UPDATE (9/07/11):
Lenovo has made good on their promise and released an updated BIOS which resolves this problem almost entirely. Almost–the performance on a 65W adapter is still slightly lower than that of that on a 90W adapter if enough power consumption applies (my system, for instance, with 8 GB of RAM, an mSATA SSD, standard 320 GB 7,200 RPM HDD, and i7-2620M requires a 90W for the very upper echelon of performance).
Nevertheless, this is certainly a major improvement over the previous situation. Be sure to snag this BIOS update if you’re an X220 owner!
Link: http://forum.lenovo.com/t5/X-Series-ThinkPad-Laptops/X220-cpu-throttling-problem/m-p/529503#M30602
Read on the for the full dirt on the original problem.
As much as we love our X220, in case you haven’t heard, over the last couple of days it’s become apparent that all is not well in the world of this amazing little ultraportable. Namely, if you happened to shell out the extra couple hundred bucks for that nifty i7, and you do anything that significantly stresses your GPU, you’re likely not getting anywhere near the performance you’ve paid for.
Here’s the situation: When the laptop is performing at a level which stresses both CPU and GPU heavily (such as with many modern 3-D games such as Civilization V), the performance is severely gimped. Sometimes it even becomes so serious that the throttling continues permanently until the system is rebooted.
The details
To replicate this behavior on your own system, you need two tools (thanks to Notebookreview’s Petrov for pointing out the initial behavior and the testing needed to unearth it). Those are:
Once you’ve got both of those, it’s easy to provoke. First, make sure your system’s power profile is set to Maximum Performance (this is obviously very important!). You can select the power profile easily from within the Lenovo Power Manager, or, alternatively, if you don’t have that installed, you can go through Windows 7’s Power Settings.
Next, execute HWiNFO32 and drag the System Summary window to the right side of the screen so that it is visible throughout the testing, like so:
HWiNFO32’s main window; currently, all is well
Now open OCCT and click the GPU : OCCT button on the left hand side. Then, click the gigantic green ON button, leaving all settings as they are by default. After one minute of monitoring the idle system, OCCT’s GPU testing window opens, and a wormhole-like effect appears to stress both GPU and CPU:
OCCT and the CPU/GPU extreme throttling
Maybe it’s hard to read those numbers, but here’s what they say: GPU 650 MHz, CPU (Cores 0 and 1) 797 MHZ. If you recall from the recent Intel spec sheets, that’s half of the GPU’s turbo frequency and one-fourth of the CPU’s turbo of 3.2 GHz. In other words, really slow.
Worse yet, under many conditions, the throttling will actually continue indefinitely, with the CPU/GPU never actually clocking back to where it ought to be. The only solution here is a cold reboot (that is, from the power off state).
You might initially be thinking that this is related to heat, but here’s the thing: that possibility has been ruled out. Thanks to monitoring temperatures via programs like HWiNFO32, it’s clear that system temperatures have nothing to do with this behavior. The throttling can occur at any temperature, even those as low as the mid 50’s C. So what could possibly be to blame here?
The power adapter factor
Here’s where it gets even stranger. What’s not currently being reported by most outlets is the strange combination of factors which leads to this behavior. So far, here are the conditions which produce the throttling:
- CPU and GPU must both be heavily stressed (as shown above)
- The Lenovo 65W power adapter must be attached to the system (the 90W adapter does not cause this behavior, but instead produces other problems (more on that in a moment)
- The CPU and GPU never actually reach their target turbo frequencies for more than a split second. What occurs is that the CPU first overclocks to 3.2 GHz, after which the GPU overclocks to 1300 MHz and immediately provokes a drop in both speeds to the above-shown abysmal levels.
Strange situation indeed, and dare we say, one which reeks of possible (hopefully not) power-related firmware problems. It could be a voltage issue where not enough power is reaching the CPU under such conditions to provide for a throttle-free experience, but that doesn’t explain why the behavior evaporates while on battery power. Hopefully this means that it would be possible for a firmware update to reapportion the power in such a way that the CPU receives the proper needed amount and the battery is charged at a slightly slower rate (or something?).
The UltraSlim 90W adapter and related (major) touchpad problems
And that’s not the end of the Lenovo X220 power adapter weirdness: if you happen to have a 90W UltraSlim adapter handy, try plugging it in and then using your touchpad for a while. You’ll notice that the pointer becomes horribly jumpy to the point where it’s practically unusable. A support rep commented to us that the “90 watt adapter is completely incompatible with the X220”, but knowing a little bit about technology immediately reveals that this should never be the case if the system configuration is healthy. Providing more power when less is needed will not hurt the system; it only takes what it can use. Not to mention that the 90W adapter is still listed as a compatible accessory everywhere at Lenovo (including in the tabook) and that it’s actually sold alongside the system as an option.
Now try running the OCCT throttling test with the UltraSlim 90W adapter attached (if you can manage to get the pointer to operate to that point–or you can always use an external mouse). The throttling behavior is absent entirely, and everything works as it should (except for the terrible touchpad issues that this causes). Finally, try using your touchpad while the OCCT GPU test is running: it works perfectly well. What the hell?
What the hell indeed. Lenovo’s support forums have been set on fire with this issue, along with another (earlier reported) one about fan misbehavior and noisy whistling qualities. It sounds like a pretty serious power-related problem (both this and the GPU/CPU throttling, which certainly seem to be inextricably related). Hopefully Lenovo steps up to the plate and can rectify both issues via a BIOS update or some sort of parts replacement–otherwise, they may be seeing an awful lot of i7 X220s returning to their doorstep in the coming weeks.
If you’re having this problem, head on over to Lenovo’s official forums and let them know your plight. Thanks to everyone at the NotebookReview Forums for digging into the details of this problem.
Update 6/3/2011: Lenovo’s forum moderators have chimed in to inform us that they are aware of the problem and are working on a fix. This will likely come in the form of a BIOS update. The timeline could be several weeks, however.
Update 6/28/2011: To date, Lenovo has released one BIOS update (1.17) which claims to correct the perma-throttling, but leaves the larger issue of repeated downclocking to 800 MHz / 650 MHz as yet uncorrected.
We’ll continue following this story and keep you updated.
Steve Schardein is a Managing Editor at DigitalChumps.com and also owner/technology specialist at Louisville, KY-based Triple-S Computers, specializing in intensive system repair and malware removal.
