unRAID OS version 6.5.3-rc1 available - TESTERS NEEDED


Recommended Posts

I recall that one time that changes  were made to the   kernel preemption mode that the low end CPU had some issues.  I have not seen anyone who is running one of these systems actually report in on their experiences.  So if you are running a Sempron, old Celeron, Intel Atom  processors, and others of the ten year old vintage, please give this 6.5.3-rc1 an evaluation.  You want to have a look at file transfers of high bit video as that is where the problems were that time..

Link to comment

Parity check, no change:

 

2018-05-01, 12:45:20 8 hr, 45 min, 19 sec 126.9 MB/s OK 0
2018-04-01, 12:45:09 8 hr, 45 min, 8 sec 127.0 MB/s OK 0
2018-03-01, 12:45:19 8 hr, 45 min, 18 sec 126.9 MB/s OK 0
2018-02-01, 12:45:24 8 hr, 45 min, 23 sec 126.9 MB/s OK 0
2018-01-01, 12:42:56 8 hr, 42 min, 55 sec 127.5 MB/s OK 0
2017-12-08, 23:30:57 9 hr, 23 min, 5 sec 118.4 MB/s OK 0
2017-11-01, 12:41:52 8 hr, 41 min, 51 sec 127.8 MB/s OK 0
2017-10-01, 12:41:32 8 hr, 41 min, 31 sec 127.9 MB/s OK 0
2017-08-29, 09:30:27 8 hr, 41 min, 18 sec 127.9 MB/s OK 0
2017-08-01, 08:42:14 8 hr, 42 min, 13 sec 127.7 MB/s OK 0
2017-07-01, 08:42:41 8 hr, 42 min, 40 sec 127.6 MB/s OK 0
2017-06-01, 08:42:43 8 hr, 42 min, 42 sec 127.6 MB/s OK 0
2017-05-01, 08:42:43 8 hr, 42 min, 42 sec 127.6 MB/s OK 0
2017-04-01, 08:42:38 8 hr, 42 min, 37 sec 127.6 MB/s OK 0
2017-03-01, 08:42:44 8 hr, 42 min, 43 sec 127.6 MB/s OK 0
2017-02-01, 08:45:14 8 hr, 45 min, 13 sec 127.0 MB/s OK  
2017-01-01, 08:45:01 8 hr, 45 min 127.0 MB/s OK  
2016-12-02, 05:40:30 8 hr, 43 min, 42 sec 127.3 MB/s OK  
2016-10-28, 18:22:40 8 hr, 46 min, 32 sec 126.6 MB/s OK
Link to comment

One minor bump on mine, I have a thumbprint scanner hooked to a USB3 hub, that hub is connected to a USB3 Controller, that controller is booted as connected to UnRaid, but then is disconnected and attached to a Win10 VM...  First boot after the upgrade, this thumb scanner didn't connect correctly, but resetting it by unplugging it and reconnecting it to the hub fixed it...  4 full system reboots after this didn't have this issue (no noticeable issues during these boots)

 

I have noticed significant boosts in VM initialization times, as well as a small boost to general VM/system responsiveness...

 

I have attached the diagnostics for that first boot just in case...

qw-diagnostics-20180523-1205.zip

Edited by Warrentheo
Link to comment

Just installed 6.5.3rc1 and can report significant decrease in my Windows 10 VM POST time.  I allocate 8 of 12 logical cores and 16 GB of memory and VM would always take about 30 seconds to POST and become to boot and CPU would spike on all those cores when it did.

 

I have an 8700K with a Z370 motherboard with 32 GB RAM if that helps any.

 

Thank you for the improvement!  I just thought it was normal cause it's a VM configured with a lot of CPU and memory. :)

Edited by nickp85
Link to comment
Just installed 6.5.3rc1 and can report significant increase in my Windows 10 VM POST time.  I allocate 8 of 12 logical cores and 16 GB of memory and VM would always take about 30 seconds to POST and become to boot and CPU would spike on all those cores when it did.
 
I have an 8700K with a Z370 motherboard with 32 GB RAM if that helps any.
 
Thank you for the improvement!  I just thought it was normal cause it's a VM configured with a lot of CPU and memory. smile.png
Do you mean decrease in post time?
Link to comment
Slight but welcome performance boost for my VM on a Haswell CPU. Haven't noticed any .adverse side effects.
 
M/B: ASUSTeK COMPUTER INC. - H87I-PLUS
CPU: Intel® Core™ i7-4770 CPU @ 3.40GHz
HVM: Enabled
IOMMU: Enabled
Cache: 256 kB, 1024 kB, 8192 kB
Memory: 16 GB (max. installable capacity 16 GB)
 
OSX VM, OVMF, Q35-2.10, 4 of 8 logical cores, 8GB RAM, GT710 passed through
 
6.5.2

To Clover boot screen: 18s
To fully booted VM: 55s

Geekbench results: single 4278, multi 8444

 

6.5.3-rc1

To Clover boot screen: 9s
To fully booted VM: 37s

Geekbench results: single 4325, multi 8494

 

Geekbench compare before/after: https://browser.geekbench.com/v4/cpu/compare/8399440?baseline=8399110

Link to comment

Not sure if this is user error or a bug with docker.


If you install a docker, and change the name - the docker is still accessible, but you lose the edit and webui buttons.

 

Reproduce: install "Grafana" docker, rename to "grafana".

I thought it'd be 'neat' to organise mine all to the same, however in doing so I lost all control to everything. 

Link to comment
1 hour ago, marshy919 said:

Not sure if this is user error or a bug with docker.


If you install a docker, and change the name - the docker is still accessible, but you lose the edit and webui buttons.

 

Reproduce: install "Grafana" docker, rename to "grafana".

I thought it'd be 'neat' to organise mine all to the same, however in doing so I lost all control to everything. 

Probable cause for the devs:  The user-template gets deleted when this happens, and doesn't get recreated.

 

Work-Around @marshy919  Remove the container's you want to rename.  Reinstall via Apps, and rename them at that time.  System is fine when the name changes during initial install.

Link to comment

To add some numbers to my marked VM boot improvement... here's the numbers with 6.5.3rc1.  Unfortunately don't have numbers from 6.5.2 but it was a long time to POST.  I'd guess at 30 seconds?

 

Time to Tiano logo: 7.72 seconds

Time to Windows login: 12.82 seconds

Total boot time: 20.55 seconds

Edited by nickp85
Link to comment

I'm seeing a huge decrease with SeaBios as well.

 

Host

Intel Xeon E5-2699 v4

Asrock X99 Taichi

32GB ECC Ram

 

Guest - Linux Mint 18.3

SeaBios

6 CPU Cores / 12 Threads

12GB RAM

Passed Through:

AMD Radeon HD 6870 (Video & Audio)

Onboard Audio

Onboard USB 3.1 controller

2x USB 3.0 PCI-E x1 controllers

 

Boot to Linux Mint Login

 

unRAID 6.5.1             7 minutes, 56 seconds

unRAID 6.5.3-rc1:     2 minutes, 52 seconds

 

I didn't try 6.5.2, i can if needed.

Link to comment

I was having this issue with a Windows 10 VM. The more cores I assign the slower it gets, sadly upgrading to 6.5.3-rc1 did nothing to help the issue. It's not on startup (or maybe it is, but it's undetectable), but it's CPU performance while using the VM.

 

I posted about it here: 

 

 

I followed all the instructions. I have an i5-8400 with 16GB ram (6 cores, no hyperthreading). Isolated cores 0 and 1 to unraid. If I assign only 1 core (core 5), the Excel macro I'm testing solves in 3:19, but if I use 4 cores (2,3,4,5) then it takes 4:02. It's the same on 6.5.2 and 6.5.3-rc1.

 

I suspect I'm not the only one, since I wouldn't have known anything was wrong if I hadn't tested with a complex macro since otherwise the difference is too negligible to tell. This is pretty disappointing since I just upgraded MB, CPU and RAM to be able to work on a Windows VM and keep my unraid server. But I get comparable performance with my notebook.  

 

This is the first time I've run a VM in unraid, so there could definitely be something I'm missing, so if anyone has any ideas I'd love to test them out. Alternatively, I'd love to see people do some Excel benchmarks with 1 and then multiple cores to see if I'm the only one with the problem (unfortunately I can't share the macro since it's stuff from my work). 

 

Edit: After some further testing it turns out Excel is the culprit. In an unbelievable twist disabling multi core calculations actually results in faster calculations in Excel on any computer (and for all the macros I've tested).

Edited by crazygambit
Link to comment
1 hour ago, crazygambit said:

Edit: After some further testing it turns out Excel is the culprit. In an unbelievable twist disabling multi core calculations actually results in faster calculations in Excel on any computer (and for all the macros I've tested).

 

One problem with multi-core computations is how much synchronization that is required. If too much synchronization is needed, then the code may run slower with multiple cores than without synchronization using a single core.

 

Multiple cores works best when the individual cores gets independent tasks that do not depend on each other - then each core can run as fast as possible (or at least as fast as the total available memory bandwidth allows).

Link to comment
5 hours ago, pwm said:

 

One problem with multi-core computations is how much synchronization that is required. If too much synchronization is needed, then the code may run slower with multiple cores than without synchronization using a single core.

 

Multiple cores works best when the individual cores gets independent tasks that do not depend on each other - then each core can run as fast as possible (or at least as fast as the total available memory bandwidth allows).

I get that, but if it detects that while trying to synchronize multiple cores performance is going about half as fast as using only one core, maybe it should just give up and prioritize one core. You have more resources, but decide to use them so inefficiently that now it's taking significantly longer, then you're not very good at your job, let me tell you. This happens on both 64 and 32 bit versions btw. And it performs the fastest if I assing 4 cores to the Windows VM, but let Excel only use 1.

Link to comment
15 minutes ago, crazygambit said:

I get that, but if it detects that while trying to synchronize multiple cores performance is going about half as fast as using only one core, maybe it should just give up and prioritize one core. You have more resources, but decide to use them so inefficiently that now it's taking significantly longer, then you're not very good at your job, let me tell you. This happens on both 64 and 32 bit versions btw. And it performs the fastest if I assing 4 cores to the Windows VM, but let Excel only use 1.

That is a classic symptom of a workload that only uses a single core efficiently in that performance of such applications decrease as the core count increases.   In such cases you get higher performance by using a processor with less cores but with faster individual cores.   The assumption on most modern systems is that there is enough going on in parallel that overall system responsiveness improves even if individual tasks can suffer slightly.   It is also one of the reasons why the PC world has not seen performance increases in line with core counts going up.  

Edited by itimpi
Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.