64BLOG
Occassional, direct and not politically correct.
Saturday 17 January 2009, 1:36 PM
Which comes first Driver or OS?
So, the question is have you moved, and if so where did you move? Server, Desktop or laptop or are you a techno-hugger and done all three?
Personally, I left it at server. I wanted my last laptop to use Vista 64 Bit, but the manufacturers advice was well you can have it, but most of the features of your laptop won't work as we don't have drivers. This was similar to an experience I had with XP on 64Bit, the support for drivers was pretty dire.
Hardware manufacturers are kind of stuck because they don't want to spend development money on drivers used by a small minority of users, while at the same time, users won't move over to a technology until it's fully supported.
I would love to have Vista 64-Bit on my desktop, as I have a really annoying embedded resource problem that just won't go away and I am sure that Vista 64 would resolve it.
So, the wires are no longer chattering about Windows Vista, instead we are being treated to a heap of pre-marketing about Windows 7. There is a beta, and you can play with it and see if it's any good. What I haven't heard is any speak of 32 or 64 bit versions, is there just one version and its 64 Bit? That would definately accelerate the 64 Bit acceptance.
Problem is, the next line is logically 128-Bit and these processors exist in mainframes today, but will they make it to the desktop? Indeed there is talk from the usual chip vendors that this is feasible, but would compatibility be needed with 32 and 64 Bit? I kind of think that the multicore has been far from exhausted yet and it still has much to offer, before stepping up another power of 2.
Discuss?
PreviousComments on this post
It depends on who you ask. My experience with 64 bit Linux is that OS, drivers and applications largely come all together, and it's the proprietary 64 bit plugins that you are left waiting for. However, having said that, both Adobe and Sun seem to be delivering 64 bit versions of their Flash(alpha?) and Java plugins respectively now, so things are no longer as bad as they were. I recently migrated a Sony VAIO laptop from 32 bit Vista to 64 bit Fedora 10, and the entire process was almost completely painless. Even NVIDIA had stepped up to provide a 64 bit graphics driver.
Mind you, the only realistic use I can see for 64 bit PCs at the moment is to support enough RAM to create a virtual machine or two, so the prospect of 128 bit PCs in the future is leaving me remarkable underwhelmed.
Funny thing is, vista 64 has a 32 bit run time, the real problem is that V64 will only work with signed drivers. Many vendors would have to wait in a lengthy queue for such a signature. So really it's microsoft holding back the deployment of 64 bit computing.
Hey Roger,
Wasn't this the idea that made Vista 64 so secure? My understanding of this is that theoretically it's true, but it comes at such a cost, manufs have avoided it and IIRC the process has been abandoned for W7. Thinking on this I can't think of many third party drivers that have been replaced by malware, but I am prepared to be corrected here. My experience is that malware goes after core windows components.
The 32-bit RT isn't that to allow WOW support? The core is 64 Bit. I must admit the driver (sorry!) for me is the memory support.
I think the memory support for 64 bit machines is a great opportunity being missed right now. I mean, just imagine the multi-tasking joy of a rig with 128GB of ram on board.
It was from Leo Laporte and Paul Thurrots windows weekly that I learned the 32bit run time allowed for full compatibility
Driver signing, although a nice idea, has not worked well enough to prevent chaos and active ignorance. Most users ignore the warning or "click-through" and install drivers from whatever web site they could get a listing on. In this area Microsoft has my sympathy. How does anyone make a working and public security process to ensure that the driver software that is connected to the heart of your operating system isn't malevolent?
Nearly every time I install a new device in XP Pro I would get the message "This is an unsigned....etc" and these are on drivers RELEASED by Microsoft! It still happens with SP3 on the system. So even Microsoft can't comply with its own policy. My guess is that Signed Driver support has gotten better on Vista and Win7 but just barely better. I am still hearing and reading complaints about things that don't work under Vista.
The scary part is that assuming that a cracker gets the user to download a "new" driver to replace the old one and its "click-once" installed, the driver is considered completely trusted. Even in Vista, its not tested again after being accepted by the user/admin.
How could you program a viral or worm signature sweep on a driver that might be changed every few months? I've seen updated Intel drivers for the same chip-set or IO device released every 3 to 4 months. They are also in the habit of releasing one driver "kit" that will update dozens of different chips all in one package. Nvidia and ATI practice software driver releases pretty much in the same fashion.
I think the world missed the driver signing initiative that Microsoft put forward, instead the world saw it as a distinct money grab by MS to make vendors sign their wares and pay for certificates. I have a code signing certificate and it cost me $400 didn't actually break the bank.
I also think there are some shoddy device drivers out there, responsible for a whole pile of BSODs but MS always gets the blame for this. I didn't think MS wrote any drivers other than plain old vanilla drivers that are just enough to get things working, I kind of figured that for all the bells and whistles of a piece of hardware to be enabled it would need the manufacturers driver. Doesn't Microsofts hardware experience extend to keyboards and Mice?
You raise a scary point indeed, but isn't click-once the domain of managed code? Is managed code used to write device drivers? I wouldn't have thought it was fast enough.
I definately think the device driver opportunity has been missed, and for that matter, signed applications. Although, I am sure a security policy could be set up to only execute signed applications, but the headaches this would cause for the infrastructure guys.....
