HD For Indies

Anand's the man.....again.

http://www.anandtech.com/IT/showdoc.aspx?i=2772

A performance review of Woodcrest vs other big boys.

Woodcrest flops his big chod out and urinates copiusly on the competition.

These numbers should serious mop any switcher winger.

The best move Apple has made in decades. Best move I made in jumping on the Apple cart!!!

Lovin' my MBP
 

I was about to comment already on the MOSR rumour, but skipped it for a reason or another. This is not meant as criticism as such, but to clear my own thoughts as well.

Core 2 Extreme, to my understanding, does not equal to "Woodcrest". It is Conroe . What (speculatively, drawing from the past) sets Woodcrest apart is multi-socket support that has not traditionally been offered with non-Xeon chips. It is notable though that AMD is slightly changing the rules by introducing 4x4, but if Intel is to follow or not is unclear. I would not count seeing multi-socket Conroes though.

Given the previews and benchmarks, it seems the Core 2 is very, very nice way to go. There is a potential achilles heel though; memory bandwidth and latency. This is also interesting as the result of having multiple prosessor sockets is not only the doubling of cores - it is also doubling the number of front side buses that connect these sockets to the chipset and memory. For example, with Opterons, once you have two processors you effectively almost double the memory bandwidth. I have yet to see benchmarks that test this reliably on Woodcrests. We know that Blackford (the chipset) should have dual FSBs, but 2CPU for example only saw very minor improvements in memory bandwidth. This might have been a shortcoming of the test program. I have not seen any other tests of that sort.

Okay, why should we care? Honestly, I'm not sure if we should care, really. 2K / HD image handling does require quite hefty amount of bandwidth though, but I can't say if it is a bottleneck for most of us (rules that apply to storage do not apply here; you can't count that stream is x MB/s and then assume it to be enough. Even in a simple case, transfering the data to memory, then to processors, back to memory and finally to graphics card, it is needed to multiply that stream with some value n > 1.). GPU effects do matter as well in this equation.

In any case, interesting times ahead. Also with pricing, I am anxious to see how Apple prices the new towers and how those models are configured.
 

MOSR, which I've read for longer than I'd like to admit, has a fuzzy track record. They write their pages in a strange sort of marketing-speak and often seem to be in the greater realm of speculation.

As to pricing Apple seems to hang on tight to it's price structure between pro and consumer models, with a few drops and increases every now and then.

So I'd think the new Pro Intel tower hardware would be roughly what you would expect based on the current models, but with the usual boost in tech (and a whole new architecture).

A few things are still nagging for pro hardware too – there aren't any third-party PCI Express video cards you can buy just yet and PCIe SATA cards are still getting warmed up with only two models out. So I can't just buy a better video card for Motion for my Quad right now. ATI hasn't shipped one (I prefer them to NVidia as far as color quality), most likely because the installed PCIe base of Macs is pretty niche right now.

Along with video cards there is probably lots of potential hardware lag with all the upgrade options pro hardware is used to, especially with legacy stuff that is pretty common with a lot of pro workstations. I'm mostly thinking of software drivers from third parties, since any conversion and bug-fixing is usually painful for users on deadlines.

Despite all that this transition is moving fast. Remember the 68000 to PPC migration? That wasn't so smooth as far as the time it took. The Intel move so far has been pretty amazing, but I'd love to have things level out enough that I could not have to consider architecture transition issues while I'm trying to crank out video projects.