Good point on the DTK benchmarks. I looked and can't find anything official. This thread is entertaining though: http://community.thefoundry.co.uk/discussion/topic.aspx?f=34&t=3836 given that when the G5 Quad dropped it was, as indicated in the announcement, after the transition was announced, and it was (as shown in the benchmarks linked from that thread) over twice as fast as every other G5 shipping.
"killing it" may be a little bit excessive, but the OptiPlex 260/270 seem to be able to get a score of about 1200 with a 2.4GHz chip: http://browser.primatelabs.com/geekbench2/2497253 and the 270 was announced in April 2004. (They are essentially the same platform and used most of the same CPUs.) The particular result is from the "SX270" which is going to be the most thermally constrained of the 270 family and is designed to be mounted behind a 17-inch LCD display.
The best collation of Geekbench scores I've seen is the Processor Browser, but the main Processor Browser excludes powerPC benchmarks, it would seem, because it focuses on GeekBench 3 results.
A dual 2GHz G5 from June 2004 can get about 1922 on geekbench2, compared to that (midrange) Pentium 4, single-socket, ExcelBox, from 2004 which gets 1200 on the same test.
So, a dual socket, dual CPU workstation with a 500-800 watt power supply does a little under twice as a cheap ExcelBox with a 300W PSU.
Although frequency is different, I would argue that if a contemporary 3GHz Intel machine (say, a 270 with a 3GHz chip) can score better than the G5, then that's fair game because they're contemporary and the 3GHz Pentium 4 chip is likely to still be pulling less than the G5, and of course there would have been 3GHz dual-socket systems.
Here's some anecdotes from the time about the DTK, which yes was a 3.6GHz single core with HT, "almost matching or outperforming on some benchmarks" the Quad. http://forums.macrumors.com/showthread.php?t=131111
One note on expansion: THe G5s did have some expansion, I often downplay it a lot but they had three PCI slots and two 3.5-inch disk bays. The physical space needed for the type of separation Apple was going for as well as the heatsinks needed to cool the G5s was sort of massive. It's a fairly good efficient design, considering how much heat there is to disperse in the highest end models, even if it's not exactly space-efficient.
"killing it" may be a little bit excessive, but the OptiPlex 260/270 seem to be able to get a score of about 1200 with a 2.4GHz chip: http://browser.primatelabs.com/geekbench2/2497253 and the 270 was announced in April 2004. (They are essentially the same platform and used most of the same CPUs.) The particular result is from the "SX270" which is going to be the most thermally constrained of the 270 family and is designed to be mounted behind a 17-inch LCD display.
The best collation of Geekbench scores I've seen is the Processor Browser, but the main Processor Browser excludes powerPC benchmarks, it would seem, because it focuses on GeekBench 3 results.
A dual 2GHz G5 from June 2004 can get about 1922 on geekbench2, compared to that (midrange) Pentium 4, single-socket, ExcelBox, from 2004 which gets 1200 on the same test.
So, a dual socket, dual CPU workstation with a 500-800 watt power supply does a little under twice as a cheap ExcelBox with a 300W PSU.
Although frequency is different, I would argue that if a contemporary 3GHz Intel machine (say, a 270 with a 3GHz chip) can score better than the G5, then that's fair game because they're contemporary and the 3GHz Pentium 4 chip is likely to still be pulling less than the G5, and of course there would have been 3GHz dual-socket systems.
Here's some anecdotes from the time about the DTK, which yes was a 3.6GHz single core with HT, "almost matching or outperforming on some benchmarks" the Quad. http://forums.macrumors.com/showthread.php?t=131111
One note on expansion: THe G5s did have some expansion, I often downplay it a lot but they had three PCI slots and two 3.5-inch disk bays. The physical space needed for the type of separation Apple was going for as well as the heatsinks needed to cool the G5s was sort of massive. It's a fairly good efficient design, considering how much heat there is to disperse in the highest end models, even if it's not exactly space-efficient.
