Hi,
Wao, u had a vip keynote pass and yet went home to spend time with family, guess u needed some rest after the Mac Pro project.
Yeah, can't have everything.
Isn't the lead engineer supposed to be at WWDC explaining and interacting with developers?
Although not the same, I did spend two WWDC's talking with developers at the party they hold at the Apple campus at the end of WWDC. Just like when you visited, for many developers it is their first visit to Apple and they are all taking photos of the campus.
Surprisingly, the developers in general do not hit us for "unannounced stuff" or "will you ever do X feature". They actually tend to tell us how cool the stuff is which is quite flattering.
At WWDC itself, though, most of the developers there are software. Its not a lot of hardware. There are hardware developers for the Mac, of course (PCI Express cards). But since most of the Mac hardware is about open standards (otherwise proprietary, i.e. no one's coming out with new Bluetooth modules) a lot of the development is really software. And I can stay out or woo and aah at the software guys and their development stuff.
Haha. Think I read somewhere tt the Airport antenna is at the bottom of the machine. But isn't tt going to affect the signal quality as the incoming radiation will be shielded by the large mass of aluminium?
It is. The affect depends highly on the degree of reflections - yes, there is no direct line of sight, but you know, signals can reflect off the floor. I half wonder if people will put alum. / steel foil under their Mac Pro's to improve reception!
Looking at test data, though, you might be surprised but the reception is no worse than the prior PowerMac (its actually better).
There should not be any reasonable issue using Airport on the Mac Pro; if it works with a MacBook Pro, it should work with a Mac Pro.
Another boggling question in my mind is whether Mac Pro will support nvidia SLI dual graphics set-up, because this dell machine (1st option under graphics card) using the same 5000X chipset does? Though, from what I know, the 5000X is not SLI compliant so how did Dell manage to do it... Also, will any standard PCI express graphics card bought from retail stores work in the Mac Pro or only the 3 cards offered by Apple will work?
Although nVidia has some specialised hardware for SLI, the basic ingredients are the ability to have two fat (x8 or x16) PCI-E lanes and software support. So from a hardware standpoint it looks possible. The details are whether nVidia (or ATI) have done additional hooks to speed up traffic between the cards (for example, the nVidia cards have an "SLI Connector" that provides dedicated card-to-card traffic). Naturally a large part of this puzzle is drivers and OS - who is going to write the software to allow applications to use it and would this software be only practical for games, or for the greater set of 3D applications (bear in mind the graphics framework in OS X is quite different from DirectX, for example).
Any standard PCI Express card bought from retail stores will work in the Mac Pro with proper firmware and drivers. The PCI Express standard designates the card will identify itself with a particular device ID after which the OS will match dev id and load the driver. In the case of a graphics card, a driver needs to be loaded in EFI in order to display graphics even before the OS loads (i.e. the white background, the Firewire Target Disk Mode graphics). Graphics cards are a problematic issue as it takes a considerable amount of software work to come up with the final shipping-quality firmware and drivers.
It is theoretically possible for a PCI Express graphics card to contain firmware for both Windows/BIOs and Mac/EFI but in reality this is unlikely given the extra expense to develop EFI drivers. (there may also be a space issue fitting both sets of codes on the ROM chip on the graphics card)
Moreover, you need an OS X driver and again due to how closely nVidia or ATI guards their IP, they won't let anyone else but themselves write the driver. Finally in order to write the EFI firmware and OS X driver they need to work closely with Apple too. (in particular the drivers are often not installed separately but for ease for the customer, they are bundled with the OS X version - i.e. the OS X version that comes with the Mac Pro has built-in drivers for the 3 cards).
Even if someone wrote the driver, etc. there is still some need to test. A lot of operations in OS X like sleep/wake and startup require fairly complex operations done by both the firmware and the driver. It is a given that the first attempt to write a driver will have kernel panics, black screen hangs, graphics corruption, etc. due to various bugs in the driver.
In order to iron out this bugs, and build confidence in the combination of hardware and software, a lot of testing over thousands or hundreds of thousands of hours have to be done.
The end result though is pleasantly since ATI and nVidia tend to make their cards quite similar (within a family), a driver written for one card may work for another card (i.e. the nVidia 7300GT and Quadro FX 4500 drivers are quite similar, so is the Radeon X1900 and the Radeon X1600 driver). Still, there exists the tricky issue of the firmware on the card so because of that its probably unlikely you can ever slot in a PC card and have it work unless you also reflash the Mac to use BIOS instead of EFI (which, is again a major undertaking that we don't ever expect an end-user to figure how to because then you can't boot OS X which seems to defeat the purpose of buying a Mac.)
Oh. And just to let you know. the 3 cards offered by Apple are customised for Apple - they are of higher quality than their PC counterparts. (hmm... interesting, so dun use the poorer PC versions )
Another observation I made and am wondering about: Shouldn't the Mac Pro allow up to 32GB of ram instead of the 16GB advertised ( 4GB FB-DIMMs are already available too) and the chipset supports up to 64GB.
You are correct that the chipset supports up to 64 GB using 8 * 8 GB FB-DIMMs. (or using 16 DIMMs of 4 GB each).
It just happens that the 4 GB and 8 GB are so expensive and produce so much heat that we don't think most customers will get to them yet.
Therefore, to limit our test matrix, we've only extensively tested up to 2 GB FB-DIMMs. We never sell anything that isn't extensively tested.
Since the large ram heat sinks are not a standard for the Xserve, won't the ECC FB-DIMMs run too hot in the xeon Xserve which is more cramped inside with less air flow compared to the Mac Pro?
Well, the Xeon XServe has 7 counter-rotating (so dual-bladed) fans that run at a higher rpm to provide more air flow, actually, than the Mac Pro which has 4 fans that run at a lower rpm for quieter operation. A server is allowed to be noisier than a workstation.
Guess thermal issue for the ram is more serious now since they have the AMB chips which gives off considerable amount of heat. So, will it be ok if rams without the heat sinks are used in the Mac Pro?
Yes. The AMB chips aren't THAT hot; they are about 3-6W which is nothing compared to say a processor. However, when you add them up, that's a lot, and of course it all needs cooling regardless.
I don't think I've seen any FB-DIMM RAM without some sort of heat sinking on the AMB.
I've seen some FB-DIMMs where only the AMB has a square heat sink and the DRAM chips are not heat-sinked - which is still just about ok. Most FB-DIMMs today have a heat sink known as a heat-spreader (you could call ours "finned heat sinks"). The spreader works, but obviously the fins have more surface area and dissipate heat better.
The rule of thumb is if its > 1W, it needs a heat sink.
It is ok to have no heat sink because the temperature management is done by monitoring the temperature of the AMB (which has a temperature sensor in it - you can read this detail in one of our tech notes). As long as the temp sensor works, we should be sufficiently cooling the DIMMs but possibly at an expense of acoustics as the fan noise increases beyond what can be attained with efficiently-heat-sinked RAM.
And at some limit, of course, the fans may be maxed out but the DIMMs still overheating. In that case, there will be a fail-safe protection scheme that will shut down the system if the RAM exceeds a safe temperature.