pstorkoh
Contributor
Contributor

VMWare Workstation 11 poor host performance

I've used VMware Workstation for years since I'm frequently called upon to run multiple Server VMs as part of my consulting practice.  After using Workstation 10 for the last couple years I recently upgraded to Workstation 11.  Within a few days after the upgrade I began to notice hesitation while typing, mousing or scrolling in both the host and guest environments.  System activity is not higher than normal during these pauses and nothing ever hangs permanently.  But it makes the system very difficult to use.  Today I downgraded from Workstation 11 back to Workstation 10 and the problems immediately went away after rebooting.  I'm running on a Lenovo W-520 with 3 internal SSD drives and 32 GB RAM on Windows 8.1 Pro, so its not a lack of resources.  I've been running 6 VMs concurrently using about 24 GB total memory in Workstation 10 for several hours this afternoon with no hesitations.  So it really seems to be something in Workstation 11 causing the issue.  Anyone else seen similar issues?

Tags (1)
59 Replies
MCastagno
Contributor
Contributor

Hi,

I'm having the same issue on a Fujitsu Workstation H730 (Intel i7-4510MQ, 32Gb ram, 1Tb SSD Disk) for a Workstation 11 Guest VM (the only installed) configured with 16Gb and 1 processor with 4 cores (main CPU is 1 processo, 4 cores dual thread).

Host and Guest are Win 7 64 Professional.

Host Graphics are Intel HD4600 and NVIDIA Quadro K1100M with 2Gb of dedicated memory.

With Workstation 9 I haven't had any problem, with 11 I've the same "hesitations", seems like a very slow interaction with graphics cards.

The system becomes even more slow when using it via remote desktop.

I've already disabled the "core parking" feature on host and the problem has been reduced (at the first startup the guest sittuation was horrible..)

On host I've tried to "switch" the preferred GPU from Intel to Nvidia for Workstation but nothing changes.

Waiting to see if we're the lonely persons with this problem....

0 Kudos
jhorsman
Contributor
Contributor

Jumping on this thread. I am using a homebrew 5000 with an i7-3770k and 32 gigs of RAM. I used to be able in WS 10 use 17 VM's concurrently with SMB shares on a raid 6 on a NAS  (Crazy, but it works.) with jumbo frames and with mainmem.useNamedFile = "FALSE" in config.ini to eliminate vmem IO. With WS 11 I am having problems with just keeping 4 VMs up without VMware Workstation not locking up and unlocking. (I know it isnt a "Supported" Method) But it is telling me there is performance problems.. I am going to proceed with a roll back until next point release and wait and see if there is improvements. Just jumping in to let you guys know yall aint alone in this one.

0 Kudos
rharry
VMware Employee
VMware Employee

please post the vmware.log

Can you also try downgrading the VM's hardware version and see if that makes a difference (VM->Manage->Change Hardware Compatibility; select compatibility with Workstation 10)

0 Kudos
jhorsman
Contributor
Contributor

You got it.. Including two logs.. One under HW 10 and one under HW 11. I used Server 2012 R2 VM as reference for the log. Tools are current as well.

0 Kudos
rharry
VMware Employee
VMware Employee

is there a difference in behavior between HWv11 and HWv10? (i.e. still using WS11)

0 Kudos
jhorsman
Contributor
Contributor

Working on converting all the windows based VM's back to HW10 right now to see.. But The 2012R2 was faster on HW 10 then 11, So will report back in 10 minutes.

0 Kudos
jhorsman
Contributor
Contributor

Looks like it is back to normal. I am now running like I was before in WS10.. HW11 Has more overhead perhaps?

0 Kudos
rharry
VMware Employee
VMware Employee

Thanks. HWv11 exposes some new capabilities. I suspect this is an issue with 3D graphics. I believe there is already an internal bug report. I'll alert the relevant developers.

bgh89928
Enthusiast
Enthusiast

I use i7-2600k with 32GB ram and runs 6 VMs 24/7.

I upgraded to workstation 11 from 10 about a month ago and I did not have any performance issue like you and many others.

However, I have seen many users complaining about workstation 11's performance.  And I figured out that they are all using Intel's 3rd generation or newer generation CPU like i7-4790 or i7-5930 and more...

I think workstation 11 is not fully optimized for those newer CPUs or maybe some newer CPU's instructions maybe bottlenecking the workstation's performance.  This is something that VMware technical team must take seriously and start investigating the issue.  I myself will build new server soon enough with i7-5930 but I do not wish to have any performance issue at all.

0 Kudos
piggyz
Enthusiast
Enthusiast

Hardware 11 seems to be evil! Now I'm asking why they push it if it is not ready 😞

I said in another post Archlinux seems to work with no issue with hardware 11, then I just discovered even it, even if it run, it is A LOT more slower compared to the same machine at hardware 10 level.

Still, whats the point of hardware 11 even if also my Windows 7 guest virtual machine (they said in the Release not of Workstation 11 the problem could be only under Linux guests) run a lot slower under hardware revision 11?

0 Kudos
MCastagno
Contributor
Contributor

Hi Guys,

just to let You know...

Reverted the guest from 11 to 10 hardware and using it from about 5 hour and it seems that all performance problems, following v11 migration, disappeared.

@RHarry. Before downgrading Vm hardware, I was having a lot of performance problems also using the VM through Windows Remote Desktop connection to host, do You think that 3D graphics was involved too?

0 Kudos
rharry
VMware Employee
VMware Employee

possibly, one other experiment you could try is to upgrade to HWv11 but also disable 3D graphics (VM->Settings->Display; uncheck 3D graphics) and see what the behavior is.

0 Kudos
jhorsman
Contributor
Contributor

To be honest, all my VM's have 3d turned off with the exception of one client OS. I am wondering though. A new version was pushed out this morning and I installed it. Does it make any headway into the issue or do I need to find out?

0 Kudos
pcguy999
Contributor
Contributor

I installed Server 2012 Stnd trial in VMWare Workstation 11.0 on an old AMD 3000+ w Windows 7 x64 w 5GB RAM. This server is not configured with any additional services at all but is fully updated. I upgraded to 11.1.0 in the hope the the performance and boot up of the Server would be more responsive. It was slow as molasses and there is no 3D acceleration enabled. I did a benchmark timing wise on this install. It takes 1:28 to get to the CTL ALT DEL screen and 3:37 to get to the Server Manager screen to show up.

I cloned the install and changed the VMWare compatibility version from 11 to 10 and the times were 1:01 and 3:55 respectively.

I also have the other free virtual machine software installed on this computer. In this software I have the same Server 2012 Stnd trial server installed but this install is configured w Active Directory, DNS and DHCP as well. Its boot time is 1:39 and 2:02 respectively. The last timing is 1:30 to 2 minutes faster than VMWare Workstation.

All three of these installs are configured as 1 CPU 2 GB of memory for the Server. The VMware installs are set for 60GB hard while the other virtual machine is using a 25GB hard drive.

0 Kudos
admin
Immortal
Immortal

HWv11 introduces several new virtual CPU features:

CPUID.01H.ECX.TSC_DEADLINE [bit 24]

CPUID.07H.EBX.TSC_ADJUST [bit 1]

CPUID.07H.EBX.BMI1 [bit 3]

CPUID.07H.EBX.HLE [bit 4]

CPUID.07H.EBX.AVX2 [bit 5]

CPUID.07H.EBX.BMI2 [bit 8]

CPUID.07H.EBX.INVPCID [bit 10]

CPUID.07H.EBX.RTM [bit 11]

CPUID.07H.EBX.FP_SEGMENT_ZERO [bit 13]

CPUID.07H.EBX.RDSEED [bit 18]

CPUID.07H.EBX.ADX [bit 19]

CPUID.07H.EBX.SMAP [bit 20]

CPUID.(EAX=0DH,ECX=01H).EAX.XSAVEOPT [bit 0]

CPUID.80000001H.ECX.ABM [bit 5]

Of these, your i7-3770K CPU supports only TSC_DEADLINE (which we would synthesize anyway) and XSAVEOPT (which we hide on Ivy Bridge CPUs for reasons associated with EVC).  We also synthesize TSC_ADJUST, even though your physical CPU doesn't have it.  In addition, the HWv11 vCPU has 42 physical address bits (2 more physical address bits than HWv10) and it supports more Windows Enlightenments than HWv10.

To see if the problem is related to virtual CPU differences, you can add the following options to a HWv11 configuration file:

hypervisor.cpuid.v3 = FALSE

monitor.phys_bits = 40

cpuid.1.ecx = "----:----:0---:----:----:----:----:----"

cpuid.7.ebx = "----:----:----:----:----:----:----:--0-"

The first line removes the new HWv11 Windows Enlightenment support.  The second line reduces the number of physical address bits back to the 40 bits supported under HWv10.  The third line removes the TSC_DEADLINE feature.  The fourth line removes the TSC_ADJUST feature.  If these configuration options don't solve the problem, then we will have to look at HWv11 platform changes.

admin
Immortal
Immortal

Would you mind posting a vmware.log file from the Workstation 11 VM on the AMD 3000+?  If this is a first-generation AMD-V CPU, which I think it is, then we don't support hardware-assisted virtualization on this part.  If VMware Workstation is using binary translation and the other hypervisor is using AMD-V, that would probably explain the performance gap.

0 Kudos
jhorsman
Contributor
Contributor

Understood. I will proceed with trying the alterations out. I am running a back-up of the entire environment right now since I will alter all of them (I mainly see issues when I scale) Let me know if there is anything in particular you would like from me. I will have results with in 24 hours.

0 Kudos
jhorsman
Contributor
Contributor

Reporting back. It is working as designed now with the tweaks in place with HW11 in the same environment I reported the issue in. (I since moved to linux and instead of SMB, I am using NFS, much faster with network I/O.). I am including another log file in case you wanted it. I do appreciate the time given for this to be looked at. I will file away the vmx tweaks since that helped the inital reported problem with HW11.

0 Kudos
admin
Immortal
Immortal

If you still have that environment, can you check to see if the hypervisor.cpuid.v3 setting alone is enough to fix the problem?

0 Kudos