Perhaps someone should "obviously" provide better instructions next time.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Looking for first to provide a MBP i7 Vega 20 CPU Thermal Test
- Thread starter Trebuin
- Start date
- Sort by reaction score
Perhaps someone should "obviously" provide better instructions next time.
It's a Mac OS, the UNIX based shells are one of its greatest strengths, it is really worth to know at least basics of the commands and syntax. If you don't feel like clicking too much you can just type this:
for i in {1..12} ; do yes >/dev/null & done
And after you're done:
killall yes
Yep, I'm pretty sure dev null isn't working anymore. I can't say why, but we need to find another program.
[doublepost=1543993109][/doublepost]
Yes, I should...I forgot to mention, that you need to run separate terminal windows, one per instant. When you're done, you just close those terminal windows & it will kill the command.
/dev/null is an empty device, it is the default redirection for emails to Dell technical support if you're not corporate customer. The command is 'yes', the '>/dev/null' just redirects the output of the command ('y' character) to a black hole. That's it. And it still works fine if you want to load the CPU.
So, I'm a bit late to the party, but by popular demand:
Methodology: run Unigine Valley Benchmark while simultaneously forcing 100% CPU uptime by spawning 12 yes processes.
Results: the screenshot shows the state after 10 minutes of running. The CPU has stabilised at 2.7Ghz and 30@ of power consumption, the GPU has also stabilised at 30W power consumption. The FPS has dropped by about 30%.
Discussion:
- The throttling is very minimal, the CPU — even with GPU drawing 30Watts — stably operates at frequency just 5% below nominal. This is a stark contrast to my previous i9 MBP which would clock down below 2Ghz in such scenario.
- The CPU seems to be about 10% faster than my previous i9 MBP. I don't know whether its luck of the draw or whether its because of power delivery and control system has been improvements.
- It seems that the MBP now operates as a true combined TDP system (this is different from earlier models!). When CPU does not need much power, GPU can consume up to 50Watts, and the other way around. The combined TDP appears to be around 70Watts, I tried different usage scenarios and that is the figure I've seen repeatedly. It makes sense, since that is what the power adapter can provide (when you add other components to the mix).
Final thoughts:
First of all, this thing is a monster. I am very impressed by the performance this thin chassis can put out. The GPU is very capable and far surpasses anything on the market in that power bracket (I will run 3dmark later, must to the office first). If my interpretation of combined TDP operation is correct, you will also get optimised performance under hybrid usage scenarios, just like the Intel/AMD Kaby Lake G+Vega M system — power is diverted to the GPU or the CPU as needed.
[doublepost=1544002576][/doublepost]
Its the same as any other MacBook Pro in the last 10+ years... If the system is mostly idling, its cool and inaudible. If you run something that demands high performance, it will get hot and eventually loud. Depends on the software you run. Has been very quiet so far in the few hours that I've been using Xcode and writing a paper. Has been also rather loud when I was building Clang or running the stress test
Methodology: run Unigine Valley Benchmark while simultaneously forcing 100% CPU uptime by spawning 12 yes processes.
Results: the screenshot shows the state after 10 minutes of running. The CPU has stabilised at 2.7Ghz and 30@ of power consumption, the GPU has also stabilised at 30W power consumption. The FPS has dropped by about 30%.
Discussion:
- The throttling is very minimal, the CPU — even with GPU drawing 30Watts — stably operates at frequency just 5% below nominal. This is a stark contrast to my previous i9 MBP which would clock down below 2Ghz in such scenario.
- The CPU seems to be about 10% faster than my previous i9 MBP. I don't know whether its luck of the draw or whether its because of power delivery and control system has been improvements.
- It seems that the MBP now operates as a true combined TDP system (this is different from earlier models!). When CPU does not need much power, GPU can consume up to 50Watts, and the other way around. The combined TDP appears to be around 70Watts, I tried different usage scenarios and that is the figure I've seen repeatedly. It makes sense, since that is what the power adapter can provide (when you add other components to the mix).
Final thoughts:
First of all, this thing is a monster. I am very impressed by the performance this thin chassis can put out. The GPU is very capable and far surpasses anything on the market in that power bracket (I will run 3dmark later, must to the office first). If my interpretation of combined TDP operation is correct, you will also get optimised performance under hybrid usage scenarios, just like the Intel/AMD Kaby Lake G+Vega M system — power is diverted to the GPU or the CPU as needed.
[doublepost=1544002576][/doublepost]
Its the same as any other MacBook Pro in the last 10+ years... If the system is mostly idling, its cool and inaudible. If you run something that demands high performance, it will get hot and eventually loud. Depends on the software you run. Has been very quiet so far in the few hours that I've been using Xcode and writing a paper. Has been also rather loud when I was building Clang or running the stress test
This was an i7?So, I'm a bit late to the party, but by popular demand:
Methodology: run Unigine Valley Benchmark while simultaneously forcing 100% CPU uptime by spawning 12 yes processes.
Results: the screenshot shows the state after 10 minutes of running. The CPU has stabilised at 2.7Ghz and 30@ of power consumption, the GPU has also stabilised at 30W power consumption. The FPS has dropped by about 30%.
Discussion:
- The throttling is very minimal, the CPU — even with GPU drawing 30Watts — stably operates at frequency just 5% below nominal. This is a stark contrast to my previous i9 MBP which would clock down below 2Ghz in such scenario.
- The CPU seems to be about 10% faster than my previous i9 MBP. I don't know whether its luck of the draw or whether its because of power delivery and control system has been improvements.
- It seems that the MBP now operates as a true combined TDP system (this is different from earlier models!). When CPU does not need much power, GPU can consume up to 50Watts, and the other way around. The combined TDP appears to be around 70Watts, I tried different usage scenarios and that is the figure I've seen repeatedly. It makes sense, since that is what the power adapter can provide (when you add other components to the mix).
Final thoughts:
First of all, this thing is a monster. I am very impressed by the performance this thin chassis can put out. The GPU is very capable and far surpasses anything on the market in that power bracket (I will run 3dmark later, must to the office first). If my interpretation of combined TDP operation is correct, you will also get optimised performance under hybrid usage scenarios, just like the Intel/AMD Kaby Lake G+Vega M system — power is diverted to the GPU or the CPU as needed.
[doublepost=1544002576][/doublepost]
Its the same as any other MacBook Pro in the last 10+ years... If the system is mostly idling, its cool and inaudible. If you run something that demands high performance, it will get hot and eventually loud. Depends on the software you run. Has been very quiet so far in the few hours that I've been using Xcode and writing a paper. Has been also rather loud when I was building Clang or running the stress test
Well, you ran pretty much the same thing as that first benchmark...your i9 stayed at a much higher clock, thought...& the temps sure show it. I'm still interested how the i7 will perform. The other benchmark likely used a more aggressive CPU loading technique & that may account for the lower clock.
Thanks for sharing, is it 70 or 60? (looking at your screenshot). Could you make a breakdown of individual power sensors?
That seems to be recurring observation of the new i9 owners, but I don't see any changes that would explain it, CPU side is identical, power delivery and cooling. Can you hit 1100 on cinebench? I personally think that Apple just tighten up the quality control and makes sure there is proper contact between baseplate and die, so you have higher chance on getting a 'good' unit. Could you turn off TM backup, kill everything that's running the background and just let your laptop sit idle for 15 minutes? Keep an eye on the CPU watts, below 0.5W Core (which is pretty amazing on its own) my 2.2/555x both show 30C flat in such scenario (proximity sensors), and there is really no reason for any other config to be any different - due to different components, and not QC. Are the idle fans still at 2k?
You just witnessed identical machines producing vastly different results, you will not get any valuable information from i7 test. When I was playing with stress tests I was ending up in a completely different state depending on how I would load the machine (in which order, and in what increments) even though the end result was 100% utilization on both CPU and GPU. And this was the same physical computer.
I will be running this again in a few hours.....now before I got **** on again, is there any other specific settings you want in Heaven? I was running it in full screen then clicking on Benchmark with the default settings (Medium).
You could run it in Extreme preset, actually this may be interesting, since the GPU will be limited to around 30W it will be in the 555x power range. Only problem Vega shows enormous improvements in old Unigine benchmarks, which was always AMD Achilles heel, looks like they solved whatever issue that was causing AMD to lag so much behind Nvidia in it, but still will be easy to see the relative performance drop. I think there should be more than 30% drop going from 50 to 30W.
No, generic settings on heaven will work. If you knew what REPPEG did on post 2, that would help, but I think we'll be fine. I would do a second one to match the i9 done just a couple posts up...I know he should have done the dev/null commands as that is actually less stressful on the cpu. He ran the Valley benchmark. I wouldn't mess with the graphics settings just so you have a standard setting that matches everyone else.
If you want to have a comparable result use Extreme preset. Without it the benchmark on first start will default to custom settings which are different for each system. And by comparable results I mean the actual score, if you want to just load up the GPU the settings don't matter, just make sure vsync is disabled.
Please ensure that during the test you maintain exactly 19C and 49% humidity. Please provide third party validation. Thanks.
I kinda hate you but on the other hand i also kinda appreciate your snarky attitude.
Yes, that's correct. It's not that nobody has the time or interest in doing it, but nobody (apart from yourself of course) actually knows how to install Windows 10, unfortunately.
I think it's because very few own the i7 vs the i9. And they have to have a purchased copy of Windows 10 and probably don't have it.
It doesn't take any experience to install Windows 10.
3dMark11 Performance GPU in 1280x720 please (could compare to notebookcheck scores). And breakdown of graphics/physics scores in Firestrike, TimeSpy and Skydiver would be great, native retina desktop resolution, default settings.Calm your slippers, its coming. I have a day job, after all
The idea is to have the settings as close to the same between tests. If you end up switching graphics tests to something completely different, not only are we looking at different environmental temps, but different GPU temps driven by significantly different demands. Ramping up the settings to max is a good idea to make sure it stays the same, but we are talking the same system with a different cpu. The custom settings should be the same. As far as I know, they should have a default setting due to the fact that the software is a benchmarking software, not gaming software.
Ok, I finally managed to install Windows 10 (no, I do not want to send all the info about my browsing patterns to you Microsoft, and no, I certainly don't want to send you each and every of my keystrokes!!) and run 3DMark.
My conclusion: the difference between Vega Pro 20 and 1050 Ti is more or less comparable to the difference between the 1050 Ti and the 560X.
Config: MBP 2018, i9, 16GB RAM, Vega Pro 20, stock bootcamp AMD drivers
Here are the results:
Time Spy:
Score: 2763
Graphics: 2623
Time Spy Extreme:
Score: 1293
Graphics: 1195
Fire Strike:
Score: 7573
Graphics: 9147
Physics: 8777
Fire Strike Extreme:
Score: 3793
Graphics: 4021
Physics: 9093
Skydiver
Score: 21296
Graphics: 27100
Physics: 10679
Comparing these to 2018 Dell XPS (GTX 1050 Ti) from notebookcheck:
Time Spy: MBP is 15% faster
Fire Strike: MBP is 23% faster
Comparing these to 2018 Razer Blade XPS (GTX 1060) from notebookcheck:
Time Spy: Blade is 42% faster [1]
Fire Strike: Blade is 23% faster
[1] https://www.reddit.com/r/razer/comments/8n2gec/stock_1060_156_time_spy_and_fire_strike_scores/
[doublepost=1544201337][/doublepost]
Sorry, I don't own 3dMark11 and I don't rally feel like spending $20 on it... I run the standard 3dmark tests using the standardised settings in order to make it comparable. Why would you want to run them on native resolution? You can't really compare the results to anything then...
My conclusion: the difference between Vega Pro 20 and 1050 Ti is more or less comparable to the difference between the 1050 Ti and the 560X.
Config: MBP 2018, i9, 16GB RAM, Vega Pro 20, stock bootcamp AMD drivers
Here are the results:
Time Spy:
Score: 2763
Graphics: 2623
Time Spy Extreme:
Score: 1293
Graphics: 1195
Fire Strike:
Score: 7573
Graphics: 9147
Physics: 8777
Fire Strike Extreme:
Score: 3793
Graphics: 4021
Physics: 9093
Skydiver
Score: 21296
Graphics: 27100
Physics: 10679
Comparing these to 2018 Dell XPS (GTX 1050 Ti) from notebookcheck:
Time Spy: MBP is 15% faster
Fire Strike: MBP is 23% faster
Comparing these to 2018 Razer Blade XPS (GTX 1060) from notebookcheck:
Time Spy: Blade is 42% faster [1]
Fire Strike: Blade is 23% faster
[1] https://www.reddit.com/r/razer/comments/8n2gec/stock_1060_156_time_spy_and_fire_strike_scores/
[doublepost=1544201337][/doublepost]
Sorry, I don't own 3dMark11 and I don't rally feel like spending $20 on it... I run the standard 3dmark tests using the standardised settings in order to make it comparable. Why would you want to run them on native resolution? You can't really compare the results to anything then...