A number of laptop owners have found to their surprise that despite having machines whose processors support hardware virtualisation, the vendor has chosen not to allow this feature to be enabled in the BIOS. Sony has come under particular fire for this, though they are not apparently alone in disabling the feature.
This has come to prominence recently because - until it announced a patch the other day - Microsoft required hardware virtualisation support to run 'XP mode' in Windows 7.
Since I had just purchased additional memory so that my Dell Inspiron 1520 could run VMs, it occurred to me to wonder whether Virtualbox was actually using hardware virtualisation. Sure, I'd enabled the feature when setting up the VMs and I knew the laptop processor - a T7250 - supported hardware virtualisation (aka VT-x), but was it really enabled?. Up till now I'd just assumed that since Virtualbox was allowing me to select hardware virtualisation, and the processor supported it, that hardware virtualisation was actually being used
WRONG!
A quick google led me to two utilities - one from Microsoft, called havdetectiontool.exe, and a second, securable.exe, from Gibson Research.
These both reported that hardware virtualisation was 'locked off', so despite VirtualBox pretending to use it, it wasn't.
Luckily Dell are not one of the vendors who have locked the feature out; it can be found under the POST options setting in the BIOS.
Once enabled, both utilities reported hardware virtualisation was active and my VMs certainly appear somewhat more responsive. Tests elsewhere have suggested the overhead of software virtualisation is between 25-40%, so it is an important feature if you intend to use VMs. This also implies that XP mode - when the patch is released - will perform considerably more slowly on some machines without hardware virtualisation, although it will run.
Intel's decision to randomly enable VT-x across their processor range is an odd one. With the recent surge in interest in virtualisation, the fact that you have to carefully check which processor supports it and which does not, then double-check that the machine vendor hasn't locked it out anyway in the BIOS, makes this a minefield for the unwary purchaser. Contrast this with AMD, where just about every processor since the Sempron has hardware virtualisation support. Of course, I don't know if any machine vendors have then chosen to lock this out. A quick check on my old Asus M2V motherboard running an Athlon 6400X2 revealed that virtualisation was definitely enabled, which was reassuring.
A particularly baffling - and frustrating - Intel example can be found in the CULV processor pair, the U4100 and SU7300, which apart from a small cache difference, are almost identical. Except that the SU7300 supports VT-x and the U4100 doesn't.
It's hard to get definitive prices for these processors; they aren't sold to end users and the Intel site, which doesn't list prices, has 'buy now' links to Avnet and Arrow which both come up with unknown product codes. Nor does searching these sites for U4100 and SU7300 get you anywhere.
However, some googling suggests that both processors sell for $289 in qty 1000. That information doesn't come from the Intel site, however, and may not be accurate. If the price is accurate, though, its baffling that Intel provide these two, almost identical processors.
What I do know, though, is that the excellent little Samsung X120 subnotebook costs over £100 more equipped with the SU7300 than it does with the U4100. Since both processors run at the same clock rate and there is only a small cache difference (3m vs 2m), the other main differentiator is the support for hardware virtualisation. This seems an odd thing to differentiate on for a subnotebook, which might not be a very attractive virtualisation platform in the first case. So what Intel's up to is somewhat of a mystery. Possibly the idea is that 'consumers' - how I hate that word! - will assume that something called an SU7300 will be *much* faster than something called a U4100. It's hard to say, but either way, Intel really ought to rationalise its product line.
Another concern is simply that the CULV products, with their 5-10W TDP, absolutely trounce the Atom. The Samsung X120 can be picked up in the UK for £399, with 3G of RAM and a 250G hard drive, Wifi and Bluetooth. This makes it hard to justify the cost of some of the higher-end netbooks, which even with a more modern Atom processor (Pinetrail) are still going to be left in the dust by these CULV processors, albeit that they may offer a slightly lower overall power consumption and longer battery life.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment