Slashdot Mirror
The "Vista-Capable" Debacle Spreads To Acer
N!NJA writes in with a Register story on a lawsuit filed against Acer for selling Windows Vista on an underpowered notebook. Of course anybody can sue for anything; it will be interesting to see if this action goes forward in the courts. "With a lawsuit filed Wednesday in San Francisco, California, two residents of Fostoria, Ohio seek damages and relief from the world's third-largest computer maker after purchasing a sub-$600 Aspire notebook that included Windows Vista Premium and a gigabyte of shared system and graphics memory. In its official "recommended system requirements," Microsoft recommends that an additional 128MB is required to run the Premium incarnation of its latest desktop operating system. ... Microsoft says that the Premium, Business, and Ultimate editions of Vista will run on 512MB systems — with certain OS features disabled. In the beginning, Redmond called these 'Vista Capable' machines, and it's facing a separate lawsuit over this potentially misleading moniker."
Probably even with shared graphics memory, resulting in something like 448Meg usable? Windows XP SP0 and SP1 ran on 256Meg RAM, SP2 seems to need 512Meg RAM, SP3 seems to need a bit more (but I never tried taht one on low-memory machines). Vista on such a machine? Eeeuh.... I don't think so.
That said, they seem to have paid quite a lot of money to get a RAM upgrade.
Linux runs fine tough on such "low-memory" (I had harddisks smaller than that, like 20Meg!) machines.
Ahhh...the great dumpster continuum. Many a free computer will be found there. -- sowth (748135)
It works fine with 512... Its just incredibly slow!
Aero offloads the GUI onto your graphics card if it is capable of DirectX 9. It provides a faster, tear free interface, and if you notice DWM.exe (Desktop Window Manager) uses only 0-1% of CPU during use.
If you disable Aero and fall back to GDI, DWM.exe will disappear, and explorer.exe instead takes the load, usually using 1-5% of my CPU (at least on this machine).
In general, you should get better performance if you have a decent video card. If you are using the desktop anyways, why not utilize the GPU?
A couple of considerations:
1. Vista uses more GPU ram with each window. If you have a shared memory GPU, it's conceivable that it would be too slow when you start opening many windows. Or if your GPU just doesn't have a lot of RAM.
2. Maybe your GPU isn't as power efficient as using the CPU for rendering the windows. Battery life could be affected.
3. Windows 7 with driver model 1.1 uses a constant amount of GPU ram for any amount of windows (steaming in textures instead of keeping them loaded). It also re-enables GDI 2D HW acceleration which was disabled in Vista, but available in XP. Windows 7 also accelerates Cleartype text.
Posting AC as I'm moderating.
There's "use" and there's "use", and I'm not sure which one you meant. If you're normally a Linux user and used to the way it uses swap, or if you're a tech that's just not familiar with the swap strategy Windows uses, it can get confusing, but see the next paragraph. However, if you mean the machine isn't seeing all the RAM and therefore not using it at all, that's different. One cause of that may be a 32-bit machine with more than 3 gigs RAM, due to the PCI device address space normally found at the top of the 4-gig 32-bit address space. Less RAM than that, perhaps it's as others have mentioned elsewhere, that it's a shared memory machine and the graphics are using part of it.
But as I mentioned, Windows has a much different swap strategy than Linux, but also one that confused me years ago before I ever switched to Linux, as it makes a sort of sense but isn't intuitive. On Windows, as long as swap is enabled, the OS will use it, preemptively preswapping, trying to keep a lot of stuff both in memory and in swap if it can, so if the memory taken by an app is needed by something else, the VMM can simply dump the physical memory copy since it's already in swap, instead of having to spend extra time swapping it out just when the user is waiting for whatever needs the memory to load. This means it's using swap long before it actually has to, to keep the extra copy there just in case. In fact, if swap is enabled, Windows at least /used/ (9x versions, whether it still does I haven't the foggiest) to break if it couldn't do this. I discovered this the hard way at one point when I installed a new CDROM that took the drive letter I /had/ been using for dedicated swap. Fortunately I knew enough about how the 9x versions worked that I was able to boot to DOS mode and mcedit the hidden msdos.sys or whatever file (it has been awhile, I'm no longer sure what file it was) and point the swap at a different device.
The effect of this is somewhat unintuitive. On Windows (again, 9x, but I believe it still applies today), if swap is enabled at all, it must effectively be the size of the physical memory at LEAST, before you get any "extra" virtual memory at all. If you have a swap LESS than the size of your physical memory, Windows may not be able to use the full physical memory at all. Either swap must be made bigger, preferrably double RAM or larger, or on big memory machines that don't really need swap anyway, it should be disabled entirely.
Linux (and presumably most OSs) work more intuitively and rather differently, but theoretically, maybe a bit slower when people are loading something new into physical meory. Swap is always additive. If you have a 2 gig physical memory and a half gig swap Linux machine, it'll let you use 2.5 gig, NOT the half-gig you might be stuck with on Windows. Also, it doesn't so aggressively pre-swap, tho adjusting the swappiness setting [1] can change that. Swappiness defaults to 60. Higher numbers cause it to favor swapping apps out to keep cache, lower numbers encourage it to favor keeping apps and flushing cache, once available memory is full of cache and apps.
I'm running kernel/md RAID, here, 4-spindle, RAID-6 for my main system so two-way striped (with two-way parity), with a swap partition set on each of the four spindles as well, with all four set to the same priority, so swap effectively runs RAID-0. Thus, the 4-way swap is actually faster than (re)loading stuff off the effectively 2-way striped RAID-6 (the other two being parity stripes), so I set swappiness high, 100, thus encouraging the kernel to swap out apps and keep cache data, because swapping in the apps out of the 4-way data striped swap will be faster than rereading the cached data off the two-way data striped RAID-6 main system. That of course is rather the opposite behavior most users, on a no-RAID single-spindle disk, will likely want. They may prefer to set swappiness much lower, say 20, or even 0.
---
[1] Swapp