III SSD's running 6 Gb/s that I "seem" to have some photoscenery issues.  Now please, bear in mind, I'm new to photoscenery so maybe what I'm seeing is totally normal.  Right now, my Fiber_Frame_Time_Fraction is set to the default .33 and my LOD_Detail is set to 8.5.  At 10,000 feet in a Jet going 360 Knots I didn't notice any problems much.  Those problems became much more noticeable at 5000 feet in a jet going 360 Knots.  I notice that the city surrounding the airfield I take off from is pretty blurry...and sort of gradually "crisps up".  Things gradually sharpen but it takes a while.  If I fly over the same area for a while...it sharpens.  But if I keep moving in a straight line its entirely possible that it will never sharpen before I've past it which defeats the purpose.

 

So next I tried setting my Fiber_Frame_Time_Fraction to .99 like Rob described.  My results of that were inconclusive becuase it really dropped my frame rate alot.  While sitting on the ground at the end of the runway awaiting takeoff my frame rate had dropped to the high teens to the low twenties.  Flying around it was generally in the teens or low twenties.  Things were pretty choppy.

 

So next I tried setting my Fiber_Frame_Time_Fraction to .46...a value somewhere between the default of .33 and Rob's .99.  It did seem like the terrain sharpened better and faster, and it was "better" but still not quite right I don't think.

 

The way this relates to my hardware question is....

 

It looks like the bottleneck is my FPS....the rendering of those textures.  That sounds like a CPU/Video Card thing.  If thats the case....then maybe its not going to make a difference whether I'm running at SATA III or SATA II speeds off the SSD because the FPS indicates that the video card/CPU is already handling all it can?  Thus...maybe using those 3 Gb/s SATA II ports is the ticket for any further scenery SSD's?

 

again...if thats a stupid question...please forgive me.  Learning here.

Yes, the actual work associated with the TRIM happens inside the SSD, but it does require support from the OS and SATA controller drivers to run.....

 

This is why I was saying that TRIM support is not just dependent on the SSD itself.

Agree, but "dependent on" is not the same as "function of". But for the record, when a TRIM command is issued, the SSD can implement it anyway it likes or do nothing at all ... but it does take ALL 3 elements of the process. From my discussions with OCZ engineers, their Vector SSDs support TRIM but pretty much ignore it and utilize their own internal controller to manage GC real time.

Thus...maybe using those 3 Gb/s SATA II ports is the ticket for any further scenery SSD's?

Your tuning experience is pretty much the same as everyone else's ... your computer needs to be addressed as a whole unit, not just any single component like the SSD. Data (going from A - your SSD to B displayed as a pixel on your monitor) is just like car traffic, if you go from 4 lanes down to 2 lanes the cars start to back up and have to slow down. If you increase the number of cars on the road (aka more Data) , the further the cars get backed and the slower everything is.

 

With FSX, what you are looking to do is manage the flow of Data minimizing the delays. In order to do this, you have to taken into account the path of the Data in every step of the way from point A to point B. Point A and Point B may have different starting points and ending points (sound for example is a different point B).

 

As far as scenery visuals go, the more you zoom OUT the better it looks. You actually do NOT want to apply AA to scenery textures as this will introduce problems ... for example SMAA does not differentiate between geometric edges and texel edges. And further more, sampling an image/texture that has high frequency components with a grid of lower frequency results in aliasing artifacts (these can appear as white speckles in FSX).

 

Anyway, the topic of AA can get very very deep -- so to summarize, you'll just have to experiment as it's rare to have any two PC's exactly the same (and being "close" is not the same as "exact" - doesn't take much of a change in hardware/OS variance to produce very different visuals). So I have no answer for you and no magic bullet settings.

 

However, if you are going to buy new SSD, then I would opt for the best you can afford ... if you ever do step up other components you'll be constantly moving in progressive direction.

Thanks Rob.  Much appreciated