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.
Geekbench 5 Released With Improved Benchmark Tests, Dark Mode Support, and More
- Thread starter MacRumors
- Start date
- Sort by reaction score
Strange. Geekbench 5 shows much lower benchmark scores than Geekbench 4. I even ran it 2x. Ideas?
Attachments
Last edited:
Samsung will probably “tune” it so that their phones show higher scores in benchmarks as they always seem to do.
An unbelievably low score with Geekbench 5 compared to 4. Unless that’s how it’s supposed to be now.
That can’t tune (cheat) enough to make up for that much of a drop. It just shows the superiority of Apple architecture compared to Samsung.
didn't the article say that they changed the memory algorithm to better reflect memory performance? hence, not directly comparable
According to other comments, looks like the scores in one version aren't comparable to scores in the other version. A Geekbench 4 score of 1000 ≠a Geekbench 5 score of 1000. That's what I don't like about the general Geekbench scores: the scores mean nothing without proper context. With other benchmarks, you have a definite score like "X MB/s transfer speed" or "Y frames/sec)." That's way more meaningful to users than a simple number score in Geekbench. However, Geekbench does have a breakdown of scores with more meaningful things.
Edited for clarity
Last edited:
As with any benchmark, the scores are only comparable using the same version. It’s wise of them to use a noticeably different scoring system, since it prevents confusion between Geekbench 4 and 5 results.
I would really like them to explain some issues like the difference in benchmark scores between windows and Linux on the same configurations, why are Linux scores almost always higher? is this an OS difference?, if so are the benchmark programs properly calibrated? Is Linux more easily tweaked to get better scores? Are the Linux programmers better than the windows guys and hence more efficient code? Why do two CPU bound tests score differently on different OSs?
Why are the memory speeds not reported correctly for a lot of machines?
Finally, can we lose the outliers on tests that are meaningless? No one really believes an i9 scores 1000? Not sure what the people are doing when they are running Geekbench, but the crappy scores really skew averages and should be eliminated from the averaging programs. Hey I learned that in statistics, not just making it up
[doublepost=1567521308][/doublepost]
LOL, have you ever looked at the scores? try this link: http://browser.geekbench.com/v5/cpu/7965
also, it seems folks failed to pickup on the change of baseline. Geekbench 4 uses a baseline of 4000 for a particular CPU, Geekbench 5 yes a baseline of 1000 for a Intel Core i3-8100
Why are the memory speeds not reported correctly for a lot of machines?
Finally, can we lose the outliers on tests that are meaningless? No one really believes an i9 scores 1000? Not sure what the people are doing when they are running Geekbench, but the crappy scores really skew averages and should be eliminated from the averaging programs. Hey I learned that in statistics, not just making it up
[doublepost=1567521308][/doublepost]
LOL, have you ever looked at the scores? try this link: http://browser.geekbench.com/v5/cpu/7965
also, it seems folks failed to pickup on the change of baseline. Geekbench 4 uses a baseline of 4000 for a particular CPU, Geekbench 5 yes a baseline of 1000 for a Intel Core i3-8100
Last edited:
Are you sure it is version 5 not 4? Free as well. Czech Republic.
iPhone 5s 32 GB:
GB 5 222/390 lol
GB 4 usually 1400/2500
But ratio single/multi is the same 0,56-7
And I can see why Apple cut iPhone 6 from iOS 13. A8 Metal scores is 4-5 times lower then A9.
https://browser.geekbench.com/ios-benchmarks/
Last edited:
If you do a detail search on a computer, the memory clocks are way off. if you look at a Mac with a core i9 http://browser.geekbench.com/v5/cpu/8270. you say the memory is supposedly clocked at 1200 MHZ? That is either misreported, or showing something other than the clock speed
Same, it’s strange
Attachments
I received the same kind of results. 5 shows much lower. Also noticed that the Battery test is gone
Last edited:
Many other popular computing benchmarks has the same problems you complained. Cinebench also use arbitrary unit “cb” to represent the final score, and those “cb” aren’t exactly the same between version R15 and newer R20.
You can easily extract useful informations and matrix from Geekbench by looking at and comparing between devices using its subcategories.
So how do we compare older results with the ones that come out of version 5?
If that can’t be done, then that would make it pretty much useless for me.
Edit: Ok. Right. We can’t and thus is it useless for me.
3dmark for example. And No, geekbench does not work on multiple platforms with the same software. There are simply 2 different versions.
Just so much BS. Apple uses the ARM aritecteture for it's processors. Yes they give them sexy names (A-series!, Bionic-series!. Matrix-series!) but all Apple does is take a reference design (an ARM chip that has to do everything) and REMOVE the parts that Apple has no business in.
BTW "compared to Samsung" it's a chip everbody can use; it's f@@k all to do with Samsung
