MacBookPro (8,1) sluggish and kernel_task hogging the CPU

Hi,

I have a MacBook Pro 8.1, an early/mid 2011 13" MBP with a 2.7Ghz 2*2 core i7 and 12Gb RAM running OS X 10.9.5 with the latest updates that until recently used to be plenty fast for just about anything I threw at it. It still is, but it seems that this is only as long as I'm not doing anything video-intensive. When thing get sluggish and I launch the top command, I always see the kernel_task eating between 200 and 300% CPU.

Until recently, I could easily run a VM in VirtualBox or Parallels Desktop 9 with 4Gb RAM allocated, and not suffer particular and unexpected slow-downs in other tasks. Today, I tried to recreate a VM running off a USB-3 disk mounted via Belkin's Thunderbolt Express Dock, both in VirtualBox and in Parallels. That failed (I'm sure I managed before, but with the disk mounted via one of the MBP's own USB-2 ports), and after quitting Parallels I started noticing the tell-tale signs of system hogging.

This for instance is with a PC-BSD VM booting in VirtualBox (with scrolling as if over a dial-up connection...):

[code]

Processes: 297 total, 14 running, 6 stuck, 277 sleeping, 1312 threads 21:01:34

Load Avg: 16.42, 11.16, 7.31 CPU usage: 18.6% user, 81.93% sys, 0.0% idle SharedLibs: 1060K resident, 0B data, 0B linkedit.

MemRegions: 73304 total, 5323M resident, 165M private, 1205M shared. PhysMem: 10G used (2149M wired), 1878M unused.

VM: 744G vsize, 1026M framework vsize, 0(0) swapins, 0(0) swapouts. Networks: packets: 14625496/18G in, 7467926/4743M out.

Disks: 3110577/96G read, 634417/78G written.

PID COMMAND %CPU TIME #TH #WQ #PORT #MRE MEM RPRV PURG CMPRS VPRV VSIZ PGRP PPID STATE UID FAULTS

0 kernel_task 259.8 02:06:24 113/4 0 2 N/A 975M N/A 0B 0B N/A N/A 0 0 running 0 31747

10797 VirtualBoxVM 50.1 01:20.30 34/2 0 319 N/A 437M N/A 68K 0B N/A N/A 10797 10794 running 505 163634+

277 WindowServer 11.4 11:30.87 4 0 660 N/A 162M N/A 33M 0B N/A N/A 277 1 sleeping 88 1501536+

372 Dock 10.1 01:15.28 4 1 271 N/A 77M N/A 132K 64K N/A N/A 372 273 sleeping 505 499710+

[/code]

A minute or so after quitting another VirtualBox VM (Ubuntu 15.04) that I left running for a while I'm still seeing this (and I can barely type):

[code]

Processes: 290 total, 14 running, 2 stuck, 274 sleeping, 1434 threads 21:18:38

Load Avg: 39.88, 31.18, 20.76 CPU usage: 16.39% user, 83.60% sys, 0.0% idle SharedLibs: 96M resident, 0B data, 20M linkedit.

MemRegions: 72889 total, 4988M resident, 159M private, 1139M shared. PhysMem: 10G used (1666M wired), 2087M unused.

VM: 728G vsize, 1312M framework vsize, 0(0) swapins, 0(0) swapouts. Networks: packets: 14632687/18G in, 7474719/4745M out.

Disks: 3130270/97G read, 643572/80G written.

PID COMMAND %CPU TIME #TH #WQ #PORT #MRE MEM RPRV PURG CMPRS VPRV VSIZ PGRP PPID STATE UID FAULTS

0 kernel_task 304.7 02:49:55 111/4 0 2 N/A 842M+ N/A 0B 0B N/A N/A 0 0 running 0 31747

10900 Google Chrom 13.9 00:48.12 10/3 1 96+ N/A 143M+ N/A 0B 92K N/A N/A 1597 1597 running 505 57057+

277 WindowServer 9.0 12:33.28 6/1 2 629 N/A 156M N/A 37M- 0B N/A N/A 277 1 running 88 1573465+

1628 Google Chrom 7.6 31:10.10 13/1 1 104+ N/A 301M+ N/A 0B 132K N/A N/A 1597 1597 running 505 487633+

1607 Google Chrom 6.4 12:05.72 5 1 116 N/A 49M+ N/A 0B 0B N/A N/A 1597 1597 sleeping 505 975889+

[/code]

Any idea what's going on here? Is this somehow related to my use of the Express Dock (1st gen.) , which mostly serves to connect my 1920*1080 external display (doesn't appear to be)? (I did notice that that resolution seems to stress the Intel HD 3000 GPU more than the 1680*10xx panel I had before.)

Is something going south on me (RAM, maybe?) Have there been recent 10.9 updates that cause this? I'd think of the recent EFI update, but I'd already noticed the symptoms before.

One thing I've had I think with all 10.9 versions I used is recurrent crashes of the SystemUIServer. Those seem benign enough though.

Additional observation: switching spaces via the keyboard shortcuts does (sometimes) show more glitching than it usually does, even when I don't have Google Chrome running with its whole lot of windows and tabs (many of which are suspended through the "Great Suspender" extension, btw).

Rebooting doesn't solve a lot, and there's nothing in the system logs that seems to be relevant and coincide with the issue when it triggers. Earlier, just after a reboot, I switched to and from another ICC profile for the external screen, and after that switching spaces was smooth as I'd expect it to be, but it's impossible to say if there was a causal effect, or if whatever was going before on simply finished. I also I tried doing an SMC reset following the guidelines for Mac portables without removable battery. I can't tell if that actually did anything; for one, the Energy settings that are supposed to be reset by the operation weren't.

Kind of makes me wonder if I shouldn't try to go back to running 10.6.8 for a while (with all the MacPorts-related inconveniences that would entail) ...

BTW: there's a heatwave here, even when hardly doing anything CPU temp sits above 70°C ... but that's never been an issue before. Curiously, the temp. doesn't rise when kernel_task hits the roof!.

Thanks for any advice/thoughts...