Honestly, for the sort of gpu in these laptops it's pretty much academical whether link is 16x or 8x, it will have a tiny impact on anything.
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.
gpu link speed
- Thread starter bigins
- Start date
- Sort by reaction score
honestly I can't remember. it came in 4 days ago and I think it might have been 10.15.4?
Oh. okay. From my case, I checked on specs from 5500M, it was listed as x8 PCIe 4.0. Weird thing is when I purchased my 16inch with 5500M and run on 10.15.3, it was x16 on system info.
My guess, it could be a mistake from System Info before update, and after update it change back to x8 as correct info?
[automerge]1594894095[/automerge]
Radeon Pro 5600M supposed to be "x16 PCIe 4.0" based on specs. In conclusion Apple has slowing down the graphics performance.
Attachments
Last edited:
The issue and potential controversy is that it used to be x16 but was reduced to x8 in a software update.
I just ran Geekbench graphics tests and can confirm a lower result than I had previously. I previously had over 28,000 Metal score (possibly well over 28,000) with the 5500M (8 GB) and it is now just shy of 27,000. This is a similar reduction to what I observed in Unigine Valley.
You can check any number of 16" MacBook Pro results to confirm that the 5500M used to achieve 28,000+. My MacBook can't have degraded itself that much in a relatively short period of time, especially considering it hasn't even left the house yet (thanks to COVID-19).
I'm not saying this is "gate"-worthy (yet), but it definitely warrants further inquiry or investigation as to why the apparent change was made and what the supposed benefit is.
I just ran Geekbench graphics tests and can confirm a lower result than I had previously. I previously had over 28,000 Metal score (possibly well over 28,000) with the 5500M (8 GB) and it is now just shy of 27,000. This is a similar reduction to what I observed in Unigine Valley.
You can check any number of 16" MacBook Pro results to confirm that the 5500M used to achieve 28,000+. My MacBook can't have degraded itself that much in a relatively short period of time, especially considering it hasn't even left the house yet (thanks to COVID-19).
I'm not saying this is "gate"-worthy (yet), but it definitely warrants further inquiry or investigation as to why the apparent change was made and what the supposed benefit is.
Last edited:
Just ran Geekbench 5 several times on my week-old MacBook Pro (compute, Metal, 5500M 4GB "x8"). Results range from 26,861 to 30,632. Not a very reliable benchmark...
This seems to be one of those topics that can "only" be answered by an Apple Engineer. I doubt that the dGPU in the 16" MacBook Pro was ever actually configured to use x16 PCIe lanes. This is most likley a glitch in System Information vices an actual "gate." My high-speed desktop gaming PC only runs its dGPU at x8, as it uses a consumer Core i7-8700K, which like the CPUs in the 16" MBP are limited to 16 PCIe lanes in the CPU package. So logically speaking, the MacBook Pro would never be configured to use an x16, as that would leave none left for the rest of the system. There are some lanes avaialble from the Platform Controller Hub (PCH), but those are "slower" due to latency than the CPU native ones, are often not used for highspeed devices like TB, GPUs, or NVme. I assume that the NVMe controllers in the MBP are actually using lanes from the PCH, so that the TB and dGPU can run as quickly as possible.
Good luck!
Rich S.
Good luck!
Rich S.
As pointed out by @LinkRS, Intel only offers 16x PCIe lanes in their CPUs. There are more lanes offered by the chipset, but they are connected to the CPU by a 4x connection, so it's a bit of a cheat. Only the "native" CPU lanes offer high performance. If I remember correctly, Apple uses 8 of them for the GPU, 4x for the SSD and the rest for TB. This is a very reasonable compromise that still gives you fast graphics but does not sacrifice the rest of the system. By the way, they have been using this setup for years.
As to why some users used to see "16x link" before, my guess is also a bug in the reporting software.
As to why some users used to see "16x link" before, my guess is also a bug in the reporting software.
There does appear to be an issue as I don't see 16x anymore as well. There was definitely 16x displayed. Now I've gotten curious as it made me think some stuff was disabled since I believe when i had the 16x displayed I had nothing else on the laptop. My benchmarks are also much higher then than now so something changed. Now I use a thunderbolt to display connector for my external and also a hub. Given the choices it would be cool to get full bandwidth when available but I understand the compromise.
Thanks for this. just tried again and I got 28,500, so I’m happy nothing has really changed from a performance perspective.
Weird that it fluctuates so much when they were both run from an essentially “cold boot” status.
Geekbench 5 does this a lot. I'm suspecting it fluctuates a while lot more than 4 did.
Either way, this is a fine performer and I'm not concerned. Wattage overage (with external) is another thing but still happy.
Either way, this is a fine performer and I'm not concerned. Wattage overage (with external) is another thing but still happy.