Respectfully GPU speed does impact drawing speed and basic graphic functions, but if you open lots of parts and large assemblies continually during a session nothing, not even CPU speed impacts retrieval speed more than hard drive speed and RAM,(swapping). For example I have an old PIII 450 Mhz with 256 meg of ram, a 32 meg Open GL card, running with a 64 bit (long PCI slot) Dual channel SCSI HBA with twin 160 Mb/sec 10K RPM SCSI drives on each channel. I can pull a 4300+ part assembly off disk in about 8+ minutes, the same assembly on a 1.8 Ghz laptop with a 4000 RPM drive and 512 meg of ram with a 64 meg GPU takes close to one hour. Both machines are running Win2K. True the laptop spins images faster and repaints dwgs faster but getting the drawing into memory is painfully slow on the laptop compared to the old PIII. Obviously, rebuild times are slower on the PIII but if you had a similar processor speeds, RAM, GPU's in the workstation and the laptop, the laptop will always get nailed on drive throughput speed (unless you get a 7200 RPM drive)and that you pay for every time you open the files no matter the graphics speed.
Side note: The new Serial ATA (SATA) drives that appearing (as well as SATA equipped MB's) will seriuosly blur the performance line between medium to high end SCSI and IDE. Maxtor has recently a 10K RPM SATA drive that when installed on the new ASUS MB with SATA onboard the sustained reads and writes to the drive are now in the realm of SCSI 3. Speeds 500 Mb's/sec throughput are not far off for this protocol. And the price point makes SCSI less and less attractive as most new workstation motherboards PCI slots are only 32 bit (short slots), not 64 which defeats high end SCSI HBA and precludes getting the top performance out of them.