I’m looking to upgrade my DAW hardware and was considering a Mac Pro. Given that the cost of more cores is considerable in this product line, is it worth going for more cores at the detriment of chip speed or is buying a quad-core model and spending the money on video and system memory a better solution?
Spending money on video system is useless if you don’t do video-oriented work. Basic video hardware is more than enough for DAW use.
So if not, it’s all about CPU power vs RAM quantity. This is hard question, because it depends on your DAW usage profile. But there is a basic rule:
- If you use lots and lots of sample-base VSTi:s, investing to memory is a good choise.
- If you run loads of algorithm-based VSTi:s, investing into CPU power is a smart thing to do.
- If neither, save the money or use it on beer.
- If both, well … bad luck.
Thanks Jarno. How about cores vs processor speed? The Mac Pro comes stock with either a 3.7 GHz 4-core or a 3.5GHz 6-core at a substantially higher cost. All other things being equal, are the extra cores worth the increased investment despite the reduced chip speed?
Are there any stats on how much difference memory speed actually makes to DAW/sampler performance?
It is an area of which I am fairly ignorant and so I would like to get some idea. As you know, technology that seems to be a quantum leap when considered in isolation (like SSDs) doesn’t necessarily make anywhere near the same difference when it gets included with other components doing real stuff.
I would imagine that while higher speed memory would perform proportionally better when testing bulk reads and writes using CPU memory move instructions, I suspect that doing real-world stuff, the CPU and DMA controllers might be giving memory a lot of chances to buffer up, so that improvements become more marginal.
One test that might give some idea is re-loading up a large patch for a sampler library, mainly because it is not also loading the CPU with plugin calculations. The first load can be ignored because that is limited by the disk access time, whereas subsequent loads are doing it from some sort of cache. There must be no other disk activity between loads, otherwise the cache might be partially flushed, which is required if testing load times for drives.
The test would be comparing different memory modules loaded from cache. If the cache is in computer memory, it should show up differences in DRAM better. If it is on the drive, it might be too slow to allow differences to show up.