Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[glynor]

Can you specify a bit more what is wrong with the log? I followed in the wiki (or so I thought) to enable logging on the same page you linked too, and then sent over the log created. Did I send the wrong log or something?

The Zip file you posted contains only the most recent log made by MC.  The Log package created by the Report a Problem button should contain at least three items:

* The current log
* The previous log
* The output of the Reporter function

(And a crash dump if any recent crashes occurred.)

I assume you zipped it manually.

[gvanbrunt]

I think what glynor means is that if you followed the directions he linked to it should create a zip file that contains several files. Some of that information is important to help sort this out.

[glynor]

I think what glynor means is that if you followed the directions he linked to it should create a zip file that contains several files. Some of that information is important to help sort this out.

Yes.

If clicking the Report a Problem button created the Zip file Elvis133 posted before, then something is seriously borked.  I think he zipped it himself because of the filename and the contents.

[flac.rules]

Yeah; i zipped the file myself. Do you only need the diagnostic information about the computer itself, or also the previous log? The reason i ask (and also why i zipped the file manually), is that by exporting the log like that you don't get the part where it uses a lot of time to quit too.

[glynor]

The reason i ask (and also why i zipped the file manually), is that by exporting the log like that you don't get the part where it uses a lot of time to quit too.

That's exactly why the Previous log file is included. 

In fact, I'd say about 80% of the time I look at these, the "good stuff" is in the Previous Log, not the current one, for this exact reason.  That's always the one I open first, for sure.  In this particular instance, I did also want the Reporter output, because we're tracking a few different Hang issues that all seem like they could be related.  So, looking for things like matching driver versions on affected systems could be useful.

Since JRiver can't reproduce the issues, we have to figure out under what conditions they happen, or it is hopeless.

Also, it is possible (though perhaps unlikely) that the long-shutdown times are a result of an actual crash of a particular component, and not just a deadlock.  In this case, we could get a Crash Dump included.  The Crash Dump is gold, because they can open it in the debugger and see exactly what MC was doing in RAM at the moment it died.  So, we're basically praying for one of those.

Always use the Report a Problem button.  Use these steps:

1. Open MC and reproduce the issue.
2. If needed (if it crashes or locks), re-open MC.
3. Immediately go to Help > Logging and click the Report a Problem button.

[flac.rules]

New log

21.22.15 Program start
21.24.30 30% Cpu starts
21.26.28 GUI Hang
21.28.36 GUI recovery
21.29.04 Log created.

[Matt]

New log

21.22.15 Program start
21.24.30 30% Cpu starts
21.26.28 GUI Hang
21.28.36 GUI recovery
21.29.04 Log created.

If the GUI only hangs a few seconds, doesn't that just mean something got a little busy but finished?

I wouldn't think that's a big deal.

[JimH]

Try running the benchmark under Help in MC, then paste the results here.

[flac.rules]

If the GUI only hangs a few seconds, doesn't that just mean something got a little busy but finished?

I wouldn't think that's a big deal.

The timestamps are hour;minute: second, so it hangs about 2 minutes.

[flac.rules]

Try running the benchmark under Help in MC, then paste the results here.

This is the score, I have no clue if the numbers are normal or not.

=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 17,922 seconds
    Single-threaded floating point math... 11,083 seconds
    Multi-threaded integer math... 3,815 seconds
    Multi-threaded mixed math... 1,657 seconds
Score: 551

Running 'Image' benchmark...
    Image creation / destruction... 0,445 seconds
    Flood filling... 1,311 seconds
    Direct copying... 0,598 seconds
    Small renders... 4,885 seconds
    Bilinear rendering... 1,887 seconds
    Bicubic rendering... 0,633 seconds
Score: 2254

Running 'Database' benchmark...
    Create database... 0,465 seconds
    Populate database... 4,424 seconds
    Save database... 0,976 seconds
    Reload database... 0,275 seconds
    Search database... 3,627 seconds
    Sort database... 3,212 seconds
    Group database... 2,024 seconds
Score: 1433

JRMark (version 20.0.27): 1413

[glynor]

New log

21.22.15 Program start
21.24.30 30% Cpu starts
21.26.28 GUI Hang
21.28.36 GUI recovery
21.29.04 Log created.

Thanks for the log.  I'll look at it later tonight.

I can say this, though, just briefly looking...

JRMark (version 20.0.27): 1413

If you just did that benchmark, then you are running a build from two weeks ago.  It is pointless to diagnose an issue that might not exist anymore.  Please update and try again.  It probably won't help, and I'll still look at the logs, but... That's really step one.  Every time.  If you aren't trying with the latest one, we don't know if it is fixed or not already.

If your benchmark was just old and you pasted it in from before, then ignore me.  

[flac.rules]

Thanks for the log.  I'll look at it later tonight.

I can say this, though, just briefly looking...

If you just did that benchmark, then you are running a build from two weeks ago.  It is pointless to diagnose an issue that might not exist anymore.  Please update and try again.  It probably won't help, and I'll still look at the logs, but... That's really step one.  Every time.  If you aren't trying with the latest one, we don't know if it is fixed or not already.

If your benchmark was just old and you pasted it in from before, then ignore me.  

Well, it is the latest stable version according to the program But I tried the newest one, no difference, the problem was also there with 19.

[glynor]

Well, it is the latest stable version according to the program

Stable doesn't mean what you think it means.  It basically means "less frequently updated".  Stable in terms of less frequent updates, and older.  That roughly equates to more thoroughly tested, but not always.  They do try to pick builds that don't generate all kinds of problems (or where big new features are included right away).  But...

As it says right at the beginning of the Troubleshooting Guide:

The version of MC available on the main external JRiver Download Page is not always, or even typically, the most recent build. The versions offered there are generally older builds that have proved reasonably trouble-free for most users. The version there is the same version that will be seen by the built-in Auto Update system in Media Center. Updates to this build only happen occasionally, if for no other reason than it isn't annoying that it pops up the "Update Me" dialog as often.

However, there are also public builds available on Interact. These generally come every week or so (though this timing fluctuates throughout development).

<snip>

It is essential to update to the most recent publicly available build of Media Center before reporting the issue (or trying additional steps to correct it). Otherwise, if the issue has been already addressed, you are only wasting everyone's time.

Jim has posted a detailed description of the way they do development in the past, but... Essentially think of the "Stable" channel as the "Everything is fine and don't bother me as much" channel, and the "Latest" as the "I want the newest stuff right away" channel.  If you are experiencing a problem, especially one like this where they're trying to address it and collect information, always test with the latest build.

It isn't to ignore your issue.  It is to make sure time is used appropriately.  I can't tell you how many times I've gone down the rabbit hole and spent hours upon hours of my own free time helping someone with an issue, when the issue in question was already solved (sometimes weeks ago).  No better way to get me to rage-quit than that.  

And, if the problem isn't improved, it you can always roll back if you prefer.  Sometimes, even if it isn't improved, if it gets worse or changes in some way, that teaches us something.

I'll say this, though... If you've had the problem since MC 19, then something is going on with your system, or you are just trying to push it too hard.  This...

JRMark (version 20.0.27): 1413

Is on the low side.  Not dismal, and MC should run on it okay, but it will have limits.  Transcoding video, for example, is going to be very hard on that box.  High-end audio features may not work reliably, or may cause too much overhead.

If your CPU is maxxed out, then it is maxxed out, and your choices are to: (1) do less or (2) improve the hardware.

[glynor]

For a frame of reference, here's my server:

Running 'Math' benchmark...
    Single-threaded integer math... 3.629 seconds
    Single-threaded floating point math... 2.412 seconds
    Multi-threaded integer math... 1.211 seconds
    Multi-threaded mixed math... 0.723 seconds
Score: 2382

Running 'Image' benchmark...
    Image creation / destruction... 0.485 seconds
    Flood filling... 0.376 seconds
    Direct copying... 0.473 seconds
    Small renders... 1.197 seconds
    Bilinear rendering... 0.908 seconds
    Bicubic rendering... 0.565 seconds
Score: 5494

Running 'Database' benchmark...
    Create database... 0.236 seconds
    Populate database... 1.070 seconds
    Save database... 0.218 seconds
    Reload database... 0.057 seconds
    Search database... 0.906 seconds
    Sort database... 1.139 seconds
    Group database... 0.711 seconds
Score: 4958

JRMark (version 20.0.35): 4278

It is a new-ish, high end machine, but it isn't brand new (the CPU is a generation and a half old).  It's a Core i7 Ivy 3770 running at stock with 16GB of RAM and an SSD boot volume.

You don't need anywhere near that much for basic use, but if you're trying to be fancy... YMMV.

[JimH]

JRMark (version 20.0.27): 1413

That's an OK machine, not a very powerful one.  It might be part of your problem.

[flac.rules]

Seems unlikely that the hardware specs in themselves are the problem.  

It is a brand new quad core Xeon at 3.2 Ghz with 16 Gb of memory.  If it's a pure performance issue, something is wrong, whether with the hardware or the software.

[JimH]

That's interesting.  

Xeon's are usually used for servers, if I'm not mistaken.  glynor will have more to say on this, I'm sure.

Your results are as if one core is carrying the load.

I have a pretty good, not great, notebook with an i5 that has a JRmark of 2700.

=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 5.074 seconds
    Single-threaded floating point math... 3.386 seconds
    Multi-threaded integer math... 3.036 seconds
    Multi-threaded mixed math... 1.981 seconds
Score: 1410

Running 'Image' benchmark...
    Image creation / destruction... 0.643 seconds
    Flood filling... 0.494 seconds
    Direct copying... 0.778 seconds
    Small renders... 1.689 seconds
    Bilinear rendering... 1.857 seconds
    Bicubic rendering... 1.283 seconds
Score: 3262

Running 'Database' benchmark...
    Create database... 0.526 seconds
    Populate database... 1.638 seconds
    Save database... 0.405 seconds
    Reload database... 0.078 seconds
    Search database... 1.338 seconds
    Sort database... 1.259 seconds
    Group database... 0.842 seconds
Score: 3533

JRMark (version 20.0.33): 2735

[glynor]

Seems unlikely that the hardware specs in themselves are the problem.  

It is a brand new quad core Xeon at 3.2 Ghz with 16 Gb of memory.  If it's a pure performance issue, something is wrong, whether with the hardware or the software.

Something is badly broken, probably driver-wise, if those are the specs.  You should be way, way above those numbers.

[JimH]

Bios setting?

[glynor]

Running 'Math' benchmark...
    Single-threaded integer math... 17,922 seconds
    Single-threaded floating point math... 11,083 seconds
    Multi-threaded integer math... 3,815 seconds
    Multi-threaded mixed math... 1,657 seconds
Score: 551

This should be way, way, way higher with a current gen (or even the last few gens) of Xeons at 3.2GHz.

[JimH]

It's as if it's running on one core.

[glynor]

Bios setting?

Could be, though only if it is artificially limited or something.

More likely, with a math score like that, it is throttling, or throwing errors left and right.

[glynor]

It's as if it's running on one core.

Look at the single-core Math benchmark.

It does have a (marked) improvement in the multithreaded test, so it is getting way better performance from the multiple cores/threads.  But it is getting creamed in single-threaded performance.  I assumed it was a Core 2 Duo class or maybe even atom, from the scores.

[glynor]

The CPU is a Xeon E3-1225.  That's a Sandy Bridge based Xeon that was launched in 2011.  It seems to be overclocked a bit, as they were sold at 3.1GHz, though that could be just the default turbo boost (and probably is), though if you have any "auto-overclocking" features enabled in your BIOS, then you need to turn them off.

Something is totally borked.  I'll write up some hardware testing you can do to try to isolate the cause.  But MC's math benchmark is pretty "close to the metal".  That means, it isn't running right.  We'll verify it with some 3rd Party benchmarks, but... Yeah, that's not good.

If it is CPU performance like that, drivers are less likely to help (it should blow those scores out of the water with just plain Windows installed and default drivers).  It could be a particularly nasty driver or piece of hardware attached, but it is more likely a hardware problem (heat, RAM, Power Supply, etc).

[glynor]

Also, should ask...

You said it was new.  The CPU was launched Q2 2011 and is EOL.  Where did it come from?

Nevermind.  I figured it out.

[gvanbrunt]

The low benchmark could also be a result of a server OS. I'm going to try it on a server 2012 R2 at work with some decent but 4 year old hardware and see how it performs.

[glynor]

I made one mistake above, the Xeons in that are more likely newer versions of the E3-1225.  I didn't realize they re-used those model numbers with different versions on the low end Xeons, so I didn't look.

That could be a 1225v2 (Ivy, stock 3.2GHz) or a 1225v3 (Haswell-E from last year).  One of those two probably makes more sense since it's listed at 3.2GHz, and that matches with the later versions.

The low benchmark could also be a result of a server OS. I'm going to try it on a server 2012 R2 at work with some decent but 4 year old hardware and see how it performs.

Unless it is wildly misconfigured for MC, then I can't see how it'd have that big of an impact on the Math scores.  Easy enough to check, with a few basic benchmarks.

Download Hyper PI.  It doesn't require installation, you can just unzip the folder directly to your desktop and run it.
Run the 1M and 32M tests on all cores.

Assuming it is an Ivy or Haswell CPU, the 1M test should complete in around 11 seconds (give or take a second or so) on all cores.  The 32M test should complete in around 9.5 - 10.5 minutes on all cores.  If your numbers are wildly off of that, then something is throttling or mis-configuring that CPU badly (or else it is busted).

If that doesn't reveal an obvious, immediate issue, then it is worth running Prime95 on it and checking the CPU temps, as well as doing a RAM test.

[flac.rules]

It is a 1225v3, 3,2 Ghz is stock, 3,6 Ghz Turbo

I checked with hyper-PI, 1 M was 12,3 Seconds, 32M was 10 minutes 22 seconds. . I have another server with the same CPU and  MS server 2012 R2 (though the "essentials" version, not the full one as the computer with the problems have), it has a different mobo and PSU, but RAM from the same manufacturer (Its the one called server A in the earlier posts).  It ran HyperPi in about the same time, 11,8 seconds. However the JRiver benchmarks give a quite different result (they are pasted in the bottom here)

For fun i tried to check the super-pi-numbers with MC running, it didn't make much of a difference, however if i reproduced the 25-30% CPU-load-condition, and ran super-PI, 3 of the cores used about the same time as before, and the last one around doubled its time. But i guess that is also normal behaviour if another process uses up about a 4th of the CPU power.



=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 7,691 seconds
    Single-threaded floating point math... 4,601 seconds
    Multi-threaded integer math... 2,374 seconds
    Multi-threaded mixed math... 1,331 seconds
Score: 1188

Running 'Image' benchmark...
    Image creation / destruction... 0,344 seconds
    Flood filling... 0,848 seconds
    Direct copying... 0,486 seconds
    Small renders... 1,833 seconds
    Bilinear rendering... 0,845 seconds
    Bicubic rendering... 0,480 seconds
Score: 4549

Running 'Database' benchmark...
    Create database... 0,396 seconds
    Populate database... 3,669 seconds
    Save database... 0,723 seconds
    Reload database... 0,197 seconds
    Search database... 1,824 seconds
    Sort database... 1,535 seconds
    Group database... 1,840 seconds
Score: 2111

JRMark (version 20.0.27): 2616

[JimH]

Are you running Windows Server 2012 on the problem machine?

[flac.rules]

Are you running Windows Server 2012 on the problem machine?

Server 2012 R2 Datacenter

(the other one has Server 2012 R2 Essentials)

[glynor]

It is a 1225v3, 3,2 Ghz is stock, 3,6 Ghz Turbo

I checked with hyper-PI, 1 M was 12,3 Seconds, 32M was 10 minutes 22 seconds.

Running it at the same time as MC's benchmark isn't useful, in this case (except maybe for stability testing).  The 32M score you reported looks OK, but the 1M score is on the low end for that CPU (most of them I looked up were faster).

However, you only reported one.  Was that the average?  I was interested to see the variability between cores.  What were the times for each core?

Are you running Windows Server 2012 on the problem machine?

That's a good question.  Please bear in mind, and I mentioned this before, configuring Windows Server to run MC well is not simple.  It will not run well on a vanilla out-of-the-box Windows Server installation.  At the very least, it requires the Desktop Experience to be installed, and it generally isn't supported.

So, if it is running on Windows Server, what have you done to ensure that it is configured correctly?

[flac.rules]

Running it at the same time as MC's benchmark isn't useful, in this case (except maybe for stability testing).  The 32M score you reported looks OK, but the 1M score is on the low end for that CPU (most of them I looked up were faster).

However, you only reported one.  Was that the average?  I was interested to see the variability between cores.  What were the times for each core?

That's a good question.  Please bear in mind, and I mentioned this before, configuring Windows Server to run MC well is not simple.  It will not run well on a vanilla out-of-the-box Windows Server installation.  At the very least, it requires the Desktop Experience to be installed, and it generally isn't supported.

So, if it is running on Windows Server, what have you done to ensure that it is configured correctly?

It wasnt the same time as MCs benchmark, it was just with MC running (just in case something with running the program causes low benchmark-scores)

The reported scores are average, without MC running, the cores had pretty similar times. I can run it again when i get home to get exact numbers. However it seems that the very low numbers seen in MCs benchmark is not mirrored by the HyperPI-test, which according to you might be below average, but still nothing close to half or less of the expected performance as shown in the MC-benchmarks.

[pfm555]

I am having the same issue of the program hanging. i have I-7 with 16 gb memory.I have a new install of windows 8.1 with all updates for windows. I have AMD 6850 with latest updates. i tried to upload the log but it is 852 kb and 700 is the max for upload. I don't know how to make it smaller.

=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 4.535 seconds
    Single-threaded floating point math... 2.553 seconds
    Multi-threaded integer math... 1.336 seconds
    Multi-threaded mixed math... 0.969 seconds
Score: 2023

Running 'Image' benchmark...
    Image creation / destruction... 0.673 seconds
    Flood filling... 0.711 seconds
    Direct copying... 0.989 seconds
    Small renders... 1.810 seconds
    Bilinear rendering... 1.216 seconds
    Bicubic rendering... 0.817 seconds
Score: 3540

Running 'Database' benchmark...
    Create database... 0.322 seconds
    Populate database... 1.743 seconds
    Save database... 0.374 seconds
    Reload database... 0.055 seconds
    Search database... 1.349 seconds
    Sort database... 1.204 seconds
    Group database... 0.854 seconds
Score: 3644

JRMark (version 20.0.35):

[glynor]

i tried to upload the log but it is 852 kb and 700 is the max for upload. I don't know how to make it smaller.

I don't know if this thread is the place to do it, since your problem isn't likely directly related to Elvis133's issue (which I think we've nailed down is due to poor single-threaded performance within MC on his machine).

But if the log is large like that, it is great because that almost certainly means it contains a crash dump.  The Logging wiki article has suggestions for how to make a public link to the log if it is too big to attach.  I'd post it over here:
http://yabb.jriver.com/interact/index.php?topic=93174.0

[glynor]

It wasnt the same time as MCs benchmark, it was just with MC running (just in case something with running the program causes low benchmark-scores)

Gotcha.

The reported scores are average, without MC running, the cores had pretty similar times. I can run it again when i get home to get exact numbers.

No need, as long as you are sure there weren't any big outliers.

I agree with your assessment, generally.  That 1M time is about 1 full second beyond what I expected (~9% deficit), but there may be other processes and drivers on your machine that have that kind of impact.  The MC benchmark score is not just 9% off.  It is something like 30% of what it should be.

So, that brings us to something that is interfering with MC specifically.  This is pretty likely to be Windows configuration, though there are some other options.

Please answer this:

Please bear in mind, and I mentioned this before, configuring Windows Server to run MC well is not simple.  It will not run well on a vanilla out-of-the-box Windows Server installation.  At the very least, it requires the Desktop Experience to be installed, and it generally isn't supported.

So, if it is running on Windows Server, what have you done to ensure that it is configured correctly?

[flac.rules]

The desktop experience is installed, anything other in particular you had in mind, AFAIK the kernel is pretty similar with win 8, main differences being other restrictions with number of connections, number of cores and so on, and a general setting to prioritize background tasks more.

Although that beeing said, it should be no problem, even for an atom to stream music from my library without hanging, i have no problem with a similar task on a weak older 1,3 Ghz dual-core (and the first minutes MC only uses about 1% CPU, which is more in line with what it should be, and it does show that the pure hardware performance alone isn't the only problem, even if it turns out be be a lot below what it should be).

Could it be that having 4 sound cards somehow causes trouble?

[pfm555]

Sorry Glynor. I thought this  was the Hangs investigation post.

[glynor]

The desktop experience is installed, anything other in particular you had in mind, AFAIK the kernel is pretty similar with win 8, main differences being other restrictions with number of connections, number of cores and so on, and a general setting to prioritize background tasks more.

Windows Server has a number of other limitations and changes from the desktop version of Windows.  For example:
* It handles power management completely differently.
* The driver rules are different (only 64-bit signed drivers are allowed)
* SEHOP and DEP are enabled by default for all processes.
* IE has many enhanced security features which can impact third-party applications that use the IE engine (like MC)
* Portions of DirectX (including Xinput and XAudio) are not included, and must be installed manually.

In general, Windows Server is NOT designed to be used as a workstation.  It is possible to do, but you better really know what you are doing, and be willing to diagnose issues yourself.  That's why we're saying it isn't supported.  You're doing something that Microsoft explicitly says you shouldn't do, and makes it difficult to do.

Although that beeing said, it should be no problem, even for an atom to stream music from my library without hanging, i have no problem with a similar task on a weak older 1,3 Ghz dual-core (and the first minutes MC only uses about 1% CPU, which is more in line with what it should be, and it does show that the pure hardware performance alone isn't the only problem, even if it turns out be be a lot below what it should be).

The poor performance in the benchmark is a symptom of the problem, as are the hangs.  If you fix one, you will almost certainly fix both.  That math performance number (which is what MC is seeing internally) is dismal.  For the record, I'm getting this on my Macbook Pro (an older Sandy Bridge i7) while running MC within Windows in a VM with only 2 CPU threads assigned:

Running 'Math' benchmark...
    Single-threaded integer math... 4.984 seconds
    Single-threaded floating point math... 3.288 seconds
    Multi-threaded integer math... 5.676 seconds
    Multi-threaded mixed math... 3.268 seconds
Score: 1104

Running 'Image' benchmark...
    Image creation / destruction... 0.339 seconds
    Flood filling... 0.597 seconds
    Direct copying... 0.885 seconds
    Small renders... 1.858 seconds
    Bilinear rendering... 2.679 seconds
    Bicubic rendering... 1.726 seconds
Score: 2721

Running 'Database' benchmark...
    Create database... 0.333 seconds
    Populate database... 1.459 seconds
    Save database... 0.195 seconds
    Reload database... 0.058 seconds
    Search database... 1.457 seconds
    Sort database... 1.657 seconds
    Group database... 1.078 seconds
Score: 3447

JRMark (version 20.0.35): 2424

[gvanbrunt]

Is Datacenter licensed? That pretty expensive software for home use. Because if there is an activation hack, that may very well cause all kinds of issues.

My brother felt the need to "save some money" in the past and I washed my hands of his computers while he used those. Constant strange issues, freezes etc. He called all the time for help and eventually I came to the conclusion that his "money saved" was my "money spent", so I refused to help any more. Since he stopped using such things he hasn't had those issues and I hardly ever get a call.

[flac.rules]

Is Datacenter licensed? That pretty expensive software for home use. Because if there is an activation hack, that may very well cause all kinds of issues.

My brother felt the need to "save some money" in the past and I washed my hands of his computers while he used those. Constant strange issues, freezes etc. He called all the time for help and eventually I came to the conclusion that his "money saved" was my "money spent", so I refused to help any more. Since he stopped using such things he hasn't had those issues and I hardly ever get a call.

It is licenced yes.

[flac.rules]

Windows Server has a number of other limitations and changes from the desktop version of Windows.  For example:
* It handles power management completely differently.
* The driver rules are different (only 64-bit signed drivers are allowed)
* SEHOP and DEP are enabled by default for all processes.
* IE has many enhanced security features which can impact third-party applications that use the IE engine (like MC)
* Portions of DirectX (including Xinput and XAudio) are not included, and must be installed manually.

In general, Windows Server is NOT designed to be used as a workstation.  It is possible to do, but you better really know what you are doing, and be willing to diagnose issues yourself.  That's why we're saying it isn't supported.  You're doing something that Microsoft explicitly says you shouldn't do, and makes it difficult to do.

The poor performance in the benchmark is a symptom of the problem, as are the hangs.  If you fix one, you will almost certainly fix both.  That math performance number (which is what MC is seeing internally) is dismal.  For the record, I'm getting this on my Macbook Pro (an older Sandy Bridge i7) while running MC within Windows in a VM with only 2 CPU threads assigned:

I am not using it as a workstation, I am using it as a server It doesn't even have a monitor connect too it in normal circumstances. It could be that the enhanced IE security is somewhat related i guess, I can try to change the browsing engine (And I seem to remember disabling enhanced security mode for IE on server B, where MC seems to run quite normal)

[JimH]

Server 2012 R2 Datacenter

(the other one has Server 2012 R2 Essentials)

Have you tweaked it in any way?  Reduced services, etc?

[flac.rules]

Have you tweaked it in any way?  Reduced services, etc?

It is very close to vanilla, it has some functions installed in addition to what comes out of the box, straight from install, like File storage services, and Hyper-V, but it is quite sparse on that front too.

[gvanbrunt]

So here are the benchmarks from the server at work:

=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 4.427 seconds
    Single-threaded floating point math... 2.635 seconds
    Multi-threaded integer math... 2.405 seconds
    Multi-threaded mixed math... 1.395 seconds
Score: 1749

Running 'Image' benchmark...
    Image creation / destruction... 0.735 seconds
    Flood filling... 1.263 seconds
    Direct copying... 2.557 seconds
    Small renders... 2.117 seconds
    Bilinear rendering... 1.468 seconds
    Bicubic rendering... 0.962 seconds
Score: 2417

Running 'Database' benchmark...
    Create database... 0.424 seconds
    Populate database... 1.951 seconds
    Save database... 0.194 seconds
    Reload database... 0.061 seconds
    Search database... 1.921 seconds
    Sort database... 1.493 seconds
    Group database... 1.111 seconds
Score: 3005

JRMark (version 20.0.27): 2390

This is on a 4 year older server that is running backup software at the time of the benchmarks. So it should have marks much lower than yours. However the math benchmark is higher.

[glynor]

Media Center is a user-land application. You ARE using it as a workstation if you are running MC.  That was my point.

[gvanbrunt]

To add to what Glynor said about server OS's. I do maintain them for a living, so I know what I'm talking about. I would never use one to run MC. I have access to extra licenses etc for free but instead pay for windows 8 to run my main MC computers at home. I'm not saying this is your issue, but I do believe you will get less performance from a server OS as opposed to workstation one.

Just because something runs "headless" (without a monitor) does not mean it should use a server OS. For example we also have some expensive monitors that run in the field to take acoustical measurements and are "headless" most of the time. They cost about 30k each. They run Windows 7 not Server 2008 because they depend on A/V which doesn't work well on server O/S. In fact, the software that runs on them won't even install on a server OS as the manufacturer knows there are issues running on that platform.

[JimH]

It is very close to vanilla, it has some functions installed in addition to what comes out of the box, straight from install, like File storage services, and Hyper-V, but it is quite sparse on that front too.

Anything other than stock could be suspect.

I asked because on some forums, people go a little crazy over shutting down any little piece of the OS that they can, and that certainly can cause problems.

[flac.rules]

Anything other than stock could be suspect.

I asked because on some forums, people go a little crazy over shutting down any little piece of the OS that they can, and that certainly can cause problems.

The things installed in addition to a vanilla server 2012 are

.NET 4.5
Media Foundation
Hyper-V Management Tools
User interfaces and infrastructure (This includes desktop experience)
Windows Powershell
WOW64 support.

[flac.rules]

I checked with the chromium web-browser instead of IE, no difference, still the same problem.

I also installed a Virtual install of win8 on the server, installed MC and ran the benchmark on the virtual machine got the following numbers

=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 6,779 seconds
    Single-threaded floating point math... 2,629 seconds
    Multi-threaded integer math... 2,620 seconds
    Multi-threaded mixed math... 1,473 seconds
Score: 1407

Running 'Image' benchmark...
    Image creation / destruction... 0,348 seconds
    Flood filling... 0,383 seconds
    Direct copying... 0,472 seconds
    Small renders... 1,265 seconds
    Bilinear rendering... 0,860 seconds
    Bicubic rendering... 0,488 seconds
Score: 5765

Running 'Database' benchmark...
    Create database... 0,205 seconds
    Populate database... 5,016 seconds
    Save database... 2,959 seconds
    Reload database... 1,674 seconds
    Search database... 2,297 seconds
    Sort database... 1,658 seconds
    Group database... 2,676 seconds
Score: 1304

JRMark (version 20.0.27): 2825

I guess not a stellar result, but the virtual machine got much better results than the host.

I noticed the install of MC triggered an installation of .net 3.5 on the vitual machine , is this needed? It is not installed on the host OS which runs MC normally.

[glynor]

MC does not require .NET at all to my knowledge.  If so, it would be one of the third-party tools that it uses, but I don't think even that.

If your server still has IE in locked down mode, that would be a good first place to look.  I can see that causing such problems with all of the activities it does that it might have multiple threads stuck (which could, maybe, explain the oddball benchmarking scores).

Those (original, server) benchmark scores are odd.  The multithreaded math test is roughly in-line with what I'd expect from the results (a little low, but in the ballpark).  The single-threaded test, however, is dismal.  If the single-threaded task is running in the main thread (which would make sense as a good benchmark for actual performance) then perhaps that explains both the hangs and the test score.

The VM test is about where I'd expect it to be.