Slow(er) tether speed on newer MacBook Pro
Today, myself and another tech, ran some semi-informal tests on tethering speeds, heat throttling, and export times in C1 v12 between a 2015 15†MacBook Pro, and a 2017 15†MacBook Pro, and ran into a surprising results; the 2015 beat the 2017 in tether speeds.
SPECS: These machines are both middle spec-ish for their time. *EDIT* Specs listed below in a separate response post
1. The 2017 being a 2.9 Intel Core i7 with 16GB ram, Radeon Pro 560 4GB, and 512GB space with a recent clean install of Mojave, with no software running in the background except temperature apps.
2. The mid 2015 was running Sierra, but I don’t know the exact specs.
INFORMAL TEST 1: Using a Nikon D810 with a solid USB3.0 tether cable, we shot through the buffer of 28 shots, and then used a stop watch to determine how long it would take to clear the buffer, and render in C1. We repeated this test consecutively, until the computer had approximately 500 shots (in order to raise the temperature of the computers and induce throttling, and also to try to simulate shooting situations) and then repeated it with the same equipment on the other computer to compare results. We then repeated it 2 more times.
*Both computers had a C1 session set up identically, with the preview size being the same, and hardware acceleration being the same across both (and later tested with and without hardware acceleration)
**Auto Select When Ready was enabled on both computers
***Graphics Switching was initially turned off for both computers
****Most known viewing conditions were the same
TEST 1 RESULT: This was not expected. The 2015 MacBook Pro was consistently clearing the buffer in 14-19 seconds (hovering around 19 seconds after the first few rounds), whereas the 2017 cleared it in 17-28 seconds (hovering around 27 seconds after the first few rounds). Temperatures of the 2015 MacBook would hover around 85-90 degrees, whereas the 2017 was a cooler 65-70 with a few spikes to 80 degrees.
*We tested this with an powered extension (no difference), with hardware acceleration turned off and on, using different USB-C to USB-A adapters for the 2017 MacBook Pro, different ports, restarts, black frames, and full data frames... the 2017 never once beat the 2015 in tethering speed. We also turned off FileVault on the 2017 MacBook Pro, didn’t seem to make any difference there either.
TEST 1 CONCLUSION: I’m not entirely sure what to conclude. By all accounts, I would expect the 2017 to be faster (or as fast as the 2015 if the cable/camera is the bottleneck). Thermal throttling does not seem to be an issue, as it actually ran quite a bit cooler overall than the 2015. My best guess is that Mojave with APFS may be to blame over the Sierra Mac OS Extended, and the way it handles files, but I can’t be sure. Any ideas?
FURTHER THOUGHTS: I’ve had quite a few issues with slow capture using C12 and Mojave, whereas anecdotally it would feel a lot faster when using older computer systems with C10-11, 12 being hit or miss. A common issue I get with the newer computers and C12, is that unlike the steady stream of shots coming in like I’m used to, no shots will come in, and suddenly a burst of files will come in, followed by a lag, before another burst of images come in.
This difference is massive, especially when multiplied over thousands of shots throughout the day, moving quite a bit, checking critical focus and exposure, and waiting for art director to approve a shot, whilst complaining about the delay to the photographer.
INFORMAL TEST 2: We exported the 500 images as high res TIFF files.
TEST 2 RESULT: This was as expected. The 2017 MacBook Pro was miles ahead of the 2015 MacBook Pro, it wasn’t even close. We didn’t take the time, but I think it finished nearly twice as fast
SPECS: These machines are both middle spec-ish for their time. *EDIT* Specs listed below in a separate response post
1. The 2017 being a 2.9 Intel Core i7 with 16GB ram, Radeon Pro 560 4GB, and 512GB space with a recent clean install of Mojave, with no software running in the background except temperature apps.
2. The mid 2015 was running Sierra, but I don’t know the exact specs.
INFORMAL TEST 1: Using a Nikon D810 with a solid USB3.0 tether cable, we shot through the buffer of 28 shots, and then used a stop watch to determine how long it would take to clear the buffer, and render in C1. We repeated this test consecutively, until the computer had approximately 500 shots (in order to raise the temperature of the computers and induce throttling, and also to try to simulate shooting situations) and then repeated it with the same equipment on the other computer to compare results. We then repeated it 2 more times.
*Both computers had a C1 session set up identically, with the preview size being the same, and hardware acceleration being the same across both (and later tested with and without hardware acceleration)
**Auto Select When Ready was enabled on both computers
***Graphics Switching was initially turned off for both computers
****Most known viewing conditions were the same
TEST 1 RESULT: This was not expected. The 2015 MacBook Pro was consistently clearing the buffer in 14-19 seconds (hovering around 19 seconds after the first few rounds), whereas the 2017 cleared it in 17-28 seconds (hovering around 27 seconds after the first few rounds). Temperatures of the 2015 MacBook would hover around 85-90 degrees, whereas the 2017 was a cooler 65-70 with a few spikes to 80 degrees.
*We tested this with an powered extension (no difference), with hardware acceleration turned off and on, using different USB-C to USB-A adapters for the 2017 MacBook Pro, different ports, restarts, black frames, and full data frames... the 2017 never once beat the 2015 in tethering speed. We also turned off FileVault on the 2017 MacBook Pro, didn’t seem to make any difference there either.
TEST 1 CONCLUSION: I’m not entirely sure what to conclude. By all accounts, I would expect the 2017 to be faster (or as fast as the 2015 if the cable/camera is the bottleneck). Thermal throttling does not seem to be an issue, as it actually ran quite a bit cooler overall than the 2015. My best guess is that Mojave with APFS may be to blame over the Sierra Mac OS Extended, and the way it handles files, but I can’t be sure. Any ideas?
FURTHER THOUGHTS: I’ve had quite a few issues with slow capture using C12 and Mojave, whereas anecdotally it would feel a lot faster when using older computer systems with C10-11, 12 being hit or miss. A common issue I get with the newer computers and C12, is that unlike the steady stream of shots coming in like I’m used to, no shots will come in, and suddenly a burst of files will come in, followed by a lag, before another burst of images come in.
This difference is massive, especially when multiplied over thousands of shots throughout the day, moving quite a bit, checking critical focus and exposure, and waiting for art director to approve a shot, whilst complaining about the delay to the photographer.
INFORMAL TEST 2: We exported the 500 images as high res TIFF files.
TEST 2 RESULT: This was as expected. The 2017 MacBook Pro was miles ahead of the 2015 MacBook Pro, it wasn’t even close. We didn’t take the time, but I think it finished nearly twice as fast
0
-
A 2015 Ferrari will still beat a 2017 Prius.
Ya gotta know the specs to determine /anything/ from your tests0 -
[quote="photoGrant" wrote:
A 2015 Ferrari will still beat a 2017 Prius.
Ya gotta know the specs to determine /anything/ from your tests
2015 Ferrari
2.8GHz quad-core Intel Core i7 (Turbo Boost up to 4.0GHz) with 6MB shared L3 cache, I believe it's a Haswell processor
— 938 Single-Core Score | 3526 Multi-Core Score on Geekbench
16GB of 1600MHz DDR3L onboard memory
2.5GHz - 1TB PCIe-based flash storage (and I believe his is actually 512)
Intel Iris 2.5GHz - 512GB and AMD Radeon R9 M370X with 2GB of GDDR5 memory
2017 Prius
2.9GHz quad-core Intel Core i7, Turbo Boost up to 3.9GHz, with 8MB shared L3 cache
907 Single-Core Score | 3251 Multi-Core Score on Geekbench
16GB of 2133MHz LPDDR3 onboard memory (only option)
512GB PCIe-based onboard SSD (middle tier option)
2.9GHz
Radeon Pro 560 with 4GB of GDDR5 memory and automatic graphics switching (top option)
Intel HD Graphics 630
Apart from the 2015 having turboboost speeds being 0.1GHz faster, and having a slight edge in single core Geekbench result (I'm not sure if Capture One can utilise multi-core, because if so, then mine takes an edge) the 2017 has the edge in components and speed everywhere else, and beats the 2015 in every other task, except for tethering, which the 2015 performs a massive 42% faster on average, with GPU acceleration disabled or enabled on the 2015. I mean that's not even close.0 -
[quote="JNOR" wrote:
[quote="photoGrant" wrote:
A 2015 Ferrari will still beat a 2017 Prius.
Ya gotta know the specs to determine /anything/ from your tests
Apart from the 2015 having turboboost speeds being 0.1GHz faster, and having a slight edge in single core Geekbench result (I'm not sure if Capture One can utilise multi-core, because if so, then mine takes an edge) the 2017 has the edge in components and speed everywhere else, and beats the 2015 in every other task, except for tethering, which the 2015 performs a massive 42% faster on average, with GPU acceleration disabled or enabled on the 2015. I mean that's not even close.
Everythings rings true but the buffer clears quicker on the older macbook pro which by any means has a less complex usb stack and a less abhorrently oppressive OS (security wise)...
I'd suggest trying to get HIgh Sierra on both machines, bare install and run the test again. Either way I'm interested0 -
would be interested in capture one 10 / 11 if they run quicker or slower as well 0
投稿コメントは受け付けていません。
コメント
4件のコメント