What Causes "snappiness" in Performance

[ces]

I am testing Asrock E350M1 . Got it for 350zł (about 87 euro , $120) . It feels better than atom in overall "snappines". Much colder than Ion and feels faster.

What causes "snappiness" in performance? Other than hard drive IO from from an SSD.

What other hardware related factors can contribute to increase "snappiness" in performance of a computer, or contribute to a decrease in snappiness?

I have been told that most elements of performance of video display cards and chips are related to 3D performance. But there seem to elements of snappiness that aren't so easy to identify (in order to speed up performance).

For example, I have noticed that when my Idrive software is actively backing up my system. The following occurs:
1. Task manager shows very little impact on CPU usage
2. Launching an MS Word file from windows explorer can take 10 to 15 seconds, from an SSD
3. Launching an MS Word file from MS Word is almost instantaneous
4, Launching a PDF file from either windows explorer or from Adobe is almost instantaneous

There is some kind of processing that is slow when idrive is operating. It isn't my SSD. It isn't my CPU. But something is getting overtaxed.

There are no doubt other aspects of snappiness or lack thereof that are not fixed with a $300 CPU, coupled to a $300 SSD and a $300 video card.

Can anyone tell me what they are? What is falling through the cracks? What the benchmarks are missing?

Someone must know or have some suspicions.

[Fire-Flare]

How much RAM do you have and what's its speed?

[ces]

How much RAM do you have and what's its speed?

4GB of Ram. But the problem is there even when I am only using 1/2 a GB of memory.

I don't remember, but it was the fastest readily available stock speed memory I could get that was on the P7H55 Pro list of compatibility tested memory on the QVL sheet.

I just don't think it is the memory.

[frenchie]

The bus maybe ? It is busy tranferring packets at full speed using all the available resources and as soon as you try to do something else, even if it doesn't show up on the CPU graphs, it throttles down... ? Just guessing...

[ces]

The bus maybe ? It is busy tranferring packets at full speed using all the available resources and as soon as you try to do something else, even if it doesn't show up on the CPU graphs, it throttles down... ? Just guessing...

What is the solution, a better motherboard? A move to an 1155 motherboard? Or?

[ces]

Are there any other performance gating factors, of any nature or type, that are not generally covered by reviews?

[Fire-Flare]

Try using perfmon.exe or resmon.exe They'll show utilization graphs.

[ces]

Try using perfmon.exe or resmon.exe They'll show utilization graphs.

perfmon.exe measures all kinds of stuff, most of which I can't understand. And the is complicated by the problem that it is hard to see which graph line represents which variable.
What do you recommend I do with it?

[Fire-Flare]

Try using perfmon.exe or resmon.exe They'll show utilization graphs.

perfmon.exe measures all kinds of stuff, most of which I can't understand. And the is complicated by the problem that it is hard to see which graph line represents which variable.
What do you recommend I do with it?

Click the green + button and add anything you believe is related to the bottleneck. Then have it monitor while that bottleneck occurs and look for a spike in the graph.

[tim851]

For example, I have noticed that when my Idrive software is actively backing up my system. The following occurs:
1. Task manager shows very little impact on CPU usage
2. Launching an MS Word file from windows explorer can take 10 to 15 seconds, from an SSD
3. Launching an MS Word file from MS Word is almost instantaneous
4, Launching a PDF file from either windows explorer or from Adobe is almost instantaneous

How scientific were your above oberservations? Did you repeatedly try it, is it a reproducable phenomenon?

3 and 4 are counter-indicative of your proposed issue, so the question is really: why was 2 so slow?

Snappiness has always also been a software thing. Example: some video games look like sh*t and perform like it too, other games look like fairydust and run well on ancient hardware. Another example: why was Vista so much less performant than 7? Here too, it could simply be a Windows issue.

Generally speaking, if you perform some intense file system operations, like backup (copying), defragmenting or un/packing archives, all kinds of interferences can arise. Software can lock certain folders for access, e.g. Idrive could be preventing access to folders it is backing up and there are usually timeouts written into Windows and its applications.

I don't think there are any hardware related bottlenecks in this situation. Buses are clocked so high these days, with low latencies and enormous bandwidths, that it is really rare that hardware becomes a bottleneck in a desktop usage scenario.
It's not like only a couple of years ago, when just about everything was attached to the measly PCI-bus, which itself was attached to a laughable front-side bus before signals finally reached high-frequency CPUs, that had next to no memory bandwidth, and high latencies and shared interrupt lines did all they could to quash what little was left of performance.

[dorin_nicolaescu]

ces,

do you have an antivirus running? Is windows search indexing enabled?

I am thinking of processes and/or services that interfere with real-time file system acceess...

Dorin.

[CTT]

As far as I can tell (from personal experience) all trivial tasks are I/O-bound on modern PCs.
CPUs (and motherboards) aren't really a factor and eye-candy (if you're into that sort of things) it's handled with relative ease even by integrated graphics.

My Athlon X2 (underclocked to) 1 GHz and Athlon II X4 2.6 GHz (both with SSDs) feel the same way as far as "snappiness" is concerned (for everyday home/office tasks). Of course when it gets to encoding a media file it's a whole different story.

That being said, if you have an SSD in your configuration (and preferably enough memory), I think it comes down mostly to the configuration of the operating system, but I guess that's a topic for another forum section.

[ces]

[Snappiness has always also been a software thing.

I would propose that everything is always a software issue..... and also a hardware issue... unless either is broken. Doubling the speed of either should double the speed of the system all things equal.

You can pick away minor exceptions with what I just said... but it is basically so.

[ces]

Software can lock certain folders for access,

That I know is not the problem... unless it is concurrently locking up about 40,000 files all at once.

[ces]

do you have an antivirus running?

Yes, but it doesn't affect Word's speed of brining the file up. It just slows down when I use windows explorer to open that same file. And the problem disappears when iDrive continues to run, but is not at that time backing up files.

[ces]

That being said, if you have an SSD in your configuration (and preferably enough memory), I think it comes down mostly to the configuration of the operating system, but I guess that's a topic for another forum section.

That may well be... but it is a peculiar and unique problem. The same process vectored through windows explorer taking 15 seconds but when done using just MS word alone taking something under half a second. The crude solution you would think would be speeding up the hardware bottleneck... which should be easier to identify than the software bottleneck... but neither seem to be readily identifiable.

[ces]

It's not like only a couple of years ago, when just about everything was attached to the measly PCI-bus, which itself was attached to a laughable front-side bus before signals finally reached high-frequency CPUs, that had next to no memory bandwidth, and high latencies and shared interrupt lines did all they could to quash what little was left of performance.

So you don't think the hardware part of the bottleneck has anything to do with the bus or memory?

For the problem to occur, the software has to be saturating some hardware process, doesn't it? It is operating on hardware. So even if it brings the hardware to a crawl, if you can get the hardware to crawl twice as fast, it should get the process to go twice as fast, right?

[dorin_nicolaescu]

Yes, but it doesn't affect Word's speed of brining the file up.

Who knows how that AV works... Turning it off for a while would be an easy way to identify if that the problem or not.

It just slows down when I use windows explorer to open that same file. And the problem disappears when iDrive continues to run, but is not at that time backing up files.

Which proves it's a software problem, rather than a hardware limitation, like bus speed or so.

The crude solution you would think would be speeding up the hardware bottleneck... which should be easier to identify than the software bottleneck... but neither seem to be readily identifiable.

Sure, but spending big $$$ to get a two-fold increase in hardware performance, when you might have a 100-fold perfomance penalty because of a software misconfiguration, might not the smartest way to solve this problem.

Microsoft's Process Monitor can help you trace down file locks.
http://technet.microsoft.com/en-us/sysi ... s/bb896645

[shadestalker]

I'm disappointed people are using the term "snappiness" as if we've all agreed on exactly what it means. Maybe the OP can explain whether this term means a perception of how much work is done over time, immediacy of UI responsiveness, some other factor(s) or a combination thereof.

2. Launching an MS Word file from windows explorer can take 10 to 15 seconds, from an SSD

I take this to mean that, with MS Word not running (and not having just been running, so it's not mostly cached already) you are double-clicking a doc file in MS Explorer. What would be the possible causes of this being slower than

3. Launching an MS Word file from MS Word is almost instantaneous

which I take to mean that you are running MS Word and using File -> Open or such to open the exact same file from the same place on disk.

Assuming in 2 that Word is not already running, you're incurring startup time for the application. It seems unlikely you'd overlook this simple thing, so let's go ahead and assume that in both 2 and 3 Word is already running.

When you double-click a file in explorer, how is that different from using File -> Open within an application? I don't pretend to know, but my guess is that it's possible for explorer shell extensions or such to insert themselves into the process of opening a file, determining its default handler and going about handing off the file to that handler. Have you looked at how many different processes or services may be involved in opening a file via explorer on your system? Have you tested this on a system with nothing loaded except necessary drivers and MS Office? How would that setup differ from what you are testing under now?

[tim851]

I don't know ces, people are trying to be helpful here but you ignore most of the tips of what you should do or what it could be and instead keep formulating your weird ideas.

Two out of three of your tasks were completed instantly, it's obvious that 1) "two out of three ain't bad" (Meat Loaf, 1977) and 2) it's NOT a hardware problem. It seems to be a Windows Explorer problem. That's software.

And it's not like there is a whole lot of influence you have on the hardware side after all. The buses are predetermined anyway. The only things you can choose yourself are CPU type and speed, memory speed and latency, graphics card and hard disk. Since you seem to be set on "improving on your hardware", here's my suggestion:

- get the EVGA SR-2 mainboard (~600$)
- get two Xeons, obiously the E5690 (~1700$ each)
- overclock the sh*t out of them, just to be sure
- get a 24-port SATA-card for PCIe 16x (~500$)
- get 24 Vertex 3 SSDs, RAID-0 (~6000$)
- get 4 GTX580s and put them in Quad-SLI (~2400$), your issue is guaranteed not a graphics issue, but ey...

Now you've assembled the fastest computer you can buy, without going cluster (next step?). You've maximized every aspect that you have actual control over. And who knows, maybe that'll solve your issue. Or maybe not. Maybe for some reason Windows actually calls a timer of 10 seconds. And the ability to speed up time is only anounced for next years processors.

To put it in World of Warcraft terms:

You're a Level 30 warrior. Somebody poisons you, the poison permanently reduces your strength by 50%. Now you can go ahead and spend a lot of experience points on levelling your strength back to where it was before - or you can just find the damn antidote!

[ces]

I don't know ces, people are trying to be helpful here but you ignore most of the tips of what you should do or what it could be and instead keep formulating your weird ideas.

I am trying to just figure out what and why this is happening. That's all. It does have two very simple solutions that do not require an understanding of the problem: (a) just turn off the idrive when I am using the computer, or (b) just use MS word to load word files.

What I am attempting to do is to figure out not only what is happening with my particular situation... but also if there are other situations where hardware is a gating factor to performance, but it isn't obvious.

So it is not my intention to offend when I turn down your fish... I am just trying to learn how to fish, is all.

Does that make sense to you?

[tim851]

but also if there are other situations where hardware is a gating factor to performance

Once again, if Word is loading the files fine and Windows Explorer is not, there is an interference with Idrive and the explorer, not a general hardware problem. Word isn't loading the file through a secret interface (the word bus) or anything. When you double-click a file in Windows Explorer, the Windows shell opens the application the file format is mapped to and tells the application to load it. But for some reason, in your specific scenario, that takes a while. Reasons, remotely diagnosed, could be Word startup time, a timeout (bug), Explorer scanning the file first (for whatever reason), the Explorer having (or respecting) a lower priority in the processor scheduling ... whatever.

If you want to learn to fish - this is it. Finding out why Windows Explorer is slower is like finding out why the fish aren't biting for you but for your friend who sits next to you and uses the same rod and bait. You'd know that the solution is not getting a better rod.

[Arbutus]

RE: "For example, I have noticed that when my Idrive software is actively backing up my system. The following occurs:
1. Task manager shows very little impact on CPU usage
2. Launching an MS Word file from windows explorer can take 10 to 15 seconds, from an SSD"

Task Manager default settings show what CPU horsepower that is left after the core of Windows gets 'first dibs'. Turning on 'Show Kernel Times' gives a clue to this. Many of the constantly running background backup systems, like Rebit/Seagate Replica consume a lot of system resources.

[ces]

Task Manager default settings show what CPU horsepower that is left after the core of Windows gets 'first dibs'. Turning on 'Show Kernel Times' gives a clue to this.

How do I do that? I remember setting that to display on another computer, but I'm not seeing that option on this version of taskmanager.

[HFat]

People have already explained that the 10-15 seconds delay you wrote about is not a hardware issue.

That leaves the "snappiness" question. The observation you quoted in your first post is useless. As has been noted, "snappiness" is subjective and undefined. The comparison hasn't been done scientifically and was reported very poorly if it was done at all.
So here are a few possible explanations for the "snappiness" observation you quoted:
-the difference is made up
-the difference is sincerely reported but imaginary
-the difference is due to unreported differences in software or hardware between test setups
-one of the CPUs compared benchmarked much better than the others (E-350 vs. single-core Atom would do that)
-the CPUs compared perform comparably in multi-threaded benchmarks but the E-350 feels a bit "snappier" because of its superior single-threaded performance (benchmarks show it)

If there is a performance difference between two hardware product, you can capture it with an objective benchmark (you just need to use the right one). If you can't pin the difference down to an objective number, chances are that the supposed difference is actually an artifact of unscientific testing.

[ces]

duplicate

[ces]

the difference is made up

You got me there. There is no getting past you

-the difference is sincerely reported but imaginary

I got me there. There is no getting past me.

People have already explained that the 10-15 seconds delay you wrote about is not a hardware issue.

You can't have one without the other. Performance is always a combination of hardware and software, unless one of them isn't working at all (broken or not functioning at all). Think. Think hard. Think about the sound of one hand clapping. There you go. I knew you could do it.

That leaves the "snappiness" question. The observation you quoted in your first post is useless. As has been noted, "snappiness" is subjective and undefined. The comparison hasn't been done scientifically and was reported very poorly if it was done at all.

Seconds (or subunits thereof), as units of measurement, have always worked for me. What do you use? Or are you saying a number with greater precision, like 12.578531267 seconds would have been more useful to your diagnostic process?

-one of the CPUs compared benchmarked much better than the others (E-350 vs. single-core Atom would do that)
-the CPUs compared perform comparably in multi-threaded benchmarks but the E-350 feels a bit "snappier" because of its superior single-threaded performance (benchmarks show it)

I think that makes a lot of sense. A lot of sense.

If there is a performance difference between two hardware product, you can capture it with an objective benchmark (you just need to use the right one).

Agreed.

If you can't pin the difference down to an objective number, chances are that the supposed difference is actually an artifact of unscientific testing.

Huh?

[tim851]

Performance is always a combination of hardware and software, unless one of them isn't working at all

I'm gonna try this one more time:

You say Word loads the file instantaneously. Conclusion: software + hardware seem capable of instantaneous execution.
You say Explorer needs 10-15 seconds. Conclusion: since the hardware stayed the same, the software must be lacking.

Think. Think hard. Think about the sound of one hand clapping.

That's the sound of one hand slapping against a forehead. It's what you've been hearing in this topic a lot...

[HFat]

The scientific method isn't about excessive precision. It's about sufficient precision. And a lot more besides, like repeating the observations and changing the conditions that could impact the result.
If the loading times are like this (seconds):
10
10
10
11
10
That would look like a timeout.
But if they're like this:
8
11
9
12
13
Then it would look more like an I/O issue (disk or network possibly if this Idrive software is transfering stuff on a slow link or something). Windows has trouble prioritizing I/O under load and some SSDs seem to have a very poor latency under load (when running Windows anyway). A quick look at the appropriate Windows monitoring tool could rebut this hypothesis.
Things which might be worth checking and reporting properly before wasting too much time would include:
-load times for a small text file associated with Notepad or some other small program
-any difference in load times when MSWord is open or closed prior to double-clicking an MSWord file in Explorer
-any difference in load times when typing something like "msword filename" at the console as opposed to double-clicking
-load times when Idrive is idle
In any case you could say the software is broken. It's not a hardware issue. It would be inefficient but you could throw hardware at it if it's a disk issue. If it's a timeout or a network issue, no amount of hardware will fix it. But fixing the software would definitely fix it.

[ces]

That's the sound of one hand slapping against a forehead. It's what you've been hearing in this topic a lot...

TIM851 All is not lost. There is one thing you and I can probably agree on... it is that one of us has critical thinking skills substantially superior to that of the other.