Well, the ol' lab Windows PC is more or less fixed up now.
Found a compatible stick'o'RAM to upgrade it from 512+128M to a full gig.
Even at that... running AVR Studio 6 (which I'd never tried before) on it is slow with a capital SLOW.
May be partly due to UltraVNC, but not entirely.
Think an Athlon XP 2600+ with 1GB of RAM is a bit less computer than Atmel's developers have these days?
I suppose I might try installing the old version of AVR Studio, too, and see if it's any faster. Um: now that I've upgraded the firmware on my Dragon, and maybe it won't play nice with the old software...? Well, worth a try anyway.
Update: well, whaddya know! Last AVR Studio I downloaded prior to 6 was 4.18; I hadn't tried 5 at all. So, there'd been plenty of opportunity for bloat to get ahead of me.
Now downloading the last v4 from the archives, and also v5.
Gee, I wonder how new it needs to be to support a tiny2313A. Or that other tiny that's on one of the other boards. At least I know that the old, decently-fast version works with a mega168, which is what goes on the third AVR-based board in the group.
Update 2: AVR Studio 4.19 can't find my installation of the latest WinAVR, so I installed 4.18, which can... and which supports the parts I figure on using at the moment. Also, it helpfully downgrades the Dragon's firmware.
Unfortunately, performance is still wretched. I should maybe try it with VNC turned off, and see if that's somehow the problem. Eventually, I may need to toddle on down to Fry's (or virtually toddle to Newegg) and spend yet more of my business capital on a replacement low-end Windows PC (where this year's low end is several steps above 2004's low end).
Come to think of it, I should probably do that before Windows 8 conquers all. Maybe, even, a refurbished little XP box would make sense... but (a) crap, the local Micro Center, where I'd most likely buy such a thing, is closed, and (b) I should probably be using Windows 7 in the lab anyway, so as to replicate the weird serial-port problems that have been cropping up in the toaster company's lab.
(Cheapest plausible option @ Newegg seems to be about $300, with fairly decent specs, and hardware virtualization support, which is a nice capability to have, even if the included RAM is a bit light for actually using it.)
Meanwhile, the cheap workaround for the sluggish performance is to do the editing on the workstation and just use the lab machine for device programming & debugging, which should be OK as long as I'm not doing a lot of debugging. This has the added benefit of letting me use a familiar editor (JEdit, under Linux so middle-button paste (usually) works).