Comments

apit34356 wrote on 6/1/2006, 8:23 PM
Coursedesign, Blackmagic is a very aggressive and smart company, and AM2 is AMD's pathway for all future/current design for highend and mass production. AMD has signup just about every OEM/reseller. All highend NLE resellers, ( that use AMD cpus) are already on the bandwagon or trying to sign up. Blackmagic would be passing on a huge market and they are a not a passive company. The real question with Blackmagic is Apple's influence, can Apple's new duo-core demand be high enought to slow new product.

Actually, x2 has now exceeded 5000 rating and AMD has stated 5600 rating is in the queue. x2 5600 samples are available for the luckly few. A small redesign has boost thru-put and kept heat under 135 watts. My understanding of the redesign was associated with the core interface with the cache for x4 and beyond designs.
Coursedesign wrote on 6/1/2006, 10:22 PM
It took BMD quite a while to approve an AMD mobo.

I bought a Supermicro workstation when that was just about the only thing they would accept, and watched with some discreet glee as others were suffering with "almost the same hardware" while I was 100% trouble-free always.

They will approve AM2 mobos someday, the question is when and which one.

135 W is in Prescott territory.... Yuck! Means noise or expensive cooling.

AMD also has a more expensive "low power" AM2 family now that starts at around 35 W, with maintained performance.

They're also coming out with a true x4 4-core CPU in the spring.

At about the same time Intel is coming out with a 4-core CPU that is made up of 2 dual-core CPUs, so it doesn't have full speed communication between all cores, even assuming they had the superior internal bus architecture AMD uses, which of course they don't.
apit34356 wrote on 6/1/2006, 11:41 PM
"135 W is in Prescott territory..", Yea, but that was 135W( these parts will be seconds) was worstcase performance, target median is 115W, low side 91W at max thru-put. The AMD x4 will be a screamer if the cache & memory subsystem can management the thru put rate.

The low power by AMD or Intel is exciting. But Intel's design has always been weak in the instruction pipeline decoding, way too much circuitry executing "what ifs", generating way too much heat. Intel's M chip was probably their best overall design, but then the design work came out of an Israel company, not Intel directly.
Jay-Hancock wrote on 6/2/2006, 3:23 PM
Most of the "gotchas" being discussed here probably result from mixing old, legacy components with new components. That can happen when doing a piecemeal upgrade. Be aware that migrating a low-speed, legacy card onto a new, high-speed architecture can impose a bottleneck and can limit future compatibility as hardware and operating systems continue to develop. You may save a few bucks in the short run, but it can limit the impact of your speedy new parts and it can cause obsolescence problems later.

The intended use of the buyer (epirb) should be a guide. Personally, I don't see any reason to shy away from motherboards that use PCI Express. If your prime, #1 consideration is holding on to legacy PCI-33 audio cards that have known issues when coexisting with PCI Express, then maybe it's advisable to shy away from PCIe. But I wouldn't steer clear of it unless you have a really compelling reason to avoid it.

So what's the tradeoff if you get a mobo with no PCI Express on it. What'll happen?

Well, if you ever want to migrate this fast new system to a 64 bit operating system (like Windows XP x64 or Windows Vista), you are probably out of luck. ATI and Radeon have probably released their last AGP cards, and might not invest R&D to make new 64 bit drivers for them. And the same might hold true for a variety of other legacy components.

And if if falling prices for SATA drives and external enclosures convinces you to beef up your storage, you probably won't have a built-in RAID controller and you'll have to sink more money into an expensive legacy PCI-33 RAID card that only utilizes a fraction of the speed potential of your new SATA RAID drives.

And, speaking about speed, the PCI-33 bus is like a raw, unswitched ethernet hub. Every component divides the bandwidth. With PCI express, every component has its own bandwidth free and clear. Do you really want to buy a fast new system and then choke its I/O subsystem?

Thus you won't be able to upgrade your operating system, and you won't get great speeds from cheap external RAID storage. These are some of the tradeoffs you should consider.

If you really want a top speed new system AND a DAW that uses old PCI cards, you may be wise to have two different computers. That way you'll get maximum speed from one and all the DAW capabilities from the other. But if your only thought is to save a few bucks by recycling an AGP video card (when you could cheaply buy a new dual-head video card), I wouldn't advise it.

Also consider another thing. One big advantage of modern motherboards is they have so many built in components that you no longer need to upgrade a whole bunch of cards. The built-in NIC is a good example. Built in firewire. And the built-in sound card, if you're not needing something special. Many of them even have built-ijn graphics card (though I don't recommend this because it uses up CPU power). The point is, upgrading a $200 mobo is a lot easier than the old days when you had to replace half a dozen components. So why resist the newer, faster architectures? Why impose bottlenecks on your new, fast machine? If you're wanting a brand new, fast machine, I don't suggest doing a half-way upgrade. Use modern parts unless you have a compelling reason not to.
epirb wrote on 6/4/2006, 8:14 PM
Thanks again for all the great advice guys.