Subject: [Commentary] Next Gen Macs....The Case for 6+ slot Macs
From: Doug Stringham (by way of Sarge)
Date: Mon, 7 Dec 1998 13:42:12 -0800

    While Apple's move to increase bus speeds, support 64-bit cards, add Firewire and USB is a great move on their part, the two things that they are at issue here (i.e. only three slots and using UltraDMA instead of SCSI) are REAL ISSUES for digital video.

First off I will concede that only a very small percentage of end-users will ever need more than three slots ever. It's a fact.... after all, no need for additional sound cards for end users or joystick serial cards, etc. But in professional video and audio editing we use PCI cards galore, in fact, sometimes six slots ISN'T enough even then.

I will address each issue regarding this argument and point out the reasons as to why a six bus system and SCSI on board are critical not only to pro video/film and audio editing solutions (I'm not talking about small hobbyist studios or set-ups but real video/film production facilities).

THE SIX PCI BUS ISSUE:

#1 - CURRENT SETUP

Currently I am using a PowerTower Pro 225 with six PCI slots. I have two 21" monitors (one Twin Turbo card for each for a total of two), two Media 100 cards (to support xr features), an Adaptec 3940UW dual channel UW SCSI card, and a 10/100 ethernet card. This comes to a total of SIX SLOTS! And I could actually use an additional slot to support more drives that I desperately need (two minutes of video on Media100xr = 1 gig HD space). I've been considering pulling my 10/100 card just to get the additional drive space I need for larger projects and holding media for when we are juggling multiple projects on the same machine. Also, moving over to the Yosemite board means that I have to use 9 additional SCSI addresses to use my existing drives, peripherals that are being supported by the Apple's current internal/external SCSI busses..... I don't have these 9 addresses available as my current 3940 is completely loaded down with RAID drives... (yes.. that IS all channels used!)

#2 - NULLIFYING THE EXPANSION CHASSIS ARGUMENT

The argument that is always brought to me is "Then why not get an expansion chassis?" To put matters to rest, I've already researched this route thoroughly and this is the problem plain and simple....bandwidth. It isn't just the fact I need more slots to plug cards into, but the slots have LIMITED BANDWIDTH. A six bus Macintosh actually has THREE SLOTS x 2.... meaning that there is one PCI controller for for three slots each having it's own bandwidth. The demands for the SCSI controller and the Media100 cards are so high that I can't place them both on the same PCI bus... one has to be assigned to one PCI controller and the other card to the second PCI bus controller. If I were to place them under the control of a single PCI controller, I would experience dropped frames, freezing for a few frames and loss of synchronization... all critical issues in video, film and audio.

Also, an expansion chassis is basically a PCI card that plugs into A SINGLE SLOT. I now have to push the bandwidth of three slots into one, not including the bandwidth the controller has to take care of for the two remaining open slots.

Let me address the argument that is now sure to surface... "but I read that Media100 has approved two chassis' that will work without these problems". True, for the moment... but let me tell you where this industry is going and do some basic math to point out the weaknesses in using expansion chassis' for PCI cards in this industry. Media100 has a solution that will work for CURRENT systems using CURRENT video technologies. Essentially 2:1 compression algorithms using the NTSC standard (640x486 and 720x486 D1 resolutions). Using a set-up like this and Media settings of 300KB/sec (2:1 compression) one can work with an expansion chassis (though the drives do take somewhat of a performance hit) and create/edit video using Media100. But, the industry is rapidly (and I mean very rapidly) moving to uncompressed video and, more importantly, HDTV (high-def TV). These both have HUGE bandwidth and drive space requirements as well as tremendous CPU loads.... current six bus systems aren't even fast enough to support uncompressed HDTV even with the bandwidth split across two PCI controllers.

Now the math:

Media100xr requires a MINIMUM of 26 MB/sec solid. This is using current 2:1 compression schemes. If Media100 were to move to uncompressed video today, it would require a throughput of roughly 50 MB/sec to maintain video stream without hiccupping. This means that Media100 needs 50 MB/sec throughput AND the drives need 50 MB/sec bandwidth as well. Current 33Mhz 32-bit PCI slots have about a 260Mbps throughput (that's bit.. not byte). High-definition television (uncompressed) has a whopping 1.5Gbps throughput requirement! These numbers don't even reflect the additional bandwidth required to run our two Twin Turbo cards or 10/100 card on board.

.....MORE SLOTS VIA A EXPANSION CHASSIS DOES NOT MEAN MORE BANDWIDTH!!!

Here's the bottom line (and it hurts). Not only have many industry professionals realized the bandwidth limitations of a three bus system and the bandwidth limitations of current six bus systems (in regards to HDTV), but they've already moved onto the PC for HDTV solutions now. There isn't a single high-def editing system on the Macintosh, but I can count numerous solutions on the PC... why?... because the PC has already addressed this issue with systems sporting 12 high-speed PCI slots with four PCI controllers on board. The systems are set-up to handle the bandwidth requirements of HDTV. If Apple doesn't create a true media system here in the future, hardcore Mac addicts like myself will find our hand forced and have to move to Windows NT.... something I've been fighting desperately. If Apple loses it's high-end market share, I can almost guarantee that it will have a trickle-down effect into the mid-range and consumer level systems. The fact is that the high-end has driven the consumer end, and benefited the consumer as well. Technologies like QuickTime and QuickTime VR were created with the professional in mind.... and consumers have had the fortunate spin-off benefit from these.

My solution is this. Apple needs to create a true workstation class machine.... not for the everyday Mac user, but for those in the media fields (i.e. film/video). They need to create a system that rivals their PC counterparts not for image, but because they need to be able to compete in this arena before the Mac is relegated to mid-level video editing and finally consumer level only. For me money is not an issue, I'd be willing to shell out $12,000+ for a workstation class Mac that has features like 6 64-bit 100Mhz PCI slots (12 would be best), multiple CPU slots for parallel processing (as I just finished a 50 hour AE render for a PBS special, parallel multi-processing is something I'm dying for), dual-channal UW2 SCSI, four-port Firewire (400Mbps), USB, ADB, 100B ethernet, and 133Mhz+ bus speeds... all with the ease of use and functionality of the MacOS.

To let you all know just how real this issue is, Media100 and AVID have been begging Apple to release a six-bus system because of these exact issues. Apple hasn't moved in that direction and, from a spokesperson from AVID, they don't intend to.... not good news for my company.

Speed is my livelihood and bandwidth is critical to production. To assume that your situation is the same as mine is wrong. Fact is, how long do you think Apple will last if they lose the high-end video production market?.... they are at a risk of losing it right now..... just visit NAB in March and see the REAL MARKET... then you will understand what issues are at hand.

THE ULTRA DMA ISSUE

The second major issue I have with the new Yosemite boards is the dropping of SCSI. Many people are saying "so what?... so you take a small performance hit... but it costs you less". Hmmm.... didn't we as Macmarines always argue that "you always get what you pay for". Isn't that part of the reason that Mac's have cost more in the past? Because Apple kept to top technologies to keep their systems in the forefront of the PC? By moving to UltraDMA we have introduced several new issues/problems. First off, unlike SCSI....UltraDMA places a large percentage of it's load onto the CPU. "Big deal... with my screaming fast G3 it doesn't make a dint in performance really"... hmmmmm.... ever wonder why PC's have been more prone to file failures/corruption than the Mac?.... well... the CPU has always been intertwined with the ATA/DMA controller, and when it locks up during a file save/read, the file runs a high risk of corruption. Fact is, UltraDMA is a consumer solution to demands of "we want more speed for less money"... essentially they've cut the corners to give you more bang for the buck, but at the cost of stability, speed and overall reliability. Secondly, UltraDMA performance/reliability in a RAID situation is majorly problem-prone. Visit last months New Media article where they compare UltraDMA with Ultrawide SCSI-2...... (http://www.newmedia.com).... they conclude that though UltraDMA is fast and comes in a close second behind UW2.... it is problem-prone and the recommend that professionals not use it. UltraDMA for the consumer is find and dandy for me, but I need SCSI, Fiber-channel or a Firewire solution for reliability, burst speeds and throughput. Also, the fact that I have 9 SCSI devices that need to migrate to the new system in addition to my RAID drives means that I either have to give up 9 drives (approx. 81 gigs of usable HD space) or give up the peripherals (which several contain client critical support files, etc.). Thus the dilemma.

Let's not replace SCSI with UltraDMA... but with a Firewire alternative.

I hope this clarifies the argument for six+ PCI slots on the Macintosh and the problems with UltraDMA. If you have further questions or comments... or if I need to clarify it more clearly... please feel free to contact me directly.

Thanks,

Doug Stringham www.impulsestudios.com