Contribute
Register

Geekbench 6 released

Z370 AORUS Gaming 3 (i5-9600 & RX 580)
Single-core: 1518
Multi-core: 5856
Vulcan: 47772
OpenCL: 44511

‘MacPro7,1’ (i5-9600 & RX 580)
Single-core: 1464
Multi-core: 5698
Metal: 47013
OpenCL: 42705

GA-EP43-DS3R (Q8400 & HD 7850)
Single-core: 407
Multi-core: 1117
Vulcan: 17872
OpenCL: 19549

‘MacPro6,1’ (Q8400 & HD 7850)
Single-core: 401
Multi-core: 1004
Metal: 22586
OpenCL: 18729

iPad mini (6th generation)
Single-core: 2134
Multi-core: 5365
Metal: 19591

iPhone XR
Single-core: 1292
Multi-core: 2042
Metal: 9245

iPad (6th generation)
Single-core: 844
Multi-core: 1060
Metal: 4544

MacBookPro9,2 (i5-3210M)
Single-core: 236
Multi-core: 778
Metal: *frozen
OpenCL: 532

MacBook4,1 (T8300)
Single-core: 328
Multi-core: 550
 
Last edited:
Overall, it looks like on GB6, the scores on single-core on Intel CPUs have gotten higher but score on multi-core is much worse while the Apple silicons score higher on both as compared to GB5. Never mind on this. Looks to be the case for MacBook Air tho.

Screenshot 2023-02-14 at 9.04.26 AM.png
 
Last edited:
Based on the GB6 home blurb, it reads that the workloads have been revamped, but the scale factors have not changed.

Unfortunately for users everywhere, details on what makes the workloads more meaningful (presumably) are not summarized.

If anyone comes across a good writeup, please post.

This update will lead to a lot of churn for GB Corp which is maybe good for them, but as they've released another instance with its own performance context this results in confusion for me (as an everyman) because my previous reference points of reference have just been reset, all too common for the computing industry.

So far GB versions have for me been informed by a thread of plausible ignorance at the GB Corp as to what constitutes a good general benchmark, which they were evolving and refining. They could wave their hands and that was ok with me.

But since 5, GB is so well established as a reference that any question of refinement must address specific details of workloads:

Why are the new metrics more relevant?!

They may well be. But for casual comparisons going from 5 -> 6 just raises the curtain on how subjective the matter.

If the scores change without clear understanding of why, a benchmark is useless. The point is that it's supposed to stay the same so you can see what else is changing.
Well, GB is the defacto general benchmark!

"I just upgraded my system: I went from GB5 to GB6 and overall performance went up 20%"

or

"ZOMG I just switched to GB6 and my system is 20% slower, what do I do?"

It reminds me of when Apple switched from storage reporting from powers of 2 units, GiB, to powers of ten units, GB, and had to try to explain why data usage suddenly got bigger.

1 GB means 1,000,000,000 bytes
1 GiB means 1,073,741,824 bytes

Apple had been using the suffix of "GB" but counting in GiB.

So if you had 1,073,741,824 bytes of files on your drive, which was reported as 1 GB, after a macOS update the same usage was reported as 1.07 GB. Hey you're short-changing me on drives! A 1 TB drive holds less than a 1 TiB drive.

I hate computers

This sort of thing has been an ongoing point of consumer legal activism. In 80s there were class action lawsuits over "missing inches" in TV screens that were advertised as 25 inches diagonal but actually measured as 24.5.

I've heard that the speed of light is subject to occasional revision even after it was discovered to be a universal constant, so there's no end to grief.
 
The creator of GB in his own words


Q: The scores from Geekbench 6 can’t be compared to the scores from Geekbench 5 since it’s a completely different benchmark. What about when it comes to versions like Geekbench 5.1 and 5.2? Are the scores always comparable?

A:
In the past, 3.0 wasn’t comparable with 3.1, and 4.0 wasn’t comparable with 4.1. While we’re able to catch a lot of issues before the software is released, we do miss things and get feedback from people after the software is already live. We then take that feedback and fix the mistakes within the first month or two.

So whether Geekbench 6.0 will be comparable with 6.1 is hard to say right now, but the following versions like 6.2 and 6.3 should be comparable since we’re mainly adding support for new hardware.
 
Last edited:
NOTE—GB6 will take the time to run the test then refuse to show the results if it can't phone home.

Here's a compare for my build... Just ran both

i9-11900K / RX6600XT (Metal) / Ventura 13.2
—GB5:
1844 - 11382 - 87393
—GB6: 2095 - 10721 - 118479

(screenshots)


GB5:

GB5 i9-11900K CPU.jpg GB5 i9-11900K RX6600XT Metal.jpg




GB6:

GB6 i9-11900K CPU.jpg GB6 i9-11900K RX6600XT Metal.jpg
 
Last edited:
I had a bunch of stuff running when I ran the tests...

Screenshot 2023-02-14 at 6.04.57 PM.png

Screenshot 2023-02-14 at 6.04.39 PM.png
 
I had a bunch of stuff running when I ran the tests...

Interesting to see the big bump of GB6 single score for AppleSi single over Intel where the was pretty much score parity in GB5, as this the most criritcal category for marketing.

Read the origin story of PrimateLabs founder and think about how bread is buttered...

My key takeaway is that (in so many words) //PrimateLabs chose new workloads that represent what Apple customers tend to do with their devices...// -and- //We have no real interest in metric fidelity across versions//. In fact the CEO prefers to reserve the right to dork around with metrics behind the curtain in the future.

My feeling is that Geekbench is debasing itself as a benchmark and becoming kit for Apple's industry segment.

I also have no doubt that AppleSi does well in the selected workloads...You don't invest a gadjillion bucks in fabrication of exotic innards for a trillion$ personal computer market without doing diligence on the design. So please don't misunderstand my gripe... I am saying that the tool is being adjusted to please a certain segment of the industry, where the CEO is forthcoming and pleased to represent this orientation.

But for me it comes down to a tool that had a spitting chance to be a general bench-marking tool is becoming a validator of preexisting design assumptions.

I don't mean to underplay what a complex subject this... But PrimateLabs itself is on record with its stance.

I have a big bine to pick, because this evolution is playing out the same way that U.S. MSM produces articles that only question world affairs from perspective of their industrial interest groups.

I have a lot more to say about the problematics of the GB workload orientation, but that's for another post.
 
Back
Top