@leman This is strange. I just ran GB5 with Activity Monitor open and it never went above 1.2GB memory usage and memory pressues was always in the green. I wonder if this effect is due to folks running GB with alot of other stuff in the background - which is easier to get away with on a 32GB machine ?
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.
16" MBP Geekbench scores so far show big differences
- Thread starter intrepidcase
- Start date
- Sort by reaction score
@leman This is strange. I just ran GB5 with Activity Monitor open and it never went above 1.2GB memory usage and memory pressues was always in the green. I wonder if this effect is due to folks running GB with alot of other stuff in the background - which is easier to get away with on a 32GB machine ?
high quality posts as always leman
A little odd the 16 inchers are lagging tho and a bummer
I did get a 16 incher and it flys.
Unfortunately my script does not grab the GPU details.
[automerge]1574410536[/automerge]
I also find it strange, it should not have any impact. Possibly a benchmarking artefact. It is possible they are not doing proper "warmup" when starting some tests. I don't know. Geekbench is not a good benchmark. It is unreliable and all over the place. We are only talking about it because that's all we have right now
[automerge]1574410597[/automerge]
Yes, I find it odd too. But then again its Geekbench, everything is possible with Geekbench. My 16" just arrived and I will try to test out the thermals. Unfortunately I don't have a 2019 model with 8-core CPU to test it agains... so I'll. have to use my 2018 6-core Vega Pro 20 machine.
I think it's easy to do that just by changing the search string. Seems to me that model number "MacBookPro15,3" refers to Vega models only and "MacBookPro15,1" are models with Polaris graphics.
Although GPU information is not visible for CPU benchmark results, when you search for "MacBookPro15,3" within compute results you can see that these are in fact 2018/2019 models with Vega graphics:
https://browser.geekbench.com/v5/compute/search?utf8=✓&q=MacBookPro15,3
a new test on barefeats - https://barefeats.com/16-inch-macbook-pro-adobe.html
they say "The 2.4GHz model exhibited an advantage ranging from 6% to 14%. With only a 4% CPU core frequency gain (and 4% Turbo Boost gain) over the 2.3GHz model, the faster times could be coming from the increase in main memory and GPU video memory."
so wondering if there is a significant difference between 2.3 @ 16GB vs 2.3 @ 32GB ram in the geekbench scores so far. not an expert, but maybe apple is using 1 "channel" (aka slot) for 16gb ram and 2 x 16gb channels for 32gb ram???
they say "The 2.4GHz model exhibited an advantage ranging from 6% to 14%. With only a 4% CPU core frequency gain (and 4% Turbo Boost gain) over the 2.3GHz model, the faster times could be coming from the increase in main memory and GPU video memory."
so wondering if there is a significant difference between 2.3 @ 16GB vs 2.3 @ 32GB ram in the geekbench scores so far. not an expert, but maybe apple is using 1 "channel" (aka slot) for 16gb ram and 2 x 16gb channels for 32gb ram???
A reasonable hypothesis. If you buy a motherboard with dual channel DIMM slots you want to always fill the slots for Channel A, Slot 0 (16GB) first. And then Channel B, Slot 0 (16G + 16GB) for optimal performance. And then fill Channel A, Slot 1 and Channel B, Slot 1 (16GB + 16GB + 16GB + 16G) to increase memory.
I do this on my deskside system to get my 64GB of memory.
Just compared Geekbench 5 for my new 16" / 8-core 2.4 GHz / 32 GB to my old 2016 / 4-core 2.9 GHz / 16 GB
2016: 855 / 3609
2019: 1184 / 7361
Or +38% / +103%
Very happy with the performance, a big difference for me with Visual Studio in VMWare Fusion, much, much snappier. Happy customer
2016: 855 / 3609
2019: 1184 / 7361
Or +38% / +103%
Very happy with the performance, a big difference for me with Visual Studio in VMWare Fusion, much, much snappier. Happy customer
Yes there is, but it rather tells something about the quality of geekbench as a representative benchmark.
I have a number of issues with that. First of all this bit "with only a 4% CPU core frequency gain (and 4% Turbo Boost gain) over the 2.3GHz mode" is not accurate. These are not real clocks, they are just marketing values that represent certain limits in the CPU configuration. My 2.3 16" model for example runs at very steady 3.1 Ghz across 8 cores under sustained 100% load. The advertised clocks don't mean much here.
For their productivity software tests, the RAM and VRAM will most certainly make a difference. What I am confused about is why Tomb Raider is so much faster.
Very unlikely. I just run a quick and dirty RAM bandwidth benchmark and for larger buffer sizes I am getting around 38000 MB/sec which is consistent with DDR4 dual channel (theoretical maximum is 42000 MB/sec). Single-channel RAM would perform slower here.
[automerge]1574449604[/automerge]
We are talking about a laptop though
2.3ghz 32gb 5500M 8GB
1154
Single-Core Score
7269
Multi-Core Score
https://browser.geekbench.com/v5/cpu/650520
1154
Single-Core Score
7269
Multi-Core Score
https://browser.geekbench.com/v5/cpu/650520
Last edited:
interesting. it would be good to get an apples-to-apples comparison of a 2.3 32/8gb vs a 2.4 32/8gb. maybe it's all about the ram, at least as far as geekbench goes.
Essentially same as the 2.4/32GB. Makes sense since thermals will prevent either model from hitting max boost anyway.
I'll consider the $150 hit an early adopter tax rather than buying a performance advantage (my goal was to buy a 2.3/32GB/55008GB/2TB, but the 2.4 was available in stores on day 1)
I think the jury is still out on that one. If you only get a 4-5% improvement in benchmarks that lines up with nominal clock speed will you see it that way?
If I got the 4-5% increase, I would at least consider the upgrade meaningful, even if not worth the money. But even going in I expected a 0% real world performance difference outside of sticking the laptop in a freezer.
Do we have any results on the 5500M 4GB vs 8GB on the 2.3 i9? Trying to see if the extra 4GB is making a huge difference.
Ive noticed with all these Geekbench screenshots for the 16" that all of them show the memory frequency @1333 MHz; shouldn't they reflect the current 2666 MHz?
1333 MHz is actually the true frequency. Because it's DDR (Double Data Rate) you have to double the clock speed to get the equivalent speed which is 2666MHz. This is a RAM thing and not Mac specific. Here is an example from my windows PC:
You can see it shows 1,799.9MHz (1800MHz really) the RAM is actually a 3600MHz kit. 1,800 x 2 = 3,600.
I think the scores will vary at the beginning as peoples machines are still setting up and they want to test their computer early..
Anyway I have had mine for a week now and it has settled down. I am extremely happy with it for my workload. Here is the score from today.
Anyway I have had mine for a week now and it has settled down. I am extremely happy with it for my workload. Here is the score from today.
I thought I waited until my new computer had finished indexing, etc., but man I was wrong. I was seeing highly variable scores, 6600-7200 on the 2.4 i9. It was a combination of things. My mail app was still downloading messages, that was a silly oversight, but also my IT installs a ton of security software that I can't 100% disable. I uninstalled the remote screen share and chat, because for some reason that was running constantly despite being off. Then we have a Norton product that runs in the background a ton, SymDaemon. Its always like 20% of a thread, but spikes to 100% for a while.
Anyway, its a real lesson in how little it takes to give pretty wide scores. That SymDaemon kicks on, and Geekbench will be 10% lower. But if its not, which I guess its settled some, I get very consistent results now. 7150-7250 now. I'd probably score a bit better if some of this stuff could be easily turned completely off. But just simple stuff like leaving safari open, drops the score by 100 points or more.
Yes, they gotten a lot better lately.... I can do a complete uninstall of the programs and put them back, but that’s hassle and not worth it for just a slightly better benchmark.
can we get some results with multiple runs back to back.. for instance.. my mid 2018 MacBook Pro i7 dropped over 200 points between the first run and the 4th run.. due to thermals.. it never could boost for longer than 3 seconds without having to throttle