- Last Updated on Thursday, 07 November 2013 12:32
Problems I ran into are worth mentioning too, so you know what you may encounter during your build.
First of all, the G.Skill memory came with fans to cool the modules, but there was no way I could install them with the Noctua CPU cooler in place.
Second issue during the install was the fans of the Chenbro drive cages, they were far too loud for my liking. Starting the machine created a noise that makes you feel at home on O'Hare during the morning rush hour. So I had to exchange them for Noctua NF-R8 fans and reversed their airflow to blow out air, with the specific consideration that we have two cats, so blowing out the air prevents sucking in cat hairs around the disks.
It is of course nice to start a new machine with the latest OS available, Windows 8 64 Enterprise, but the drawback is that a lot of drivers and utilities are still missing. No printer drivers for my printers, no Symantec Endpoint protection, no PerfectDisk, etc. Also lacking are all the Asus utilities like SSD caching, but I expect them to be available in the next weeks, with the release of Windows 8 for the public.
Meanwhile I now know that of the total cost of this new system, 50% went into the disk setup, 19% into the case, PSU and drive cages, 20% into CPU, memory and motherboard, 7% into the video card and the rest into miscellaneous stuff like fans, cooling paste, connectors and the like. I know, I am very much a disk oriented person, so with around 50% of my total investment going into the disk setup should not come as a big surprise, but it may mean that for a more modest disk setup, one should still count on spending around 30% or more of the total budget on the disk setup.
Was it worth it?
The system is very fast. My old system was not really slow, but this system is way faster. On the
it has now achieved one of the top ranks and it could have been even faster, but that would mean reverting to artificial measures. I tested with the settings that I'm comfortable with for editing, not extreme overclocking or the use of RAM disks for the sake of better benchmark results, and stuff like that. I know a score of 7000+ is relatively easy to achieve, but would not reflect my system accurately for daily use.
The Geekbench results are not too bad either:
In addition I tested the system with our PPBM5 benchmark, running CS6 and it took first place with a score of 111 seconds by a wide margin to the runner up at currently 123 seconds, but 41% better performance. If CS6 had not put a performance penalty on MPEG2-DVD encoding by more than 200% over CS5, this score could have been the first one to break the magic 100 second barrier. And if the AME Queue would have approached the speed of Direct Export instead of being over 400% slower, a score close to 61 seconds or even less would have been feasible. When doubling the Disk I/O time-line to 2 hours, a Direct Export takes only 14 seconds and the AME Queue of the same 2-hour time-line takes 76 seconds. Can you imagine the overhead AME carries?
Here are the RPI figures as of today (they are dynamically updated from the PPBM5 database and sorted by Total Time scores). We have added a vertical line showing the Top 10% of all observations.
The Areca ARC raid controller is superb, with sequential read transfer rates around 3,556 MB/s and write transfers around 3,966 MB/s, it is the fastest I have yet seen. Even with practical testing of time-line exports from PR transfer rates in exesss of 1,800 MB/s are quite normal, even with the PR overhead.
Even in the benchmark test of HD Tune Pro is performs pretty decently, with sustained transfer rates of around 2,317 MB/s and no fill-rate degradation:
However, since benchmark results only tell part of the story, I also ran the Crystal Disk Mark benchmark and it showed these rather nice figures:
It is a disappointment that different tests show such differing results. Isn't benchmarking all about replicable results, that show consistency? Unfortunately the AJA Disk test shows again quite different figures,
Tremendous write speeds, but disappointing read speeds in comparison to the Crystal Disk Mark Test or the HDTune Pro File Benchmark results. So which test to believe? I honestly don't know. All I know is that all results are way better than I expected from the outset and I think it does not really matter whether write speeds are around 4.0 or 3.5 or even 3.9 GB/s as these tests show. It is darn fast.
Looking at 'Bang-for-the-Buck', BFTB it failed miserably. 'Miserably' may be too strong a wording since with CS6 my old system would have had a score of around 185 seconds and from that perspective the performance improvement is sizable but still far too expensive for what it delivers. Effectively, if we look at the RPI (Relative Performance Index) scores, there is only a 24% improvement over the old system that still ranks # 7 on the overall list. And even though this was the first Win8 system in the PPBM5 benchmark, rest assured that Win8 is not faster than Win7 for editing, only in start-up.