OFF and SSD

[almccoyjr 7]

Just for fun, see how SuperFetch in W7 works using Task Manager.

 

Run two tests that last one hour each. One test with SuperFetch enabled and the other with SuperFetch disabled. Don't do anything; just let the pc run. Reboot between tests.

 

Make note of the allocated ram and the services running in Task Manager for each test. I think you'll be surprised . Here's a hint: despite W7's "efficiency" it still leaks and depending on the services, leaks badly.

 

plug_nickel

 

ps..I always use Alacrity on extreme lean when simming and only enable Superfetch sometimes depending on the program.

Edited June 11, 2013 by almccoyjr

[+Lothar of the Hill People 6]

 I'd certainly admit a dose of sarcasm, Lothar, but you have to admit many of those things I said are certainy implied by what you've said.  Logically, if 16G marks the point beyond which SuperFetch makes SSD performance look sluggish, well, what the heck would be the point in putting an SSD in a system with 16G+ RAM?

 

(Because it's still going to be faster, that's why...but never mind that for a moment...)

 

That's kinda my point, gobs and gobs of RAM are less valuable on a pure SSD system, while fast working RAM is more valuable when long-term storage is slow but cacheable. It's a trade-off, with diminishing returns in either direction. And not everyone can afford to go in both directions.

 

If solid state tech was as fast as RAM, or non-volatile RAM cost-effective, you wouldn't need two types of storage. Just read and write everything out of superfast perma-RAM! Too bad the EEPROM thing never worked out, I guess.

 

But at the end of the day, RAM is much faster than SSDs. The more you can read and work out of RAM, the better. If fast SSDs can feed the even faster RAM to be cached, even better still. But that's a big if...

 

So, here's the thing: The performance figures you see in the sig (as well as a lot of others I fool with, but you don't see in the sig) are all programs.  I've not disabled SuperFetch.  RAM, we all agree, is faster than even the fast Revodrive SSD...

 

So, I get another 8G of RAM, slap it in the rig here, run the same several programs enough times, I should fairly soon see my SSDs performance get even faster...

 

I will have gone well past the 8G that isn't enough to give the increase.  Even if the SSDs are doing their best, the RAM's still faster...and the drive testing apps I use are all programs which would be recognized, cached and made faster by SuperFetch.

 

Even better, I think...since I have two different types of drives (but both solid state media) I ought to see some comparative differences once the caching kicks in, as well....

 

Oh, man, this is fun...I'm almost giddy here...*whew*  I gotta go get a beverage

 

I'm confused about your use of SuperFetch with the SSDs. From what I've heard you have to disable SuperFetch while any SSDs are installed in your system, as part of what it does to speed the caching is reorganize how data is written to drives, kinda like defragging. And this is not good for the lifespan of SSDs. Perhaps this has changed with Win8? Something you might want to look into more. Cause forcing SuperFetch to be disabled is the real problem with SSDs in high-RAM machines.

 

But assuming you can SuperFetch the SSDs, once it's trained you should see performance improve if it's reading the data out of RAM rather than having to get it off the solid-state drives themselves. Leave your machine on (reboot only when necessary for Windows Update), and its predictive modeling and giant cache will soon make itself felt.

 

All I know is 8GB wasn't enough for me--typically use nearly that much memory just with the stuff I have open, and used to have to start grinding the swap file to launch a game or something. But now with the added RAM not just to avoid virtual memory but to make use of SuperFetch, launching a game or large application that's cached is a snap. Only problem is crappy console ports that make you sit through five minutes of logos even though their no longer tied to slow console hardware.

 

Looking forward to hearing about your experience.

 

I wonder if we might impose on you - strictly in the name of science - to do a little experiment.  I do hate to ask, but, given the discussion we've had here, I'd welcome any actual specific testing to verify.

 

Good luck with the upgrades guys! It is a good chance to do relatively controlled before and after experiments.

 

Just for fun, see how SuperFetch in W7 works using Task Manager.

 

Run two tests that last one hour each. One test with SuperFetch enabled and the other with SuperFetch disabled. Don't do anything; just let the pc run. Reboot between tests.

 

Make note of the allocated ram and the services running in Task Manager for each test. I think you'll be surprised . Here's a hint: despite W7's "efficiency" it still leaks and depending on the services, leaks badly.

 

plug_nickel

 

ps..I always use Alacrity on extreme lean when simming and only enable Superfetch sometimes depending on the program.

 

Not sure what you mean about memory leaks. The amount of Free memory is supposed to decrease slowly over time as more stuff is cached in RAM in the background. Notice the amount of Cached memory goes up accordingly, filling with all the pre-fetched data ready to be read instantly or freed for other uncached applications if necessary, while the amount of Available RAM stays about the same.

 

This is a feature, not a memory leak! But this is what freaked everyone out and had supposedly tech-savy people disabling their SuperFetch. Totally "Free" memory is wasted, and using it to cache likely-needed data is a major OS innovation. 64-bit Win7 (and Win8 presumably) absolutely soars with gobs of RAM, because it's actually using it.

 

And what's Alacrity?

[hood 2]

Tamper, I would be happy to take part in your experiment as long as my rapidly deteriorating brain cells are capable of handling it. The new extra 8g of RAM should arrive in the next few days.The SSD will have to wait a while until a few more shekels become available. I await your instructions.

[almccoyjr 7]

Not sure what you mean about memory leaks. The amount of Free memory is supposed to decrease slowly over time as more stuff is cached in RAM in the background. Notice the amount of Cached memory goes up accordingly, filling with all the pre-fetched data ready to be read instantly or freed for other uncached applications if necessary, while the amount of Available RAM stays about the same.

 

This is a feature, not a memory leak! But this is what freaked everyone out and had supposedly tech-savy people disabling their SuperFetch. Totally "Free" memory is wasted, and using it to cache likely-needed data is a major OS innovation. 64-bit Win7 (and Win8 presumably) absolutely soars with gobs of RAM, because it's actually using it.

 

And what's Alacrity?

From Windows 95 SP2 thru W7, Microsoft has steadily decreased the amount of ram that services cache for potential use. When the OS kernal is ready, "free" ram is allocated to any and all services whether it's required or not.

 

And yes memory is supposed to go down "over time" as more "stuff" is cached; however, if you trace that useage, you'll find that many background services don't relinquish all the ram that's allocted to them when they're not in use.

 

The cache/allocation alogorithm MS uses in W7 is the most efficient to date but it's still flawed, hence the term "memory leaks" continues to exist. It's most apparent when the pc is booted and just left to run for a period of time. I've seen as many as 68 services caching "free" ram after inital boot and as little as 45 when disabling SuperFetch under the same startup scenario. And to note, the ram being reported to be free and consistenly shown to be available are, politely, very misleading. MS has has noted this in many publications, as have many noted ITT and "geek" forums.

 

Disabling SuperFetch, depending on the software being run and the background services being employed will actually increase performance; however, your mileage will most definitely vary going down this road.

 

Alacrity is probably the most comprehensive appliction used to control what services continue to run, when they run and when to automatically start them back up. On this forum, one of the members, Homeboy, wrote an excellent how-to tutorial on Alacrity. Combine this with Black Viper's website and you can have a very lean pc when needed, BUT take care on what services to control or........oops!

[Siggi 10]

This thread just made me go order another 8gb of RAM.

 

And yes, I have SSDs.

Edited June 11, 2013 by Siggi

[+Lothar of the Hill People 6]

And yes memory is supposed to go down "over time" as more "stuff" is cached; however, if you trace that useage, you'll find that many background services don't relinquish all the ram that's allocted to them when they're not in use.

 

The cache/allocation alogorithm MS uses in W7 is the most efficient to date but it's still flawed, hence the term "memory leaks" continues to exist. It's most apparent when the pc is booted and just left to run for a period of time. I've seen as many as 68 services caching "free" ram after inital boot and as little as 45 when disabling SuperFetch under the same startup scenario. And to note, the ram being reported to be free and consistenly shown to be available are, politely, very misleading. MS has has noted this in many publications, as have many noted ITT and "geek" forums.

 

??? That's how it's supposed to work! Cache the stuff needed by services so they can run instantly without a hiccup when needed, rather than wait for them to repeatedly reload stuff. Only when some uncached program requests a bunch of memory does SuperFetch need to release stuff that's been cached.

 

The less RAM that's "free" and the more that's being used to cache stuff, the better!

 

Disabling SuperFetch, depending on the software being run and the background services being employed will actually increase performance; however, your mileage will most definitely vary going down this road.

 

Alacrity is probably the most comprehensive appliction used to control what services continue to run, when they run and when to automatically start them back up. On this forum, one of the members, Homeboy, wrote an excellent how-to tutorial on Alacrity. Combine this with Black Viper's website and you can have a very lean pc when needed, BUT take care on what services to control or........oops!

 

Disabling SuperFetch will not increase performance but instead decrease it dramatically. Running stuff out of RAM is vastly faster than off HDD or even SDD, more than enough to make up for the actual processor cycles of the service itself. The more RAM you have, the more you'll notice the difference.

 

In Vista SuperFetch would grind the disk a bit filling the cache for the first time right after boot, which indeed could slow launching apps in just the first few minutes after booting. But Win7 is much smarter about how it fills the cache after first boot, without this noticeable performance hit.

 

Maybe if you're running 512MB like the crippled POS Vista machine with which Dell ripped off my ignorant mother, you'd benefit more from using a USB key as ReadyBoost storage than enabling SuperFetch. But for anyone on a modern multi-gig gaming machine, the ONLY reason to disable SuperFetch is if it's thrashing your SSD drives.

 

Most of the so-called tech "advice" on the web regarding SuperFetch is outdated garbage or pure FUD.

[Tamper 9]

I feel I should apologize, and offer an explanation to make something clear (or try):

 

Folks often don't get my sarcasm.  Easy enough to understand, especially in a 'written-only' environment like this forum.  I forget that people don't necessarily know me, and might not appreciate when I'm being sarcastic.

 

Let me make something perfectly clear:  I bought the extra 8G of RAM for exactly the reasons I cited: It's gotten cheaper, I had a coupon; I had considered adding memory anyway, and Father's Day is literally coming up.

 

At the same time, let me make something else perfectly clear:  I expect absolutely zero performance increase from adding this memory, at least due to any caching algorithm of itself.

 

I have a "slow" hard disk that I intentionally use to load any/all programs that don't require peak performance in terms of texture loading, etc.  I've checked, and although I never turned off SuperFetch, the service isn't running on my machine.  Well, if that means I'm not caching a hard disk because I have SSDs, guess what:  I load what I do on that drive precisely because it doesn't require top-end performance.

 

TBH I still haven't figured out why SuperFetch is off; my best guess is installing RAID driver software, having RAID set up in an UEFI BIOS, and/or the array controller(s) recognizing SSD media has set the service to "Manual".  I am absolutely certain I didn't turn it off, yet it's off.  Another mystery, I suppose, for another time.  (There are some reports of Windows not recognizing SSDs properly and therefore not disabling SuperFetch etc by default...but I have always believed that, being as my SSDs are all in RAID arrays, any operational enhancements would need to be made manually.  It appears that's not true.)

 

WIndows, by design, is supposed to turn off SuperFetch and the other caching tricks where SSDs because (get this) the Engineers at Microsoft understood that SSD performance means all that caching isn't necessary.  It was designed for conventional drives.  Imagine that:  Even with all the wonders reported about how great it is, SuperFetch is turned off by design, by the people who invented it, if you have an SSD.

 

Now, time for a common sense question:  What does this tell you about the comparative performance of the two?

 

I'm not using SuperFetch.  And I only have 8G of RAM, and my poor old SSDs, so much slower than RAM...and running things from that old platter based drive must really be killing me.

 

Thing is, nope, none of that's true.  Somehow, with my less than preferred amount of memory and SuperFetch not even running, I have no stutters and my system is responsive and snappy.  It boots fast, apps load fast, no stutters...

 

But, as long as I had to admit I only had 8G of RAM, you see, I was excluded from being able to say I was sure SuperFetch didn't do what a fast storage subsystem can.  8 freakin' gigs of memory ain't enough, right?

 

Well, now I have addressed that.  We shall see.

 

I do not recommend anyone buy RAM because they have been lead to believe it will rival the performance of having a truly fast storage subsystem.  I didn't buy RAM because I believe that, and I hope no one else does either.

 

In fact, I pretty much bought the RAM because it's necessary to test this with, and what I expect is to find it doesn't offer that much of an advantage (if any).

 

Finally, I have enough computers at home that I can always use the memory in another machine, but not everyone can say that.

 

Folks, I encourage you to use common sense here:  If you find it unbelievable that a $75 memory upgrade will accomplish what others wind up spending $200 on, congratulations - you've obviously learned to recognize, accept and apply the common sense rule about things that "sound too good to be true".

 

Lothar, we'll just have to agree to disagree.  I enjoy technical pursuits but I'm also a busy person.  I hope one day to post in follow-up after testing. but it will take time and it's honestly not a priority for me ATM.  If your experience is that 8G isn't enough RAM, then I'd suggest maybe taking a hard look at what you're running on your machines (knowing full well you have no intention of listening to my advice, mind you).  And I sincerely mean every bit of this in the nicest possible way - I'm sorry if it seems an affront; I'm just being honest.

 

I believe everyone has a right to decide what makes sense for themselves, and should do so.  I'm posting this now in the hope that people will think for themselves which option(s) make the most common sense to them.

 

Again, if my actions (in sarcasm) caused anyone to make a poor decision here, I deeply regret it, and felt compelled to set the record straight.

 

If you buy memory, I am not saying you'll regret it.  I am saying if you buy memory with the idea that it will somehow replace a fast storage subsystem, you'll probably wind up realizing exactly how adequately it replaces fast storage.  And then, later on, you'll still probably wind up buying fast storage, anyway (shekels being always a factor, naturally).

 

My opinion remains:  If you have a 64-bit OS, a decent CPU and video card, 8+G of memory, only conventional hard disk(s), and some money to spend on an upgrade, find a good sale on an SSD.

Edited June 12, 2013 by Tamper

[Tamper 9]

hood, I can briefly describe the testing I had in mind:  It works out well you're 'shekel-challenged' *lol* for lack of a more polite term :)

 

This way, you get the RAM first, and that's precisely the order I would have suggested:  Using only the RAM first, check to make sure SuperFetch is running (there's a service and a registry entry, I will be glad to post directions if need be).  For the most part if the service is running, SuperFetch is running.

 

Do some testing now, prior to adding the new memory,  Of particular note would be anything that seems to suffer from 'loading issues', like stuttering, or momentary pauses which you sometimes see in the normal course of using your machine.  Web sites aren't the best for this, because the delays are often in the content or provider, not necessarily your machine.  Not just games, either - if you can, use some 'productivity apps' like Office (I can tell you where to get a very good one, legitmately free if you don't have one).

 

If SuperFetch is on, and you've followed your own typical usage patterns with regard to what you're running at what time of day/what day of teh week, you're pretty much testing SuperFetch in the truest, 'real-world' sense.

 

Then, add your new memory.  Notice the same factors as before, trying your best to follow the same patterns you normally would (SuperFetch, after all, is intended to be 'adaptive').

 

One caveat (and another reason I'm not crazy about the caching schemes) is it's all in memory.  If you reset or shut down, the cache has to re-load.  It should still theoretically know, better as you go, what to load based on your behavior, but it still has to load once you restart.  My own recommendation here is to just do what you would normally do, during both test periods, because the technoclogy isn't worth much if it can't work as you need it to.

 

After I dunno, maybe a week or two with the new RAM, maybe you can report back your perceptions,   I'm hoping to do a more empirical, measured test myself, but your testing can help determine the impact from an anecdotal, 'real-world' perspective, which has value of itself.

 

Hopefully it's not too much, largely it's just doing what you do.  Many thanks to you for at least offering to help test.

 

There may be flaws in what I'm suggeting, but if so, I'd imagine someone will be fairly quick to say so - and I'd welcome it, so long as it doesn't unreasonably burden anyone willing to test voluntarily.

 

Thanks :)

[hood 2]

Tamper, your instructions are so clear that I believe even I will be able to manage them.It's is really what I am doing now, but with notes.My new RAM will take anything up to a week to arrive but I will start immediately to note the hiccups and coughs that you describe.

Here's to some interesting results.

[Tamper 9]

I appreciate your willingness to help, hood.  Just make sure SuperFetch is on.  If you can indulge us, the thing to do would be, after the memory is installed and running for a week or so (once you've had a chance to gauge whether SuperFetch is helping or not), then go to the SuperFetch service and turn it off.  (You can do this, at will, and it doesn't hurt anything)

 

If the SuperFetch memory caching was working as advertised, when you turn it off you should note a very dramatic difference, immediately, in the responsiveness of your computer.  Since it's a change for the worse, it should be very obvious to you.

 

Of course, it takes patience and time to do this, but there's not a lot to actually "do"; hopefully it's not a burden.  Once it's all said and done, you'll have all the evidence anyone could ask for as to whether SuperFetch is appropriate to your needs - and I hope, some helpful info for the rest of us, too!

 

Thanks again, and best of luck!

[MudWasp48 3]

Hey Tamper,

Did you change anything with your pagefile and what is your opinion of doing so when using ssd(s)? I find the whole topic a little confusing.

 

Never thought of superfetch before, mine is set to manual

[Tamper 9]

Good question, MudWasp - first, let me say how fascinating I think it is that, since you're not running SuperFetch, you noticed the elimination of stutters by using an SSD.  In fact, I'd bet (if you had a default Windows install) you were running Superfetch, right up until you installed the SSD which (depending on how it was installed) may have been why Windows disabled SuperFetch.  Again, I have a 'default' Windows install other than adding SSDs, and I did not turn off SuperFetch, but it's set to manual on my machine as well.

 

If all this is the way I'm seeing it, then it means you had noticeable stutters before, when SuperFetch was enabled, but they went away when you got your SSD.

 

Tell me:  Were you running a default Windows install on a conventional platter drive before the SSD, and definitely noticed stuttering, etc?

 

Here's what Microsoft had to say about W7 and SSDs; (http://blogs.msdn.com/b/e7/archive/2009/05/05/support-and-q-a-for-solid-state-drives-and.aspx?Redirected=true) - dated, but probably still accurate:

 

Should the pagefile be placed on SSDs?

Yes. Most pagefile operations are small random reads or larger sequential writes, both of which are types of operations that SSDs handle well.

In looking at telemetry data from thousands of traces and focusing on pagefile reads and writes, we find that

• Pagefile.sys reads outnumber pagefile.sys writes by about 40 to 1,
• Pagefile.sys read sizes are typically quite small, with 67% less than or equal to 4 KB, and 88% less than 16 KB.
• Pagefile.sys writes are relatively large, with 62% greater than or equal to 128 KB and 45% being exactly 1 MB in size.

In fact, given typical pagefile reference patterns and the favorable performance characteristics SSDs have on those patterns, there are few files better than the pagefile to place on an SSD.

 

Personally I turned off paging files on all but the conventional drive in my machine, primarily out of concern for writes as Lothar describes.  I note that even Microsoft doesn't come out and say having the pagefile there won't hurt the SSD over time.

 

Even if there is wear leveling, I thought, I'm really only concerned with the SSDs ultra fast random reads, and therefore there's no real need to incur a lot of writes anyway.  I basically wanted to copy a game to the SSDs, and from then on, only read from that drive, as much as possible.  So I don't have a pagefile on any of the SSDs I use.

 

Again, I'd be very interested to hear others' input.

 

I forget, MudWasp, how much memory you have - but you could also do some SuperFetch testing, especially if you have 16G+.  Go turn the Service on, and you should see a significant improvement, if SuperFetch works as advertised.

[MudWasp48 3]

Thanks Tamper,

I was running windows 7 home premium when I first installed my ssd. I did a fresh install using my windows 7 home premium disk.. I have no clue if superfetch was running. At the time I was playing x-plane 9.7 and FSX with lots of addons. There was 32 GB of ram on windows 7 hp which only used 16GB, or at least listed 16gb as usable.

 

After getting the ssd up and running, which was pretty dang easy, I fired up FSX. Now I always had "microstutters" in certain heavy addon scenery areas when flying into them. I repeated one of my favorite flights and they where gone. Same thing for X-plane.

 

Now since then I've scrapped windows 7 home premium and did a fresh install of windows 7 pro, which uses all 32 GB.

 

Today I'm do to recieve a 10,000 rpm wd velociraptor. If you have any experiement ideas for it let me know. I was intending to put FSX and X-plane 9.7 on it. There is also a 2 TB wd black 7200 coming. A various times I've had hdds installed in addition to the ssd. Been repairing some desktop computers for family members lately, mostly dust damage, and put all but one to use in their machines. My case has a hdd swap dock and I kept one 7200rpm 500GB sata 3 for a backup. Now I can put that in another machine.

 

and this cpu cooler...

 

http://www.zalman.com/global/product/Product_Read.php?Idx=416

[Tamper 9]

MudWasp, I think your experience really tells the tale.  It may not seem obvious, but here's what I see (and, whatever you do, be sure to correct anything I say out of order here, because the order in this case is everything):

 

First, you had an existing, default Windows install on a conventional hard disk.  This undoubtedly means Windows would be running SuperFetch (I have confirmed this is default recently through research, and first-hand on the machines I have running W7 on conventional, platter drives).

 

You also had 16G RAM available to the OS, which is established in this thread as the 'cutoff' more or less, for the improvements SuperFetch supposedly provides.

 

While running the conventional drive, which would've had SuperFetch enabled by default, and having 16G RAM, you had "...always had "microstutters" in certain heavy addon scenery areas when flying into them."

 

Then, you got an SSD and did a clean install.  You've confirmed that the SuperFetch service is set to 'Manual', which is completely consistent with Windows installing to a (properly recognized) SSD.  This effectively disables SuperFetch - by design, per Microsoft.

 

And, presto - no more stutters.

 

To review:

a. Using Conventional hard disk, default Windows install w/SuperFetch - stutters.

b. Using SSD, default Windows install w/SuperFetch disabled - stutters gone.

(Same system; and at the time, same OS and same 16G accessible system RAM.)

 

Even though I do intend to continue research into the matter, and encourage others to; as well as continued discussion, I'd say your case proves the point as clearly and concisely as necessary.

 

Thanks for the continued information.

Edited June 14, 2013 by Tamper

[hood 2]

Tamper,

Thanks for the instructions.Yes,I have checked and SuperFetch is running.I have been making notes of my computer's habits and am looking forward to the arrival of my extra 8G RAM.I hope to have some details in a couple of weeks. I must say I do not have a lot of stutters but I will make all my slides as demanding as I can and get into the biggest furball I can in OFF. I only hope I last long enough to make some worthwile findings.

Cheers

[Tamper 9]

Goodness, hood...you're right!  I guess in all the excitement, I have completely lost perspective.  I hadn't taken into account the inherent risk in exposing one's pilots to increased activity.  You are truly a brave man indeed, sir, to face such peril with selflessness and devotion to duty, in the name of science!!

 

I can almost envision the citation now!  "Having encountered overwhelming odds, and confronting his own, almost certain demise, Lt Hood did indeed guide his craft squarely into the face of danger, in order to enhance the chances of having stutters..."

 

Your commitment reflects the greatest credit upon yourself, the air service, and the Empire, sir!

 

(I do indeed hope this doesn't become one of those posthumous affairs...*ughh*)

 

MudWasp offers a hint, though:  TBH I'm not sure which - landscape/terrain or aircraft - have the sort of textures that are more likely to cause this effect.  I honestly would go with landscape, simply because there's more of it, it is literally everywhere, no matter where you go, and it extends (if you'll forgive me) as far as the eye can see.

 

I know I've heard No Man's Land associated with heavy system loads/performance hits for as long as I've had OFF.  However, I've often associated this with the magnificent way OFF has truly recreated the war over the lines - the massive amount of material, men and movement causes an entire system to be loaded (and noticeably lower framerates, but still fairly consistent).

 

I also have, myself, often associated the stutters with other aircraft coming into range - it's very obvious when it happens, so much so that I have been able to associate this with the arrival on-scene (if you will) of particular model.  I'm not the oly one who has noticied this, either.  The distinct impression I get is that, as the other craft get closer, of course, higher levels of detail must be revealed for each craft.  This continues until the point at which the system can't get info from the hard disk to the video adapter fast enough.  It hiccups or pauses, ever-so-slightly, while trying to retreive all the extra data.  A lot like pauses you sometimes see in a YouTube video when the playback rate exceeds the download rate.

 

What's significant about this is that it's not necessarily accompanied by sustained, consistently lowered framerates.  It's just a small, momentary bump - you notice it, but generally by the time you think to look at a framerate counter, the effect is diminished and the framerate 'average' buries the distinctive glitch.

 

So, I'm not sure which is more likely...but I guess we could say if you use both higher A/C counts and 'busier' terrain/surroundings, that's probably the highest odds you have of getting stutters.

 

In any event, I thank you again for your sacrifice and contribution.

Edited June 14, 2013 by Tamper

[MudWasp48 3]

 

MudWasp, I think your experience really tells the tale.  It may not seem obvious, but here's what I see (and, whatever you do, be sure to correct anything I say out of order here, because the order in this case is everything):

 

 

 

HeyTamper,

You have all the details correct. I also made sure my settings in FSX and X-Plane were identical, same for graphics card tweaking software. I actually had the info written down on paper. At that time I did not own OFF

 

 

Hood,

I've a suggestion which would not risk your pilot's life. Try some "free flights" in heavily dense scenery areas. My suggestion is creating a British pilot in the RNAS that has his squad based near London and do some low altitude sight seeing over the city. Another idea is to create identical dog fights in quick combat with as many planes as possible. The "pilot never dies" box can be ticked to preserve the longevity of the pilot for futher testing. Might want to set the weather manually to reduce variables.

 

 

Anyone,

Lothar mentioned in another thread that he believed the max setting for either scenery or terrain was 4, but wasn't sure which one.

Anyone know for sure which it is?

Edited June 14, 2013 by MudWasp48

[Tamper 9]

MudWasp,  It's terrain which should not be set higher than 4, per http://www.overflandersfields.com/FAQ.htm

 

HTH :)

 

Great ideas BTW for testing,

[MudWasp48 3]

Thanks Tamper!

Had a hunch that was it, kinda remember reading it, but couldn't remember where.

[Tamper 9]

hood,

 

One other thing, if you please:  I know we talked about your memory, but I don't recall hearing much about the balance of your setup (CPU, graphics, etc).  This is of particular interest, as you indicate not having a lot of stutters to begin with - I'd like to try to establish why that is.  To that end, if you could also include your OFF slider settings, it would be most helpful and appreciated.

 

These 'microstutters' I've learned are difficult to quantify, some people I think just plain don't notice them (no offense intended to anyone, of course).  For me, it just absolutely destroys immersion to not have what I would call 'fluid' video.  Not just high frames rates, as I explained earlier, but smoothness.  Sometimes they are so minute as to make them unnoticed.

 

When I've seen them, I would say that, as a *very* rough estimate - and even then it depends on a lot of factors - I would usually experience a small stutter or a few during a QC session.  It's not that the graphics are 'choppy' the entire time, either...just every so often there's a distinct pause; genuinely a "split-second" as they say, and it's over just as fast as it came about; the whole thing's usually done in less than a second.

 

Maybe that will help to clarify what I mean when I'm talking about (micro)stutters.

Edited June 14, 2013 by Tamper

[almccoyjr 7]

Hey Tamper,

 

I use SuperFetch whenever I'm on the PC except for gaming; Alacrity shuts it and about 48 other services down. I don't use a SSD, but several WD Raptors in different size configurations and I don't experience micro/stutters or slow downs.

 

I honestly don't see any benefit what so ever in a service that tries to "predict" what cache to preload when running a campaign. That being said, SuperFetch does come in handy when running strict or prescripted training scenarios.

 

So, regarding the test. Why not use Quick Combat and setup a couple of different flights or use some of the existing missions for the test? Incorporate ProcMon to log everything that's being accessed and you should be able, over a short period of time, to log simming, services and memory useage to go along with a subjective view.

 

Just some thoughts.

 

plug_nickel

[hood 2]

Tamper,My rig as follows

Motherboard    Asrock Z77 Extreme 4

CPU                 Intel 3570K 3.4 G

Vid Card          MSI HD7850 2G

HDD                WD Caviar Black 500G

RAM                Corsair 2 x 4G vengeance 1600

PSU                Corsair HX650 650Watt 80+ Gold

Optical Drive  Samsung Black Internal Sata DVD+RW

Monitor          Samsung Syncmaster SA300 22.5 Inch

Case              Bitfenix Raider Black Mid Tower

 

My slidewrs are on 54453. OFF runs around 59/61 FPS and about mid 40's/mid50's on the ground

 

Thanks all for the info on torturing OFF. I will try all suggestions.

Cheers 

[hood 2]

Tamper

I have noticed the very occasional stutter in OFF when low down in a very active situation.I have done much better in FSX with everything turned up, when I get a lot of stuttering.I cannot notice any difference with SuperFetch on or off.I am still running with 8 gig and don't expect my new 8 gig for a few days.I will continue taking notes.

Cheers

[Tamper 9]

Hi Al, as always your suggestions are very relevant and helpful.  I'd have to let the other folks here doing the testing decide how detailed they can afford to make it.

 

Myself, while I'll probably try to work in something more objective in due course, like measurements of some sort(s), I am also of the mind that this is a subjective kind of arrangement by nature.  What I mean is, while I usually want to 'see the math', I've often heard others contend that regardless of what the numbers say, their own observations tell them that (insert whatever subjective findings are being discussed here).

 

What I've decided to do, at least initially, is to install the memory I got (just in time for Father's Day!), enable SuperFetch, and check the task manager/process list/resource monitor figures to see what changes in the way the extra memory is used with SUperFetch on v off (fairly objective observation).  At the same time, I intend to 'just run it' to see if it "feels" any different (obviously subjective) for a period of time; let it train itself around me and what I do (because that's what it's supposed to do).  Then, for the most dramatic comparison, I plan to shut it off, click-boom.

 

I thought about it, and considered that it might be too difficult to notice/observe/follow the change SuperFetch made, particularly during the time it supposedly 'learns' - if I had poor results, it could be I didn't give it time to adapt or whatever.  But, I figure that by letting it do whatever it should for a reasonable time (~couple weeks), then pulling the plug fairly abruptly, I should be able to see the most pronounced difference.  By design, SuperFetch will supposedly improve more over time, so it's not an accurate test to start measuring/observing right after you first turn it on.

 

I also thought to test while basically doing what I would normally, because even if it's subjective, that is the exact environment in which SuperFetch must provide improvement, if it's going to.  My reasoning here is that If it isn't going to help me within the way I use my machine, what good is it?  Again, I may well decide later on to set up/use carefully structured scenarios, to compare the same elements over and over.  However, within this thread, it's been discussed that SuperFetch should have ample ability to cache everything I can throw at it (assuming adquate memory), so I shouldn't necessarily have to follow a strict regimen to see improvement.  While duplicating tests is obviously valid methodology, it's not at all the way anyone uses a computer day to day.  So this is one case where I'm as interested in subjective observations as I am the subjective - at least at first.

 

I also cheated a bit, I purchased memory for my youngest's machine - it currently has 6G, being an X58 chipset.  I plan to install the extra 6G and let him 'test' without knowing what or why.  I already interviewed him, without letting on what it was about, and I can do the same thing with him: Let it run awhile, then shut it off abruptly and note what happens.  I work on, and ask him questions about, his machine all the time, so there's nothing unusual about this to him.  Being a teenager, he's far too concerned with the games themselves to pay attention to how much memory's in the machine - IOW, he won't have reason to suspect memory's been changed, and he damn sure does't know or care what SuperFetch is.

 

For all the objections I can hear, I do realize there is some *ahem* "non-standardized" methodology here, but do keep in mind, I do support and maintenance on computer equipment every day for a living, and have for almost 30 years now.  I know fairly well how to interview users to see if a change 'worked'; and I know how to vet out folks who are imagining things.  *lol* We used to make a big, very visible spectacle out of doing absolutely nothing to an equipment while convincing the onlooking users we were clearly changing something...and then we'd get the biggest laughs out of some of the stories we'd get back over the next few days.  Some would say it was running much better; some would insist it had problems after we 'messed with it'.  In fact, it got to where we could *predict* who was going to complain and who would actually be more accurate in reporting back - thus proving our suspicions that it had more to do with circumstance and the individual users than anything else *hehe*

 

So there it is.  More to follow :)

[Tamper 9]

hood,  hello sir and thanks for getting back with us.  I am glad you have noticed stutters (sounds rude, but you know what I mean *lol*) - this is the very core of what we're discussing here, so it helps that you have actually now seen the behavior we're concerned with.

 

As far as your testing, the only thing I might note is that SuperFetch does take time to work - truth is, I don't know how much.  I base some of my reasoning of testing for 'weeks' because I gathered SuperFetch is aware over time of your computer usage day-by-day and acts accordingly.  So, I thought, let it settle in and 'learn' as it wants to, while you do...what you do :)

 

In that regard, it might be best to leave it on for a while before switching back.  I'm still very early in recording values from the task manager and resource monitor, so I can't even say yet that turning it on most recently has had any effect.  Gotta give it time, though :)