Geekbench 5 is out. Post MBP results!

[CreeptoLoser]

Geekbench 5 is out. Post MBP results!

Gotta add that these guys are still crippling the Mac GPU results. The scores used to be the same as Windows and then last year without any changes on the user side the macOS GPU results came sliding down.

 

[alias99]

2019 13in 1.4ghz, 512ssd, 16gb Ram

Single Core - 921
Multi Core - 3913
Metal - 6709
OpenCL - 7115

Ran them all back to back so I'm sure there is a bit more left in the tank.

 

[MrGunnyPT]

2019 13in 1.4ghz, 512ssd, 16gb Ram

Single Core - 921
Multi Core - 3913
Metal - 6709
OpenCL - 7115

Ran them all back to back so I'm sure there is a bit more left in the tank.

That 4 cores setup is pretty good.. Should have gone for that instead of the 13" 2017.. But hey that's what you get when you need a computer at the last minute.

 

[Coyote2006]

836 Single-Core Score
3382 Multi-Core Score

MacBook Pro (15-inch Retina Late 2013)
Intel Core i7-4850HQ
Topology 1 Processor, 4 Cores, 8 Threads
Base Frequency 2.30 GHz

4897 OpenCL Score

 

[donawalt]

MacBook Pro (15-inch, 2019)
2.4 GHz Intel Core i9
Radeon Pro Vega 20

Single Core - 1253
Multi Core - 7025
Metal - 22406
OpenCL - 23815

 

[mmackinven]

MacBook Pro (15-inch 2018)
2.9GHz i9
32gb RAM
Radeon Pro 555x (was supposed to be the 560X...)

Single core: 1206
Multi core: 5536
Metal: 11714
OpenCL: 11997

 

[venom600]

Single-Core Score: 894
Multi-Core Score: 3364
OpenCL Score: 10441
Metal Score: 10840

MacBook Pro (15-inch Retina Mid 2015)
Intel Core i7-4870HQ, 2.5Ghz
16GB DDR3 RAM, 1TB SSD

 

[liudayu]

MacBook Pro (15-inch, 2019)
2.3 GHz Intel Core i9
Radeon Pro 560X

Single Core - 1253
Multi Core - 6549
Metal - 15316
OpenCL - 17023

 

[DaakuMaujii]

MacBook Pro (15-inch, mid 2014)
2.8 GHz Intel Core i7-4980HQ
Intel Iris Pro graphics 1538MB
16GB DDR3 RAM, 256GB SSD

Single core: 879
Multi core: 3407
OpenCL: 4884
Metal: 531

CPU: https://browser.geekbench.com/v5/cpu/34241
OpenCL: https://browser.geekbench.com/v5/compute/17656
Metal: https://browser.geekbench.com/v5/compute/17694

I ran the OpenCL (score of 5014) and Metal (score of 529) benchmarks a second time, allowing some time in between tests, but the results did not improve. That's the power of integrated graphics! Even hardware video decoding makes the fans spin like crazy.

 

[FlyingDutch]

MacBook Pro 13" (2019)

Single Core: 985
Multi Core: 4254

Not too bad...

 

[smbu2000]

Tried out the CPU test on my 2018 15" MBP
2.9GHz i9 (6 core)/32GB/560X/2TB

Single Core: 1259
Multi Core: 5847

https://browser.geekbench.com/v5/cpu/38315

 

[tpivette89]

Mid-2012 15" 2.3ghz i7 with 16GB of RAM



For some reason, the RAM frequency is only half of what is actually installed.

 

[TheRealAlex]

Sad to see anyone not having like a 2018 MacBook Pro is slower than an 2018 iPad Pro

My results


https://browser.geekbench.com/v5/cpu/55928

 

[_Kiki_]

Mid-2012 15" 2.3ghz i7 with 16GB of RAM

View attachment 856111

For some reason, the RAM frequency is only half of what is actually installed.

your ram speed is normal, it's a DDR (Double Data Rate) so 800MHz times 2, so you have 1600MHz

 

[PROFESS0R]

Mid 2019 15" MacBook Pro, 2.4Ghz 8-core i9, 32Gb Ram 4Tb SSD, Vega 20

Single Core 1151
Multi Core 6739
Open CL 24012
Metal 25157

And for some additional information, Blackmagic Design Disk Speed Test:

Write: 2746.6 MB/s
Read: 2664.9 MB/s

 

[genomez]

MacBook Pro (15-inch, 2019)
2.4 GHz Intel Core i9
32 GB 2400 MHz DDR4
Vega 20, 1TB

Single Core: 1281
Multi Core: 7275
Open CL: 24081
Metal: 24017

 

[jlgolson]

MacBook Pro (15-inch, 2019)
2.3 GHz Intel Core i9
32GB
Vega 20 4GB, 512GB

Single Core: 1071
Multi Core: 6801
Open CL: 25627
Metal: 24487

 

[cynics]

MacBook Pro (13-inch, 2018)
i5-8259U 2.3 GHz
8gb RAM
Intel Iris Plus 655 integrated graphics

Single Core: 884
Multi Core: 3902
OpenCL: 6490
Metal: 6354

Out of curiosity I check my old iMac too....interesting results...

iMac (27-inch, 2013)
i5-4670 3.4 GHz
32gb RAM
Nvidia Geforce GTX 775M 2gb VRAM

Single Core: 911
Multi Core: 2736
OpenCL: 3816
Metal: 10822

I ran OpenCL again just to verify Nvidia's poor performance with OpenCL back when the 775M was relevant.

 

[smbu2000]

Tried out the CPU test on my 2018 15" MBP
2.9GHz i9 (6 core)/32GB/560X/2TB

Single Core: 1259
Multi Core: 5847

https://browser.geekbench.com/v5/cpu/38315

Same 2018 15" i9 MBP listed above.
Tested a bit more and had a slight improvement on Multi Core: 5897

https://browser.geekbench.com/v5/cpu/95150

Also tested using Win10 and Single Core was generally the same (highest score was 1279), but Multi Core was slightly higher on average. My highest Multi Core score was 6080.

https://browser.geekbench.com/v5/cpu/94791

 

[venom600]

iMac (27-inch, 2013)
i5-4670 3.4 GHz
32gb RAM
Nvidia Geforce GTX 775M 2gb VRAM

Single Core: 911
Multi Core: 2736

That doesn’t make sense. The multi core score should be roughly equal to the single core score x however many cores the processor has. 2736 sounds like a three core processor, not a quad core. Do you get that result after retesting?

 

[nylon]

MacBook Pro (15-inch 2016)
Intel Core i7-6700HQ 2.6 GHz
16gb RAM
Radeon Pro 460 (4GB) & Intel HD 530

CPU:
Single core: 842
Multi core: 3365

Radeon 460 (4GB):
Metal: 13634
OpenCL:14097

Intel HD 530:
Metal: 4268
OpenCL: 4518

 

[smbu2000]

That doesn’t make sense. The multi core score should be roughly equal to the single core score x however many cores the processor has. 2736 sounds like a three core processor, not a quad core. Do you get that result after retesting?

That would be it with near perfect scaling/no throttling, but that doesn't happen.

My 6 core 2018 15" has a single core score of 1258, multiplying that by the number of cores (6) then I should have a multi score of 7548, but in reality my multi core score is 5939, which would mean I have a 4.7 core CPU only looking at my single core score for reference.

My 12 core 2013 Mac Pro had a 776 single core score on the 1 run I ran on it. Multiplying by 12, then I should have a multi score of 9312, but my actual score was 7663, which would be just under 10 cores (9.8) with perfect scaling.

 

[cynics]

That doesn’t make sense. The multi core score should be roughly equal to the single core score x however many cores the processor has. 2736 sounds like a three core processor, not a quad core. Do you get that result after retesting?

That would be it with near perfect scaling/no throttling, but that doesn't happen.

My 6 core 2018 15" has a single core score of 1258, multiplying that by the number of cores (6) then I should have a multi score of 7548, but in reality my multi core score is 5939, which would mean I have a 4.7 core CPU only looking at my single core score for reference.

My 12 core 2013 Mac Pro had a 776 single core score on the 1 run I ran on it. Multiplying by 12, then I should have a multi score of 9312, but my actual score was 7663, which would be just under 10 cores (9.8) with perfect scaling.

It's a combination of two things.

Amdahl’s law highlighting how the individual test in the benchmark aren't (or can't be) parallelized perfectly across all cores. Cryptography and deep learning/machine learning benchmarks won't parallel compute well where as compressions and rendering test typically will. You can actually look at the Geekbench 5 results and compare individual single core test to their multi core counterpart to see what can leverage more cores better.

Encryption single and multi core score

Compression single and multi core score

The i5-4670 specifically is older by todays standards and lacks instruction sets for acceleration for certain task. This is why my i5-8259U in my MBP has a similar single core score using a lower frequency and less power.

Throttling is a always a concern of course however with the i5-4670 I'm hitting computational limits well before power and thermal limits of 84w (max is around 45-50w) and 90-100c (max is around 65-70c).

Like most benchmarks the results are factored from individual specific benchmarks. While Geekbench claims....

That doesn't mean those individual task are optimized as well as a program that utilizes them. But mostly it means that higher overall results doesn't make the best CPU for your specific application. An Intel Core series CPU that is faster than an Intel Xeon CPU in benchmarks doesn't mean much as both will shine at specific things.