Rick Lang
Forum Replies Created
-
[Andre van Berlo] “CPU upgrade prices(there were estimates on some site, can’t remember which)
4 core +$0. ($294 on intel.com)
6 core. +$250. (583 on intel.com)
8 core +$1250 (1723 on intel.com)
12 core. +$2500. (2950 on intel.com)”Linear upgrades for the other components but the CPU options, using Intel’s prices, do leave me wondering why the jump from $583 for the 6-core to $1,723 for the 8-core. A while ago I looked at each configuration on the Intel site, but I don’t quite grasp why such a large increase in cost? Anyone have any insight into the merits of the 8-core versus the 6-core? Do you feel it will be justified in the overall machine performance gain?
Rick Lang
iMac 27” 2.8GHz i7 16GB
-
The Apple Cinema HD 23” display was originally $3,499 according to Mactracker. I paid less than that if memory serves me but I remember the monitor cost $300 more than the outfitted PowerMac G5 Dual 2 GHz I bought with it. That monitor is still my secondary display off my iMac.
Rick Lang
iMac 27” 2.8GHz i7 16GB
-
Walter Soyka:
“Perhaps someone could devise and publish a more real-world, footage-based test could supplement BruceX? What would you want to see tested in such a benchmark?”That is such a difficult question to answer. Benchmarks normally use specific subsets of the computers being tested and (thanks to Samsung’s ingenuity and total lack of ethics) we know machines can be tuned to perform a given benchmark better than they will perform under real world use. Forgetting that issue, benchmark configurations are always very limited and you are still left wondering, “Well what if I doubled this capacity or selected that option or used this device?” So relying on a few differently configured machines with a few predictable fixed benchmarks does limit the usefulness of the benchmark. They are better than nothing but not a perfect answer to “What should I buy?”
I’d like to see benchmarks that exercise more of the components in a Mac Pro. Specific tests to do that are certainly difficult to design in ways that will really exercise the components. For instance if you had the default 256 GB flash drive, current benchmarks I suspect would happily run using a small fraction of that, but when you design a test to use more than the defaults, you might get interesting results. Still you want the test to be reasonably accurate and representative of real world workload. So do you exercise a video file that fits easily within the default system flash drive or work on file sizes that you know will have to reside on external storage? Another example, your test may exercise all cores available or exercise only one or four cores. Cinebench uses all cores available I believe and that is very useful but it would be informative to have tests that are limited to four cores to see how the models compare under those constraints. As you know, not all apps do use every core available and I suspect the 6-core machine could beat a 12-core machine in those tests although the 12-core wins in Cinebench of course. So maybe I’m looking for tests in which you have options or parameters to control the nature of your benchmark.
And a way to crowd source the results by collecting a lot of information about the variously configured machines being tested. If the results were presented as a score, the way Geekbench rates your CPU and memory performance, it would be interesting to show a scatter diagram and as you hover over the results, a pop up would show you details of the machine that achieved those scores. Just from our discussions on the Cow, it is apparent more than a few people are looking at the same configurations and as people buy and then test those machines, other people should be able to find those same machines they too are now considering. I expect I shall look at a variety of different benchmark sites to come to some conclusions as well as read what early adopters have to say. Honestly, it is the anecdotal evidence that can be the most convincing.
Sorry, Walter, I don’t have a succinct and satisfying answer to your question. Benchmark testing is challenging and relying on benchmark results can be dangerous. When I used to benchmark mainframe computers in the 70s, we actually did that by taking our core applications and data and seeing how the machines ran our real world stuff. That was trustworthy. The vendors had generally never encountered such a laborious approach, but it worked (and I got to see a lot of the United States going to those test sites). The best test is the one that does your work and so far with workstations that means you buy first and test when you get it. It would be interesting to be able to tune a benchmark to be like your projected workload and then have it run on a remote benchmark site. Of course no other site is going to be able to duplicate some of the fabulous things you personaly do Walter! How about we just wait for your results?
Rick Lang
iMac 27” 2.8GHz i7 16GB
-
[Tom Sefton] ” I’d love to see some benchmarks at some point soon, but they will no doubt arrive.”
Understandably due to limited resources, I think a lot of benchmarks will only show two or three comparisons. It’s the brave early adopters who will provide most of the valuable feedback.
I like the concept of Geekbench where the data is gathered from everyone posting to their site. Just too bad it is limited to evaluating CPU and memory performance (although those results will be interesting too). Would be great to have Geekbench’s concept expanded to include all the optional components in the Mac Pro as people reported their own results with their individual configurations. Of course the difficulty there would be the variety of external storage setups that would have a bearing I wonder if Barefeats handles crowd-sourced results.
Oh, well. Life’s a gamble, no way around it. Let the games begin.
Rick Lang
iMac 27” 2.8GHz i7 16GB
-
[Gary Huff] “I’ve decided to go with, “If it’s worth doing in the first place, it should be worth doing right.””
Good one. Good standard to bear.
Rick Lang
iMac 27” 2.8GHz i7 16GB
-
Gary Huff:
‘The problem is fighting against, “Good enough.”‘Good point, Gary. Try a counter-punch of “Be the best you can be.” Hopefully you’ll be blessed with either clients with a better perspective, you just better clients.
Rick Lang
iMac 27” 2.8GHz i7 16GB
-
Craig Seeman:
“4 and 6 core coming in December. Those aren’t going to be delayed. That those two are listed and 8 and 12 aren’t has to indicate “something.” Waiting on pricing decisions wouldn’t be a delay of course but part shortage would be.”I think it means the 4-core and 6-core with their corresponding pre-configured options will be available off the shelf and initially ship within 24 hours. All other options on the 4-core and 6-core will be considered build-to-order as will the additional 8-core and 12-core CPU options and will have longer ship times depending upon the options selected.
Rick Lang
iMac 27” 2.8GHz i7 16GB
-
[Walter Soyka] “[Rick Lang] “Looking forward to aggressive pricing from Apple to make a statement that they intend to be competitive and to make a dent in the market.”
Competitive with whom, and a dent in what market?”
I had in mind competitive with the higher-end PCs and the market that those PCs attract. By competitive I don’t imply that they have to be cheaper but be in the ballpark.
I do feel that Apple is not going to dominate on specs with the higher-end PC machines, but it remains to be seen how close they’ll get in real use due to their tight integration of hardware (and tuning of selected Pro Apps software) and what value their new Mac Pro price-performance will be perceived to have.
Once we start getting feedback from actual use in January, I think the threads here on the COW will be very enlightening. I distrust Apple’s ad copy on the new Mac Pro, but I do trust you and others here to present balanced and insightful (and entertaining) reports to the community.
Rick Lang
iMac 27” 2.8GHz i7 16GB
-
Marcus, with the optimization of the Pro Apps for the new Mac Pro, the existing behaviour of the apps that Andy described earlier may be changing. Aperture isn’t changing much I don’t think but we know FCP X has undergone significant performance enhancements to take advantage of the new dual GPUs, Other Pro Apps will likely see performance changes but perhaps not at the launch date.
SSD and PCIe Flash
There’s an old rule of thumb for system disks to ensure you have at least 10% free space. For one thing, your virtual disk needs free disk space to write new memory pages. But in reality you can practically fill a hard disk used for any other purpose without ill effects. All ‘traditional’ SSDs do require enough free space to do their data management housekeeping. SSDs that are sold as 480 GB devices for example already have about 6% of their free space set aside so you never see it. SSDs that are sold as 512 GB devices for example likely don’t have any free space kept in reserve so you want to be very cautious about filling them up and certainly you shouldn’t fill such a device if it is your boot disk. Now the PCIe flash drives probably follow the same principles: if it’s your system drive, don’t fill it up.RAM
More is better within reason. In an ideal world, you wouldn’t be writing any memory pages out to your system drive and all the programs you need to use concurrently will run fine in real memory. Some minor exceptions won’t be a concern but if your virtual memory is much larger than that, you are suffering performance hits. In the old days when the virtual memory was on a spinning hard disk, it was a very big hit potentially. But in the new Mac Pro as well as the MacBook Air for example, the PCIe system flash drive is much faster and virtual memory likely isn’t as great an impact. Still obviously better to have the real memory you anticipate you will need. FCP X and DaVinci Resolve seem to run well in 16 GB RAM, but I don’t know the impact of 4K video on the overall requirements and hence feel 32 GB is better. By the way, the memory is four-channel access so Apple is frankly wrong to even offer that low-end 12 GB configuration. Configurations that use all four memory slots with identical sticks will run at the rated speed. That 12 GB will be slow. Intel must have a white paper on it for the new memory, but I don’t think it is worth the effort to find it. I’m sure everyone will be smarter than Apple’s marketing department and buy either 16 GB, 32 GB or 64 GB memory configurations.GPU
We are going to have to readjust our thinking regarding Apple’s use of GPUs now that they are supporting dual Firepros in the new Mac Pro and tuning their apps for that. Having dual GPUs sounds very good and hopefully the implementations in the software will take full advantage of both at some point if they are available. Time will tell. I know Blackmagic Design has commented that DaVinci Resolve “screams” on the new Mac Pro. That’s been discussed in another thread and “screams” is a relative word comparing the Mac Pro to previous Macs and the verdict is out about how it will stack up against high-end PCs that can run multiple Titans.CPU
again, we need to wait to see how the new versions of the Pro Apps behave. And even if some apps like FCP X make extensive use of the dual GPUs, there will be other apps that continue to use the CPUs. Over time I expect that will change.External Storage
Don’t forget that all the potential of the new Mac Pro can be humbled by expeternal storage that doesn’t keep up! Enter Thunderbolt 2 to squeeze more performance from external storage devices. I had thought something like the 6-bay Promise Pegasus2 R6 would suffice for 4K video but someone recently vaguely warned about 6-bay RAIDs. I don’t know what the concern was but will think about an 8-bay RAID. If anyone can explain why 6-bay RAIDs should be avoided, I’d appreciate hearing from you. I think the goal posts have moved to sustaining at least 1 GB/s of incompressible data (video) over Thunderbolt to the Mac Pro. Again, love to hear other opinions.Rick Lang
iMac 27” 2.8GHz i7 16GB
-
Darren Roark:
“Peter Chamberlain gave an overview of Resolve 10 and mentioned that in an update it would support GPU acceleration for Red files once the new Mac Pro is released. It would make a lot of sense that fcpx would have this as well. If that’s the case, it makes the top end GPU more enticing than getting more cores on the CPU.”Good information. I agree with you and Andy, given the budget can handle it, the dual D700 is highly desirable and personally I’ll sacrifice some RAM (down to 32 GB) and CPU cores (down to 6 cores) if needed. So weary of not having a great GPU. Sacrificing RAM and cores doesn’t mean the tasks can’t be done, they just take longer. Finding you have an inadequate GPU, can mean you aren’t going to be able to do the task. Now I don’t plan to be doing 16 angles of 4K multicam, but you know someone will want to do it.
As for working with Red files, that’s for others to value. I’m hoping that FCP X will work with (compressed 4K) CinemaDNG raw files from the Blackmagic Production Camera. If not there is of course the workflow of using DaVinci Resolve at the front and back end to handle raw debayer and rendering final deliverables with FCP X using any flavour of ProRes generated by Resolve. For some things though, it would be better if FCP X did CinemaDNG.
I think many people feel there will be substantial enhancements to the next version of FCP X and it likely will be a paid update to 10.1.0. I also expect that Mavericks will have a point increase as well to include further support for the new Mac Pro and perhaps other accessories such as a 4K screen which hopefully will finally include support for display graphics beyond 8-bit such as true 10-bit. Maybe a trivial thing in a way but a welcome improvement. I just can’t imagine how Apple could put the design energy they have done into the Mac Pro without adding a 4K 10-bit colour display option. It doesn’t have to be a retina screen to be useful but display technology is evolving and it may be the best time for Apple to upgrade their displays. I haven’t calculated the bandwidth of supporting 3 4K 10-bit screens but if it fits within the Thunderbolt 2 threshold, then in my opinion they will do it soon.
Edit: there is hope–did the calculations and it should be feasible to support 3 4K 10-bit colour displays (just not on a single Thunderbolt 2 controller; requires two streams on two controllers plus one on HDMI as Apple has said. Here’s my calculations assuming one 4K 10-bit 30 fps video… UHD and 24 or 25 fps will be less. The bottleneck could be getting the video data fast enough from external storage as one video would take about a gigabyte per second! Apple’s performance claims use UHD, 3480×2160, so my calculations are perhaps overstating the requirements but let’s say that will allow for the audio data I left out.
4096×2160 = 8,847,360 pixels per frame
X 30 bits per pixel = 265,420,800 bits per frame
X 30 frames per second = 7,962,624,000 bits per second for one display which I think HDMI 1.4 supports plus audio.Rick Lang
iMac 27” 2.8GHz i7 16GB